quantgap 0.1.0__py3-none-any.whl

quantgap/probes/data/probes_v1.jsonl

@@ -0,0 +1,160 @@
+{"seq_id": 0, "group": "general", "kind": "text", "lang": "en", "text": "The Androphagi were an ancient Scythian tribe whose existence was recorded by ancient Greco-Roman authors.\n\nThe Androphagi were closely related to the Melanchlaeni and the Budini.\n\nName\nThe name  is a Latinisation of the ancient Greek name  (), which meant \"Man-Eaters.\" This name is a descriptive one based on this tribe's practice of cannibalism, and their own tribal name is unknown.\n\nLocation\n\nThe Androphagi lived in the region to the east of the middle Dnipro river, especially in the valley of the Sula and some smaller rivers.\n\nThe neighbours of the Androphagi were the Neuri to the west and the Scythians to the south.\n\nHistory\n\nOrigin\nThe Scythians originated in the region of the Volga-Ural steppes of Central Asia, possibly around the 9th century BC, as a section of the population of the Srubnaya culture containing a significant element originating from the Siberian Andronovo culture. The population of the Srubnaya culture culture was among the first truly nomadic pastoralist groups, who themselves emerged in the Central Asian and Siberian steppes during the 9th century BC as a result of the cold and dry climate then prevailing in these regions.\n\nDuring the 9th to 8th centuries BC, a significant movement of the nomadic peoples of the Eurasian Steppe started when another nomadic Iranic tribe closely related to the Scythians from eastern Central Asia, either the Massagetae or the Issedones, migrated westwards, forcing the early Scythians to the west across the Araxes river, following which some Scythian tribes had migrated westwards into the steppe adjacent to the shores of the Black Sea, which they occupied along with the Cimmerians, who were also a nomadic Iranic people closely related to the Scythians.", "messages": null, "license": "CC BY-SA 4.0", "source": "wikipedia:en:Androphagi", "greedy": true, "greedy_prompt": null}
+{"seq_id": 1, "group": "general", "kind": "text", "lang": "en", "text": "The Staatliches Bauhaus (), commonly known as the , was a German art school operational from 1919 to 1933 that combined crafts and the fine arts. The school became famous for its approach to design, which attempted to unify individual artistic vision with the principles of mass production and emphasis on function. Along with the doctrine of functionalism, the Bauhaus initiated the conceptual understanding of architecture and design.\n\nThe Bauhaus was founded by architect Walter Gropius in Weimar. It was grounded in the idea of creating a Gesamtkunstwerk (\"comprehensive artwork\") in which all the arts would eventually be brought together. The Bauhaus style later became one of the most influential currents in modern design, modernist architecture, and architectural education. The Bauhaus movement had a profound influence on subsequent developments in art, architecture, graphic design, interior design, industrial design, and typography. Staff at the Bauhaus included prominent artists such as Paul Klee, Wassily Kandinsky, and László Moholy-Nagy at various points.\n\nThe school existed in three German cities—Weimar, from 1919 to 1925; Dessau, from 1925 to 1932; and Berlin, from 1932 to 1933—under three different architect-directors: Walter Gropius from 1919 to 1928; Hannes Meyer from 1928 to 1930; and Ludwig Mies van der Rohe from 1930 until 1933, when the school was closed by its own leadership under pressure from the Nazi regime, having been painted as a centre of communist intellectualism. Internationally, former key figures of Bauhaus were successful in the United States and became known as the avant-garde for the International Style.\n\nThe changes of venue and leadership resulted in a constant shifting of focus, technique, instructors, and politics. For example, the pottery shop was discontinued when the school moved from Weimar to Dessau, even though it had been an important revenue source; when Mies van der Rohe took over the school in 1930, he transformed it into a private school and would not allow any supporters of Hannes Meyer to attend it.\n\nTerm and concept", "messages": null, "license": "CC BY-SA 4.0", "source": "wikipedia:en:Bauhaus", "greedy": true, "greedy_prompt": null}
+{"seq_id": 2, "group": "general", "kind": "text", "lang": "en", "text": "Bacterial conjugation is the transfer of genetic material between bacterial cells by direct cell-to-cell contact or by a bridge-like connection between two cells. This takes place through a pilus. It is a parasexual mode of reproduction in bacteria.\n\nIt is a mechanism of horizontal gene transfer as are transformation and transduction although these two other mechanisms do not involve cell-to-cell contact.\n\nClassical E. coli bacterial conjugation is often regarded as the bacterial equivalent of sexual reproduction or mating since it involves the exchange of genetic material. However, it is not sexual reproduction, since no exchange of gamete occurs, and indeed no generation of a new organism: instead an existing organism is transformed. During classical E. coli conjugation the donor cell provides a conjugative or mobilizable genetic element that is most often a plasmid or transposon. Most conjugative plasmids have systems ensuring that the recipient cell does not already contain a similar element.\n\nThe genetic information transferred is often beneficial to the recipient. Benefits may include antibiotic resistance, xenobiotic tolerance or the ability to use new metabolites. Other elements can be detrimental and may be viewed as bacterial parasites.\n\nConjugation in Escherichia coli by spontaneous zygogenesis and in Mycobacterium smegmatis by distributive conjugal transfer differ from the better studied classical E. coli conjugation in that these cases involve substantial blending of the parental genomes.\n\nHistory\nThe process was discovered by Joshua Lederberg and Edward Tatum in 1946.\n\nMechanism \n\nConjugation diagram\n\n Donor cell produces pilus.\n Pilus attaches to recipient cell and brings the two cells together.\n The mobile plasmid is nicked and a single strand of DNA is then transferred to the recipient cell.\n Both cells synthesize a complementary strand to produce a double stranded circular plasmid and also reproduce pili; both cells are now viable donor for the F-factor.", "messages": null, "license": "CC BY-SA 4.0", "source": "wikipedia:en:Bacterial conjugation", "greedy": true, "greedy_prompt": null}
+{"seq_id": 3, "group": "general", "kind": "text", "lang": "en", "text": "The zebrafish (Danio rerio) is a freshwater fish belonging to the minnow family (Cyprinidae) of the order Cypriniformes. Native to India  and South Asia, it is a popular aquarium fish, frequently sold under the trade name zebra danio (and thus often called a \"tropical fish\" although both tropical and subtropical). It is also found in private ponds.\n\nThe zebrafish is an important and widely used vertebrate model organism in scientific research. Zebrafish has been used for biomedicine and developmental biology. The species is used for studies, such as neurobehavioral phenomena. It is also used for psychological reasons such as abuse, cognitive, and affective disorders. The species are used to study and observe behavioral research.\n\nTaxonomy \nThe zebrafish is a derived member of the genus Brachydanio, of the family Cyprinidae. It has a sister-group relationship with Danio aesculapii. Zebrafish are also closely related to the genus Devario, as demonstrated by a phylogenetic tree of close species.\n\nDistribution\n\nRange \nThe zebrafish is native to freshwater habitats in South Asia where it is found in India, Pakistan, Bangladesh, Nepal and Bhutan. The northern limit is in the South Himalayas, ranging from the Sutlej river basin in the Pakistan–India border region to the state of Arunachal Pradesh in northeast Indian. Its range is concentrated in the Ganges and Brahmaputra River basins, and the species was first described from Kosi River (lower Ganges basin) of India. Its range further south is more local, with scattered records from the Western and Eastern Ghats regions. It has frequently been said to occur in Myanmar (Burma), but this is entirely based on pre-1930 records and likely refers to close relatives only described later, notably Danio kyathit. Likewise, old records from Sri Lanka are highly questionable and remain unconfirmed.", "messages": null, "license": "CC BY-SA 4.0", "source": "wikipedia:en:Zebrafish", "greedy": true, "greedy_prompt": null}
+{"seq_id": 4, "group": "general", "kind": "text", "lang": "en", "text": "Çevik Bir (born 1939) is a retired Turkish army general. He was a member of the Turkish General Staff in the 1990s. He took a major part in several important international missions in the Middle East and North Africa. He was born in Buca, Izmir Province, in 1939 and is married with one child.\n\nHe graduated from the Turkish Military Academy as an engineer officer in 1958, from the Army Staff College in 1970 and from the Armed Forces College in 1971. He graduated from NATO Defense College, Rome, Italy in 1973.\n\nFrom 1983 to 1985, he served at SHAPE, NATO's headquarters in Belgium. He was promoted to brigadier general and commanded an armed brigade and division in Turkey. From 1987 to 1991, he served as major general, and then was promoted to lieutenant general.\n\nAfter the dictator Siad Barre’s ousting, conflicts between the General Mohammed Farah Aidid party and other clans in Somalia had led to famine and lawlessness throughout the country. An estimated 300,000 people had died from starvation. A combined military force of United States and United Nations (under the name \"UNOSOM\") were deployed to Mogadishu, to monitor the ceasefire and deliver food and supplies to the starving people of Somali. Çevik Bir, who was then a lieutenant-general of Turkey, became the force commander of UNOSOM II in April 1993. Despite the retreat of US and UN forces after several deaths due to local hostilities mainly led by Aidid, the introduction of a powerful military force opened the transportation routes, enabling the provision of supplies and ended the famine quickly. He was succeeded as Force Commander by a Malaysian general in January 1994.\n\nHe became a four-star general and served three years as vice chairman of the Turkish Armed Forces, then appointed commander of the Turkish First Army, in Istanbul. While he was vice chairman of the TAF, he signed the Turkish-Israeli Military Coordination agreement in 1996.", "messages": null, "license": "CC BY-SA 4.0", "source": "wikipedia:en:Çevik Bir", "greedy": true, "greedy_prompt": null}
+{"seq_id": 5, "group": "general", "kind": "text", "lang": "en", "text": "Outside TV (formerly RSN Television) is a sports-oriented cable and satellite television network based on Outside magazine. The network features programming related to various outdoor activities and the lives of those who engage in them. High-definition programs appear on the company's cable, satellite and broadband providers’ sports and entertainment offerings.\n\nHistory\nOutside TV was the result of a complete re-branding of the existing Resort Sports Network, the national television network that specialized in creating and distributing outdoor-lifestyle content to premier vacation destinations throughout the country.\n\nAs of June 2010, Outside TV was in 110 resort markets representing 61 million potential viewers.\n\nOutside TV has a corporate office in Westport, Connecticut, and a main office in Portland, Maine.  Its sales office is in the Graybar Building at 420 Lexington in New York City.\n\nOutside TV was founded by publisher Lawrence Burke and founding executive producer and executive vice president Les Guthman in 1994.  Over the next ten years, it produced the Outside Television Presents TV series, whose production Farther Than the Eye Can See, captured blind climber Erick Weihenmayer's historic ascension to the summit of Mount Everest.  Into the Tsangpo Gorge, produced by director and expedition leader Scott Lindgren, achieved the first whitewater descent of through the 18,000-ft.-deep Tsangpo Gorge (Yarlung Tsangpo Grand Canyon) in Tibet.  Into the Tsangpo Gorge aired on NBC Sports in May 2002 and was Outside Magazine'''s cover story in July 2002.\n\nIn July 2013, Outside TV entered into a new multi-year distribution agreement with the National Cable Television Cooperative (NCTC), representing more than 950 different cable providers and thousands of local systems nationwide. The addition of NCTC to Outside Television's other core distribution partners such as Comcast Xfinity makes the independent network available to more than 40 million homes.", "messages": null, "license": "CC BY-SA 4.0", "source": "wikipedia:en:Outside TV", "greedy": true, "greedy_prompt": null}
+{"seq_id": 6, "group": "general", "kind": "text", "lang": "en", "text": "Belle Haven Consultants was a for-profit organization founded in 1997 by former Heritage Foundation president Edwin Feulner and Heritage Foundation Asia policy expert Ken Sheffer. Feulner's wife, Linda Feulner, later took her husband's place as a partner until 2001, when she became a paid senior adviser in the firm.\n\nHistory\n\nBelle Haven, The Heritage Foundation, and the Alexander Strategy Group (ASG) shared the same office in Hong Kong, located in Suite 401 of the Baskerville House office building in Central Hong Kong. Belle Haven was a subcontractor, then was purchased by ASG partner Edward Stewart, co-owner of the firm along with long-time associate Beth Allison Cave. In 2006, ASG closed as a result of the Jack Abramoff lobbying scandals.\n\nBy the end of 2001, Belle Haven hired ASG for help \"promoting and advocating Malaysia's positive investment climate and business opportunities\" in connection with a company called PK Baru Energy. A new group called the US-Malaysia Exchange Association also hired ASG for support \"enhancing the bilateral relationship between Malaysia and the US.\"  Megat Junid, an associate of then-prime minister Mahathir Mohammed, said in a 2004 interview that he organized Malaysia Exchange after talks with Edwin Feulner.\n\nIn 2001, House majority leader Tom DeLay and three other congressmen traveled to Malaysia with their spouses on a trip officially sponsored by Heritage. Heritage senior fellow and former U.S. Senator Malcolm Wallop, who went on the trip, told Time magazine that Belle Haven's financial involvement was more important to the trip than Heritage's.  In the following months, more congressmen made their way to Kuala Lumpur, the capital of Malaysia and senior Malaysian officials began beating a path to Washington, D.C., an interchange that climaxed with Mahathir visiting the White House in May 2002, which was his first state visit in eight years. Though in past years Heritage had been publicly critical of Mahathir, Feulner hosted a dinner reception on that visit to honor the prime minister.", "messages": null, "license": "CC BY-SA 4.0", "source": "wikipedia:en:Belle Haven Consultants", "greedy": true, "greedy_prompt": null}
+{"seq_id": 7, "group": "general", "kind": "text", "lang": "en", "text": "The Royal Air Force's Logistics Command was a command formed to provide logistics support for the RAF.\n\nHistory\nThe Command was formed on 1 April 1994 and its role was to provide logistics support to the RAF. The formation of Logistics Command resulted from the Government's PROSPECT study which was aimed to achieve a 20% reduction in the UK armed forces' headquarters staff to match the previous 'Options for Change' front-line cuts.  This provided the mandate to create a centre of excellence in logistics management within the RAF with the task of delivering the best standards of support for the front-line whilst at the same time achieving significant reductions in cost.\n\nThe Command brought together most of the logistics functions of the Air Member for Supply and Organisation with those of RAF Support Command.  These included the Maintenance Group Defence Agency, Support Command Communication and Information Systems, the Radio Introduction Unit, and the Central Servicing Development Establishment.  Logistics Command was headquartered across two sites, RAF Brampton and RAF Wyton; on the later base 4 purpose-built open-planned pavilions were built.  Its motto was Sustentamus ut Bellent which means 'We sustain that they may fight'.\n\nThe Strategic Defence Review (SDR) in 1997 marked the beginning of a process of radical and far-reaching modernization of the way the Armed Forces conducted defence activities including the creation of unified logistics support.  Many of the RAF's innovations in Logistics Command were incorporated into the SDR's 'Smart Procurement Initiative'.\n\nLogistics Command was disbanded on 31 October 1999 and thereafter the majority of its functions were subsumed by the tri-Services Defence Logistics Organisation which stood up formally on 1 April 2000.\n\nAir Officers Commanding-in-Chief\nAir Officers Commanding-in-Chief were:\n 1 April 1994 – 8 March 1996 Air Chief Marshal Sir Michael Alcock GCB KBE DSc FEng FIMechE FRAeS\n 8 March 1996 – 11 July 1997 Air Chief Marshal Sir John Allison KCB CBE FRAeS\n 11 July 1997 – 30 April 1999 Air Marshal Sir Colin Terry KBE CB BSc(Eng) FRAeS FRSA FILog FCGI", "messages": null, "license": "CC BY-SA 4.0", "source": "wikipedia:en:RAF Logistics Command", "greedy": true, "greedy_prompt": null}
+{"seq_id": 8, "group": "general", "kind": "text", "lang": "en", "text": "Amanda Marshall is the 1995 debut album by Canadian singer Amanda Marshall. The album peaked at number four on the RPM Albums Chart and has also been certified Diamond by the CRIA with over 1,000,000 copies sold in Canada, making it Marshall's best-selling album of her career. In the United States, the album charted at number 156 on the Billboard 200 and sold over 350,000 copies. It was particularly successful in Norway, where it reached number one and received a Platinum certification. The album sold 2 million copies worldwide.\n\nSeven songs from the album were released as singles: \"Let It Rain\", \"Birmingham\", \"Fall from Grace\", \"Beautiful Goodbye\", \"Dark Horse\", \"Sitting on Top of the World\", and \"Trust Me (This Is Love)\". \"Birmingham\" is Marshall's highest-charting song in Canada, peaking at number three. \"Fall from Grace\", \"Dark Horse\", and \"Sitting on Top of the World\" all reached the top five. \"Let It Rain\" is her most successful international hit, reaching number three in Norway and the top 30 in Australia and New Zealand; it was certified Gold in Norway.\n\nTrack listing\n \"Let It Rain\" (Kristen Hall) – 4:33\n \"Birmingham\" (Dean McTaggart, David Tyson, Gerald O'Brien) – 5:21\n \"Fall from Grace\" (Marc Jordan, Kim Bullard) – 4:20\n \"Dark Horse\" (Amanda Marshall, David Tyson, Dean McTaggart) – 5:37\n \"Beautiful Goodbye\" (David Tyson, Chistopher Ward) – 5:17\n \"Sitting on Top of the World\" (Amanda Marshall) – 4:19\n \"Last Exit to Eden\" (Dean McTaggart, David Tyson) – 5:24\n \"Trust Me (This is Love)\" (Dean McTaggart, David Tyson) – 4:59\n \"Let's Get Lost\" (Amanda Marshall, Christopher Ward) – 4:14\n \"Promises\" (John Capek, Marc Jordan) – 5:22\n\n2023 Deluxe Remastered Edition bonus tracks\n\"Don't Let It Bring You Down\" (Neil Young) – 3:35\n\"This Could Take All Night\" (Linda Thompson, Steve Dorff) – 3:09\n\"Birmingham (Live Acoustic)\" (Dean McTaggart, David Tyson, Gerald O'Brien) – 4:33\n\nPersonnel\n Amanda Marshall – vocals, background vocals\n Kenny Aronoff – drums, percussion\n Leland Sklar – bass guitar\n Tommy Byrnes – electric guitars\n David Wipper – acoustic guitars, mandolin\n Bob Mann – acoustic guitar\n Diana DeWitt – background vocals", "messages": null, "license": "CC BY-SA 4.0", "source": "wikipedia:en:Amanda Marshall (album)", "greedy": false, "greedy_prompt": null}
+{"seq_id": 9, "group": "general", "kind": "text", "lang": "en", "text": "Bruce Maccabee (born May 6, 1942) is an American optical physicist formerly employed by the U.S. Navy, and a  ufologist.\n\nBiography \nMaccabee received a B.S. in physics at Worcester Polytechnic Institute in Worcester, Mass., and then at American University, Washington, DC, (M.S. and Ph.D. in physics).  In 1972 he began his career at the Naval Ordnance Laboratory, White Oak, Silver Spring, Maryland; which later became the Naval Surface Warfare Center Dahlgren Division.  Maccabee retired from government service in 2008.  He has worked on optical data processing, generation of underwater sound with lasers and various aspects of the Strategic Defense Initiative (SDI) and Ballistic Missile Defense (BMD) using high power lasers.\n\nUfology \nMaccabee has been interested in UFOs since the late 1960s when he joined the National Investigations Committee on Aerial Phenomena (NICAP) and was active in research and investigation for NICAP until its demise in 1980. He became a member of the Mutual UFO Network (MUFON) in 1975 and was subsequently appointed to the position of state director for Maryland, a position he still holds. In 1979 he was instrumental in establishing the Fund for UFO Research (FUFOR) and was the chairman for about 13 years. He presently serves on the National Board of the Fund.\n\nHis UFO research and investigations (which, he often stresses, are completely unrelated to his Navy work) have included the Kenneth Arnold sighting (June 24, 1947), the McMinnville, Oregon (Trent) photos of 1950, the Gemini 11 astronaut photos of September, 1966, the Tehran UFO incident of September 1976, the New Zealand sightings of December 1978, the Japan Airlines (JAL1628) sighting of November 1986, the numerous sightings of Gulf Breeze UFO incident, 1987–1988, the \"red bubba\" sightings, 1990-1992 (including his own sighting in September, 1991), the Mexico City video of August, 1997 (which he deemed a hoax), the Phoenix lights sightings of March 13, 1997, 2004 Mexican UFO incident and many others.", "messages": null, "license": "CC BY-SA 4.0", "source": "wikipedia:en:Bruce Maccabee", "greedy": false, "greedy_prompt": null}
+{"seq_id": 10, "group": "general", "kind": "text", "lang": "en", "text": "Hollie Smith (born 17 November 1982) is a New Zealand soul singer-songwriter based in Auckland, New Zealand. Her four solo albums Long Player, Humour and the Misfortune of Others, Water or Gold, and Coming In From The Dark have all reached number one on the RIANZ albums chart, making her one of the most successful female New Zealand artists of the 21st century.\n\nEarly years\nSmith attended Auckland's Willow Park Primary School, Takapuna Normal Intermediate and Rangitoto College.\nIn 1999, as a 16-year-old, Smith made the album Light From a Distant Shore after winning Best Female Vocalist at the National Jazz Festival of NZ. This album of Celtic music was produced by her stepfather, Steve McDonald. One of these early songs with McDonald, featuring Smith, would eventually be sampled for a track by US rap artist DMX for his album Year Of The Dog... Again.\n\nCareer\nIn 2003, Smith moved to Wellington singing with TrinityRoots. She recorded an album Home, Land and Sea and toured with the band. She also featured on Fat Freddy’s Drop's track ‘Hope’ and performed on Concord Dawn and Recloose’ albums.\nIn 2005, the 5-track Hollie Smith EP was released, featuring her own compositions. In 2005, the single \"Bathe in the River\", written by Don McGlashan and featuring Smith, was released. The song, from the soundtrack of the New Zealand feature film, No. 2, was a significant hit, spending 22 weeks in the NZ Top 10 and went on to win the APRA Silver Scroll.\n\nIn 2005, Smith signed a multi-album record deal with Manhattan Records, part of the Blue Note Label Group. Hollie would then record in New York City with producer and keyboard legend James Poyser (The Roots), Adam Blackstone (Kanye West, Justin Timberlake) and Jeff Lee Johnson (Erykah Badu, Aretha Franklin). The deal then collapsed upon the sale of Manhattan Records/EMI to Universal, which forced the singer to use the profits of her first album release, Long Player, released in 2006, debuting at #1 and going on to sell Gold, to get out of the deal with Blue Note. During this time she opened for artists Bob Dylan and Coldplay on the New Zealand legs of their tours.", "messages": null, "license": "CC BY-SA 4.0", "source": "wikipedia:en:Hollie Smith", "greedy": false, "greedy_prompt": null}
+{"seq_id": 11, "group": "general", "kind": "text", "lang": "en", "text": "Punnathurkotta  is a fort and former palace located in Kottapadi, about 3 km  from the Guruvayoor Sree Krishna Temple, in Thrissur District of Kerala State in South India.\n\nDescription\n\nPunnathurkotta was once the palace of a local ruler, but the palace grounds are now used to house the elephants belonging to the Guruvayoor temple, and has been renamed Anakkotta (meaning \"Elephant Fort\"). There were 86 elephants housed there, but currently there are about 59 elephants. The elephants are ritual offerings made by the devotees of Lord Guruvayurappa. \n\nDescribed as a 'Palace for Elephants,'  this facility is also used to train the elephants to serve Lord Krishna as well as to participate in many festivals that occur throughout the year. The oldest elephant is around 82 years of age and is called 'Ramachandran'. The rituals of Gajapooja (Worshipping Elephants) and Anayoottu (Feeding Elephants) are observed here, as an offering to Lord Ganesha. The legendary elephant \"Guruvayur Keshavan\" was housed here. \n\nThe compound also has a naalu kettu, a traditional rectangular home with a central courtyard, which belonged to the Punnathur Raja.  It presently houses a training school for Papans (Mahout). \n\nThis complex also contains a temple dedicated to Lord Shiva and Bhagavathy. Some scenes in the famous Malayalam movie \"Oru Vadakkan Veeragatha\" (starring Mammooty) were filmed at this location. The visiting hours are 9.00 AM to 5.00 PM. The entry fee is ₹ 20.00 per adult. An extra Rs 25 is charged for to use a camera inside the complex. Nowadays  camera usage is not allowed in Punnathurkotta.\n\nElephant Camp at Punnathurkotta", "messages": null, "license": "CC BY-SA 4.0", "source": "wikipedia:en:Punnathurkotta", "greedy": false, "greedy_prompt": null}
+{"seq_id": 12, "group": "general", "kind": "text", "lang": "en", "text": "\"Ricky\" is a 1983 song by \"Weird Al\" Yankovic, duetting with voice actress Tress MacNeille. It is a parody of the 1982 song \"Mickey\" by Toni Basil, which itself, is a cover of Mike Chapman and Nicky Chinn's \"Kitty\" recorded by Racey. The song focuses on the sitcom I Love Lucy, and ends with a segment of the theme from the sitcom.\n\nA music video was created for the song; it was Yankovic's first. According to the liner notes in the compilation DVD \"Weird Al\" Yankovic: The Ultimate Video Collection, \"it was arguably the first comedy video ever shown on MTV\".\n\nTrack listing\n\"Ricky\" – 2:35\n\"Buckingham Blues\" – 3:11\n\nMusic video\nThe music video was directed by Janet Greek. It was filmed mostly in black and white and parodies the television sitcom I Love Lucy. Yankovic plays the role of Ricky Ricardo, complete with affected Cuban accent. Yankovic appears, minus his mustache, glasses, and curly hair, to more closely resemble Desi Arnaz. He is also seen briefly in several scenes as himself with his band, playing the accordion and sporting his normal mustache and curly hair. Tress MacNeille plays the role of Lucy, giving an impression of Lucille Ball's somewhat raspy voice and her usual shticks such as her unique way of crying.\n\nNear the end of the video, a cheerleader can be seen dancing in the crowd. This is a direct reference to the music video for \"Mickey\".\n\nThe video (and song) ends with a segment of the I Love Lucy theme played on guitar rounding out with Ricky on accordion, followed by Ricky doing his trademark \"Huah! Huah! Huah!\" laugh. Dr. Demento makes a cameo appearance at the very end.\n\nChart positions\n\nSee also\nList of singles by \"Weird Al\" Yankovic\nList of songs by \"Weird Al\" Yankovic\n\nReferences\n\nExternal links\n, official upload of the music video\n\n1983 singles\n1983 songs\nI Love Lucy\nMale–female vocal duets\nScotti Brothers Records singles\nSongs about fictional male characters\nSongs about television\nSongs with lyrics by \"Weird Al\" Yankovic\nSongs written by Mike Chapman\nSongs written by Nicky Chinn\n\"Weird Al\" Yankovic songs", "messages": null, "license": "CC BY-SA 4.0", "source": "wikipedia:en:Ricky (song)", "greedy": false, "greedy_prompt": null}
+{"seq_id": 13, "group": "general", "kind": "text", "lang": "en", "text": "The Visitation is a 1999 contemporary Christian novel by Frank Peretti.  Taking place in the fictional wheat town of Antioch, located in eastern Washington, The Visitation is told in first-person narrative by the protagonist, a former minister named Travis Jordan who struggles to reconcile his former pastoral life with that of a present-day false Messiah.\n\nPlot summary\nCentered on the life of Travis Jordan, The Visitation begins when miracles, ranging from a healing, weeping crucifix to sights of Jesus in the clouds, start occurring, giving way to the arrival of a man who calls himself Brandon Nichols.  Nichols begins healing people; giving a man who lost the use of his legs in the Vietnam War the ability to walk, and performing various other \"healings\".  Most of the townspeople — who are portrayed as disillusioned, post-Pentecostal farmers — begin to believe in Nichols as a Messiah.\n\nBrandon Nichols begins to hold \"revival meetings\" on a large ranch outside of town every Sunday, and many churchgoers in town stop going to Sunday morning mass/services and instead listen to Brandon talk and watch him \"heal\".  It is at this point that Nichols arouses the ire of one of the local ministers, Kyle Sherman.  Enlisting the help of Travis Jordan, he seeks to prove that the so-called Brandon Nichols is not in fact a \"better\" Christian Messiah, but a puffed-up egomaniac using occult powers.\n\nIn the end, the team (along with the help of a few others) uncover a host of pseudonyms and a hefty helping of deception surrounding Nichols' past. Startling parallels are revealed with the life story of Travis Jordan, all of which come to light as the story progresses.\n\nParody\nNumerous \"in jokes\" are to be found within the book, most of them referencing some Pentecostal belief or practice.  For example, the entire life story of Travis Jordan is one of disillusionment with idyllic Charismaticism.  The entire book is rife with slightly humorous stories of \"leg-lengthening services\" and long nights of wailing in tongues.", "messages": null, "license": "CC BY-SA 4.0", "source": "wikipedia:en:The Visitation (novel)", "greedy": false, "greedy_prompt": null}
+{"seq_id": 14, "group": "general", "kind": "text", "lang": "en", "text": "Olga Talamante (born 1950) is a Chicana political activist and the past executive director of the California-based Chicana/Latina Foundation.\n\nEarly life\nTalamante was born in Mexicali, Baja California, Mexico, to Eduardo and Refugio Talamante. Her father was Mexican and her mother was a Mexican-American from Lompoc, California. At the age of eleven, she and her family moved to Gilroy, an agricultural community.\n\nShe learned English and excelled in her studies, being elected president of her sophomore class and vice president of her senior class at Gilroy High School. She became a naturalized U.S. citizen, and attended the University of California, Santa Cruz, graduating with a degree in Latin American studies.\n\nActivism\nThere she became active in the anti-Vietnam War peace movement and the Chicano Movement. While doing a field study in Chiapas, Mexico, she came into contact with many Argentines who told her of recent leftist political successes in their country. After graduating, Talamante decided to go to Argentina, and arrived shortly after the election of the Peronist Justicialist Party candidate for president Héctor José Cámpora. She arrived in Azul, Buenos Aires Province, and began working for Juventud Peronista, a poverty-relief agency, in one of the city's poorest sectors.\n\nAfter the death of Juan Perón, who had resumed control of the government of Argentina after Cámpora's resignation, the Peronist party split into left- and right-wing factions, with the conservatives supporting the government of Isabel Perón, which banned political assembly.\n\nImprisonment\nIn November 1974, Talamante was arrested in Azul, Argentina for political activity, and subsequently imprisoned and tortured.\n\nThe Olga Talamante Defense Committee petitioned the United States Congress and State Department for her release. By the time she was freed on March 27, 1976, Talamante had become nationally known. She returned to the San Francisco Bay Area, where she began working for the Argentine Commission for Human Rights.\n\nIn her capacity as a representative of the Argentine Commission for Human Rights, she was involved in the case of United States vs. Horacio Daniel Lofredo..", "messages": null, "license": "CC BY-SA 4.0", "source": "wikipedia:en:Olga Talamante", "greedy": false, "greedy_prompt": null}
+{"seq_id": 15, "group": "general", "kind": "text", "lang": "en", "text": "María de Salinas, Baroness Willoughby de Eresby (c. 1490 – 1539) was an English noblewoman and courtier from Spain. She was a confidante and lady-in-waiting to Catherine of Aragon, Queen of England.\n\nFamily background\nMaría was the daughter of Martín de Salinas (d. 1503), and Josefa González de Salas, who were members of the royal household in Castile, and probably related to the royal family. She had a younger sister, Ines, born in 1491, who died sometime after she did.\n\nAs a lady-in-waiting\nThe exact date that María became a maid-of-honour to Catherine of Aragon is unknown, but is thought to have been in 1501. María remained unmarried until 5 June 1516, when she married the English nobleman William Willoughby, 11th Baron Willoughby de Eresby, and they had one child, Catherine, who succeeded to her father's barony Grimsthorpe Castle was granted by King Henry VIII to the de Eresby family on the occasion of María's marriage. Henry esteemed María so much that in 1522 he named a ship HMS Mary Willoughby in her honour.\n\nIn 1511 María became the godmother to Mary Brandon, the daughter of Charles Brandon, 1st Duke of Suffolk and his second wife Anne Browne. Lord Willoughby died in 1526; the Duke of Suffolk made María's daughter, Catherine, his ward shortly thereafter, marrying her as his fourth wife in 1533.\n\nThe death of Catherine of Aragon\nMaría was a devoted friend to Catherine of Aragon. In August 1532, shortly before Catherine's marriage to Henry was annulled, she was ordered to leave Catherine's household and not to make any attempt to communicate with her. In September 1534, when Catherine's health deteriorated, María begged permission to visit her but was denied. On 5 January 1536 she forced her way into Kimbolton Castle to see Catherine, having again been refused permission to visit. Catherine died in her arms two days later.", "messages": null, "license": "CC BY-SA 4.0", "source": "wikipedia:en:María de Salinas", "greedy": false, "greedy_prompt": null}
+{"seq_id": 16, "group": "general", "kind": "text", "lang": "en", "text": "Jonathan Michael Gledhill (14 February 1949 – 1 November 2021) was an English Anglican clergyman. He was the Bishop of Southampton from 1996 to 2003, and the 98th Bishop of Lichfield from 2003 to 2015.\n\nEarly life\nGledhill was born on 14 February 1949, in Windsor, Berkshire, England. He was educated at Keele University and the University of Bristol. He trained for the ministry at Trinity College, Bristol. In 2007 he was awarded an honorary doctorate by Keele University in recognition of his \"outstanding contribution to the Church and to the people and the County of Staffordshire.\"\n\nOrdained ministry\nGledhill was curate at All Saints', Marple, (Diocese of Chester) from 1975 to 1978 and from 1978 to 1983 was priest in charge of St George's Folkestone. From 1983 to 1996 he was Vicar at St Mary Bredin, Canterbury and a tutor at the Canterbury School of Ministry. From 1988 to 1994 he was also the Rural Dean at Canterbury and from 1992 to 1996 an honorary canon of Canterbury Cathedral.\n\nFrom 1995 to 1998, Gledhill was a member of the General Synod. From 1996 to 2003 he was the suffragan Bishop of Southampton. From 1997 he chaired the Anglican-Old Catholic International Co-ordinating Council and was chairman of the Governing Body of the National College of Evangelists.\n\nGledhill was the 98th Bishop of Lichfield.  He was enthroned in Lichfield Cathedral on 15 November 2003 and he retired on 30 September 2015. Following his retirement he moved back to Canterbury where he lived and had permission to officiate in its Diocese.\n\nGledhill published Leading a Local Church in the Age of the Spirit.\n\nPersonal life\nGledhill's wife Jane, was a university lecturer and lay reader; they had one daughter and one son. \n\nHe announced he had Parkinson's disease, shortly before his retirement from ministry in 2015. Gledhill died on 1 November 2021, at the age of 72.\n\nStyles\n The Reverend Jonathan Gledhill (1975–1992)\n The Reverend Canon Jonathan Gledhill (1992–1996)\n The Right Reverend Jonathan Gledhill (1996–2021)\n\nReferences\n\n1949 births\n2021 deaths\nAlumni of Keele University\nAlumni of the University of Bristol\nBishops of Lichfield\nBishops of Southampton\nCanons of Canterbury", "messages": null, "license": "CC BY-SA 4.0", "source": "wikipedia:en:Jonathan Gledhill", "greedy": false, "greedy_prompt": null}
+{"seq_id": 17, "group": "general", "kind": "text", "lang": "en", "text": "\"Let's Go\" is a song by American rock band the Cars, written by Ric Ocasek for the band's second studio album, Candy-O (1979). A new wave rock song, the song's hook was inspired by the Routers. The song's vocals are performed by bassist Benjamin Orr.\n\n\"Let's Go\" was released in 1979 as the debut single from Candy-O on Elektra Records. The single was a chart success, reaching number 14 in the United States and charting in multiple other countries. It has since appeared on several compilation albums and has seen critical acclaim.\n\nIt was the 100th video to be played on the first day of MTV on August 1, 1981.\n\nComposition\n\"Let's Go\" was described by Brett Milano as \"another double-edged anthem\" in the liner notes for Just What I Needed: The Cars Anthology.\n\nThe song's signature hook is a series of claps followed by a shouted \"Let's go!\", which is derived from the 1962 song \"Let's Go (Pony)\" by the Routers, as well as a simple synth melody played by Greg Hawkes, using the Sync II lead preset (or a slight variation of it) from a Prophet-5 synthesizer. The song tells the story of a 17-year-old girl and her budding interest in \"the nightlife\".\n\nRelease\n\"Let's Go\" was released as the debut single from Candy-O in June 1979. The song's B-side is a non-album track titled \"That's It\" that features Benjamin Orr on lead vocals. The single peaked at number 14 on the Billboard Hot 100 singles chart in the US, making it the first song by The Cars to reach the Billboard top 20. The song was an even bigger success elsewhere, hitting the top 10 in multiple countries. In Canada, the track reached #5, and remains the Cars' highest ever charting single in that country. Similarly, \"Let's Go\" peaked at #6 in Australia, where it remains The Cars' highest charted hit.\n\nTwo follow-up singles from Candy-O, \"It's All I Can Do\" and \"Double Life\", were released after \"Let's Go\". Although \"It's All I Can Do\" was a minor hit, \"Double Life\" failed to chart.\n\nReception", "messages": null, "license": "CC BY-SA 4.0", "source": "wikipedia:en:Let's Go (The Cars song)", "greedy": false, "greedy_prompt": null}
+{"seq_id": 18, "group": "general", "kind": "text", "lang": "en", "text": "Alabama's 3rd congressional district is a United States congressional district in Alabama that elects a representative to the United States House of Representatives. It is based in east-central Alabama and encompasses Calhoun, Chambers, Cherokee, Clay, Cleburne, Lee, Macon, Randolph, Russell, St. Clair, Talladega, and Tallapoosa counties, and parts of Chilton County. Cities in the district include Phenix City, Talladega, Tuskegee, and Auburn. Prior to the most recent redistricting cycle, the 3rd district had included parts of the state capital city of Montgomery in Montgomery County.\n\nAt the federal level, the district is strongly Republican-leaning, though not quite as strongly as some of the other districts in the state. John McCain carried the district in 2008 with 56.21% of the vote while Barack Obama won 43.04% of the vote.\n\nThe district is currently represented by Republican Mike Rogers and was once represented by Bob Riley, the former Governor of Alabama.\n\nElection results from statewide races\n\nList of members representing the district\n\nRecent election results\nThese are the results from the previous ten election cycles in Alabama's 3rd district.\n\n2002\n\n2004\n\n2006\n\n2008\n\n2010\n\n2012\n\n2014\n\n2016\n\n2018\n\n2020\n\n2022\n\nSee also\n\nAlabama's congressional districts\nList of United States congressional districts\n\nReferences\nSpecific\n\nGeneral\n \n \n Congressional Biographical Directory of the United States 1774–present\n\nExternal links\nCNN coverage of the 2000 election\nCNN coverage of the 2002 election\nCNN coverage of the 2004 election\nCNN coverage of the 2006 election\n\n03\nCalhoun County, Alabama\nChambers County, Alabama\nCherokee County, Alabama\nClay County, Alabama\nCleburne County, Alabama\nCoosa County, Alabama\nLee County, Alabama\nMacon County, Alabama\nMontgomery County, Alabama\nRandolph County, Alabama\nTalladega County, Alabama\nTallapoosa County, Alabama\nRussell County, Alabama\nConstituencies established in 1823\n1823 establishments in Alabama\nConstituencies disestablished in 1841\n1841 disestablishments in Alabama\nConstituencies established in 1843\n1843 establishments in Alabama\nConstituencies disestablished in 1861\n1861 disestablishments in Alabama", "messages": null, "license": "CC BY-SA 4.0", "source": "wikipedia:en:Alabama's 3rd congressional district", "greedy": false, "greedy_prompt": null}
+{"seq_id": 19, "group": "general", "kind": "text", "lang": "en", "text": "Laser surgery is a type of surgery that uses a laser (in contrast to using a scalpel) to cut tissue.\n\nExamples include the use of a laser scalpel in otherwise conventional surgery, and soft-tissue laser surgery, in which the laser beam vaporizes soft tissue with high water content.\n\nLaser surgery is commonly used on the eye. Techniques used include LASIK, which is used to correct near and far-sightedness in vision, and photorefractive keratectomy, a procedure which permanently reshapes the cornea using an excimer laser to remove a small amount of the human tissue.\n\nTypes of surgical lasers include carbon dioxide, argon, Nd:YAG laser, and potassium titanyl phosphate, among others.\n\nEffects \n  Photochemical effect: clinically referred to as photodynamic therapy. Photosensitizer (photophrin II) is administered which is taken up by the tumor tissue and later irradiated by laser light resulting in highly toxic substances with resultant necrosis of the tumor. Photodynamic therapy is used in palliation of oesophageal and bronchial carcinoma and ablation of mucosal cancers of Gastrointestinal tract and urinary bladder.\n  Photoablative effect: Used in eye surgeries like band keratoplasty, and endartectomy of peripheral blood vessels.\n  Photothermal effect: this property is used for endoscopic control of bleeding e.g. Bleeding peptic ulcers, oesophageal varices\n  Photomechanical effect: used in intraluminal lithotripsy\n\nEquipment\n\nSurgical laser systems, sometimes called \"laser scalpels\", are differentiated not only by the wavelength, but also by the light delivery system: flexible fiber or articulated arm, as well as by other factors.\n\nCO2 lasers were the dominant soft-tissue surgical lasers as of 2010.\n\nApplications\n\nSoft tissue", "messages": null, "license": "CC BY-SA 4.0", "source": "wikipedia:en:Laser surgery", "greedy": false, "greedy_prompt": null}
+{"seq_id": 20, "group": "general", "kind": "text", "lang": "en", "text": "Terry W. Virts (born December 1, 1967) is a retired NASA astronaut, International Space Station Commander and colonel in the United States Air Force.\n\nBackground and education \nVirts was born in Baltimore, Maryland, but considers Columbia, Maryland, to be his hometown. He graduated from Oakland Mills High School in 1985.  He earned a Bachelor of Science degree in Mathematics (with a French minor) from the United States Air Force Academy in 1989 and a Master of Aeronautical Science degree in aeronautics from Embry-Riddle Aeronautical University in 1997. While at the Air Force Academy, Virts attended the École de l'Air in 1988 on an exchange program.\n\nMilitary career \nVirts was commissioned as a Second Lieutenant upon graduation from the United States Air Force Academy in 1989 and earned his pilot wings at Williams Air Force Base, Arizona. He completed basic fighter lead-in training at Holloman Air Force Base, New Mexico followed by formal training in the F-16 Fighting Falcon with the 56th Tactical Fighter Wing at MacDill Air Force Base, Florida.\n\nHe was then assigned to the 31st Tactical Fighter Wing at Homestead Air Force Base, Florida. After Hurricane Andrew struck southern Florida in 1992, his squadron was moved to Moody Air Force Base, Georgia. He was later assigned to the 36th Fighter Squadron at Osan Air Base, Republic of Korea, and the 22nd Fighter Squadron at Spangdahlem Air Base, Germany.\n\nVirts was selected for Test Pilot School in 1997 at Edwards Air Force Base, California. Following graduation, he was an Experimental Test Pilot at the F-16 Combined Test Force. He has logged more than 5,300 flight hours in 40 different aircraft.\n\nNASA career \n\nTerry Virts' NASA career spanned several pivotal roles and space missions, showcasing his dedication and expertise in the field of astronautics. In the year 2000, Virts achieved a significant milestone, when he was selected as a Space Shuttle pilot by NASA. His career was marked by a series of important technical assignments and mission roles.", "messages": null, "license": "CC BY-SA 4.0", "source": "wikipedia:en:Terry W. Virts", "greedy": false, "greedy_prompt": null}
+{"seq_id": 21, "group": "general", "kind": "text", "lang": "en", "text": "Nathan Hale (September 23, 1743 – September 23, 1780) was an American Revolutionary War officer who fought in the Battle of Lexington and Concord, Battle of Bunker Hill, Siege of Fort Ticonderoga, and Battle of Hubbardton. Hale was taken prisoner by the British at Hubbardton and died in prison on September 23, 1780, at New Utrecht, Brooklyn, New York.\n\nFamily and early life \nNathan Hale was born in Hampstead, New Hampshire, son of Moses and Elizabeth (Wheeler) Hale. He was a descendant of Thomas Hale of Newbury, Massachusetts, who arrived in 1637 from Watton-At-Stone, Hertfordshire, England as the latter part of the Winthrop Fleet and Great Migration.  There is no known relation between Colonel Nathan Hale and Captain Nathan Hale, the American spy hanged by the British in 1776.\n\nIn his teens, Hale moved with his family to the area that would become Rindge, New Hampshire. He married Abigail Grout, daughter of Col. John and Joanna (Boynton) Grout of Lunenburg, Massachusetts. When the town of Rindge was organized in 1768, Hale was chosen the first constable of the town, and served as the moderator of the annual town meetings in 1773, 1774 and 1775\n\nService in the American Revolution \nIn 1774, Hale became the captain of a militia company of minutemen.  Once Hale was told of the Battle of Lexington on April 19, 1775, he and his fifty men marched to Cambridge, Massachusetts, to join the Army of Observation.\n\nOn June 2, 1775, Hale was commissioned as a captain in the 3rd New Hampshire Regiment. They fought at the Battle of Bunker Hill on June 17, 1775.  The Army of Observation (consisting of militiamen from Massachusetts, Connecticut, New Hampshire and Rhode Island) had about 2,400 men and the British had over 3,000. The colonies suffered 450 casualties and the British suffered 1,054 casualties in what has been described as a British Pyrrhic victory.\n\nHale was promoted to lieutenant-colonel of the 2nd New Hampshire Regiment on November 8, 1776. Achieving success in battle, he was again promoted to colonel on April 2, 1777. In the same year he served under Major General Arthur St. Clair at the Siege of Fort Ticonderoga in early July.", "messages": null, "license": "CC BY-SA 4.0", "source": "wikipedia:en:Nathan Hale (colonel)", "greedy": false, "greedy_prompt": null}
+{"seq_id": 22, "group": "general", "kind": "text", "lang": "en", "text": "Weldon Heyburn (born Weldon Heyburn Franks; September 19, 1903 – May 18, 1951) was an American character actor.\n\nEarly years\nThe son of Marie Pierce and United States Army Col. Wyatt G. Franks, Heyburn was most likely born in Washington, D.C. although other sources indicate Selma, Alabama or Delaware City, Delaware as the place of birth.\n\nHeyburn attended Central High School and Emerson Institute (both in Washington, D.C.), before attending the University of Alabama.\n\nIn the 1920s, Heyburn represented himself as \"the son of Charles Heyburn, judge of the United States Supreme Court\" and \"the nephew of Weldon Heyburn, the late Senator of Iowa.\" The senator's niece denied any kinship, saying in 1925 that Heyburn was \"an impostor if he persists in his claims\", with other members of the family supporting her assertion.\n\nStage\nIn the mid-1920s, Heyburn was the \"leading man in a Lansing, Mich., stock company.\" His Broadway credits include The Mystery Man (1927), Troyka (1930, Good Men and True (1935), and I Want a Policeman (1936).\n\nFilm\nIn 1931, Heyburn ventured into film. An August 31, 1931, newspaper article reported, \"After playing leading man for such stage stars as Jeanne Eagels, Lenore Ulric and Mary Boland, he has given up the stage to become a feature screen player like his roommate and fellow gridiron player, John Mack Brown.\"\n\nMilitary service\nDuring World War II, Heyburn served in the U.S. Army Signal Corps.\n\nPersonal life\nIn 1924, Heyburn married actress Phyllis Connard. They divorced October 15, 1926.\n\nHeyburn married Norwegian star Greta Nissen March 30, 1932, in Tijuana, Mexico. They met when they appeared together in The Silent Witness (1932), she as the female lead, he as an unbilled player. On October 19, 1935, Nissen went to court to have the marriage annulled, \"charging their marriage ... was illegal and violated legal witness and residence requirements.\" The annulment was granted April 30, 1936.", "messages": null, "license": "CC BY-SA 4.0", "source": "wikipedia:en:Weldon Heyburn", "greedy": false, "greedy_prompt": null}
+{"seq_id": 23, "group": "general", "kind": "text", "lang": "en", "text": "The Curtiss F7C Seahawk was a carrier-capable biplane fighter aircraft of the United States Navy Marine Corps in the late 1920s and early 1930s.\n\nDesign and development\nCurtiss' Model 43 was their first aircraft designed expressly for the Navy, rather than a modified Army type. While clearly a descendant of the P-1 Hawk, its wings were constant-chord rather than tapered, and the upper wing had a slight sweepback. The engine was a  Pratt & Whitney R-1340-B Wasp radial. Entirely fabric-covered, the top wing was framed with spruce, while the fuselage was built from a combination of aluminum and steel tubing, sufficiently strong to serve as a dive bomber as well as a fighter.\n\nOperational history\nThe prototype XF7C-1 first flew on 28 February 1927. After some modification demanded by the Navy (such as the wing sweepback), 17 production aircraft F7C-1 Seahawks were built, and entered service in the USMC's VF-5M at Quantico. In 1930 VF-9M organized the Marines' first aerobatic stunt team, \"The Red Devils\", with F7Cs featuring red painted noses. They continued in service until 1933.\n\nVariants\n\nXF7C-1: Prototype aircraft; one built.\nF7C-1 Seahawk: Single-seat fighter aircraft, main production version; 17 built.\nXF7C-2: Single F7C-1 conversion for evaluation with the  Wright R-1820-1 radial engine and large-span full-span flaps.\nXF7C-3: A demonstration prototype for China with an armament of four  M1919 Browning machine guns, I-type interplane struts, and ailerons on both the upper and lower wings rather than on just the upper wing. The type was superseded by the Model 64, F11C Goshawk.\n\nOperators\n\nUnited States Marine Corps\n\nSpecifications (F7C-1)\n\nReferences\n\nNotes\n\nBibliography\n\n Barrow, Jess C. WWII Marine Fighting Squadron Nine (VF-9M) (Modern Aviation Series). Blue Ridge Summit, Pennsylvania: Tab Books Inc., 1981. . \n Eden, Paul and Soph Moeng, eds. cover The Complete Encyclopedia of World Aircraft. London: Amber Books Ltd., 2002. .\n Jones, Lloyd S. U.S. Naval Fighters. Fallbrook, California: Aero Publishers, 1977, pp. 50–52. .\n\n1920s United States fighter aircraft\nF07C Seahawk\nCarrier-based aircraft\nSingle-engined tractor aircraft\nBiplanes", "messages": null, "license": "CC BY-SA 4.0", "source": "wikipedia:en:Curtiss F7C Seahawk", "greedy": false, "greedy_prompt": null}
+{"seq_id": 24, "group": "general", "kind": "text", "lang": "en", "text": "Climatic adaptation refers to adaptations of an organism that are triggered due to the patterns of variation of abiotic factors that determine a specific climate. Annual means, seasonal variation and daily patterns of abiotic factors are properties of a climate where organisms can be adapted to. Changes in behavior, physical structure, internal mechanisms and metabolism are forms of adaptation that is caused by climate properties. Organisms of the same species that occur in different climates can be compared to determine which adaptations are due to climate and which are influenced majorly by other factors. Climatic adaptations limits to adaptations that have been established, characterizing species that live within the specific climate. It is different from climate change adaptations which refers to the ability to adapt to gradual changes of a climate. Once a climate has changed, the climate change adaptation that led to the survival of the specific organisms as a species can be seen as a climatic adaptation. Climatic adaptation is constrained by the genetic variability of the species in question.\n\nClimate patterns\nThe patterns of variation of abiotic factors determine a climate and thus climatic adaptation. There are many different climates around the world, each with its unique patterns. Because of this, the manner of climatic adaptation shows large differences between the climates. A subarctic climate, for instance, shows daylight time and temperature fluctuations as most important factors, while in rainforest climate, the most important factor is characterized by the stable high precipitation rate and high average temperature that doesn't fluctuate a lot. Humid continental climate is marked by seasonal temperature variances which commonly lead to seasonal climate adaptations. Because the variance of these abiotic factors differ depending on the type of climate, differences in the manner of climatic adaptation are expected.\n\nResearch", "messages": null, "license": "CC BY-SA 4.0", "source": "wikipedia:en:Climatic adaptation", "greedy": false, "greedy_prompt": null}
+{"seq_id": 25, "group": "general", "kind": "text", "lang": "en", "text": "Kim Kristine Martin Hasson (born 28 February 1986) is a retired Swedish goaltender, currently working in the Linköping HC organization. With the Swedish national team she won two Olympic medals, bronze in 2002 and silver in 2006, and two IIHF World Women's Championships bronze medals, in 2005 and 2007. Martin Hasson played in the SDHL with AIK and Linköping HC, in the Russian Women's Hockey League with Tornado Dmitrov, in the NCAA Division I with the Minnesota-Duluth Bulldogs, in the J20 SuperElit with the Malmö Redhawks’ junior men's team, and in the J18 Allsvenskan with Hammarby IF's junior men's team.\n\nShe was scheduled to make her debut with the Malmö Redhawks of the HockeyAllsvenskan, the second-tier men's league in Sweden, on 8 March 2006, as the first woman to ever play for a professional men's team in Sweden but this move was vetoed by the University of Minnesota Duluth, the U.S. school that had offered her a scholarship to play hockey. This was because the NCAA, the main governing body for U.S. college sports, prohibits athletes at its member schools from having previously played in a professional league, even if they are not paid.\n\nPlaying career\n\nMinor ice hockey\nMartin's first season of organized ice hockey was at the age of 10, and she did not allow a goal during the entire 17-game schedule. Martin made her international debut in November 2000 at the Four Nations Cup. She appeared in one game, a 2–2 tie against Finland.\n\nIn club competition, she is a two-time European women's champion with AIK, in 2004–05 and 2005–06. She was also acclaimed as the top goaltender in the tournament in 2005–06.\n\nUniversity of Minnesota-Duluth\nOn 22 March 2008, Martin and the UMD Bulldogs beat the Wisconsin Badgers 4–0 at the Duluth Entertainment Convention Center arena in Duluth, MN for their fourth NCAA Division I national championship. It was just the second shutout in NCAA women's hockey championship history. Martin made 28 saves in the game and 69 total saves in the Frozen Four championship tournament and was named the tournament's Most Outstanding Player. She was also named to the All-Tournament Team along with four of her teammates.", "messages": null, "license": "CC BY-SA 4.0", "source": "wikipedia:en:Kim Martin Hasson", "greedy": false, "greedy_prompt": null}
+{"seq_id": 26, "group": "general", "kind": "text", "lang": "en", "text": "The Messenger of the Sacred Heart is a Roman Catholic periodical; the print organ of the Apostleship of Prayer, a pious association founded in nineteenth century France by the Jesuits. There are many editions in various languages, promoting devotion to the Sacred Heart of Jesus. In the late 1940s, The Messenger was banned in many Soviet Block countries.\n\nHistory\nThe Messenger of the Sacred Heart was originally founded in France as the print organ of the Apostleship of Prayer, a pious association founded in 1844 by Francois-Xavier Gautrelet, S.J. The first Messenger of the Sacred Heart was published in 1861. The Apostleship of Prayer has always operated under the auspices of the Society of Jesus. By 1941, there were seventy-two Messengers around the world published in forty-four different languages. The English and the Australian Jesuits prepared their own editions. A children's edition was also produced.\n\nEnglish Messenger\nWilliam Maher SJ (1823-1877) translated and composed the words and music to the hymn Soul of My Saviour, among other hymns. Maher served as the editor of the English Messenger until his death in 1877. Henry James Coleridge then added the Messenger to his other duties as editor of The Month. In 1884, Coleridge was able to pass the editorship of \"The Messenger\" to Augustus Dignam SJ, who also happened to be spiritual adviser to Frances Margaret Taylor, the former publisher of The Lamp. Taylor was by then Mother Magdalen of the Sacred Heart and head of the Congregation of the Poor Servants of the Mother of God. A writer, in her own right, with her experience in publishing, she helped Dignam with \"The Messenger\".\n\nIrish Messenger", "messages": null, "license": "CC BY-SA 4.0", "source": "wikipedia:en:Messenger of the Sacred Heart", "greedy": false, "greedy_prompt": null}
+{"seq_id": 27, "group": "general", "kind": "text", "lang": "en", "text": "Nigel Mitchell is a British television and radio presenter and voice-over artist.\n\nEarly life and education\n\nNigel was born in Kingston Upon Thames. He attended Shrewsbury House School in Surbiton, from where he gained a scholarship to Reed's School in Cobham, Surrey at the age of 11.\n\nTelevision career\n\nNickelodeon\nNigel started his TV career as a presenter on the kid's television station, Nickelodeon at the age of 17.\n\nITV\n\nNigel left Nickelodeon to present 'Sticky' on CITV. Sticky was a live magazine show produced by 70's TV personality Mick Robertson. The show was co-presented by Gail Porter.\n\nDisney Channel\n\nNigel went on to Disney Channel in 1997 to host Studio Disney and remained with the channel until the live shows were ended in May 2005. Studio Disney featured live games, callers, special guests and live performances.\n\nFilm 24\n\nNigel was a regular presenter on Sky's Film 24 channel looking at upcoming movies and interviewing the stars before the premieres. Interviews include Kiefer Sutherland, Reese Witherspoon and Robert Pattinson.\n\nArsenal FC\n\nNigel works with Arsenal Football Club as the pitch side presenter at Emirates Stadium. He also hosts Arsenal World.\n\nThis Morning\nNigel regularly presented The Hub on the ITV morning show\n\nRadio presenting\nInformation on the radio presenting Nigel has presented during his career.\n\nHeart\nMost recently, Nigel has covered shows on the network.\n\nBBC Radio Cambridgeshire\nNigel's broadcast across Cambridgeshire when he covered the afternoon show for Sue Dougan, Bank Holidays and the Saturday Summer Sport programme.\n\nFox FM\nNigel worked for Fox FM in Oxford during 2008.\n\nCapital Disney\nNigel hosted the breakfast show when Capital Disney launched in 2002 and throughout Capital Disney's time on air he also hosted live shows across the weekend and 'The Friday Afternoon Thing.'  Producer, Andrew Rendle, nicknamed Radio Rendle, was part of the show every weekend.\n\nKingston FM", "messages": null, "license": "CC BY-SA 4.0", "source": "wikipedia:en:Nigel Mitchell", "greedy": false, "greedy_prompt": null}
+{"seq_id": 28, "group": "general", "kind": "text", "lang": "en", "text": "The culture of Kolkata concerns the music, art, museums, festivals, and lifestyle within Kolkata. It is the former capital of India and, , the capital of the Indian state of West Bengal. Geir Heierstad writes that Bengalis tend to have a special appreciation for art and literature.\n\nArts  \n\nThe city has a long tradition of commercial theatres and group theatres. As opposed to commercial theatres, group theatres usually do not have any profit making agenda. Group theatre activists use the proscenium stage to portray some social message. The commercial theatres of the city, however, has been declining in popularity since the 1980s, and only a handful of commercial theatre productions are made, as of 2009.\n\nNotable group theatres include the Little Theatre Group, Gandharba, Calcutta Theatre, Nandikar, Bahurupee etc. and movements like the Indian People's Theatre Association (IPTA). Famous drama and theatrics personalities include Ajitesh Bandyopadhyay, Utpal Dutta, Rudraprasad Sengupta and Shambhu Mitra.\n\nArchitecture\n\nKolkata has many buildings adorned with Indo-Islamic and Indo-Saracenic architectural motifs. Several well-maintained major buildings from the colonial period have been declared \"heritage structures\"; however, others are in various stages of decay. \nEstablished in 1814 as the nation's oldest museum, the Indian Museum houses large collections that showcase Indian natural history and Indian art. Marble Palace is a classic example of a European mansion that was built in the city. The Victoria Memorial, a place of interest in Kolkata, has a museum documenting the city's history.", "messages": null, "license": "CC BY-SA 4.0", "source": "wikipedia:en:Culture of Kolkata", "greedy": false, "greedy_prompt": null}
+{"seq_id": 29, "group": "general", "kind": "text", "lang": "en", "text": "Zuni fetishes are small carvings made from primarily stone but also shell, fossils, and other materials by the Zuni people. Within the Zuni community, these carvings serve ceremonial purposes for their creators and depict animals and icons integral to their culture. As a form of contemporary Native American art, they are sold with secular intentions to collectors worldwide. Prior to the establishment of a non-Native market for fetishes, Hopi, Navajo, and other Pueblo peoples, especially at Kewa Pueblo also carved and used fetishes.\n\nDirections and typology \nThe primary non-Native source for academic information on Zuni fetishes is the Second Annual Report of the Bureau of Ethnology submitted in 1881 by Frank Hamilton Cushing and posthumously published as Zuni Fetishes in 1966, with several later reprints. Cushing reports that the Zuni divided the world into six regions or directions: north, west, south, east, above, and below. At the center of each region is a great mountain peak that is a very sacred place. Yellow mountain to the north, blue mountain to the west, red mountain to the south, white mountain to the east, the multicolored mountain above, and the black mountain below.\n\nEach direction is represented by a Prey God, or guardian animal, and are listed by Cushing as follows:\nnorth: yellow mountain lion\nwest: black bear (represented by the color blue),\nsouth: red badger,\neast: white wolf,\nabove or the sky: multicolored eagle,\nbelow or underground: black shrew (often misnamed \"mole,\" but moles do not live in the Southwest).\n\nEach prey god is the “guardian and master” of their region, with the yellow mountain lion being the elder brother of all animals and the master and guardian of all regions. Each one of these regions contains an order of all the guardian animals, but the \"guardian and master\" of a particular region is the elder brother to all animals of that region. These guardians are considered as having protective and healing powers. They are held by the priests of the medicine orders as if \"in captivity\" and act as mediators between the priests and the animals they represent.", "messages": null, "license": "CC BY-SA 4.0", "source": "wikipedia:en:Zuni fetishes", "greedy": false, "greedy_prompt": null}
+{"seq_id": 30, "group": "chat", "kind": "chat", "lang": "en", "text": null, "messages": [{"role": "user", "content": "What is a polygon?"}, {"role": "assistant", "content": "A polygon is a form in Geometry.  It is a single dimensional plane made of connecting lines and any number of vertices.  It is a closed chain of connected line segments or edges.  The vertices of the polygon are formed where two edges meet.  Examples of polygons are hexagons, pentagons, and octagons.  Any plane that does not contain edges or vertices is not a polygon.  An example of a non-polygon is a circle."}], "license": "CC BY-SA 3.0", "source": "dolly-15k:13", "greedy": true, "greedy_prompt": null}
+{"seq_id": 31, "group": "chat", "kind": "chat", "lang": "en", "text": null, "messages": [{"role": "user", "content": "How do I start running?"}, {"role": "assistant", "content": "Make sure you get comfortable running shoes and attire. Start with achievable goal in mind like a 5K race. If you never ran before, start gradually from a walk, to brisk walk, light jog aiming for 15-30mins initially. Slowly increase your running time and distance as your fitness level improves. One of the most important things is cool down and gentle stretching. Always listen to your body, and take rest days when needed to prevent injury."}], "license": "CC BY-SA 3.0", "source": "dolly-15k:14", "greedy": true, "greedy_prompt": null}
+{"seq_id": 32, "group": "chat", "kind": "chat", "lang": "en", "text": null, "messages": [{"role": "user", "content": "What are some unique curtain tie backs that you can make yourself?"}, {"role": "assistant", "content": "There are many items you can use to make a curtain tie back. Some ideas for this include a chain, a circle and pin, jute rope, wooden beaded rope, a necklack or bracelet, a door knob, a length of a leather belt, macrame rope, or a string of artificial flowers."}], "license": "CC BY-SA 3.0", "source": "dolly-15k:17", "greedy": true, "greedy_prompt": null}
+{"seq_id": 33, "group": "chat", "kind": "chat", "lang": "en", "text": null, "messages": [{"role": "user", "content": "What is a verb?"}, {"role": "assistant", "content": "A verb is an action word that describes an activity.  Examples of verbs are the following: swimming, hiking, biking, roller skating, or skydiving.  All of these single and combination word examples are tied to an activity that any living thing can do.  For example, a dog can bike ride with a human when the human is pedaling the bike.  Verb use is not limited to humans or people only but applies to all living things."}], "license": "CC BY-SA 3.0", "source": "dolly-15k:24", "greedy": true, "greedy_prompt": null}
+{"seq_id": 34, "group": "chat", "kind": "chat", "lang": "en", "text": null, "messages": [{"role": "user", "content": "What is best Playstation or Xbox"}, {"role": "assistant", "content": "Playstation and Xbox are two competing game consoles, where which is \"best\" will depend on the preferences of the gamer.  From a graphics and performance perspective they are comparable, with Sony's Playstation technically leading, though most gamers would be unlikely to notice a significant difference.  From a game library perspective, both consoles have their respective first-party titles that appeal to different demographics.  For example, Microsoft has seen great popularity due to popular title Halo being exclusively available on the console.  On the other hand, Playstation has popular exclusive titles such as God of War and Gran Turismo.  Ultimately the decision of which console is best will likely come down to which games titles the purchaser prefers."}], "license": "CC BY-SA 3.0", "source": "dolly-15k:30", "greedy": true, "greedy_prompt": null}
+{"seq_id": 35, "group": "chat", "kind": "chat", "lang": "en", "text": null, "messages": [{"role": "user", "content": "why did Syd Barrett left the Pink Floyd?"}, {"role": "assistant", "content": "Syd Barrett was one of the main members of the Pink Floyd. He has used drugs a lot and after a while he was not able to perform with the team. Even though the band gave him multiple chances, he could not keep up with the band. In the end, they had to ask him to leave the band. The main reason that he has left the team is his health problems and addiction to drugs."}], "license": "CC BY-SA 3.0", "source": "dolly-15k:32", "greedy": true, "greedy_prompt": null}
+{"seq_id": 36, "group": "chat", "kind": "chat", "lang": "en", "text": null, "messages": [{"role": "user", "content": "what is REST API"}, {"role": "assistant", "content": "REST API is an architectural style for building network APIs. In practice, most REST APIs are using JSON over HTTP. REST APIs have several properties that make them very popular.\n\nHTTP is the foundational and the most popular network protocol across the internet. By using HTTP, REST APIs can fully utilize the internet infrastructure, which greatly reduce the complexity and operational costs. For example, using HTTPS for transport security, and using URL for request routing.\n\nJSON is a simple, stable, and intuitive data format , which is not expected to change ever. This offers great ease of use and stability to REST APIs. For high performance applications, binary formats can be used instead, such as protobuf or thrift.\n\nREST APIs are often designed as CRUD APIs, where API operations are commonly expressed as Create, Retrieve, Update, and Delete operations on resources. CRUD operations are naturally mapped to HTTP methods (POST, GET, PUT/PATCH, and DELETE), and API resources can be naturally identified using URLs."}], "license": "CC BY-SA 3.0", "source": "dolly-15k:33", "greedy": true, "greedy_prompt": null}
+{"seq_id": 37, "group": "chat", "kind": "chat", "lang": "en", "text": null, "messages": [{"role": "user", "content": "What is underwriting?"}, {"role": "assistant", "content": "Underwriting (UW) services are provided by some large financial institutions, such as banks, insurance companies and investment houses, whereby they guarantee payment in case of damage or financial loss and accept the financial risk for liability arising from such guarantee. An underwriting arrangement may be created in a number of situations including insurance, issues of security in a public offering, and bank lending, among others. The person or institution that agrees to sell a minimum number of securities of the company for commission is called the underwriter.\n\nThe term \"underwriting\" derives from the Lloyd's of London insurance market. Financial backers (or risk takers), who would accept some of the risk on a given venture (historically a sea voyage with associated risks of shipwreck) in exchange for a premium, would literally write their names under the risk information that was written on a Lloyd's slip created for this purpose."}], "license": "CC BY-SA 3.0", "source": "dolly-15k:34", "greedy": true, "greedy_prompt": null}
+{"seq_id": 38, "group": "chat", "kind": "chat", "lang": "en", "text": null, "messages": [{"role": "user", "content": "Give me a list of some characteristics of a good coffee shop?"}, {"role": "assistant", "content": "Some characteristics of good coffee shops include:\n- Very good coffee, either freshly roasted locally or acquired from a quality roaster\n- High end coffee grinders, brewing machines, and a high quality espresso machine and well-trained baristas\n- Different coffee roast options (i.e. medium roast, dark roast) and brewing methods (i.e. drip coffee and pour-over) available\n- A variety of cafe fare, such as pastries and croissants, available for purchase to accompany coffee\n- Ample seating, with power outlets available by most seats."}], "license": "CC BY-SA 3.0", "source": "dolly-15k:35", "greedy": false, "greedy_prompt": null}
+{"seq_id": 39, "group": "chat", "kind": "chat", "lang": "en", "text": null, "messages": [{"role": "user", "content": "When was the wheel invented?"}, {"role": "assistant", "content": "No one knows when the wheel was first made. Archaeological evidence seems to show that the wheel was first used to make pottery in Mesopotamia about 7,000 years ago, during the Neolithic period. Later, it was used to make chariots. Chariot wheels have been found in Sumer as early as 3000 BC. The wheel is one of the most important inventions in human history. It was very useful for making pottery and getting around on land. It is also a key part of machines."}], "license": "CC BY-SA 3.0", "source": "dolly-15k:41", "greedy": false, "greedy_prompt": null}
+{"seq_id": 40, "group": "chat", "kind": "chat", "lang": "en", "text": null, "messages": [{"role": "user", "content": "Did dinosaurs have lips?"}, {"role": "assistant", "content": "Recent research suggests that the popular perception that dinosaurs had exposed teeth may be incorrect, and the current hypothesis is that the Tyrannosaurus rex did indeed have lips. A 2023 study by Dr Mark Witton of the University of Portsmouth said popular depictions of dinosaurs in movies like Jurassic Park are out of date. The study examined a large tooth that had spent over 500 days in the mouth of a Tyrannosaurus rex with no evidence of substantial wear, which is consistent with all findings of teeth found in the theropod family of dinosaurs. When contrasted with alligators and crocodiles, who have no lips and have more easily damaged teeth the theory says the evidence points to theropods, including Tyrannosaurus rex did indeed have lips. There would have been expected to be wear on the thin enamel on the teeth far more consistent with modern animals like crocodiles if dinosaurs had no lips."}], "license": "CC BY-SA 3.0", "source": "dolly-15k:43", "greedy": false, "greedy_prompt": null}
+{"seq_id": 41, "group": "chat", "kind": "chat", "lang": "en", "text": null, "messages": [{"role": "user", "content": "What is enriched air and why would divers dive with enriched air?"}, {"role": "assistant", "content": "Enriched air, also known as nitrox, is one kind of air with more than 21% oxygen. 21% oxygen is the normal air we breathe on the earth. The primary benefit of using enriched air is to extend the diver's bottom time, and reduce the need to push to no decompression limit. This advantage is really beneficial in deeper diving. \nFor example, at 80 feet of depth, normal air can last around 30 min, while enriched air of 32% oxygen can last for 45 min and 36% can last for 55 min. \nWhile enriched air is really beneficial in extending bottom time, it is important that divers get trained to use it before diving with enriched air."}], "license": "CC BY-SA 3.0", "source": "dolly-15k:52", "greedy": false, "greedy_prompt": null}
+{"seq_id": 42, "group": "chat", "kind": "chat", "lang": "en", "text": null, "messages": [{"role": "user", "content": "Write a short paragraph about why you should not have both a pet cat and a pet bird."}, {"role": "assistant", "content": "Having both a pet bird and a pet cat is not a good idea because cats will try to hunt birds, which are naturally the prey of cats. This hunting instinct is completely normal, and it isn't fair to the cat to try to train it not to attack birds, as it is simply in its nature to do so. However, letting a cat hunt your pet bird is similarly not humane, as it will be very stressful to the bird and could lead to the bird being injured. Larger birds are also not a good idea, as the bird could cause injury to the cat if the cat tries to attack it. Overall, it just isn't a good idea to have both cats and birds, since they tend not to coexist well, and pet owners should just pick one species to have."}], "license": "CC BY-SA 3.0", "source": "dolly-15k:56", "greedy": false, "greedy_prompt": null}
+{"seq_id": 43, "group": "chat", "kind": "chat", "lang": "en", "text": null, "messages": [{"role": "user", "content": "What makes a formula one car so fast?"}, {"role": "assistant", "content": "Formula One cars are one of the fastest racing cars in the world reaching speeds as high as 240 mph.\n\nThere are many aspects that make formula one cars fast, below are the top few\n\n1. Engine:  F1 cars have extremely powerful engines that can generate around 1000 bhp at up to 20,000 rpm\n2. Power-to-Weight ratio: F1 cars are also extremely light at around 800 kgs giving them a power-to-weight ratio of 12000-1400 hp/ton. For comparison, the best Porsche cars have a power-to-weight ratio of only 300-400 hp/ton.\n2. Aerodynamics: F1 cars are fitted with extremely sophisticated aerodynamics including a rear wing, front wing and body parts that help reduce aerodynamic drag for straight-line speeds while adding downforce to help around corners\n3. Tires: F1 tires are extremely large and sticky. This help the cars stick to the road as the car goes through turns very fast.\n4. Brakes: Lastly F1 brakes are very powerful and can slow down a car extremely quickly. This allows F1 cars to be at top speed for much longer before needing to slow down for taking a turn."}], "license": "CC BY-SA 3.0", "source": "dolly-15k:58", "greedy": false, "greedy_prompt": null}
+{"seq_id": 44, "group": "chat", "kind": "chat", "lang": "en", "text": null, "messages": [{"role": "user", "content": "Why do potato chip bags become stale after opening?"}, {"role": "assistant", "content": "Potato chip bags are filled with nitrogen! Many consumers think the chip companies are out there to get your money by filling three quarters of your bag with air, but it’s actually nitrogen to keep your chips fresh. This is why once you open the bag, you should finish the chips as soon as possible. There, now you have a reason to finish the whole bag of chips in one go without feeling guilty."}], "license": "CC BY-SA 3.0", "source": "dolly-15k:63", "greedy": false, "greedy_prompt": null}
+{"seq_id": 45, "group": "chat", "kind": "chat", "lang": "en", "text": null, "messages": [{"role": "user", "content": "What is the future for human?"}, {"role": "assistant", "content": "This is a very wide open question. It could be two directions. One direction is that human would just disappear like the dinosaur.  One direction is that human would become a species like a god. For the first direction, it may be caused by many situations. e.g. Nuclear war from WW3, extreme climate change. For the other direction, reproduced energy is ready, all cancers can be cured, everything can be done with robot with AI, human life is unlimited."}], "license": "CC BY-SA 3.0", "source": "dolly-15k:67", "greedy": false, "greedy_prompt": null}
+{"seq_id": 46, "group": "chat", "kind": "chat", "lang": "en", "text": null, "messages": [{"role": "user", "content": "Who invented the telephone?"}, {"role": "assistant", "content": "In 1876, Alexander Graham Bell was the first to obtain a United States patent for a device that produced a clearly intelligible replica of the human voice on a second device.\nA resolution passed by the United States House of Representatives on June 11, 2002 called for recognition of Antonio Meucci's work and contributions towards the invention of the telephone."}], "license": "CC BY-SA 3.0", "source": "dolly-15k:71", "greedy": false, "greedy_prompt": null}
+{"seq_id": 47, "group": "chat", "kind": "chat", "lang": "en", "text": null, "messages": [{"role": "user", "content": "What is a kite?"}, {"role": "assistant", "content": "A kite is a toy that can be be purchased in a store or a specialty shop.  It is a craft that is lighter an air that is tied to a string that you can fly on wind currents.  The kite has wings that react against the air to create lift enabling it to fly.  You can fly a kite at a park or an open area, and you will want to make sure there are no power lines nearby to be safe when flying the kite."}], "license": "CC BY-SA 3.0", "source": "dolly-15k:82", "greedy": false, "greedy_prompt": null}
+{"seq_id": 48, "group": "chat", "kind": "chat", "lang": "en", "text": null, "messages": [{"role": "user", "content": "Why is Santa Cruz, California a great place to live?"}, {"role": "assistant", "content": "Santa Cruz, California has one of the best climates in the continental U.S. – being close to the Pacific Ocean, temperatures stay warmer in the winters and cooler in the summers, but without the dreaded fog that nearby cities like San Francisco or Monterey experience. The small city is world-famous for surfing, and the relaxed vibe that comes with it. Natural beauty is abundant, with the mountains and beach only a few miles apart."}], "license": "CC BY-SA 3.0", "source": "dolly-15k:87", "greedy": false, "greedy_prompt": null}
+{"seq_id": 49, "group": "chat", "kind": "chat", "lang": "en", "text": null, "messages": [{"role": "user", "content": "What is investment banking?"}, {"role": "assistant", "content": "Investment banking pertains to certain activities of a financial services company or a corporate division that consist in advisory-based financial transactions on behalf of individuals, corporations, and governments. Traditionally associated with corporate finance, such a bank might assist in raising financial capital by underwriting or acting as the client's agent in the issuance of debt or equity securities. An investment bank may also assist companies involved in mergers and acquisitions (M&A) and provide ancillary services such as market making, trading of derivatives and equity securities, FICC services (fixed income instruments, currencies, and commodities) or research (macroeconomic, credit or equity research). Most investment banks maintain prime brokerage and asset management departments in conjunction with their investment research businesses. As an industry, it is broken up into the Bulge Bracket (upper tier), Middle Market (mid-level businesses), and boutique market (specialized businesses).\n\nUnlike commercial banks and retail banks, investment banks do not take deposits."}], "license": "CC BY-SA 3.0", "source": "dolly-15k:99", "greedy": false, "greedy_prompt": null}
+{"seq_id": 50, "group": "chat", "kind": "chat", "lang": "en", "text": null, "messages": [{"role": "user", "content": "How much music theory knowledge is required to compose music?"}, {"role": "assistant", "content": "Technically speaking, very little music theory knowledge is “required” to compose music. Music theory is, essentially, what musicians have learned over the course of several hundred years about how to write music that is interesting and sounds good. In other words, music theory is there to help you, not hold you back. \n\nWhen you learn music theory, what often happens is that you expand your broader understanding of music, which can be a very helpful safety net. The more music theory you know, the easier you’ll be able to distinguish between the good ideas and the not-so-good. You need to know the rules before you can break them. You need to be able to write by the book before you throw the book away. If not, you run the risk of writing music that’s not nearly as clever as you think it is, that tries to be daring and experimental but ends up just being dull and vanilla. And frankly, I just don’t see the value in ignoring all the work that centuries worth of musicians have done into knowing how music works. Putting the time into learning and really understanding it can only make you a better musician and a better composer.\n\nSo, that’s my answer. Learn at least as much theory as you need to know to understand the rules you want to break."}], "license": "CC BY-SA 3.0", "source": "dolly-15k:102", "greedy": false, "greedy_prompt": null}
+{"seq_id": 51, "group": "chat", "kind": "chat", "lang": "en", "text": null, "messages": [{"role": "user", "content": "What is the tallest mountain in the continental US?"}, {"role": "assistant", "content": "Mt. Whitney, which is 14500 ft,  is the tallest mountain in the continental US. Although if we were to include the state of Alaska, then the tallest mountain is actually Mt Denali, which is 20,320 ft. Mt. Whitney is actually only the 17th tallest mountain in all of the US, where the tallest 1-16 are all mountains in Alaska."}], "license": "CC BY-SA 3.0", "source": "dolly-15k:107", "greedy": false, "greedy_prompt": null}
+{"seq_id": 52, "group": "chat", "kind": "chat", "lang": "en", "text": null, "messages": [{"role": "user", "content": "Plan for a three day road trip for May long weekend from Westfield, New Jersey"}, {"role": "assistant", "content": "May 26: Drive to Montreal from Westfield, NJ. Time taken  6 hrs. Stay there overnight\nMay 27:  have a quick breakfast..checkout and drive to Mount Tremblant. Time: 90 minutes\nMay 27: Hike Grand Brule ( red path) on the way up and green path on the way down\nMay 27: Stroll around the Tremblant village\nMay 28: Drive to Quebec City. Time taken: 4 hrs\nMay 28: Explore the city, aquarium and go for wildlife watching\nMay 29: Drive to Montreal. Spend time exploring the city\nMay 30: Drive back to Westfield, New Jersey"}], "license": "CC BY-SA 3.0", "source": "dolly-15k:112", "greedy": false, "greedy_prompt": null}
+{"seq_id": 53, "group": "chat", "kind": "chat", "lang": "en", "text": null, "messages": [{"role": "user", "content": "How do you operate a car with a manual transmission?"}, {"role": "assistant", "content": "Through a combination of a shifter and three pedals: gas, brake, and clutch. Press down the clutch pedal with one foot, and the brake with your other foot first and then turn on the engine of the car. It is recommended to shift to neutral immediately after turning on your car and prior to departing. To shift between one of the engaged gears, typically marked 1, 2, 3, 4, 5 N (for neutral) and R (for reverse), press down on the clutch all the way to the floor and move the shifter to the desired location. Once the car is on and you're in neutral if you want to move forward, press and hold the clutch pedal to the floor and move the shifter to the \"1\" position. Slowly disengage the clutch pedal partially, and while disengaging slowly start to press down on the gas pedal. After gradually pressing more on the gas pedal, slowly remove your foot from the clutch entirely. This will start you in forward motion, and you will be restricted to the speed limit of 1st gear. You can track this limit through the RPM (revolutions per minute) gauge on the dashboard of your car. Typically you want to keep the RPMs between 1500-4000. If the RPMs start to decrease to 1500-2000 that is a sign you want to shift from a higher gear to a lower gear (i.e. 2nd gear to 1st gear), and as the RPMs increase to 4000, that is an indication you want to shift to a higher gear (i.e. 2nd gear to 3rd gear)."}], "license": "CC BY-SA 3.0", "source": "dolly-15k:113", "greedy": false, "greedy_prompt": null}
+{"seq_id": 54, "group": "chat", "kind": "chat", "lang": "en", "text": null, "messages": [{"role": "user", "content": "What is the Research Collaboratory for Structural Bioinformatics Protein Data Bank (RCSB PDB)?"}, {"role": "assistant", "content": "The Research Collaboratory for Structural Bioinformatics Protein Data Bank (RCSB PDB) is a database that provides a wealth of information about the 3D structures of proteins, nucleic acids, and other macromolecules. The database contains experimentally determined atomic coordinates for a large number of macromolecules, which can be used to study their structures, functions, and interactions. The RCSB PDB is widely used in genomics research and drug discovery, as it provides a valuable resource for understanding the structural basis of many biological processes and for designing new drugs that target specific macromolecules.\n\nIn addition to the atomic coordinates, the RCSB PDB contains a wealth of additional information about each macromolecule, including experimental methods used for structure determination, citations to relevant scientific literature, and information about biological function, sequence, and homology. The database also provides a variety of tools and resources for visualizing, analyzing, and downloading structural data, including a web-based viewer that allows users to interactively explore the 3D structures of macromolecules."}], "license": "CC BY-SA 3.0", "source": "dolly-15k:114", "greedy": false, "greedy_prompt": null}
+{"seq_id": 55, "group": "chat", "kind": "chat", "lang": "en", "text": null, "messages": [{"role": "user", "content": "What are 5 potential uses for an empty La Croix can?"}, {"role": "assistant", "content": "1. Make your bike sound like it has an engine by wrapping the can around the back tire.\n2. A percussion instrument to play along with friends in a jam session.  You could just crinkle it, hit it with a pen, or fill it with sand or debris and shake it.  \n3. A baking tool.  You can used the bottom of the can to stamp out perfectly circle, uniform cookies from a rolled-out sheet of cookie dough.\n4. A field razor.  Tear the can carefully to create an effective cutting device for emergency use.\n5. A make shift ball to throw around with friends at the park. An empty can could be a fun projectile to use in a friendly game of 3 Flies Up."}], "license": "CC BY-SA 3.0", "source": "dolly-15k:115", "greedy": false, "greedy_prompt": null}
+{"seq_id": 56, "group": "chat", "kind": "chat", "lang": "en", "text": null, "messages": [{"role": "user", "content": "Why are so many US Hollywood films made in the state of Georgia?"}, {"role": "assistant", "content": "While film in the United States as an audience artform started in New York it moved to Los Angeles for the stated reason of longer and sunnier days.  Another part of the story was a patent dispute in 1898 between Thomas Edison and his patent for the Kinetograph.  Edison stated that he effectively had a say in how films were made and wanted royalties for their use.  Moving from New York near where Edison was based to Los Angeles was a simple way to grow the industry and distance themselves from a patent war.  \n\nThis mobility at the start of the modern film era due to costs continued through the 1960's and 1970's with westerns being shot in Spain.  The 1980's and 1990's continued this trend primarily shooting in Canada or Mexico.  At the turn of the century state legislators in the United States started provided steep tax rebates.  The practical outworking of this is studios are encouraged to hire local talent and use local services when producing their film.  The ratios for these rebates is fairly complicated and have created a new industry of tax professionals serving the film industry.\n\nSo then why Georgia?  Georgia currently provides one of the highest incentive structures at 20% with minimal restrictions on talent and a very low minimum budget of $500,000."}], "license": "CC BY-SA 3.0", "source": "dolly-15k:117", "greedy": false, "greedy_prompt": null}
+{"seq_id": 57, "group": "chat", "kind": "chat", "lang": "en", "text": null, "messages": [{"role": "user", "content": "What are some tools that every homeowner should have?"}, {"role": "assistant", "content": "There is a tool designed for almost every challenge a homeowner will face, but purchasing every tool could prove to be cost prohibitive, especially if they are highly specialized for tasks that may occur infrequently.  It's best to have a general set of tools that have a wide variety of uses, including a hammer, a set of screwdrivers both flat and philips, a tape measure, a level, a stud finder, and a utility knife. A battery powered drill and impact driver can also save time, and make mounting projects easier."}], "license": "CC BY-SA 3.0", "source": "dolly-15k:121", "greedy": false, "greedy_prompt": null}
+{"seq_id": 58, "group": "chat", "kind": "chat", "lang": "en", "text": null, "messages": [{"role": "user", "content": "write a short essay on California wild fire"}, {"role": "assistant", "content": "California is known for its hot, dry summers, which increase the risk of wildfires. Every year, the state experiences dozens of fires, which can be devastating to both the environment and people's homes and businesses.\n\nIn recent years, California has seen some of the worst wildfires in its history, with tens of thousands of acres of land destroyed and hundreds of homes lost. These fires are often started by a combination of high temperatures, dry conditions, and strong winds, which can quickly spread flames and make them difficult to control.\n\nThe effects of these fires are felt not just by those directly impacted, but by everyone in the state. Smoke and ash from the fires can cause respiratory problems and other health issues, and the destruction of habitats can harm wildlife populations and disrupt entire ecosystems.\n\nDespite the challenges, California has made significant progress in fighting wildfires in recent years. The state has invested in new technologies and equipment to help firefighters respond more quickly and effectively to fires, and has increased efforts to prevent fires from starting in the first place.\n\nWhile wildfires will likely always be a risk in California, the state's proactive approach to managing them gives hope that we can continue to protect people, wildlife, and the environment from their devastating effects."}], "license": "CC BY-SA 3.0", "source": "dolly-15k:123", "greedy": false, "greedy_prompt": null}
+{"seq_id": 59, "group": "chat", "kind": "chat", "lang": "en", "text": null, "messages": [{"role": "user", "content": "What is the difference between pets and cattle?"}, {"role": "assistant", "content": "Although both are animals at the end of the day, however, the main difference comes partially in the form of perception and, historically, the relationship of humans with certain animals. Pets, e.g. dogs, cats, etc., are used more for companionship and not for food. On the contrary, cattle, e.g. cows, pigs, chickens, etc., are primarily raised for food, i.e. meat, milk, eggs, etc."}], "license": "CC BY-SA 3.0", "source": "dolly-15k:133", "greedy": false, "greedy_prompt": null}
+{"seq_id": 60, "group": "code", "kind": "text", "lang": "javascript", "text": "// Debounce: run `fn` only after `ms` milliseconds of silence.\nfunction debounce(fn, ms) {\n  let timer = null;\n  return function (...args) {\n    clearTimeout(timer);\n    timer = setTimeout(() => fn.apply(this, args), ms);\n  };\n}\n\nconst onResize = debounce(() => {\n  console.log(\"window size\", window.innerWidth, window.innerHeight);\n}, 200);\nwindow.addEventListener(\"resize\", onResize);\n", "messages": null, "license": "Apache-2.0", "source": "handwritten:javascript", "greedy": true, "greedy_prompt": null}
+{"seq_id": 61, "group": "code", "kind": "text", "lang": "c", "text": "#include <stdio.h>\n#include <stdlib.h>\n\ntypedef struct Node {\n    int value;\n    struct Node *next;\n} Node;\n\nNode *push(Node *head, int value) {\n    Node *n = malloc(sizeof *n);\n    if (!n) return head;\n    n->value = value;\n    n->next = head;\n    return n;\n}\n\nint main(void) {\n    Node *list = NULL;\n    for (int i = 0; i < 5; i++) list = push(list, i * i);\n    for (Node *p = list; p; p = p->next) printf(\"%d\\n\", p->value);\n    while (list) { Node *next = list->next; free(list); list = next; }\n    return 0;\n}\n", "messages": null, "license": "Apache-2.0", "source": "handwritten:c", "greedy": true, "greedy_prompt": null}
+{"seq_id": 62, "group": "code", "kind": "text", "lang": "rust", "text": "use std::collections::HashMap;\n\nfn word_counts(text: &str) -> HashMap<String, usize> {\n    let mut counts = HashMap::new();\n    for word in text.split_whitespace() {\n        let key = word.trim_matches(|c: char| !c.is_alphanumeric()).to_lowercase();\n        if !key.is_empty() {\n            *counts.entry(key).or_insert(0) += 1;\n        }\n    }\n    counts\n}\n\nfn main() {\n    let counts = word_counts(\"the quick brown fox jumps over the lazy dog the end\");\n    let mut pairs: Vec<_> = counts.into_iter().collect();\n    pairs.sort_by(|a, b| b.1.cmp(&a.1).then(a.0.cmp(&b.0)));\n    for (word, n) in pairs.iter().take(3) {\n        println!(\"{word}: {n}\");\n    }\n}\n", "messages": null, "license": "Apache-2.0", "source": "handwritten:rust", "greedy": true, "greedy_prompt": null}
+{"seq_id": 63, "group": "code", "kind": "text", "lang": "sql", "text": "-- Top 5 customers by total spend in 2025, with their order counts.\nSELECT c.id,\n       c.name,\n       COUNT(o.id)        AS orders,\n       SUM(o.total_cents) / 100.0 AS spend\nFROM customers AS c\nJOIN orders AS o ON o.customer_id = c.id\nWHERE o.created_at >= '2025-01-01'\n  AND o.created_at <  '2026-01-01'\n  AND o.status = 'paid'\nGROUP BY c.id, c.name\nHAVING SUM(o.total_cents) > 0\nORDER BY spend DESC\nLIMIT 5;\n", "messages": null, "license": "Apache-2.0", "source": "handwritten:sql", "greedy": true, "greedy_prompt": null}
+{"seq_id": 64, "group": "code", "kind": "text", "lang": "go", "text": "package main\n\nimport (\n\t\"fmt\"\n\t\"sync\"\n)\n\nfunc main() {\n\tvar wg sync.WaitGroup\n\tresults := make(chan int, 10)\n\tfor i := 1; i <= 10; i++ {\n\t\twg.Add(1)\n\t\tgo func(n int) {\n\t\t\tdefer wg.Done()\n\t\t\tresults <- n * n\n\t\t}(i)\n\t}\n\twg.Wait()\n\tclose(results)\n\tsum := 0\n\tfor r := range results {\n\t\tsum += r\n\t}\n\tfmt.Println(\"sum of squares:\", sum)\n}\n", "messages": null, "license": "Apache-2.0", "source": "handwritten:go", "greedy": true, "greedy_prompt": null}
+{"seq_id": 65, "group": "code", "kind": "text", "lang": "bash", "text": "#!/usr/bin/env bash\nset -euo pipefail\n\n# Keep only the newest N backups in a folder.\ndir=\"${1:?usage: prune.sh DIR [KEEP]}\"\nkeep=\"${2:-7}\"\n\nmapfile -t files < <(ls -1t \"$dir\"/*.tar.gz 2>/dev/null || true)\nif (( ${#files[@]} > keep )); then\n  for f in \"${files[@]:keep}\"; do\n    echo \"removing $f\"\n    rm -- \"$f\"\n  done\nfi\n", "messages": null, "license": "Apache-2.0", "source": "handwritten:bash", "greedy": true, "greedy_prompt": null}
+{"seq_id": 66, "group": "code", "kind": "text", "lang": "python", "text": "\"\"\"Pick the best quant that fits a memory budget (section 7.5 of the spec).\"\"\"\n\nfrom __future__ import annotations\n\nfrom dataclasses import dataclass, field\n\nfrom quantgap.models import ScoreResult\nfrom quantgap.util import GIB, human_size\n\nOVERHEAD_FACTOR = 1.1\nOVERHEAD_BYTES = GIB\nASSUMPTION = (\n    \"Memory estimate = file size x 1.1 + 1 GiB (context buffers at 2K context). \"\n    \"Real use depends on your engine and context length.\"\n)\nDAMAGE_WARN_GAP = 3  # prompts worse than the reference\n\n\ndef estimated_memory(size_bytes: int) -> float:\n    return size_bytes * OVERHEAD_FACTOR + OVERHEAD_BYTES\n\n\ndef _overlap(a: ScoreResult, b: ScoreResult) -> bool:\n    lo1, hi1 = a.metrics.kld_mean_ci95\n    lo2, hi2 = b.metrics.kld_mean_ci95\n    return lo1 <= hi2 and lo2 <= hi1\n\n\n@dataclass\nclass Recommendation:\n    sentence: str\n    chosen: ScoreResult | None\n    warnings: list[str] = field(default_factory=list)\n    not_clearly_different: list[ScoreResult] = field(default_factory=list)\n\n\ndef _name(r: ScoreResult) -> str:\n    return r.quant.label\n", "messages": null, "license": "Apache-2.0", "source": "quantgap:score/recommend.py", "greedy": true, "greedy_prompt": null}
+{"seq_id": 67, "group": "code", "kind": "text", "lang": "python", "text": "def _topk_for_sequence(model: Any, seq: RenderedSeq, k: int, device: Any) -> SeqTopK:\n    \"\"\"Forward pass for one sequence. Used for the unbatched path and for tests.\"\"\"\n    return _topk_batch(model, [seq], k, device)[0]\n\n\ndef _topk_batch(model: Any, seqs: list[RenderedSeq], k: int, device: Any) -> list[SeqTopK]:\n    import torch\n\n    max_len = max(len(s.token_ids) for s in seqs)\n    ids = torch.zeros((len(seqs), max_len), dtype=torch.long)\n    att = torch.zeros((len(seqs), max_len), dtype=torch.long)\n    for i, s in enumerate(seqs):\n        ids[i, : len(s.token_ids)] = torch.tensor(s.token_ids)\n        att[i, : len(s.token_ids)] = 1\n    with torch.no_grad():\n        logits = model(input_ids=ids.to(device), attention_mask=att.to(device)).logits\n    out: list[SeqTopK] = []\n    for i, s in enumerate(seqs):\n        mask = np.array(s.score_mask[1:], bool)  # position p scored if token p+1 is scored\n        pos = np.nonzero(mask)[0]\n        target = np.array(s.token_ids, np.int64)[pos + 1]\n        tops_i, tops_v, tgt_lp = [], [], []\n        for a in range(0, len(pos), POS_CHUNK):\n            sel = torch.from_numpy(pos[a : a + POS_CHUNK]).to(device)\n            lp = torch.log_softmax(logits[i, sel].float(), dim=-1)\n            v, ix = torch.topk(lp, k, dim=-1)\n            t = torch.from_numpy(target[a : a + POS_CHUNK]).to(device)\n            tgt_lp.append(lp.gather(1, t[:, None])[:, 0].cpu().numpy())\n            tops_i.append(ix.cpu().numpy())\n            tops_v.append(v.cpu().numpy())\n        if len(pos) == 0:\n            out.append(\n                SeqTopK(\n                    s.seq_id, pos, target, np.zeros((0, k), np.int64), np.zeros((0, k)), np.zeros(0)\n                )\n            )\n            continue\n        out.append(\n            SeqTopK(\n                s.seq_id,\n                pos.astype(np.int32),\n                target,\n                np.concatenate(tops_i),\n                np.concatenate(tops_v).astype(np.float64),\n                np.concatenate(tgt_lp).astype(np.float64),\n            )\n        )\n    return out\n", "messages": null, "license": "Apache-2.0", "source": "quantgap:pack/runner.py", "greedy": true, "greedy_prompt": null}
+{"seq_id": 68, "group": "code", "kind": "text", "lang": "python", "text": "def gather_from_logits(\n    logits: np.ndarray, target_ids: np.ndarray, gather_ids: np.ndarray\n) -> PromptLogprobs:\n    \"\"\"Shared maths for backends that hand back full logits.\n\n    logits: (n, V) raw scores, one row per scored position.\n    target_ids: (n,) the actual next token at each row.\n    gather_ids: (n, k) the reference's top-k ids at each row.\n    Uses a stable float64 log-softmax, in chunks so big vocabularies stay cheap.\n    \"\"\"\n    n, k = gather_ids.shape\n    gathered = np.empty((n, k), dtype=np.float64)\n    target = np.empty(n, dtype=np.float64)\n    top1 = np.empty(n, dtype=np.int64)\n    vocab = logits.shape[1]\n    for a in range(0, n, LOGSOFTMAX_CHUNK):\n        b = min(n, a + LOGSOFTMAX_CHUNK)\n        lp = log_softmax(logits[a:b])\n        ids = gather_ids[a:b]\n        oob = ids >= vocab  # reference ids beyond this model's vocabulary\n        g = np.take_along_axis(lp, np.where(oob, 0, ids), axis=1)\n        gathered[a:b] = np.where(oob, -np.inf, g)\n        tid = target_ids[a:b]\n        target[a:b] = lp[np.arange(b - a), np.minimum(tid, vocab - 1)]\n        top1[a:b] = lp.argmax(axis=1)\n    return PromptLogprobs(gathered, target, top1, approximate=False)\n", "messages": null, "license": "Apache-2.0", "source": "quantgap:backends/base.py", "greedy": false, "greedy_prompt": null}
+{"seq_id": 69, "group": "code", "kind": "text", "lang": "python", "text": "def print_compare(c: Console, results: list[ScoreResult]) -> None:\n    t = Table(show_header=True, header_style=\"bold\")\n    for col, j in ((\"quant\", \"left\"), (\"size\", \"right\"), (\"KLD mean\", \"right\"), (\"95% CI\", \"right\"),\n                   (\"worst 1%\", \"right\"), (\"top-1\", \"right\"), (\"ppl\", \"right\"), (\"loops\", \"right\"),\n                   (\"wrong script\", \"right\"), (\"JSON ok\", \"right\")):  # fmt: skip\n        t.add_column(col, justify=j)\n    for r in sorted(results, key=lambda r: r.metrics.kld_mean):\n        m, d = r.metrics, r.metrics.damage\n        t.add_row(\n            r.quant.label + (\" *\" if r.approximate else \"\") + (\" (quick)\" if r.quick else \"\"),\n            human_size(r.quant.size_bytes),\n            f\"{m.kld_mean:.4f}\",\n            f\"{m.kld_mean_ci95[0]:.4f}-{m.kld_mean_ci95[1]:.4f}\",\n            f\"{m.kld_p99:.3f}\",\n            f\"{m.top1_agree * 100:.1f}%\",\n            f\"{m.ppl_ratio:.3f}\",\n            f\"{d.loops}/{d.loops_ref}\" if d else \"-\",\n            f\"{d.wrong_script}\" if d else \"-\",\n            f\"{d.json_valid}/{d.json_n}\" if d else \"-\",\n        )\n    c.print(t)\n    if any(r.approximate for r in results):\n        c.print(\"* approximate (server top-N)\")\n    c.print(\"loops shows quant/original. Lower KLD is better.\")\n", "messages": null, "license": "Apache-2.0", "source": "quantgap:report/table.py", "greedy": false, "greedy_prompt": null}
+{"seq_id": 70, "group": "code", "kind": "text", "lang": "python", "text": "class RunnerInfo(_Model):\n    quantgap: str\n    torch: str | None = None\n    transformers: str | None = None\n    device: str | None = None\n\n\nclass Manifest(_Model):\n    schema_: str = Field(default=PACK_SCHEMA, alias=\"schema\")\n    model: ModelRef\n    tokenizer: TokenizerInfo\n    probes: ProbesInfo\n    chat_template_sha256: str | None = None\n    k: int\n    logp_dtype: Literal[\"float16\", \"float32\"] = \"float32\"\n    greedy: GreedyInfo\n    runner: RunnerInfo\n    files: dict[str, str] = Field(default_factory=dict)\n    created_at: str\n\n    model_config = ConfigDict(extra=\"forbid\", populate_by_name=True)\n\n    def canonical_json(self) -> str:\n        data = self.model_dump(by_alias=True, mode=\"json\")\n        return json.dumps(data, sort_keys=True, separators=(\",\", \":\"), ensure_ascii=False)\n\n    @property\n    def pack_id(self) -> str:\n        \"\"\"First 16 hex chars of SHA-256 over the canonical manifest.\"\"\"\n        return hashlib.sha256(self.canonical_json().encode(\"utf-8\")).hexdigest()[:16]\n\n    @property\n    def short_name(self) -> str:\n        return f\"{self.model.repo.split('/')[-1]}@{self.pack_id}\"\n\n\n# --------------------------------------------------------------------------- results\n\n\nclass QuantInfo(_Model):\n    source: str\n    sha256: str | None = None\n    size_bytes: int\n    label: str\n    publisher: str | None = None\n    format: str = \"gguf\"\n\n\nclass BackendInfo(_Model):\n    name: str\n    version: str\n    settings: dict[str, Any] = Field(default_factory=dict)\n\n\nclass GroupMetrics(_Model):\n    kld_mean: float\n    top1_agree: float\n    n_positions: int = 0\n", "messages": null, "license": "Apache-2.0", "source": "quantgap:models.py", "greedy": false, "greedy_prompt": null}
+{"seq_id": 71, "group": "code", "kind": "text", "lang": "python", "text": "def score_quant(\n    pack: Pack,\n    backend: Backend,\n    source: ResolvedSource,\n    settings: BackendSettings,\n    *,\n    groups: list[str] | None = None,\n    quick: bool = False,\n    do_damage: bool = True,\n    warnings: list[str] | None = None,\n    progress: Progress | None = None,\n) -> ScoreResult:\n    warnings = list(warnings or [])\n    manifest = pack.manifest\n    topk = pack.topk()\n    probes = {r.seq_id: r for r in pack.probes()}\n    seq_ids = select_sequences(pack, groups, quick)\n    check_context(pack, settings, seq_ids)\n\n    # Tokenizer check (never silently fixed).\n    fp = backend.tokenizer_fingerprint()\n    max_id = max(int(t.ids.max()) for t in topk.values() if len(t.ids))\n    try:\n        check_tokenizer(manifest.tokenizer, fp, max_id)\n    except TokenizerMismatch as e:\n        raise ScoreError(str(e)) from e\n\n    # E6: run the first sequence twice and see if the numbers move.\n    nondeterministic = False\n    first = next(s for s in seq_ids if len(topk[s].pos))\n    a = backend.prompt_logprobs(list(probes[first].token_ids), topk[first].pos, topk[first].ids)\n    b = backend.prompt_logprobs(list(probes[first].token_ids), topk[first].pos, topk[first].ids)\n    both = ~np.isnan(a.gathered) & ~np.isnan(b.gathered)\n    if both.any() and np.max(np.abs(a.gathered[both] - b.gathered[both])) > NONDETERMINISM_TOL:\n        nondeterministic = True\n        warnings.append(\"Running the same text twice gave different numbers (non-deterministic).\")\n\n    scores: list[_SeqScore] = []\n    approximate = False\n    clamps = 0\n    for i, sid in enumerate(seq_ids):\n        if progress:\n            progress(\"scoring\", i, len(seq_ids))\n        t = topk[sid]\n        if len(t.pos) == 0:\n            continue\n        pl = backend.prompt_logprobs(list(probes[sid].token_ids), t.pos, t.ids)\n        approximate |= pl.approximate\n        res = kld_topk(t.logp, pl.gathered)\n        clamps += res.clamps\n        scores.append(\n            _SeqScore(\n                probes[sid].group,\n                res.kl,\n                pl.top1 == t.ids[:, 0],\n                pl.target,\n                t.logp_target,\n            )\n        )\n    if progress:\n        progress(\"scoring\", len(seq_ids), len(seq_ids))\n\n    kl = np.concatenate([s.kl for s in scores])\n    agree = np.concatenate([s.agree for s in scores])\n    lq = np.concatenate([s.lq_target for s in scores])\n    lp = np.concatenate([s.lp_target for s in scores])\n    sums = np.array([s.kl.sum() for s in scores])\n    counts = np.array([len(s.kl) for s in scores])\n    ci = bootstrap_mean_ci(sums, counts)\n\n    by_group: dict[str, GroupMetrics] = {}\n    for g in dict.fromkeys(s.group for s in scores):\n        sel = [s for s in scores if s.group == g]\n        gk = np.concatenate([s.kl for s in sel])\n        ga = np.concatenate([s.agree for s in sel])\n        by_group[g] = GroupMetrics(\n            kld_mean=jsonable_float(float(gk.mean())),\n            top1_agree=jsonable_float(float(ga.mean())),\n            n_positions=len(gk),\n        )\n\n    if clamps > max(10, 0.001 * len(kl)):\n        warnings.append(\n            f\"{clamps} probabilities were raised to 1e-12 before taking logs. A high number \"\n            \"means something is broken (wrong tokenizer or a damaged model).\"\n        )\n    if approximate:\n        warnings.append(\n            \"approximate (server top-N): the server returned only its own top-N, so some \"\n            \"reference ids were counted in the tail bucket.\"\n        )\n\n    damage_metrics = None\n    if do_damage:\n        damage_metrics = run_damage(backend, pack, quick, progress)\n\n    metrics = Metrics(\n        kld_mean=jsonable_float(float(kl.mean())),\n        kld_mean_ci95=(jsonable_float(ci[0]), jsonable_float(ci[1])),\n        kld_p99=jsonable_float(percentile(kl, 99)),\n        kld_p999=jsonable_float(percentile(kl, 99.9)),\n        top1_agree=jsonable_float(float(agree.mean())),\n        ppl_ratio=jsonable_float(ppl_ratio(lq, lp)),\n        n_positions=len(kl),\n        n_sequences=len(scores),\n        clamp_count=clamps,\n        by_group=by_group,\n        damage=damage_metrics,\n    )\n    return ScoreResult(\n        pack_id=pack.pack_id,\n        pack_name=manifest.short_name,\n        quant=QuantInfo(\n            source=source.source,\n            sha256=source.sha256,\n            size_bytes=source.size_bytes,\n            label=source.label,\n            publisher=source.publisher,\n            format=source.format,\n        ),\n        backend=BackendInfo(name=backend.name, version=backend.version(), settings=settings.public()),\n        metrics=metrics,\n        hardware=hardware_string(),\n        quantgap=__version__,\n        created_at=utc_now(),\n        quick=quick,\n        approximate=approximate,\n        nondeterministic=nondeterministic,\n        warnings=warnings,\n    )\n", "messages": null, "license": "Apache-2.0", "source": "quantgap:score/run.py", "greedy": false, "greedy_prompt": null}
+{"seq_id": 72, "group": "code", "kind": "text", "lang": "python", "text": "def gguf_base_model(path: str | Path) -> str | None:\n    \"\"\"The `org/repo` of the model this GGUF was made from, when the file says so.\"\"\"\n    keys = {\"general.base_model.0.repo_url\", \"general.base_model.0.name\"}\n    try:\n        meta = read_gguf_metadata(path, keys)\n    except (OSError, ResolveError, struct.error, KeyError):\n        return None\n    url = meta.get(\"general.base_model.0.repo_url\", \"\")\n    m = re.search(r\"huggingface\\.co/([^/]+/[^/?#]+)\", url)\n    if m:\n        return m.group(1)\n    return meta.get(\"general.base_model.0.name\")\n\n\n# ------------------------------------------------------------------ resolve\n\n\ndef _dir_format(path: Path) -> str:\n    cfg = path / \"config.json\"\n    if cfg.is_file() and '\"quantization\"' in cfg.read_text(encoding=\"utf-8\", errors=\"ignore\"):\n        return \"mlx\"\n    return \"hf\"\n\n\ndef _check_disk(size: int, cache_dir: str | None) -> None:\n    from huggingface_hub import constants\n\n    where = cache_dir or constants.HF_HUB_CACHE\n    free = free_disk_bytes(where)\n    need = int(size * 1.1)\n    if free < need:\n        raise ResolveError(\n            f\"Not enough disk space in {where}: need about {human_size(need)}, \"\n            f\"{human_size(free)} free. Set HF_HOME to a bigger drive.\"\n        )\n    print(f\"Downloading to {where} ({human_size(size)}, {human_size(free)} free)\")\n", "messages": null, "license": "Apache-2.0", "source": "quantgap:hub/resolve.py", "greedy": false, "greedy_prompt": null}
+{"seq_id": 73, "group": "code", "kind": "text", "lang": "python", "text": "\"\"\"GGUF backend through llama-cpp-python (needs `quantgap[llamacpp]`).\n\nVerified fact (docs/DECISIONS.md, M0): `Llama(..., logits_all=True)` followed by `eval(tokens)`\nfills `llm.scores`, an array of shape (n_ctx, n_vocab). Row i holds the logits that come out\nafter reading token i, so it predicts token i + 1.\n\"\"\"\n\nfrom __future__ import annotations\n\nfrom typing import Any\n\nimport numpy as np\n\nfrom quantgap.backends.base import (\n    BackendSettings,\n    PromptLogprobs,\n    ResolvedSource,\n    gather_from_logits,\n)\nfrom quantgap.tokcheck import CHECK_STRING, TokFingerprint\n\n\nclass LlamaCppBackend:\n    name = \"llamacpp\"\n\n    def __init__(self) -> None:\n        self.llm: Any = None\n        self.settings: BackendSettings | None = None\n\n    def load(self, source: ResolvedSource, settings: BackendSettings) -> None:\n        try:\n            from llama_cpp import Llama\n        except ImportError as e:  # pragma: no cover\n            raise RuntimeError(\n                'llama-cpp-python is not installed. Run: pip install \"quantgap[llamacpp]\"'\n            ) from e\n        if source.path is None or not source.path.is_file():\n            raise FileNotFoundError(f\"GGUF file not found: {source.path}\")\n        self.settings = settings\n        kwargs: dict[str, Any] = {\n            \"model_path\": str(source.path),\n            \"n_ctx\": settings.ctx,\n            \"n_gpu_layers\": settings.n_gpu_layers,\n            \"logits_all\": True,\n            \"verbose\": False,\n        }\n        if settings.n_threads:\n            kwargs[\"n_threads\"] = settings.n_threads\n        self.llm = Llama(**kwargs)\n\n    def tokenizer_fingerprint(self) -> TokFingerprint:\n        llm = self.llm\n        ids = llm.tokenize(CHECK_STRING.encode(\"utf-8\"), add_bos=False, special=False)\n        return TokFingerprint(\n            vocab_size=int(llm.n_vocab()),\n            bos_id=int(llm.token_bos()) if llm.token_bos() >= 0 else None,\n            eos_id=int(llm.token_eos()) if llm.token_eos() >= 0 else None,\n            check_ids=[int(i) for i in ids],\n        )\n\n    def prompt_logprobs(\n        self, token_ids: list[int], positions: np.ndarray, gather_ids: np.ndarray\n    ) -> PromptLogprobs:\n        llm = self.llm\n        llm.reset()\n        llm.eval(token_ids)\n        positions = np.asarray(positions)\n        rows = np.asarray(llm.scores[positions], dtype=np.float32)\n        tgt = np.asarray(token_ids, dtype=np.int64)[positions + 1]\n        return gather_from_logits(rows, tgt, gather_ids)\n\n    def generate_greedy(self, token_ids: list[int], max_new_tokens: int) -> list[int]:\n        llm = self.llm\n        llm.reset()\n        out: list[int] = []\n        eos = llm.token_eos()\n        for tok in llm.generate(token_ids, top_k=1, top_p=1.0, temp=0.0, reset=True):\n            if tok == eos or len(out) >= max_new_tokens:\n                break\n            out.append(int(tok))\n        return out\n\n    def decode(self, ids: list[int]) -> str:\n        return self.llm.detokenize(ids).decode(\"utf-8\", errors=\"replace\")\n\n    def version(self) -> str:\n        import llama_cpp\n\n        return f\"llama-cpp-python {llama_cpp.__version__}\"\n\n    def close(self) -> None:\n        if self.llm is not None:\n            self.llm.close()\n            self.llm = None\n", "messages": null, "license": "Apache-2.0", "source": "quantgap:backends/llamacpp.py", "greedy": false, "greedy_prompt": null}
+{"seq_id": 74, "group": "code", "kind": "text", "lang": "python", "text": "class MLXBackend:\n    name = \"mlx\"\n\n    def __init__(self) -> None:\n        self.model: Any = None\n        self.tok: Any = None\n\n    def load(self, source: ResolvedSource, settings: BackendSettings) -> None:\n        try:\n            from mlx_lm import load\n        except ImportError as e:  # pragma: no cover\n            raise RuntimeError('mlx-lm is not installed. Run: pip install \"quantgap[mlx]\"') from e\n        where = str(source.path) if source.path else source.source.removeprefix(\"hf://\")\n        self.model, self.tok = load(where)\n\n    def _hf_tok(self) -> Any:\n        return getattr(self.tok, \"_tokenizer\", self.tok)\n\n    def tokenizer_fingerprint(self) -> TokFingerprint:\n        t = self._hf_tok()\n        ids = list(t(CHECK_STRING, add_special_tokens=False)[\"input_ids\"])\n        vocab = len(t)\n        try:  # the real embedding size is what matters for padding ids\n            vocab = int(self.model.model.embed_tokens.weight.shape[0])\n        except Exception:\n            pass\n        return TokFingerprint(vocab, t.bos_token_id, t.eos_token_id, ids)\n\n    def prompt_logprobs(\n        self, token_ids: list[int], positions: np.ndarray, gather_ids: np.ndarray\n    ) -> PromptLogprobs:\n        import mlx.core as mx\n\n        logits = self.model(mx.array([token_ids]))[0]\n        rows = np.array(logits.astype(mx.float32))[np.asarray(positions)]\n        tgt = np.asarray(token_ids, dtype=np.int64)[np.asarray(positions) + 1]\n        return gather_from_logits(rows, tgt, gather_ids)\n\n    def generate_greedy(self, token_ids: list[int], max_new_tokens: int) -> list[int]:\n        import mlx.core as mx\n\n        ids = list(token_ids)\n        out: list[int] = []\n        eos = self.tok.eos_token_id\n        for _ in range(max_new_tokens):  # no KV cache: slow but simple and exact\n            logits = self.model(mx.array([ids]))[0, -1]\n            nxt = int(mx.argmax(logits).item())\n            if nxt == eos:\n                break\n            out.append(nxt)\n            ids.append(nxt)\n        return out\n\n    def decode(self, ids: list[int]) -> str:\n        return self.tok.decode(ids)\n\n    def version(self) -> str:\n        import mlx_lm\n\n        return f\"mlx-lm {getattr(mlx_lm, '__version__', 'unknown')}\"\n\n    def close(self) -> None:\n        self.model = None\n        self.tok = None\n", "messages": null, "license": "Apache-2.0", "source": "quantgap:backends/mlx.py", "greedy": false, "greedy_prompt": null}
+{"seq_id": 75, "group": "code", "kind": "text", "lang": "python", "text": "class PackWriter:\n    def __init__(self, out_dir: str | Path, k: int, logp_dtype: str = \"float32\"):\n        self.out = Path(out_dir)\n        self.out.mkdir(parents=True, exist_ok=True)\n        self.k = k\n        self.logp_dtype = logp_dtype\n        self._np_dtype = np.float16 if logp_dtype == \"float16\" else np.float32\n        self._pa_dtype = pa.float16() if logp_dtype == \"float16\" else pa.float32()\n        self._buf: list[SeqTopK] = []\n        self._buf_rows = 0\n        self._shards: list[str] = []\n        self.n_positions = 0\n\n    # ---- top-k shards\n    def add_seq(self, s: SeqTopK) -> None:\n        if s.ids.shape != (len(s.pos), self.k):\n            raise ValueError(f\"ids must have shape (n_positions, {self.k}), got {s.ids.shape}\")\n        self._buf.append(s)\n        self._buf_rows += len(s.pos)\n        self.n_positions += len(s.pos)\n        if self._buf_rows >= SHARD_ROWS:\n            self._flush()\n\n    def _flush(self) -> None:\n        if not self._buf:\n            return\n        seq_id = np.concatenate([np.full(len(s.pos), s.seq_id, np.int32) for s in self._buf])\n        pos = np.concatenate([s.pos for s in self._buf]).astype(np.int32)\n        target = np.concatenate([s.target for s in self._buf]).astype(np.int32)\n        ids = np.concatenate([s.ids for s in self._buf]).astype(np.int32)\n        logp = np.concatenate([s.logp for s in self._buf]).astype(self._np_dtype)\n        lpt = np.concatenate([s.logp_target for s in self._buf]).astype(np.float32)\n        table = pa.table(\n            {\n                \"seq_id\": pa.array(seq_id, pa.int32()),\n                \"pos\": pa.array(pos, pa.int32()),\n                \"target\": pa.array(target, pa.int32()),\n                \"ids\": _fixed_list(ids, self.k, pa.int32()),\n                \"logp\": _fixed_list(logp, self.k, self._pa_dtype),\n                \"logp_target\": pa.array(lpt, pa.float32()),\n            }\n        )\n        name = f\"topk-{len(self._shards):05d}.parquet\"\n        pq.write_table(table, self.out / name, compression=_COMPRESSION)\n        self._shards.append(name)\n        self._buf, self._buf_rows = [], 0\n\n    # ---- probes and greedy\n    def write_probes(self, seqs: Iterable[RenderedSeq]) -> None:\n        seqs = list(seqs)\n        table = pa.table(\n            {\n                \"seq_id\": pa.array([s.seq_id for s in seqs], pa.int32()),\n                \"group\": pa.array([s.group for s in seqs], pa.string()),\n                \"token_ids\": _ragged_list(\n                    [np.array(s.token_ids, np.int32) for s in seqs], pa.int32()\n                ),\n                \"n_tokens\": pa.array([len(s.token_ids) for s in seqs], pa.int32()),\n                \"score_mask\": _ragged_list(\n                    [np.array(s.score_mask, bool) for s in seqs], pa.bool_()\n                ),\n            }\n        )\n        pq.write_table(table, self.out / \"probes.parquet\", compression=_COMPRESSION)\n\n    def write_greedy(self, rows: list[dict]) -> None:\n        \"\"\"rows: dicts with seq_id, group, lang, prompt_ids, ref_ids, ref_text.\"\"\"\n        table = pa.table(\n            {\n                \"seq_id\": pa.array([r[\"seq_id\"] for r in rows], pa.int32()),\n                \"group\": pa.array([r[\"group\"] for r in rows], pa.string()),\n                \"lang\": pa.array([r[\"lang\"] for r in rows], pa.string()),\n                \"prompt_ids\": _ragged_list(\n                    [np.array(r[\"prompt_ids\"], np.int32) for r in rows], pa.int32()\n                ),\n                \"ref_ids\": _ragged_list(\n                    [np.array(r[\"ref_ids\"], np.int32) for r in rows], pa.int32()\n                ),\n                \"ref_text\": pa.array([r[\"ref_text\"] for r in rows], pa.string()),\n            }\n        )\n        pq.write_table(table, self.out / \"greedy.parquet\", compression=_COMPRESSION)\n\n    # ---- manifest\n    def finalize(\n        self,\n        *,\n        model: ModelRef,\n        tokenizer: TokenizerInfo,\n        probes_set: str,\n        probes_sha256: str,\n        n_sequences: int,\n        chat_template_sha256: str | None,\n        greedy: GreedyInfo,\n        runner: RunnerInfo,\n    ) -> Manifest:\n        self._flush()\n        files = {\n            name: sha256_file(self.out / name)\n            for name in sorted([*self._shards, \"probes.parquet\", \"greedy.parquet\"])\n        }\n        m = Manifest(\n            model=model,\n            tokenizer=tokenizer,\n            probes=ProbesInfo(\n                set=probes_set,\n                sha256=probes_sha256,\n                n_sequences=n_sequences,\n                n_positions=self.n_positions,\n            ),\n            chat_template_sha256=chat_template_sha256,\n            k=self.k,\n            logp_dtype=self.logp_dtype,\n            greedy=greedy,\n            runner=runner,\n            files=files,\n            created_at=utc_now(),\n        )\n        text = json.dumps(m.model_dump(by_alias=True, mode=\"json\"), indent=2, ensure_ascii=False)\n        (self.out / \"manifest.json\").write_text(text + \"\\n\", encoding=\"utf-8\")\n        return m\n", "messages": null, "license": "Apache-2.0", "source": "quantgap:pack/writer.py", "greedy": false, "greedy_prompt": null}
+{"seq_id": 76, "group": "code", "kind": "text", "lang": "python", "text": "def check_tokenizer(pack_tok: TokenizerInfo, quant: TokFingerprint, pack_top_ids_max: int) -> None:\n    \"\"\"Raise TokenizerMismatch if the quant cannot be compared to the pack.\n\n    `pack_top_ids_max` is the largest token id seen anywhere in the pack's top-k lists and targets.\n    A quant may have extra padding ids beyond the reference vocabulary if those never appear there.\n    \"\"\"\n    # 1. vocabulary size\n    # A server cannot tell us its vocabulary size (vocab_size <= 0): skip this one check.\n    if quant.vocab_size > 0 and quant.vocab_size != pack_tok.vocab_size:\n        if quant.vocab_size < pack_tok.vocab_size:\n            raise TokenizerMismatch(\n                f\"Check 1 (vocab size) failed: quant has {quant.vocab_size} tokens, \"\n                f\"pack expects {pack_tok.vocab_size}.\"\n            )\n        if pack_top_ids_max >= pack_tok.vocab_size:  # pragma: no cover - defensive\n            raise TokenizerMismatch(\"Check 1 (vocab size) failed: pack ids exceed its own vocabulary.\")\n        # extra padding ids are fine: no pack id points at them.\n    # 2. fixed check string\n    if pack_tok.check_ids and quant.check_ids != pack_tok.check_ids:\n        a, b = pack_tok.check_ids, quant.check_ids\n        i = next((j for j in range(min(len(a), len(b))) if a[j] != b[j]), min(len(a), len(b)))\n        got = b[i] if i < len(b) else \"(end)\"\n        want = a[i] if i < len(a) else \"(end)\"\n        raise TokenizerMismatch(\n            f\"Check 2 (encode a fixed test string) failed at token index {i}: \"\n            f\"pack has id {want}, quant has id {got}. \"\n            \"These models do not use the same tokenizer, so scores would be meaningless.\"\n        )\n    # 3. BOS / EOS ids\n    for label, want, got in (\n        (\"BOS\", pack_tok.bos_id, quant.bos_id),\n        (\"EOS\", pack_tok.eos_id, quant.eos_id),\n    ):\n        if want is not None and got is not None and want != got:\n            raise TokenizerMismatch(\n                f\"Check 3 ({label} id) failed: pack has {want}, quant has {got}.\"\n            )\n", "messages": null, "license": "Apache-2.0", "source": "quantgap:tokcheck.py", "greedy": false, "greedy_prompt": null}
+{"seq_id": 77, "group": "code", "kind": "text", "lang": "python", "text": "def resolve_quant(\n    quant: str,\n    *,\n    base_url: str | None = None,\n    model_name: str | None = None,\n    compute_hash: bool = True,\n) -> ResolvedSource:\n    \"\"\"Find a quant. Downloads from the Hub when given an hf:// URL.\"\"\"\n    if quant.startswith(\"openai://\"):\n        if not base_url:\n            raise ResolveError(\"openai:// needs --base-url, for example http://localhost:8000/v1\")\n        name = quant[len(\"openai://\") :] or (model_name or \"\")\n        return ResolvedSource(\n            source=quant,\n            label=name or \"server-model\",\n            format=\"openai\",\n            extra={\"model_name\": model_name or name},\n        )\n    if quant.startswith(\"hf://\"):\n        ref = parse_hf_url(quant)\n        return _resolve_hf(ref, quant)\n    p = Path(quant).expanduser()\n    if p.is_file():\n        src = ResolvedSource(\n            source=str(p),\n            label=label_from_filename(p.name) if p.suffix == \".gguf\" else p.stem,\n            path=p,\n            size_bytes=p.stat().st_size,\n            format=\"gguf\" if p.suffix.lower() == \".gguf\" else \"hf\",\n        )\n        if src.format == \"gguf\":\n            src.base_model = gguf_base_model(p)\n        if compute_hash:\n            src.sha256 = sha256_file(p)\n        return src\n    if p.is_dir():\n        size = sum(f.stat().st_size for f in p.rglob(\"*\") if f.is_file())\n        return ResolvedSource(\n            source=str(p), label=p.name, path=p, size_bytes=size, format=_dir_format(p)\n        )\n    raise ResolveError(\n        f\"Cannot find {quant!r}. Use a local file or folder, hf://org/repo/file.gguf, \"\n        \"or openai://NAME with --base-url.\"\n    )\n", "messages": null, "license": "Apache-2.0", "source": "quantgap:hub/resolve.py", "greedy": false, "greedy_prompt": null}
+{"seq_id": 78, "group": "code", "kind": "text", "lang": "python", "text": "def dataset_card(m: Manifest) -> str:\n    lic = m.model.license or \"unknown (check the model page before sharing)\"\n    return f\"\"\"---\nlicense: cc-by-4.0\ntags:\n- {PACK_TAG}\n- quantization\npretty_name: quantgap pack for {m.model.repo}\n---\n\n# quantgap pack: {m.model.repo}\n\nPack id: `{m.pack_id}`. Made with quantgap {m.runner.quantgap}.\n\nThis pack holds the reference model's top-{m.k} next-token log probabilities on the quantgap\nprobe set `{m.probes.set}` ({m.probes.n_positions:,} positions). Use it with\n`quantgap score QUANT --pack <this dataset>` to measure how far a quantized copy of the model\ndrifts from the original, without running the original.\n\n- Reference model: `{m.model.repo}` @ `{m.model.revision}` ({m.model.dtype})\n- Reference model license: **{lic}**\n- The pack contains model outputs (log probabilities) on openly licensed texts. It does not\n  contain the model weights.\n\"\"\"\n\n\ndef push_pack(pack_dir: str | Path, repo_id: str, token: str | None = None) -> str:\n    from huggingface_hub import HfApi\n\n    pack = Pack.load(pack_dir)\n    api = HfApi(token=token)\n    api.create_repo(repo_id, repo_type=\"dataset\", exist_ok=True)\n    (Path(pack_dir) / \"README.md\").write_text(dataset_card(pack.manifest), encoding=\"utf-8\")\n    api.upload_folder(folder_path=str(pack_dir), repo_id=repo_id, repo_type=\"dataset\")\n    return f\"https://huggingface.co/datasets/{repo_id}\"\n", "messages": null, "license": "Apache-2.0", "source": "quantgap:hub/packs.py", "greedy": false, "greedy_prompt": null}
+{"seq_id": 79, "group": "code", "kind": "text", "lang": "python", "text": "def _resolve_hf(ref: HFRef, original: str) -> ResolvedSource:\n    from huggingface_hub import HfApi, hf_hub_download, snapshot_download\n\n    api = HfApi()\n    if ref.path and ref.path.endswith(\".gguf\"):\n        info = api.get_paths_info(ref.repo, [ref.path])\n        if not info:\n            raise ResolveError(f\"{ref.path} was not found in {ref.repo}.\")\n        size = info[0].size or 0\n        sha = info[0].lfs.sha256 if getattr(info[0], \"lfs\", None) else None\n        _check_disk(size, None)\n        local = Path(hf_hub_download(ref.repo, ref.path))\n        if sha is None:\n            sha = sha256_file(local)\n        return ResolvedSource(\n            source=f\"hf://{ref.repo}/{ref.path}\",\n            label=label_from_filename(ref.path),\n            path=local,\n            size_bytes=size or local.stat().st_size,\n            format=\"gguf\",\n            publisher=ref.repo.split(\"/\")[0],\n            base_model=gguf_base_model(local),\n            sha256=sha,\n        )\n    if ref.path:\n        raise ResolveError(\n            f\"{original!r}: only .gguf files or whole repos (for MLX or Transformers) are supported.\"\n        )\n    # whole repo: MLX or Transformers folder\n    files = api.model_info(ref.repo, files_metadata=True).siblings or []\n    size = sum(s.size or 0 for s in files if s.rfilename.endswith((\".safetensors\", \".bin\")))\n    _check_disk(size, None)\n    local = Path(snapshot_download(ref.repo))\n    return ResolvedSource(\n        source=f\"hf://{ref.repo}\",\n        label=ref.repo.split(\"/\")[-1],\n        path=local,\n        size_bytes=size,\n        format=_dir_format(local),\n        publisher=ref.repo.split(\"/\")[0],\n    )\n", "messages": null, "license": "Apache-2.0", "source": "quantgap:hub/resolve.py", "greedy": false, "greedy_prompt": null}
+{"seq_id": 80, "group": "code", "kind": "text", "lang": "python", "text": "def draw_chart(\n    results: list[ScoreResult],\n    out: str | Path,\n    title: str | None = None,\n    budget_bytes: int | None = None,\n) -> Path:\n    out = Path(out)\n    fig, ax = plt.subplots(figsize=(8, 5), dpi=160)\n    fig.patch.set_facecolor(SURFACE)\n    ax.set_facecolor(SURFACE)\n    rs = sorted(results, key=lambda r: r.quant.size_bytes)\n    xs = [r.quant.size_bytes / 1e9 for r in rs]\n    ys = [r.metrics.kld_mean for r in rs]\n    lo = [r.metrics.kld_mean - r.metrics.kld_mean_ci95[0] for r in rs]\n    hi = [r.metrics.kld_mean_ci95[1] - r.metrics.kld_mean for r in rs]\n    ax.errorbar(\n        xs, ys, yerr=[[max(0, v) for v in lo], [max(0, v) for v in hi]],\n        fmt=\"none\", ecolor=MARK, elinewidth=1, alpha=0.45, capsize=0,\n    )  # fmt: skip\n    front = pareto(rs)\n    ax.plot(\n        [r.quant.size_bytes / 1e9 for r in front], [r.metrics.kld_mean for r in front],\n        color=MARK, linewidth=1.5, alpha=0.5, zorder=2,\n    )  # fmt: skip\n    ax.scatter(xs, ys, s=64, color=MARK, edgecolor=SURFACE, linewidth=2, zorder=3)\n    for r, x, y in zip(rs, xs, ys, strict=True):\n        label = r.quant.label + (\" *\" if r.approximate else \"\")\n        ax.annotate(\n            label, (x, y), textcoords=\"offset points\", xytext=(7, 6), fontsize=8, color=INK\n        )\n    if budget_bytes:\n        ax.axvline(budget_bytes / 1e9, color=MUTED, linewidth=1, linestyle=(0, (4, 3)))\n        ax.annotate(\n            \"memory budget\", (budget_bytes / 1e9, 1), xycoords=(\"data\", \"axes fraction\"),\n            xytext=(-4, -12), textcoords=\"offset points\", ha=\"right\", fontsize=8, color=MUTED,\n        )  # fmt: skip\n    ax.set_xlabel(\"File size (GB)\", color=MUTED, fontsize=9)\n    ax.set_ylabel(\"KLD mean vs original (lower is better)\", color=MUTED, fontsize=9)\n    if min(ys) > 0 and max(ys) / max(min(ys), 1e-12) > 30:\n        ax.set_yscale(\"log\")\n    ax.grid(axis=\"y\", color=GRID, linewidth=0.8)\n    ax.set_axisbelow(True)\n    for side in (\"top\", \"right\", \"left\"):\n        ax.spines[side].set_visible(False)\n    ax.spines[\"bottom\"].set_color(GRID)\n    ax.tick_params(colors=MUTED, labelsize=8, length=0)\n    pack = results[0].pack_name or results[0].pack_id\n    ax.set_title(title or f\"Quality vs size  |  pack {pack}\", loc=\"left\", color=INK, fontsize=11)\n    if any(r.approximate for r in rs):\n        fig.text(0.01, 0.01, \"* approximate (server top-N)\", fontsize=7, color=FLAG)\n    fig.tight_layout()\n    fig.savefig(out, facecolor=SURFACE)\n    plt.close(fig)\n    return out\n", "messages": null, "license": "Apache-2.0", "source": "quantgap:report/chart.py", "greedy": false, "greedy_prompt": null}
+{"seq_id": 81, "group": "code", "kind": "text", "lang": "python", "text": "def _score_source(\n    src: ResolvedSource, pack, cfg: Settings, *, backend_name: str, ctx, n_gpu_layers,\n    base_url, model_name, groups, quick: bool, do_damage: bool,\n) -> ScoreResult:  # fmt: skip\n    from quantgap.hub.packs import fine_tune_warning\n    from quantgap.score.run import score_quant\n\n    name = backend_name if backend_name != \"auto\" else pick_backend_name(src, base_url)\n    settings = _settings_for(cfg, name, ctx, n_gpu_layers, base_url, model_name)\n    warns = []\n    if (w := fine_tune_warning(src.base_model, pack)) is not None:\n        warns.append(w)\n    backend = get_backend(name)\n    backend.load(src, settings)\n    prog, cb = _progress()\n    try:\n        with prog:\n            return score_quant(\n                pack, backend, src, settings, groups=groups, quick=quick,\n                do_damage=do_damage, warnings=warns, progress=cb,\n            )  # fmt: skip\n    finally:\n        backend.close()\n\n\ndef _get_pack(pack_ref: str | None, src: ResolvedSource, cfg: Settings):\n    from quantgap.hub.packs import auto_find_pack, load_pack\n\n    if pack_ref:\n        return load_pack(pack_ref)\n    return auto_find_pack(src.base_model, cfg)\n\n\ndef _resolve(quant: str, base_url, model_name) -> ResolvedSource:\n    from quantgap.hub.resolve import ResolveError, resolve_quant\n\n    try:\n        return resolve_quant(quant, base_url=base_url, model_name=model_name)\n    except ResolveError as e:\n        hint = OLLAMA_HINT if (\":\" in quant and \"/\" not in quant and \"\\\\\" not in quant) else \"\"\n        raise _fail(f\"{e}\\n{hint}\".strip()) from e\n\n\n# ---------------------------------------------------------------------------- score\n", "messages": null, "license": "Apache-2.0", "source": "quantgap:cli.py", "greedy": false, "greedy_prompt": null}
+{"seq_id": 82, "group": "code", "kind": "text", "lang": "python", "text": "\"\"\"Tokenizer compatibility check.\n\nA quant can only be compared to a pack if both use the same token ids. We check this before\nscoring and stop with a clear error if it fails. We never \"fix\" a mismatch silently.\n\"\"\"\n\nfrom __future__ import annotations\n\nfrom dataclasses import dataclass, field\n\nfrom quantgap.models import TokenizerInfo\n\n# Mixed scripts, code, numbers, emoji, odd spacing. Special tokens are checked by id, not by text.\nCHECK_STRING = (\n    \"Hello, world! 你好，世界。 नमस्ते दुनिया. مرحبا بالعالم. ¿Qué tal?\\n\"\n    \"def f(x):\\n    return {'a': [1, 2.5e-3, None]}  # 😀🚀\\n\"\n    \"  two  spaces\\ttab 12345 3.14159 https://example.com/a?b=c&d=e\\n\"\n)\n\n\nclass TokenizerMismatch(RuntimeError):\n    \"\"\"The quant's tokenizer is not the same as the pack's.\"\"\"\n\n\n@dataclass\nclass TokFingerprint:\n    vocab_size: int\n    bos_id: int | None\n    eos_id: int | None\n    check_ids: list[int] = field(default_factory=list)\n", "messages": null, "license": "Apache-2.0", "source": "quantgap:tokcheck.py", "greedy": false, "greedy_prompt": null}
+{"seq_id": 83, "group": "code", "kind": "text", "lang": "python", "text": "\"\"\"Backend adapters. Each one is imported only when it is asked for, so the default install\n(no torch, no llama.cpp) stays light.\"\"\"\n\nfrom __future__ import annotations\n\nimport platform\n\nfrom quantgap.backends.base import (\n    Backend,\n    BackendSettings,\n    PromptLogprobs,\n    ResolvedSource,\n    gather_from_logits,\n)\n\nBACKEND_NAMES = (\"llamacpp\", \"mlx\", \"openai\", \"hf\")\n\n__all__ = [\n    \"BACKEND_NAMES\",\n    \"Backend\",\n    \"BackendSettings\",\n    \"PromptLogprobs\",\n    \"ResolvedSource\",\n    \"gather_from_logits\",\n    \"get_backend\",\n    \"pick_backend_name\",\n]\n\n\ndef get_backend(name: str) -> Backend:\n    if name == \"llamacpp\":\n        from quantgap.backends.llamacpp import LlamaCppBackend\n\n        return LlamaCppBackend()\n    if name == \"mlx\":\n        if not (platform.system() == \"Darwin\" and platform.machine() == \"arm64\"):\n            raise RuntimeError(\"The mlx backend only works on Apple silicon Macs.\")\n        from quantgap.backends.mlx import MLXBackend\n\n        return MLXBackend()\n    if name == \"openai\":\n        from quantgap.backends.openai_compat import OpenAIBackend\n\n        return OpenAIBackend()\n    if name == \"hf\":\n        from quantgap.backends.hf import HFBackend\n\n        return HFBackend()\n    raise ValueError(f\"Unknown backend {name!r}. Choose one of: {', '.join(BACKEND_NAMES)}.\")\n\n\ndef pick_backend_name(src: ResolvedSource, base_url: str | None) -> str:\n    \"\"\"Guess a backend from what the quant is.\"\"\"\n    if src.format == \"openai\" or base_url:\n        return \"openai\"\n    if src.format == \"gguf\":\n        return \"llamacpp\"\n    if src.format == \"mlx\":\n        return \"mlx\"\n    return \"hf\"\n", "messages": null, "license": "Apache-2.0", "source": "quantgap:backends/__init__.py", "greedy": false, "greedy_prompt": null}
+{"seq_id": 84, "group": "code", "kind": "text", "lang": "python", "text": "\"\"\"Small helpers shared by several modules.\"\"\"\n\nfrom __future__ import annotations\n\nimport hashlib\nimport platform\nimport re\nimport shutil\nfrom datetime import UTC, datetime\nfrom pathlib import Path\n\n_SIZE_RE = re.compile(r\"^\\s*([0-9]*\\.?[0-9]+)\\s*([kmgt]?i?b?)\\s*$\", re.IGNORECASE)\n_UNITS = {\n    \"\": 1,\n    \"b\": 1,\n    \"k\": 1000,\n    \"kb\": 1000,\n    \"m\": 1000**2,\n    \"mb\": 1000**2,\n    \"g\": 1000**3,\n    \"gb\": 1000**3,\n    \"t\": 1000**4,\n    \"tb\": 1000**4,\n    \"ki\": 1024,\n    \"kib\": 1024,\n    \"mi\": 1024**2,\n    \"mib\": 1024**2,\n    \"gi\": 1024**3,\n    \"gib\": 1024**3,\n    \"ti\": 1024**4,\n    \"tib\": 1024**4,\n}\nGIB = 1024**3\n\n\ndef parse_size(text: str) -> int:\n    \"\"\"Parse '16GB', '16GiB', '16000MB' into bytes. GB means 10^9, GiB means 2^30.\"\"\"\n    m = _SIZE_RE.match(text)\n    if not m or m.group(2).lower() not in _UNITS:\n        raise ValueError(f\"Cannot read size {text!r}. Try something like 16GB, 16GiB or 16000MB.\")\n    return int(float(m.group(1)) * _UNITS[m.group(2).lower()])\n\n\ndef human_size(n: int | float) -> str:\n    n = float(n)\n    for unit in (\"B\", \"KB\", \"MB\", \"GB\", \"TB\"):\n        if n < 1000 or unit == \"TB\":\n            return f\"{n:.0f} {unit}\" if unit == \"B\" else f\"{n:.1f} {unit}\"\n        n /= 1000\n    return f\"{n:.1f} TB\"\n\n\ndef sha256_bytes(data: bytes) -> str:\n    return hashlib.sha256(data).hexdigest()\n\n\ndef sha256_file(path: str | Path, chunk: int = 1 << 20) -> str:\n    h = hashlib.sha256()\n    with open(path, \"rb\") as f:\n        while block := f.read(chunk):\n            h.update(block)\n    return h.hexdigest()\n\n\ndef utc_now() -> str:\n    return datetime.now(UTC).strftime(\"%Y-%m-%dT%H:%M:%SZ\")\n", "messages": null, "license": "Apache-2.0", "source": "quantgap:util.py", "greedy": false, "greedy_prompt": null}
+{"seq_id": 85, "group": "code", "kind": "text", "lang": "python", "text": "\"\"\"The small interface every backend adapter has to provide (section 6.3 of the spec).\"\"\"\n\nfrom __future__ import annotations\n\nfrom dataclasses import dataclass, field\nfrom pathlib import Path\nfrom typing import Any, Protocol\n\nimport numpy as np\n\nfrom quantgap.score.kld import log_softmax\nfrom quantgap.tokcheck import TokFingerprint\n\nLOGSOFTMAX_CHUNK = 64  # positions per float64 log-softmax, keeps memory flat for big vocabularies\n\n\n@dataclass\nclass ResolvedSource:\n    \"\"\"A quant that has been located on disk (or on a server).\"\"\"\n\n    source: str  # what the user typed, or the canonical hf:// URL\n    label: str  # short name, for example \"UD-Q4_K_XL\"\n    path: Path | None = None  # local file or folder; None for a server\n    size_bytes: int = 0\n    format: str = \"gguf\"\n    publisher: str | None = None\n    base_model: str | None = None\n    sha256: str | None = None\n    extra: dict[str, Any] = field(default_factory=dict)\n\n\n@dataclass\nclass BackendSettings:\n    ctx: int = 2048\n    n_gpu_layers: int = -1\n    n_threads: int | None = None\n    base_url: str | None = None\n    model_name: str | None = None\n    max_prompt_logprobs: int = 20\n    api_key: str | None = None\n    extra: dict[str, Any] = field(default_factory=dict)\n\n    def public(self) -> dict[str, Any]:\n        \"\"\"Settings that go into the result file (never the API key).\"\"\"\n        d = {\"ctx\": self.ctx}\n        if self.n_gpu_layers != -1:\n            d[\"n_gpu_layers\"] = self.n_gpu_layers\n        else:\n            d[\"n_gpu_layers\"] = -1\n        if self.n_threads:\n            d[\"n_threads\"] = self.n_threads\n        if self.base_url:\n            d[\"base_url\"] = self.base_url\n        if self.model_name:\n            d[\"model_name\"] = self.model_name\n        d.update(self.extra)\n        return d\n", "messages": null, "license": "Apache-2.0", "source": "quantgap:backends/base.py", "greedy": false, "greedy_prompt": null}
+{"seq_id": 86, "group": "code", "kind": "text", "lang": "python", "text": "\"\"\"Write a pack: parquet shards plus manifest.json, with a SHA-256 for every file.\"\"\"\n\nfrom __future__ import annotations\n\nimport json\nfrom collections.abc import Iterable\nfrom pathlib import Path\n\nimport numpy as np\nimport pyarrow as pa\nimport pyarrow.parquet as pq\n\nfrom quantgap.models import (\n    GreedyInfo,\n    Manifest,\n    ModelRef,\n    ProbesInfo,\n    RunnerInfo,\n    TokenizerInfo,\n)\nfrom quantgap.pack.reader import SeqTopK\nfrom quantgap.probes.render import RenderedSeq\nfrom quantgap.util import sha256_file, utc_now\n\nSHARD_ROWS = 40_000\n_COMPRESSION = \"zstd\"\n\n\ndef _fixed_list(values: np.ndarray, width: int, dtype: pa.DataType) -> pa.Array:\n    n = len(values)\n    offsets = pa.array(np.arange(0, (n + 1) * width, width, dtype=np.int32))\n    return pa.ListArray.from_arrays(offsets, pa.array(values.reshape(-1), type=dtype))\n\n\ndef _ragged_list(rows: list[np.ndarray], dtype: pa.DataType) -> pa.Array:\n    lengths = np.array([len(r) for r in rows], dtype=np.int64)\n    offsets = pa.array(np.concatenate([[0], np.cumsum(lengths)]).astype(np.int32))\n    flat = np.concatenate(rows) if rows else np.array([])\n    return pa.ListArray.from_arrays(offsets, pa.array(flat, type=dtype))\n", "messages": null, "license": "Apache-2.0", "source": "quantgap:pack/writer.py", "greedy": false, "greedy_prompt": null}
+{"seq_id": 87, "group": "code", "kind": "text", "lang": "python", "text": "def build_pack_from_objects(\n    model: Any,\n    tok: Any,\n    out_dir: str | Path,\n    *,\n    repo: str,\n    revision: str | None = None,\n    dtype: str = \"bfloat16\",\n    probes_version: str = \"v1\",\n    probes_file: Path | None = None,\n    k: int = 32,\n    batch_size: int = 4,\n    max_new_tokens: int = 256,\n    logp_dtype: str = \"float32\",\n    license_: str | None = None,\n    device: Any = \"cpu\",\n    device_label: str = \"cpu\",\n    progress: Callable[[int, int], None] | None = None,\n    max_len: int = 512,\n) -> Manifest:\n    \"\"\"Build a pack from an already loaded model and tokenizer.\"\"\"\n    import torch\n    import transformers\n\n    probes = load_probes(probes_version, probes_file)\n    seqs, greedy = render_probes(probes, tok, max_len=max_len)\n    writer = PackWriter(out_dir, k=k, logp_dtype=logp_dtype)\n    writer.write_probes(seqs)\n    model.eval()\n\n    order = sorted(range(len(seqs)), key=lambda i: len(seqs[i].token_ids))\n    done = 0\n    for a in range(0, len(order), batch_size):\n        batch = [seqs[i] for i in order[a : a + batch_size]]\n        for r in _topk_batch(model, batch, k, device):\n            writer.add_seq(r)\n        done += len(batch)\n        if progress:\n            progress(done, len(seqs) + len(greedy))\n\n    rows = []\n    for g in greedy:\n        ref = _greedy(model, tok, g, max_new_tokens, device)\n        rows.append(\n            {\n                \"seq_id\": g.seq_id,\n                \"group\": g.group,\n                \"lang\": g.lang,\n                \"prompt_ids\": g.prompt_ids,\n                \"ref_ids\": ref,\n                \"ref_text\": tok.decode(ref, skip_special_tokens=True),\n            }\n        )\n        done += 1\n        if progress:\n            progress(done, len(seqs) + len(greedy))\n    writer.write_greedy(rows)\n\n    return writer.finalize(\n        model=ModelRef(repo=repo, revision=revision, dtype=dtype, license=license_),\n        tokenizer=tokenizer_info(tok),\n        probes_set=probes_version,\n        probes_sha256=_probes_hash(probes_version, probes_file),\n        n_sequences=len(seqs),\n        chat_template_sha256=chat_template_hash(tok),\n        greedy=GreedyInfo(n_prompts=len(greedy), max_new_tokens=max_new_tokens),\n        runner=RunnerInfo(\n            quantgap=__version__,\n            torch=torch.__version__,\n            transformers=transformers.__version__,\n            device=device_label,\n        ),\n    )\n", "messages": null, "license": "Apache-2.0", "source": "quantgap:pack/runner.py", "greedy": false, "greedy_prompt": null}
+{"seq_id": 88, "group": "math", "kind": "text", "lang": "en", "text": "Question: Natalia sold clips to 48 of her friends in April, and then she sold half as many clips in May. How many clips did Natalia sell altogether in April and May?\nLet's think step by step.\nNatalia sold 48/2 = 24 clips in May.\nNatalia sold 48+24 = 72 clips altogether in April and May.\nThe answer is 72", "messages": null, "license": "MIT", "source": "gsm8k:train:0", "greedy": true, "greedy_prompt": null}
+{"seq_id": 89, "group": "math", "kind": "text", "lang": "en", "text": "Question: Weng earns $12 an hour for babysitting. Yesterday, she just did 50 minutes of babysitting. How much did she earn?\nLet's think step by step.\nWeng earns 12/60 = $0.2 per minute.\nWorking 50 minutes, she earned 0.2 x 50 = $10.\nThe answer is 10", "messages": null, "license": "MIT", "source": "gsm8k:train:1", "greedy": true, "greedy_prompt": null}
+{"seq_id": 90, "group": "math", "kind": "text", "lang": "en", "text": "Question: Betty is saving money for a new wallet which costs $100. Betty has only half of the money she needs. Her parents decided to give her $15 for that purpose, and her grandparents twice as much as her parents. How much more money does Betty need to buy the wallet?\nLet's think step by step.\nIn the beginning, Betty has only 100 / 2 = $50.\nBetty's grandparents gave her 15 * 2 = $30.\nThis means, Betty needs 100 - 50 - 30 - 15 = $5 more.\nThe answer is 5", "messages": null, "license": "MIT", "source": "gsm8k:train:2", "greedy": true, "greedy_prompt": null}
+{"seq_id": 91, "group": "math", "kind": "text", "lang": "en", "text": "Question: Julie is reading a 120-page book. Yesterday, she was able to read 12 pages and today, she read twice as many pages as yesterday. If she wants to read half of the remaining pages tomorrow, how many pages should she read?\nLet's think step by step.\nMaila read 12 x 2 = 24 pages today.\nSo she was able to read a total of 12 + 24 = 36 pages since yesterday.\nThere are 120 - 36 = 84 pages left to be read.\nSince she wants to read half of the remaining pages tomorrow, then she should read 84/2 = 42 pages.\nThe answer is 42", "messages": null, "license": "MIT", "source": "gsm8k:train:3", "greedy": true, "greedy_prompt": null}
+{"seq_id": 92, "group": "math", "kind": "text", "lang": "en", "text": "Question: James writes a 3-page letter to 2 different friends twice a week.  How many pages does he write a year?\nLet's think step by step.\nHe writes each friend 3*2=6 pages a week\nSo he writes 6*2=12 pages every week\nThat means he writes 12*52=624 pages a year\nThe answer is 624", "messages": null, "license": "MIT", "source": "gsm8k:train:4", "greedy": true, "greedy_prompt": null}
+{"seq_id": 93, "group": "math", "kind": "text", "lang": "en", "text": "Question: Mark has a garden with flowers. He planted plants of three different colors in it. Ten of them are yellow, and there are 80% more of those in purple. There are only 25% as many green flowers as there are yellow and purple flowers. How many flowers does Mark have in his garden?\nLet's think step by step.\nThere are 80/100 * 10 = 8 more purple flowers than yellow flowers.\nSo in Mark's garden, there are 10 + 8 = 18 purple flowers.\nPurple and yellow flowers sum up to 10 + 18 = 28 flowers.\nThat means in Mark's garden there are 25/100 * 28 = 7 green flowers.\nSo in total Mark has 28 + 7 = 35 plants in his garden.\nThe answer is 35", "messages": null, "license": "MIT", "source": "gsm8k:train:5", "greedy": true, "greedy_prompt": null}
+{"seq_id": 94, "group": "math", "kind": "text", "lang": "en", "text": "Question: Albert is wondering how much pizza he can eat in one day. He buys 2 large pizzas and 2 small pizzas. A large pizza has 16 slices and a small pizza has 8 slices. If he eats it all, how many pieces does he eat that day?\nLet's think step by step.\nHe eats 32 from the largest pizzas because 2 x 16 = 32\nHe eats 16 from the small pizza because 2 x 8 = 16\nHe eats 48 pieces because 32 + 16 = 48\nThe answer is 48", "messages": null, "license": "MIT", "source": "gsm8k:train:6", "greedy": true, "greedy_prompt": null}
+{"seq_id": 95, "group": "math", "kind": "text", "lang": "en", "text": "Question: Ken created a care package to send to his brother, who was away at boarding school.  Ken placed a box on a scale, and then he poured into the box enough jelly beans to bring the weight to 2 pounds.  Then, he added enough brownies to cause the weight to triple.  Next, he added another 2 pounds of jelly beans.  And finally, he added enough gummy worms to double the weight once again.  What was the final weight of the box of goodies, in pounds?\nLet's think step by step.\nTo the initial 2 pounds of jelly beans, he added enough brownies to cause the weight to triple, bringing the weight to 2*3=6 pounds.\nNext, he added another 2 pounds of jelly beans, bringing the weight to 6+2=8 pounds.\nAnd finally, he added enough gummy worms to double the weight once again, to a final weight of 8*2=16 pounds.\nThe answer is 16", "messages": null, "license": "MIT", "source": "gsm8k:train:7", "greedy": true, "greedy_prompt": null}
+{"seq_id": 96, "group": "math", "kind": "text", "lang": "en", "text": "Question: Alexis is applying for a new job and bought a new set of business clothes to wear to the interview. She went to a department store with a budget of $200 and spent $30 on a button-up shirt, $46 on suit pants, $38 on a suit coat, $11 on socks, and $18 on a belt. She also purchased a pair of shoes, but lost the receipt for them. She has $16 left from her budget. How much did Alexis pay for the shoes?\nLet's think step by step.\nLet S be the amount Alexis paid for the shoes.\nShe spent S + 30 + 46 + 38 + 11 + 18 = S + 143.\nShe used all but $16 of her budget, so S + 143 = 200 - 16 = 184.\nThus, Alexis paid S = 184 - 143 = $41 for the shoes.\nThe answer is 41", "messages": null, "license": "MIT", "source": "gsm8k:train:8", "greedy": false, "greedy_prompt": null}
+{"seq_id": 97, "group": "math", "kind": "text", "lang": "en", "text": "Question: Tina makes $18.00 an hour.  If she works more than 8 hours per shift, she is eligible for overtime, which is paid by your hourly wage + 1/2 your hourly wage.  If she works 10 hours every day for 5 days, how much money does she make?\nLet's think step by step.\nShe works 8 hours a day for $18 per hour so she makes 8*18 = $144.00 per 8-hour shift\nShe works 10 hours a day and anything over 8 hours is eligible for overtime, so she gets 10-8 = 2 hours of overtime\nOvertime is calculated as time and a half so and she makes $18/hour so her overtime pay is 18*.5 = $9.00\nHer overtime pay is 18+9 = $27.00\nHer base pay is $144.00 per 8-hour shift and she works 5 days and makes 5 * $144 = $720.00\nHer overtime pay is $27.00 per hour and she works 2 hours of overtime per day and makes 27*2 = $54.00 in overtime pay\n2 hours of overtime pay for 5 days means she makes 54*5 = $270.00\nIn 5 days her base pay is $720.00 and she makes $270.00 in overtime pay so she makes $720 + $270 = $990.00\nThe answer is 990", "messages": null, "license": "MIT", "source": "gsm8k:train:9", "greedy": false, "greedy_prompt": null}
+{"seq_id": 98, "group": "math", "kind": "text", "lang": "en", "text": "Question: A deep-sea monster rises from the waters once every hundred years to feast on a ship and sate its hunger. Over three hundred years, it has consumed 847 people. Ships have been built larger over time, so each new ship has twice as many people as the last ship. How many people were on the ship the monster ate in the first hundred years?\nLet's think step by step.\nLet S be the number of people on the first hundred years’ ship.\nThe second hundred years’ ship had twice as many as the first, so it had 2S people.\nThe third hundred years’ ship had twice as many as the second, so it had 2 * 2S = 4S people.\nAll the ships had S + 2S + 4S = 7S = 847 people.\nThus, the ship that the monster ate in the first hundred years had S = 847 / 7 = 121 people on it.\nThe answer is 121", "messages": null, "license": "MIT", "source": "gsm8k:train:10", "greedy": false, "greedy_prompt": null}
+{"seq_id": 99, "group": "math", "kind": "text", "lang": "en", "text": "Question: Tobias is buying a new pair of shoes that costs $95. He has been saving up his money each month for the past three months. He gets a $5 allowance a month. He also mows lawns and shovels driveways. He charges $15 to mow a lawn and $7 to shovel. After buying the shoes, he has $15 in change. If he mows 4 lawns, how many driveways did he shovel?\nLet's think step by step.\nHe saved up $110 total because 95 + 15 = 110\nHe saved $15 from his allowance because 3 x 5 = 15\nHe earned $60 mowing lawns because 4 x 15 = 60\nHe earned $35 shoveling driveways because 110 - 60 - 15 = 35\nHe shoveled 5 driveways because 35 / 7 = 5\nThe answer is 5", "messages": null, "license": "MIT", "source": "gsm8k:train:11", "greedy": false, "greedy_prompt": null}
+{"seq_id": 100, "group": "math", "kind": "text", "lang": "en", "text": "Question: Randy has 60 mango trees on his farm. He also has 5 less than half as many coconut trees as mango trees. How many trees does Randy have in all on his farm?\nLet's think step by step.\nHalf of the number of Randy's mango trees is 60/2 = 30 trees.\nSo Randy has 30 - 5 = 25 coconut trees.\nTherefore, Randy has 60 + 25 = 85 treeson his farm.\nThe answer is 85", "messages": null, "license": "MIT", "source": "gsm8k:train:12", "greedy": false, "greedy_prompt": null}
+{"seq_id": 101, "group": "math", "kind": "text", "lang": "en", "text": "Question: Jasper will serve charcuterie at his dinner party. He buys 2 pounds of cheddar cheese for $10, a pound of cream cheese that cost half the price of the cheddar cheese, and a pack of cold cuts that cost twice the price of the cheddar cheese. How much does he spend on the ingredients?\nLet's think step by step.\nA pound of cream cheese cost $10 / 2 = $5.\nA pack of cold cuts cost $10 x 2 = $20.\nJasper spent $10 + $5 + $20 = $35 on the ingredients.\nThe answer is 35", "messages": null, "license": "MIT", "source": "gsm8k:train:13", "greedy": false, "greedy_prompt": null}
+{"seq_id": 102, "group": "math", "kind": "text", "lang": "en", "text": "Question: Joy can read 8 pages of a book in 20 minutes. How many hours will it take her to read 120 pages?\nLet's think step by step.\nIn one hour, there are 3 sets of 20 minutes.\nSo, Joy can read 8 x 3 = 24 pages in an hour.\nIt will take her 120/24 = 5 hours to read 120 pages.\nThe answer is 5", "messages": null, "license": "MIT", "source": "gsm8k:train:14", "greedy": false, "greedy_prompt": null}
+{"seq_id": 103, "group": "math", "kind": "text", "lang": "en", "text": "Question: James creates a media empire.  He creates a movie for $2000.  Each DVD cost $6 to make.  He sells it for 2.5 times that much.  He sells 500 movies a day for 5 days a week.  How much profit does he make in 20 weeks?\nLet's think step by step.\nHe sold each DVD for 6*2.5=$15\nSo he makes a profit of 15-6=$9\nSo each day he makes a profit of 9*500=$4500\nSo he makes 4500*5=$22,500\nHe makes 22,500*20=$450,000\nThen after the cost of creating the movie he has a profit of 450,000-2000=$448,000\nThe answer is 448000", "messages": null, "license": "MIT", "source": "gsm8k:train:15", "greedy": false, "greedy_prompt": null}
+{"seq_id": 104, "group": "math", "kind": "text", "lang": "en", "text": "Question: The profit from a business transaction is shared among 2 business partners, Mike and Johnson in the ratio 2:5 respectively. If Johnson got $2500, how much will Mike have after spending some of his share on a shirt that costs $200?\nLet's think step by step.\nAccording to the ratio, for every 5 parts that Johnson gets, Mike gets 2 parts\nSince Johnson got $2500, each part is therefore $2500/5 = $500\nMike will get 2*$500 = $1000\nAfter buying the shirt he will have $1000-$200 = $800 left\nThe answer is 800", "messages": null, "license": "MIT", "source": "gsm8k:train:16", "greedy": false, "greedy_prompt": null}
+{"seq_id": 105, "group": "math", "kind": "text", "lang": "en", "text": "Question: In a truck, there are 26 pink hard hats, 15 green hard hats, and 24 yellow hard hats.  If Carl takes away 4 pink hard hats, and John takes away 6 pink hard hats and twice as many green hard hats as the number of pink hard hats that he removed, then calculate the total number of hard hats that remained in the truck.\nLet's think step by step.\nIf there were 26 pink hard hats and Carl took away 4 pink hard hats, the number of pink hard hats that remained is 26-4 = 22\nJohn also took away 6 pink hard hats, leaving 22-6 = 16 pink hard hats in the truck.\nIf John also took twice as many green hard hats as pink hard hats, he took 2*6 = 12 green hard hats.\nThe total number of green hard hats that remained in the truck is 15-12 = 3\nIn the truck, after some are taken, there were 3 green hard hats + 16 pink hard hats = 19 hard hats in the truck.\nAltogether, 19 green and pink hard hats + 24 yellow hards hats = 43 hard hats remained in the truck\nThe answer is 43", "messages": null, "license": "MIT", "source": "gsm8k:train:17", "greedy": false, "greedy_prompt": null}
+{"seq_id": 106, "group": "math", "kind": "text", "lang": "en", "text": "Question: It takes Roque two hours to walk to work and one hour to ride his bike to work. Roque walks to and from work three times a week and rides his bike to and from work twice a week. How many hours in total does he take to get to and from work a week with walking and biking?\nLet's think step by step.\nRoque takes 2*3 = 6 hours a week to walk to work.\nRoque takes 6*2 = 12 hours a week to walk to and from work.\nRoque takes 1*2 = 2 hours a week to bike to work.\nRoque takes 2*2 = 4 hours a week to bike to and from work.\nIn total, Roque takes 12+4 = 16 hour a week to go to and from work.\nThe answer is 16", "messages": null, "license": "MIT", "source": "gsm8k:train:18", "greedy": false, "greedy_prompt": null}
+{"seq_id": 107, "group": "math", "kind": "text", "lang": "en", "text": "Question: Tim rides his bike back and forth to work for each of his 5 workdays.  His work is 20 miles away.  He also goes for a weekend bike ride of 200 miles.    If he can bike at 25 mph how much time does he spend biking a week?\nLet's think step by step.\nHe bikes 20*2=40 miles each day for work\nSo he bikes 40*5=200 miles for work\nThat means he bikes a total of 200+200=400 miles for work\nSo he bikes a total of 400/25=16 hours\nThe answer is 16", "messages": null, "license": "MIT", "source": "gsm8k:train:19", "greedy": false, "greedy_prompt": null}
+{"seq_id": 108, "group": "math", "kind": "text", "lang": "en", "text": "Question: Bella bought stamps at the post office. Some of the stamps had a snowflake design, some had a truck design, and some had a rose design. Bella bought 11 snowflake stamps. She bought 9 more truck stamps than snowflake stamps, and 13 fewer rose stamps than truck stamps. How many stamps did Bella buy in all?\nLet's think step by step.\nThe number of truck stamps is 11 + 9 = 20.\nThe number of rose stamps is 20 − 13 = 7.\nBella bought 11 + 20 + 7 = 38 stamps in all.\nThe answer is 38", "messages": null, "license": "MIT", "source": "gsm8k:train:20", "greedy": false, "greedy_prompt": null}
+{"seq_id": 109, "group": "math", "kind": "text", "lang": "en", "text": "Question: Each bird eats 12 beetles per day, each snake eats 3 birds per day, and each jaguar eats 5 snakes per day. If there are 6 jaguars in a forest, how many beetles are eaten each day?\nLet's think step by step.\nFirst find the total number of snakes eaten: 5 snakes/jaguar * 6 jaguars = 30 snakes\nThen find the total number of birds eaten per day: 30 snakes * 3 birds/snake = 90 snakes\nThen multiply the number of snakes by the number of beetles per snake to find the total number of beetles eaten per day: 90 snakes * 12 beetles/snake = 1080 beetles\nThe answer is 1080", "messages": null, "license": "MIT", "source": "gsm8k:train:21", "greedy": false, "greedy_prompt": null}
+{"seq_id": 110, "group": "math", "kind": "text", "lang": "en", "text": "Question: Samantha’s last name has three fewer letters than Bobbie’s last name. If Bobbie took two letters off her last name, she would have a last name twice the length of Jamie’s. Jamie’s full name is Jamie Grey. How many letters are in Samantha’s last name?\nLet's think step by step.\nThere are 4 letters in Jamie’s last name, so Bobbie’s name is 4*2 +2 = 10 letters long.\nSamantha’s last name is 3 letters shorter than Bobbie’s, so there are 10 - 3 = 7 letters in Samantha’s last name.\nThe answer is 7", "messages": null, "license": "MIT", "source": "gsm8k:train:22", "greedy": false, "greedy_prompt": null}
+{"seq_id": 111, "group": "math", "kind": "text", "lang": "en", "text": "Question: Ann's favorite store was having a summer clearance. For $75 she bought 5 pairs of shorts for $7 each and 2 pairs of shoes for $10 each. She also bought 4 tops, all at the same price. How much did each top cost?\nLet's think step by step.\nShe bought 5 shorts at $7 each so 5*7=$35\nShe bought 2 pair of shoes at $10 each so 2*10=$20\nThe shorts and shoes cost her 35+20 = $55\nWe know she spent 75 total and the shorts and shoes cost $55 which left a difference of 75-55 = $20\nShe bought 4 tops for a total of $20 so 20/4 = $5\nThe answer is 5", "messages": null, "license": "MIT", "source": "gsm8k:train:23", "greedy": false, "greedy_prompt": null}
+{"seq_id": 112, "group": "math", "kind": "text", "lang": "en", "text": "Question: Mary does her grocery shopping on Saturday. She does her shopping only at a specific store where she is allowed a credit of $100, which must be paid in full before her next shopping trip. That week she spent the full credit limit and paid $15 of it on Tuesday and $23 of it on Thursday. How much credit will Mary need to pay before her next shopping trip?\nLet's think step by step.\nSo far, Mary has paid back $15 +$23=$38 of the credit.\nSo she still needs to pay $100-$38=$62\nThe answer is 62", "messages": null, "license": "MIT", "source": "gsm8k:train:24", "greedy": false, "greedy_prompt": null}
+{"seq_id": 113, "group": "math", "kind": "text", "lang": "en", "text": "Question: Ralph is going to practice playing tennis with a tennis ball machine that shoots out tennis balls for Ralph to hit. He loads up the machine with 175 tennis balls to start with. Out of the first 100 balls, he manages to hit 2/5 of them. Of the next 75 tennis balls, he manages to hit 1/3 of them. Out of all the tennis balls, how many did Ralph not hit?\nLet's think step by step.\nOut of the first 100 balls, Ralph was able to hit 2/5 of them and not able to hit 3/5 of them, 3/5 x 100 = 60 tennis balls Ralph didn't hit.\nOut of the next 75 balls, Ralph was able to hit 1/3 of them and not able to hit 2/3 of them, 2/3 x 75 = 50 tennis balls that Ralph didn't hit.\nCombined, Ralph was not able to hit 60 + 50 = 110 tennis balls Ralph didn't hit.\nThe answer is 110", "messages": null, "license": "MIT", "source": "gsm8k:train:25", "greedy": false, "greedy_prompt": null}
+{"seq_id": 114, "group": "math", "kind": "text", "lang": "en", "text": "Question: Jack is stranded on a desert island. He wants some salt to season his fish. He collects 2 liters of seawater in an old bucket. If the water is 20% salt, how many ml of salt will Jack get when all the water evaporates?\nLet's think step by step.\nFirst find how many liters of the seawater are salt: 2 liters * 20% = .4 liters\nThen multiply that amount by 1000 ml/liter to find the number of ml of salt Jack gets: .4 liters * 1000 ml/liter = 400 ml\nThe answer is 400", "messages": null, "license": "MIT", "source": "gsm8k:train:26", "greedy": false, "greedy_prompt": null}
+{"seq_id": 115, "group": "math", "kind": "text", "lang": "en", "text": "Question: Brennan was researching his school project and had to download files from the internet to his computer to use for reference. After downloading 800 files, he deleted 70% of them because they were not helpful. He downloaded 400 more files but again realized that 3/5 of them were irrelevant. How many valuable files was he left with after deleting the unrelated files he downloaded in the second round?\nLet's think step by step.\nThe number of non-valuable files Brennan downloaded in the first round is 70/100*800 = 560 files.\nThe number of valuable files Brennan downloaded in the first round is 800-560 = 240\nWhen he downloaded 400 new files, there were 3/5*400= 240 non-useful files, which he deleted again.\nThe total number of valuable files he downloaded in the second round is 400-240 = 160\nTo write his research, Brennan had 160+240 = 400 useful files to reference to write his research.\nThe answer is 400", "messages": null, "license": "MIT", "source": "gsm8k:train:27", "greedy": false, "greedy_prompt": null}
+{"seq_id": 116, "group": "multilingual", "kind": "text", "lang": "hi", "text": "आशा गणपतराव भोसले (जन्म: 8 सितम्बर 1933) हिन्दी फ़िल्मों की मशहूर पार्श्वगायिका हैं। लता मंगेशकर की छोटी बहन और दिनानाथ मंगेशकर की पुत्री आशा ने फिल्मी और गैर फिल्मी लगभग 16 हजार गाने गाये हैं और इनकी आवाज़ के प्रशंसक पूरी दुनिया में फैले हुए हैं। हिंदी के अलावा उन्होंने मराठी, बंगाली, गुजराती, पंजाबी, भोजपुरी, तमिल, मलयालम, अंग्रेजी और रूसी भाषा के भी अनेक गीत गाए हैं। आशा भोंसले ने अपना पहला गीत वर्ष 1948 में सावन आया फिल्म चुनरिया में गाया। \n\nआशा की विशेषता है कि इन्होंने शास्त्रीय संगीत, गजल और पॉप संगीत हर क्षेत्र में अपनी आवाज़ का जादू बिखेरा है और एक समान सफलता पाई है। उन्होने आर॰ डी॰ बर्मन से दूसरा विवाह किया था।\n\nजीवनी \nआशा भोसले का जन्म 08 सितम्बर 1933 को महाराष्ट्र के ‘सांगली’ जिले एक मराठी परिवार में हुआ। इनके पिता दिनानाथ मंगेशकर प्रसिद्ध गायक एवं नायक थे। जिन्होंने शास्त्रीय संगीत की शिक्षा काफी छोटी उम्र में ही आशा जी को दी। आशा जी जब केवल 9 वर्ष की थीं, इनके पिता का स्वर्गवास हो गया । पिता के मरणोपरांत, इनका परिवार पुणे से कोल्हापुर और उसके बाद मुंबई आ गया। परिवार की सहायता के लिए आशा और इनकी बड़ी बहन लता मंगेशकर ने गाना और फिल्मों में अभिनय शुरू कर दिया। 1943 में इन्होंने अपनी पहली मराठी फिल्म ‘माझा बाळ’ में गीत गाया। यह गीत ‘चला चला नव बाळा...’ दत्ता डावजेकर के द्वारा संगीतबद्ध किया गया था। 1948 में हिन्दी फिल्म ‘चुनरिया’ का गीत ‘सावन आया।", "messages": null, "license": "CC BY-SA 4.0", "source": "wikipedia:hi:आशा भोसले", "greedy": true, "greedy_prompt": null}
+{"seq_id": 117, "group": "multilingual", "kind": "text", "lang": "hi", "text": "ब्रह्मा जी के पुत्र दरीची, दरीची के पुत्र महर्षि कूर्म ( कुर्मी )कश्यप के वंश से जो सूर्यवंश हुआ उसी के इक्षवा कुल में श्री राम का जन्म हुआ इन्हें रामचंद्र भी कहते हैं, रामायण के अनुसार,महाराजा प्रजापति दशरथ और रानी कौशल्या के सबसे बड़े पुत्र, सीता के पति व लक्ष्मण, भरत तथा शत्रुघ्न के भ्राता थे। हनुमान उनके परम भक्त है। लंका के राजा रावण का वध उन्होंने ही किया था। उनकी प्रतिष्ठा मर्यादा पुरुषोत्तम के रूप में है क्योंकि उन्होंने मर्यादा के पालन के लिए राज्य, मित्र, माता-पिता तक का त्याग किया।\n\nवे भगवान विष्णु प्रजापति के अवतार माने जाते हैं।श्री राम  का जन्म वैवस्वत मन्वंतर में 23 वे चतुर्युग के त्रेता में हुआ था । उन्होंने करीब 11000 वर्ष अयोध्या का शासन किया था , श्री राम विश्वामित्र के साथ यज्ञ की रक्षा के लिए छोटे भाई लक्षण के साथ गए और ताड़का आदि राक्षस मारे उसके बाद गुरु के साथ राजा जनक के यहा उनकी पुत्री सीता के स्वयंवर में पहुंच कर शिव के धनुष को तोड़ सीता से विवाह किया , जब उन्होंने अगला राजा बनाया जा रहा था तब उनके पिता के वचन के लिए 14 वर्ष वनवास जाना हुआ जहां पर लक्ष्मण ने शुर्पनखा के नाक कान काट दिए और उसके भाई खर दूषण का राम ने वध कर दिया माता सीता का अपहरण रावण द्वारा किया गया जिसकी खोज में जटायु पक्षी जो बात करता था जिसने रावण का प्रतिकार किया था का पिता के समान अंत्येष्ठि संस्कार किया  आगे हनुमान सुग्रीव से मिले सुग्रीव ले भाई बाली का वध कर सुग्रीव को राजा बनाया", "messages": null, "license": "CC BY-SA 4.0", "source": "wikipedia:hi:राम", "greedy": true, "greedy_prompt": null}
+{"seq_id": 118, "group": "multilingual", "kind": "text", "lang": "hi", "text": "हरिशंकर परसाई (२२ अगस्त, १९२४ - १० अगस्त, १९९५) हिंदी के प्रसिद्ध लेखक और व्यंगकार थे। उनका जन्म जमानी, होशंगाबाद, मध्य प्रदेश में हुआ था। वे हिंदी के पहले रचनाकार हैं जिन्होंने व्यंग्य को विधा का दर्जा दिलाया और उसे हल्के–फुल्के मनोरंजन की परंपरागत परिधि से उबारकर समाज के व्यापक प्रश्नों से जोड़ा। उनकी व्यंग्य रचनाएँ हमारे मन में गुदगुदी ही पैदा नहीं करतीं बल्कि हमें उन सामाजिक वास्तविकताओं के आमने–सामने खड़ा करती है, जिनसे किसी भी और राजनैतिक व्यवस्था में पिसते मध्यमवर्गीय मन की सच्चाइयों को उन्होंने बहुत ही निकटता से पकड़ा है। सामाजिक पाखंड और रूढ़िवादी जीवन–मूल्यों के अलावा जीवन पर्यन्त विस्ल्लीयो पर भी अपनी अलग कोटिवार पहचान है। उड़ाते हुए उन्होंने सदैव विवेक और विज्ञान–सम्मत दृष्टि को सकारात्मक रूप में प्रस्तुत किया है। उनकी भाषा–शैली में खास किस्म का अपनापन महसूस होता है कि लेखक उसके सामने ही बैठे हैं।ठिठुरता हुआ गणतंत्र की रचना हरिशंकर परसाई ने की जो एक व्यंग्य है||\n\nशिक्षा \nउन्होंने सेमस्तार ग्लोबल स्कूल इलाहाबाद में आर. टी.एम.नागपुर विश्वविद्यालय से हिन्दी में एम॰ए॰ की उपाधि प्राप्त की।\n\nजीवन", "messages": null, "license": "CC BY-SA 4.0", "source": "wikipedia:hi:हरिशंकर परसाई", "greedy": true, "greedy_prompt": null}
+{"seq_id": 119, "group": "multilingual", "kind": "text", "lang": "hi", "text": "भागीरथी भारत के उत्तराखण्ड राज्य में बहने वाली एक नदी है। इस नदी को किरात   नदी  के नाम से भी जाना जाता है। यह देवप्रयाग में अलकनंदा से मिलकर गंगा नदी का निर्माण करती है। भागीरथी का उद्गम स्थल उत्तरकाशी ज़िले में गौमुख (गंगोत्री ग्लेशियर) है। भागीरथी यहाँ २५ कि॰मी॰ लम्बे गंगोत्री हिमनद से निकलती है। २०५ किमी बहने के बाद, भागीरथी व अलकनंदा का देवप्रयाग में संगम होता है, जिसके पश्चात वह गंगा के रूप में पहचानी जाती है। okk\n\nअवस्थिति \nभागीरथी गोमुख स्थान से 25 कि॰मी॰ लम्बे गंगोत्री हिमनद से निकलती है। यह स्थान उत्तराखण्ड राज्य में उत्तरकाशी जिले में है। यह समुद्रतल से 618 मीटर की ऊँचाई पर, ऋषिकेश से 70 किमी दूरी पर स्थित है।\n\nटिहरी बाँध \nभारत में टिहरी बाँध, टेहरी विकास परियोजना का एक प्राथमिक बाँध है, जो उत्तराखण्ड राज्य के टिहरी में स्थित है। यह बाँध भागीरथी नदी पर बनाया गया है। टिहरी बाँध की ऊँचाई 260 मीटर है, जो इसे विश्व का पाँचवा सबसे ऊँचा बाँध बनाती है। इस बाँध से 2400 मेगा वाट विद्युत उत्पादन, 270,000 हेक्टर क्षेत्र की सिंचाई और प्रतिदिन 102.20 करोड़ लीटर पेयजल दिल्ली, उत्तर प्रदेश एवँ उत्तराखण्ड को उपलब्ध कराया जाना प्रस्तावित किया गया है।\n\nसहायक नदियाँ \n रुद्रागंगा- गंगोत्री ग्लेशियर के पास रुद्रागेरा ग्लेशियर से निकलती है।\n केदारगंगा- केदारताल से निकलकर गंगोत्री में भागीरथी से मिलती है।\n जाडगंगा / जाह्नवी - भैरोघाटी नामक स्थान पर भागीरथी नदी से मिलती है।", "messages": null, "license": "CC BY-SA 4.0", "source": "wikipedia:hi:भागीरथी नदी", "greedy": true, "greedy_prompt": null}
+{"seq_id": 120, "group": "multilingual", "kind": "text", "lang": "hi", "text": "हिन्दू धर्म में पुराणों के अनुसार शिवजी जहाँ-जहाँ स्वयं प्रकट हुए उन बारह स्थानों पर स्थित शिवलिंगों को ज्योतिर्लिंगों के रूप में पूजा जाता है। सौराष्ट्र प्रदेश (काठियावाड़) में श्रीसोमनाथ, श्रीशैल पर श्रीमल्लिकार्जुन, उज्जयिनी (उज्जैन) में श्रीमहाकाल, ॐकारेश्वर अथवा अमलेश्वर (मालवा में), चिताभूमि (देवघर) में वैद्यनाथ, डाकिनी नामक स्थान में श्रीभीमशंकर, सेतुबंध पर श्री रामेश्वर, दारुकावन में श्रीनागेश्वर, वाराणसी (काशी) में श्री विश्वनाथ, गौतमी (गोदावरी) के तट पर श्री त्र्यम्बकेश्वर, हिमालय पर केदारखंड में श्रीकेदारनाथ और शिवालय में श्रीघुश्मेश्वर। हिंदुओं में मान्यता है कि जो मनुष्य प्रतिदिन प्रात:काल उठकर इन बारह ज्योतिर्लिंगों के नाम का पाठ करता है, उसके सभी प्रकार के पाप छूट जाते हैं और उसको संपूर्ण सिद्धियों का फल प्राप्त हो जाता है।\nप्रत्येक ज्योतिर्लिङ्ग का एक-एक उपलिंग भी है जिनका विवरण शिवमहापुराण की 'ज्ञानसंहिता' के 38वें एवं 'कोटिरुद्रसंहिता' के प्रथम अध्याय में प्राप्त होता है।\n\nस्थल \n12 ज्योतिर्लिंगों के नाम शिव पुराण (शतरुद्र संहिता, अध्याय 42/2-4) के अनुसार इस प्रकार हैं-\n\nसन्दर्भ\n\nटीका टिप्पणी \n\n   क.    \nसौराष्ट्रे सोमनाथं च श्रीशैले मल्लिकार्जुनम्।\nउज्जयिन्यां महाकालमोङ्कारं परमेश्वरम्॥1॥\nकेदारं हिमवत्पृष्ठे डाकिन्यां भीमशङ्करम्।\nवाराणस्यां च विश्वेशं त्र्यम्बकं गौतमीतटे।।2।।\nवैद्यनाथं चिताभूमौ नागेशं दारुकावने।\nसेतुबन्धे च रामेशं घुश्मेशं च शिवालये॥3॥॥", "messages": null, "license": "CC BY-SA 4.0", "source": "wikipedia:hi:द्वादश ज्योतिर्लिंग", "greedy": true, "greedy_prompt": null}
+{"seq_id": 121, "group": "multilingual", "kind": "text", "lang": "hi", "text": "शम्मी कपूर (जन्म: 21 अक्टूबर, 1931 निधन: 14 अगस्त, 2011) हिन्दी फ़िल्मों के एक अभिनेता थे। उन्होने जंगली, तुमसा नहीं देखा, दिल देके देखो, सिंगापुर, कॉलेज गर्ल, प्रोफेसर, चाइना टाउन, प्यार किया तो डरना क्या, कश्मीर की कली, जानवर, तीसरी मंज़िल, अंदाज़, एन ईवनिंग इन पेरिस, ब्रह्मचारी और सच्चाई जैसी हिट फ़िल्मों में काम किया।\n\nव्यक्तिगत जीवन \nशम्मी कपूर का जन्म २१ अक्टूबर १९३१ को बंबई(प्रस्तुत मुंबई) में हुआ, इनके पिता पृथ्वीराज कपूर जाने माने अभिनेता थे अभिनय इनको विरासत में मिला था। शम्मी कपूर की पहली शादी अभिनेत्री गीता बाली से हुई, लेकिन १९६५ मे चेचक की वजह से गीता की म्रत्यु हो गयी। इसके पश्चात शम्मी कपूर ने नीला देवी से शादी की। १४ अगस्त २०११ की सुबह ५:१५ को इनका निधन हो गया। वे ७९ वर्ष के थे।\n\nफिल्मी सफर \nसाल 1953 में रिलीज फ़िल्म जीवन 'ज्योति' से बतौर अभिनेता शम्मी कपूर ने बॉलीवुड में कदम रखा। शम्मी कपूर की शुरुवाती फ़िल्में जीवन ज्योति, रेल का डिब्बा और गुल सनोवर थीं। शम्मी कपूर ने मुख्य अभिनेता के रूप में 50 से अधिक फ़िल्मों में अभिनय किया है और सहायक भूमिकाओं में 20 से अधिक फ़िल्मों में अभिनय किया है। उन्होने फ़िल्म ब्रह्मचारी (1968) में अपने प्रदर्शन के लिए सर्वश्रेष्ठ अभिनेता का फ़िल्मफेयर पुरस्कार जीता है।\n\nप्रमुख फिल्में\n\nबतौर निर्देशक\n\nनामांकन और पुरस्कार \n 1969 - फ़िल्मफ़ेयर सर्वश्रेष्ठ अभिनेता पुरस्कार - ब्रह्मचारी\n\nसन्दर्भ\n\nबाहरी कड़ियाँ", "messages": null, "license": "CC BY-SA 4.0", "source": "wikipedia:hi:शम्मी कपूर", "greedy": true, "greedy_prompt": null}
+{"seq_id": 122, "group": "multilingual", "kind": "text", "lang": "hi", "text": "कवि वह है जो भावों को रसाभिषिक्त अभिव्यक्ति देता है और सामान्य अथवा स्पष्ट के परे गहन यथार्थ का वर्णन करता है। इसीलिये वैदिक काल में ऋषय: मन्त्रदृष्टार: कवय: क्रान्तदर्शिन: अर्थात् ऋषि को मन्त्रदृष्टा और कवि को क्रान्तदर्शी कहा गया है। \"जहाँ न पहुँचे रवि, वहाँ पहुँचे कवि\" लोकोक्ति प्रसिद्ध है।\n\nकुछ प्रसिद्ध कवि \n महादेवी वर्मा \n जयशंकर प्रसाद \n सूर्यकांत त्रिपाठी निराला \n सुमित्रानंदन पंत \n सुभद्रा कुमारी चौहान \n रामधारी सिंह दिनकर \n भवानी प्रसाद मिश्र\n नागार्जुन\n अज्ञेय \n हरिवंश राय बच्चन \n गोपाल दास नीरज \n गोलेन्द्र पटेल \n प्रदीप\n प्रद़्यूम्न शर्मा\n कुमार विश्वास\n\nशायर \nशायरी लिखने वाले कवि को शायर कहा जाता है। शायरी भारतीय उपमहाद्वीप में प्रचलित एक कविता का रूप हैं जिसमें हिन्दी-उर्दू भाषाओँ में कविताएँ लिखी जाती हैं।\n\nकुछ प्रसिद्ध शायर \n अमीर ख़ुसरो\n अशफाक उल्ला खाँ वारसी 'हसरत'\n दुष्यंत कुमार\n निदा फ़ाज़ली\n पण्डित राम प्रसाद 'बिस्मिल'\n बहादुर शाह ज़फ़र\n मिर्ज़ा ग़ालिब\n मीर तकी \"मीर\"\n मुहम्मद इक़बाल\n राहत इन्दौरी\n कुँवर बेचैन\n मुनव्वर राणा\n\nइन्हें भी देखें  \nकविशिक्षा\nकाव्य\nकविता\nमहाकाव्य\nकवि सम्मेलन\nकाव्यशास्त्र\nशायरी\n\nसन्दर्भ\n\nकवि\nशायर\nशायरी\nहिन्दी\nउर्दू\nसंगीत\nकवि\nकला व्यवसाय", "messages": null, "license": "CC BY-SA 4.0", "source": "wikipedia:hi:कवि", "greedy": true, "greedy_prompt": null}
+{"seq_id": 123, "group": "multilingual", "kind": "text", "lang": "hi", "text": "सुन्दरमूर्ति स्वमिगल नयन्नर (d. 824) आठवी सदी के तमिलनाडु के एक नायनमार सन्त थे। वे भगवान शिव के भक्त थे और चार तमिल साम्य अचार्यो मे से एक है।\n\nबचपन \nउनका जन्म थिरुनवलूर नाम के गांव मे हुआ। उनके बचपन का नाम \"नम्बि अरुरार\" था। उनके पिता का नाम \"सदैयार\" और माँ का नाम \"इसैग्नानि\" था। वे ब्राह्मण थे।\n\nथिरुनवलूर के राजा नारसिन्घ मुनैयार ने नम्बि को गोद लेना चाहा। सदैयार अनासक्ति से परिपूर्ण थे, उन्होने अपना बेटा खुशीपूर्वक राजा को दे दिया। और नम्बि राजपुत्र के रूप मे पले बढे।\n\nपौराणिक कथा \nदन्तकथा के अनुसार जब सुन्दरार की शादी थी तब शंकर भगवान वहाँ आए। वे एक साधु के रूप मे थे। उनके पास एक पर्चा था, जिसके अनुसार सुन्दरार के दादा ने वचन दिया था कि उनके आने वाली पीढी उनके अनुयायी रहेगे और उनकी सेवा करगे। सुन्दरार ने इसे अपनी नियती समझा और साधु के साथ तमिलनाडु का भ्रमण किया, वहाँ के मन्दिरो की सैर की। थान्जवुर के तिरूवरूर ग्राम पहुँचने पर उन्हे परवाई नाम की लडकी से पहचान हुई और बाद मे उन्होने उनसे शादी की।\n\nदन्तकथा के अनुसार उन्होने थिरुवरूर मे सब ६३ नायनमार के नाम का किर्तन किया। इन किर्तनों को तमिलनाडु मे तिरुतोन्दर तोकै के नाम से जाना गया। उनका तमिलनाडु भ्रमण जारी रहा, उन्होने भगवान शंकर के भजन लिखे और कई चमत्कारो को उनके साथ जोडा जाता है।\n\nराजा का साथ", "messages": null, "license": "CC BY-SA 4.0", "source": "wikipedia:hi:सुन्दरमूर्ति", "greedy": true, "greedy_prompt": null}
+{"seq_id": 124, "group": "multilingual", "kind": "text", "lang": "zh", "text": "{{Refimprove|time={{subst:#time:c}}}}\n通讯社（）指的是收集新闻报道并出售给订阅其新闻机构的组织，这些机构可以是报社、杂志社、或电视台等。在英语环境中，通讯社还可以写作“”、“”或“”。\n\n尽管世界各地有许多通讯社，但全球最大的三家通讯社则分别是：美联社、路透社和法新社。这三大通讯社在全球各地区均设有办事处，报道范围涵盖所有领域。这三大通讯社都遵循同一理念，即向所有订阅用户提供统一的客观的新闻内容；同时拒绝向保守派或自由派新闻机构提供额外的不同的新闻报道。英国记者兼分析家是如此阐释这一理念的。\n\n与的区别在于，报业辛迪加是将其内容出售给某个地区的具体客户，而通讯社则是出售给所有感兴趣的团体。\n\n发展历史 \n\n由于只有少数报社负担得起在总部以外的城市建立分支机构，因此它们为了获取其他城市的新闻就需要依靠通讯社。1835年，全球第一家通讯社——哈瓦斯通讯社在法国成立；随后到了1846年，美国的美联社也正式成立。哈瓦斯通讯社的两名前雇员从这里辞职之后分别于1849年在德国创立了沃尔夫通讯社和于1851年在英国创立了路透社；哈瓦斯通讯社则发展成了现在的法新社。在报道国际新闻时，这些通讯社各有侧重点。例如法新社的报道重点在法兰西帝国、南美洲和巴尔干地区，并与其他地区的通讯社共享新闻。在法国，根据哈瓦斯通讯社与法国各省省级报社所签订的标准合同，哈瓦斯通讯社每天向这些报社提供1800行电报的文字内容，而它每年则收取订阅费约1万法郎。而其他机构则负责提供专栏及虚构作品。", "messages": null, "license": "CC BY-SA 4.0", "source": "wikipedia:zh:通讯社", "greedy": false, "greedy_prompt": null}
+{"seq_id": 125, "group": "multilingual", "kind": "text", "lang": "zh", "text": "华国锋（），原名苏铸，字成九，山西交城人，中国共产党和中华人民共和国政治人物，曾任最高领导人，身兼中国共产党中央委员会主席、中华人民共和国国务院总理、中国共产党中央军事委员会主席。也曾经担任国务院副总理和公安部部长，中国共产党中央委员会副主席、中央政治局委员、常委，曾经为中国共产党和中华人民共和国的主要领导人之一。\n\n华国锋早年参加了抗日战争，并任中共交城县、阳曲县委书记等。中华人民共和国成立后，他先后担任中共湖南湘阴县、湘潭县委书记。文化大革命初期受到短暂冲击，后被“三结合”进湖南省革委会，担任中共湖南省委书记。1969年参加中共九大，成为主席团成员，并被选为第九届中央委员会委员。1970年晋升成为湖南省委第一书记、湖南省革委会主任，经毛泽东提名兼任广州军区政委与湖南省军区第一书记、第一政委。1971年3月经毛泽东提名进入中央，10月3日增补为国务院业务组成员并任副组长，分管农业、财贸。1972年1月通过国务院计委与李先念、余秋里联名起草向周恩来上报《关于进口成套化纤、化肥技术设备的报告》，最后形成1973年的“四三方案”。1972年3月经毛泽东提名，接任谢富治成为公安部部长。1973年8月在中共十大上被选为第十届中央委员、第十届中央政治局委员。1975年第四届人大后成为国务院常务副总理。2月起草《关于专案审查对象处理意见的请示报告》和准备释放的专案审查对象名单，3月17日在四届人大常委会二次会议上代表公安部做说明，会议决定对全部在押战犯实行特赦释放，并给予公民权。1976年周恩来去世后，2月被毛泽东任命为国务院代总理，4月经天安门事件被毛泽东正式任命为中共中央第一副主席、国务院总理。", "messages": null, "license": "CC BY-SA 4.0", "source": "wikipedia:zh:华国锋", "greedy": false, "greedy_prompt": null}
+{"seq_id": 126, "group": "multilingual", "kind": "text", "lang": "zh", "text": "8月5日是阳历年的第217天（闰年是218天），离一年的结束还有148天。\n\n大事记\n\n1世紀 \n 25年：身為漢朝宗室的漢景帝後裔劉秀在綠林軍的協助下推翻新朝即位，隨後建立東漢。\n\n11世紀 \n 1100年：在英格兰国王红发威廉意外逝世后，儒雅者亨利在西敏寺加冕成为英格兰国王。\n\n14世紀 \n 1305年：苏格兰民族英雄威廉·华莱士在格拉斯哥附近被英格兰军队抓获，并随即被囚送伦敦。\n\n16世紀 \n 1582年：應召前往中華傳教的葡萄牙教士利瑪竇，到達廣州香山澳（澳門），開始了天主教與西學在廣州的首次傳播。\n 1583年：英國探險家漢弗萊·吉爾伯特在紐芬蘭與拉布拉多建立第一個英國美洲殖民地聖約翰斯。\n 1600年：苏格兰发生高里阴谋事件。\n\n18世紀 \n 1772年：俄罗斯帝国、普鲁士王国及哈布斯堡君主国开始第一次瓜分波兰，以在东欧达到权力平衡。\n 1783年：日本淺間山的天明大喷发。\n\n19世紀 \n 1813年：中国严禁贩食鸦片。\n 1884年：法國艦隊司令李士卑斯率領三艘軍艦強攻雞籠，清法戰爭攻台戰事爆發。\n\n20世紀 \n 1919年：山东济南镇守使马良制造“济南血案”。\n 1949年：厄瓜多尔通古拉瓦省首府安巴托东南部发生ML 6.8的強烈地震，造成5,050人死亡。\n 1949年：英國皇家海軍軍艦「紫水晶號」抵達香港。當年4月，「紫水晶號」沿著長江運送補給物資予駐南京的英國大使館。當時中國人民解放軍正發動渡江戰役，「紫水晶號」駛到長江下游時，遭遇解放軍炮火攻擊，艦上包括艦長在內有17人喪生，20人受傷，艦身亦嚴重損毀無法航行，後來雙方多次談判才能脫險。\n 1962年：美国电影演员玛丽莲·梦露在洛杉矶因服用过量巴比妥类药物而逝世，成为著名阴谋论之一。", "messages": null, "license": "CC BY-SA 4.0", "source": "wikipedia:zh:8月5日", "greedy": false, "greedy_prompt": null}
+{"seq_id": 127, "group": "multilingual", "kind": "text", "lang": "zh", "text": "中国大陆是一个有關中國的政治及法律术语。在中华人民共和国及其特别行政区法律中，「中国大陆」（或稱為大陸、祖國大陆）是指除香港、澳門和台灣以外的中华人民共和国法理領土。在中華民國主要規範兩岸關係的法律《臺灣地區與大陸地區人民關係條例》中，「大陆地区」（或稱為中国大陆、大陆）是指除台湾地区以外的法理中華民國領土；該詞涵蓋範圍可能也包括香港及澳門，視具體情況而定。\n\n「中国大陆」這一法律术语常用于与台湾同时出现的语境。此外，“中国大陆”一词也在政经关系、地理科学、学术讨论等场合有所运用。\n\n法律上的用法\n\n中华人民共和国 \n\n1958年9月4日，全国人大常委会通过的《》中即指出，“中华人民共和国的领海宽度为12海里。这项规定适用于中华人民共和国的一切领土，包括中国大陆及其沿海岛屿，和同大陆及其沿海岛屿隔有公海的台湾及其周围各岛、澎湖列岛、东沙群岛、西沙群岛、中沙群岛、南沙群岛以及其他属于中国的岛屿。”\n\n1987年10月16日，中华人民共和国国务院办公厅发布《》，通知指出“台湾同胞回祖国大陆探亲旅游，须申请办理旅行证件。在香港地区，由中国外交部驻香港签证办事处办理，或由香港中国旅行社代办；在美国，日本或其他国家，由中国驻外使、领馆办理旅行证件。”\n\n2013年，全国人大常委会通过的《》第八十九条载明，“出境，是指由中国内地前往其他国家或者地区，由中国内地前往香港特别行政区、澳门特别行政区，由中国大陆前往台湾地区”。该法将“中国大陆”与台湾地区对应，将“中国内地”与香港特别行政区、澳门特别行政区对称，而在英文版中两词统一翻译为“”。", "messages": null, "license": "CC BY-SA 4.0", "source": "wikipedia:zh:中国大陆", "greedy": false, "greedy_prompt": null}
+{"seq_id": 128, "group": "multilingual", "kind": "text", "lang": "zh", "text": "内容管理系统（，缩写为 ）是指在一个合作模式下，用于管理工作流程的一套制度。该系统可应用于手工操作中，也可以应用到电脑或网络裡。作为一种中央储存器（central repository），可将相关内容集中储存并具有群组管理、版本控制等功能。版本控制是内容管理系统的一个主要优势。\n\n内容管理系统在物品或文案或数据的存储、掌管、修订（盘存）、语用充实、文档发布等方面有着广泛的应用。现在流行的开源CMS系统有WordPress、Joomla!、Drupal、Xoops、CmsTop等。\n\n概述 \n内容是任何类型的数字信息的结合体，可以是文本、图形图像、Web页面、业务文档、数据库表单、视频、声音、XML文件等。应该说，内容是一个比数据、文档和信息更广的概念，是对各种结构化数据、非结构化文档、信息的聚合。管理就是施加在“内容”对象上的一系列处理过程，包括收集、存储、审批、整理、定位、转换、分发、搜索、分析等，目的是为了使“内容”能够在正确的时间、以正确的形式传递到正确的地点和人。\n\n内容管理可以定义为：协助组织和个人，借助信息技术，实现内容的创建、储存、分享、应用、检索，并在企业个人、组织、业务、战略等诸方面产生价值的过程。而内容管理系统就是能够支撑内容管理的一种工具或一套工具的软件系统。\n\n内容管理系统的定义可以很狭窄，通常是指门户或商业网站的发布和管理系统；定义也可以很宽泛，个人网站系统也可归入其中。Wiki也是一种内容管理系统，Blog也算是一种内容管理系统。\n\n特點 \n 有利於為數眾多的人員協同作業的資料；", "messages": null, "license": "CC BY-SA 4.0", "source": "wikipedia:zh:内容管理系统", "greedy": false, "greedy_prompt": null}
+{"seq_id": 129, "group": "multilingual", "kind": "text", "lang": "zh", "text": "李奧納德·伯恩斯坦（，），猶太裔美國作曲家、指挥家、作家、音樂教育家、鋼琴家。\n\n生平 \n伯恩斯坦生於美國麻省劳伦斯，生名「路易斯」（Louis Bernstein），後於16歲時更為現名李奧納德·伯恩斯坦。\n\n1935年，入读哈佛大学，师从瓦尔特·辟斯顿、学习作曲，并于1939年毕业。随后进入费城柯蒂斯音乐学院学习指挥。\n\n1941年，担任波士顿交响乐团助理指挥，并师从谢尔盖·库塞维兹基和弗里兹·莱纳。\n\n1943年，临时代替生病的布鲁诺·瓦尔特上台，指挥了施特劳斯《》、舒曼《曼弗雷德》序曲、瓦格纳《纽伦堡的名歌手》序曲、米克罗斯变奏曲等而大获成功，自此在乐坛上崭露头角。当时他就已经有不少作品面世，包括《》“耶利米”、芭蕾舞剧《》、音乐剧《》等等。\n\n自1943年起频频担任纽约爱乐的客席指挥，后在1958年升任为该乐团音乐总监，成为土生土长美国人中，第一位获得指挥国际级乐团殊荣的指挥家。他在任的12年间，开创了该团的黄金时代，后来成为终身荣誉指挥，从此一举确立了第一流指挥家的名声。这12年间号称纽约爱乐的黄金岁月（1958－1969）。期间，他还和乐团拍摄了一系列电视节目《年轻人的音乐会》（Young People's Concerts），對古典音乐在美國年轻族群中的推廣，有極大的作用。\n\n1969年，伯恩斯坦自紐約愛樂退休后，後半生转往欧洲发展，原本是希望多花点时间作曲，不过显然事与愿违，多數時間都在從事教育後進，以及擔任歐洲知名樂團的客席巡迴演出指揮工作。他在欧洲最重要的功绩，是与维也纳爱乐保持长久而良好的关系，在DG留下許多經典錄音，因此维也纳爱乐授予伯恩斯坦荣誉会员。", "messages": null, "license": "CC BY-SA 4.0", "source": "wikipedia:zh:伦纳德·伯恩斯坦", "greedy": false, "greedy_prompt": null}
+{"seq_id": 130, "group": "multilingual", "kind": "text", "lang": "zh", "text": "奥地利共和国（），通稱奥地利（），是一个位於中歐的內陸國家，但在歷史上也被分類成西歐或者東歐的主權國家。奧地利与多國接壤，东面是匈牙利和斯洛伐克，南面是意大利和斯洛文尼亚，西面是列支敦士登和瑞士，北面是德国和捷克。首都兼最大城市是維也納，人口约900万。國土面積，同時由于阿爾卑斯山的存在，奧地利成爲了一個山地國，只有32%的國土海拔低於，最高點海拔。\n\n如今的奧地利是一個議會制的代議民主國家，下含九個聯邦州。奧地利是當今世界最富裕的國家之一，2012年的人均國民生產總值達到46,330美元。其人類發展指數在2014年排世界第21位。同時自1995年以來就是歐盟成員，是OECD的創始國之一。1995年簽訂申根公約，1999年接受並于2002年起使用歐元。奧地利和德國、瑞士、盧森堡均屬於德語區，在歷史上是日耳曼民族的文化發源地之一，其首都維也納曾經是神聖羅馬帝國、奧地利帝國和奧匈帝國的帝都，也是統治中歐1300餘年（650～1918年）的哈布斯堡王朝之核心，在歷史上亦和同屬中歐的匈牙利、捷克、斯洛伐克和波蘭，南歐的斯洛文尼亞、克羅地亞（兩國長期由奧地利帝國和奧匈帝國統治，人民多信奉天主教，文化也接近中歐多於南歐）有緊密的關聯。\n\n名稱 \n奧地利的德文名稱Österreich源自古高地德語Ostarrîchi，即是“東方的邊藩”，首見於996年一個記錄。該名可能是中世紀拉丁語同義詞Marchia orientalis在當地(巴伐利亞)德語方言的轉譯。另一個可能是該名源自一座山的當地稱呼，斯洛文尼亞語“Ostravica”(意即“尖峰”)，因為山脈兩面都十分陡峭。奧地利的中文譯名就是由德文名稱音譯過來。", "messages": null, "license": "CC BY-SA 4.0", "source": "wikipedia:zh:奥地利", "greedy": false, "greedy_prompt": null}
+{"seq_id": 131, "group": "multilingual", "kind": "text", "lang": "zh", "text": "凱旋門是一種獨立的紀念性結構，該建築呈現拱門形狀，是門的一種，通常為横跨在一条道路上之獨立性建築，在其最簡單的形式中，凱旋門由兩個拱門連接的巨大橋墩組成，頂部設有一個平坦的柱頂或閣樓，上面能夠安裝雕像或刻有紀念銘文。主體結構通常雕刻裝飾、浮雕和奉獻者等。更精緻的凱旋門則可能設有多個拱門。\n\n凱旋門是與古羅馬建築相關，最有影響力和最獨特的建築類型之一。凱旋門被認為古代羅馬人創建的，用於紀念戰爭勝利的重大公共事件，例如新殖民地的建立、道路或橋樑的建造、皇室成員的去世或新皇帝的登基等，至今留存的羅馬凱旋門則包括提图斯凯旋门、塞维鲁凯旋门及君士坦丁凱旋門。啟發了許多後羅馬國家和統治者以模仿凱旋門。使世界各地皆有仿照古羅馬凱旋門建造的建築物。最著名的是巴黎的凱旋門、聖彼得堡的納爾瓦凱旋門與倫敦的威靈頓拱門。\n\n凱旋門本身具有慶祝軍事成功之宣傳性質功能，因此不應與紀念門、拱門和城門混淆，例如柏林的勃蘭登堡門、紐約市的華盛頓廣場拱門或新德里的印度門。或是因紀念一個事件（例如國家獨立、戰爭傷亡）而建造的建築。凱旋門也是中世紀教堂聖壇入口上方拱門的名稱，能夠放置十字架。\n\n起源與發展\n\n羅馬凱旋門\n\n凱旋門的創立及發展通常被視為古羅馬建築的一部分。如同過古羅馬時期所興建的渡槽、橋樑、圓形劇場和圓頂一樣，凱旋門也是採用拱形原理技術而興建。羅馬人可能從伊特拉斯坎文明學習了拱門建造技術。在當時，伊特魯里亞人經常興建精心裝飾的單海灣拱門，作為通往城市的大門或門戶 ；伊特魯里亞拱門的例子在今日的佩魯賈和沃爾泰拉依舊可見。", "messages": null, "license": "CC BY-SA 4.0", "source": "wikipedia:zh:凯旋门", "greedy": false, "greedy_prompt": null}
+{"seq_id": 132, "group": "multilingual", "kind": "text", "lang": "es", "text": "Güímar es una localidad y municipio español perteneciente a la isla de Tenerife, en la provincia de Santa Cruz de Tenerife, comunidad autónoma de Canarias.\n\nToponimia \nEl municipio toma su nombre de su capital, que a su vez lo hace del antiguo reino guanche. Algunos autores traducen el término por 'ángulo, esquina, rincón'.\n\nSímbolos\n\nEscudo \nEl escudo fue aprobado por el Ministerio de Gobernación el 2 de febrero de 1928.\n\nEn su parte central, en franja oblicua, en fondo de plata, dos personajes entrelazan sus brazos simbolizando «la paz sin sangre». Uno de los brazos, desnudo, representa a Añaterve el Bueno y su reino, el otro armado y desprovisto de guantelete a Alonso Fernández de Lugo y los suyos. A ambos lados los atributos de uno y otro: la isla de Tenerife, con el Teide en erupción, emergiendo de las aguas marinas, y el castillo y el león rampante, atributos éstos de los reinos que llevaron a cabo la conquista. Rodea esta escena central una bordura de sinople con ocho rejas de arado que representan «el trabajo agrícola» y en la cúspide, presidiendo el timbre del escudo, la corona de los Reyes Católicos.\n\nPendón \nEl municipio cuenta con pendón heráldico, aprobado por la Corporación Municipal en mayo de 1966. El pendón es de color blanco con el escudo al centro.\n\nGeografía física\n\nUbicación \nSe extiende por el sector sureste de la isla, limitando con los municipios de Arafo, La Orotava y Fasnia. El límite sur es el barranco de Herques, que sirve de frontera con el municipio de Fasnia; la divisoria con el municipio de Arafo, al norte, está constituida por la aproximación a la mediana de la corriente de lava producida por la erupción volcánica de 1706, una de las erupciones históricas de Tenerife, que discurre de la cumbre dorsal de la isla al mar, aunque en este caso la trayectoria no llegó al mismo, sino que finalizó a unos trescientos metros de la costa. El lado oeste del municipio lo constituye la cordillera dorsal de la isla, cuyo máxima elevación es Izaña, importante centro de astronomía, y uno de los más antiguos observatorios meteorológicos de España.\n\nOrografía", "messages": null, "license": "CC BY-SA 4.0", "source": "wikipedia:es:Güímar", "greedy": false, "greedy_prompt": null}
+{"seq_id": 133, "group": "multilingual", "kind": "text", "lang": "es", "text": "Los Mitos de Cthulhu constituyen un ciclo literario de horror cósmico comprendido entre 1921 y 1935 por el escritor estadounidense Howard Phillips Lovecraft y complementado por otros escritores pertenecientes al Círculo de Lovecraft. Aunque muy vinculado a la ciencia ficción, el género onírico y la fantasía pura, en rigor los Mitos de Cthulhu pertenecen a la tradición del cuento de terror anglosajón. \n\nEn él se renueva el desgastado horror gótico de fantasmas y seres inmateriales en un terror realista, de seres monstruosos y desconocidos que se esconden en los parajes más oscuros de la Tierra, el tiempo y el espacio. Influido por Arthur Machen y Lord Dunsany, los Mitos exploran a ciegas la perspectiva de que bajo el mundo cotidiano y conocido se esconde una realidad prodigiosa y aterradora que acecha a la humanidad desde las tinieblas y sume en el pánico o la locura a quien osa atisbar los abismos de aquella inaprensible dimensión.\n\nPese a las diferencias en los relatos, su principio fundamental fue establecido por el propio Lovecraft:\n\nEn la actualidad sigue habiendo escritores que desarrollan sus historias en el marco de los Mitos de Cthulhu, aportando nuevos elementos (localizaciones, sucesos, criaturas y personajes) al corpus de los Mitos, como por ejemplo Francisco Javier Olmedo (con sus novelas del Ciclo del Dios de la Carne) o Marc Barqué (con sus libros titulados \"Nuevos Mitos de Cthulhu\").\n\nContenido de la obra \n\nPrecedida por destacados escritores (Lord Dunsany, Ambrose Bierce, R. W. Chambers, Arthur Machen y Algernon Blackwood) y enriquecida por diversos autores (Frank Belknap Long, Robert E. Howard, Clark Ashton Smith, Hazel Heald, Henry Kuttner, Robert Bloch, August Derleth, Ramsey Campbell, Brian Lumley) corresponde no obstante a Howard Phillips Lovecraft el protagonismo en la creación de los Mitos. A continuación se nombran, según Derleth, los trece relatos de Lovecraft pertenecientes a este ciclo literario ordenados de acuerdo a su fecha de creación:\n\n La ciudad sin nombre (1921)\n El ceremonial (1923)\n La llamada de Cthulhu (1926)\n El color surgido del espacio (1927)\n El caso de Charles Dexter Ward (1927-1928)\n El horror de Dunwich (1928)", "messages": null, "license": "CC BY-SA 4.0", "source": "wikipedia:es:Mitos de Cthulhu", "greedy": false, "greedy_prompt": null}
+{"seq_id": 134, "group": "multilingual", "kind": "text", "lang": "es", "text": "El pion (abreviatura del vocablo griego pi meson) es el nombre que conjuntamente reciben tres partículas subatómicas:  π0, π+, π−. Es un mesón y, como tal, actúa como partícula portadora de la interacción nuclear fuerte. Está compuesto por un cuark y un antiquark y es el más ligero de todos los mesones. Los piones con carga son inestables con una vida media de 26,033 nanosegundos desintegrándose en un muon y en un neutrino muónico. Los piones neutros son todavía más inestables, con una vida media de 8,4 × 10−17 segundos, y se desintegran en dos fotones en forma de rayos gamma.\n\nPropiedades básicas  \nLos piones tienen espín cero, y están compuestos por un  cuark \"up\" y otro \"anti-down\" componen el π+, mientras que un cuark \"down\" y otro \"anti-up\" componen el π−, su antipartícula. La combinación \"up-antiup\" y \"down-antidown\" constituyen el π0, el cual es su propia antipartícula.\n\nEl mesón π± tiene una masa de 139,6 MeV/c2 y una vida media de 2,6 × 10−8 segundos. La desintegración principal es en un muon y un neutrino muónico o bien en un antimuón y un antineutrino muónico en el caso de que la carga del pion sea negativa:\n\nEl π0 es un poco más ligero, teniendo una masa de 135,0 MeV/c2 y una vida media mucho más corta, de 8,4 × 10−17 segundos. La desintegración principal es la de dos fotones:\n\nEstabilidad de los núcleos \nLa mayoría de núcleos atómicos por debajo de un cierto peso atómico y que además presentan un equilibrio entre el número de neutrones y el número de protones (número atómico) son estables. Sin embargo, se sabe que los neutrones aislados y los núcleos con demasiados neutrones (o demasiados protones) son inestables.\n\nLa explicación de esta estabilidad de los núcleos reside en los piones. Aisladamente los neutrones pueden sufrir vía interacción débil la siguiente desintegración, donde un neutrón se desintegra en un protón junto con un electrón más un antineutrino electrónico:\n\nSin embargo, dentro del núcleo atómico la cercanía entre neutrones y protones hace que sean mucho más rápidas, vía interacción fuerte las reacciones:", "messages": null, "license": "CC BY-SA 4.0", "source": "wikipedia:es:Pion", "greedy": false, "greedy_prompt": null}
+{"seq_id": 135, "group": "multilingual", "kind": "text", "lang": "es", "text": "Stipa es un género de gramíneas (Poaceae), perennes y cespitosas, que comprende aproximadamente 250 especies distribuidas por todo el globo.\n\nGeneralidades \nEn las regiones esteparias de América, el género Stipa es con frecuencia dominante. Forma la Pampa seca, las antiguas Grandes Praderas de Norteamérica, las sabanas de África y Sudamérica y las estepas euroasiáticas, mediterráneas y africanas. Sus especies tienen a veces valor forrajero, o bien son perjudiciales debido a las perforaciones que sus frutos producen en los cueros de los animales. Una especie del sur de Europa y norte de África, Stipa tenacissima L., es el \"esparto\" utilizado en la fabricación de papel y cordelería.\n\nMuchas especies están adaptadas a suelos áridos, semidesiertos y estepas, a las que les dan nombre, caracterizan y se emplean para evitar la erosión de los suelos y la desertización.\nEn zonas apropiadas forman praderas densas solas o en compañía de otros grupos. Aunque en agricultura suelen ser consideradas malezas, algunas especies se emplean por la resistencia de sus fibras para realizar cuerdas, cestos y otras urdimbres. Por su gran capacidad de regeneración y su resistencia a desaparecer son un alimento importante para muchos herbívoros y caracterizan muchas praderas salvajes en zonas secas de todo el planeta. Aunque hay excepciones, conforme las zonas son más frescas y húmedas, se desarrolla otro tipo de vegetación que sustituyen poco a poco a las especies del género stipa.\n\nDescripción \nStipa incluye pastos perennes, cespitosos, de unos 30 cm pero en ocasiones de hasta dos metros y medio, frecuentemente con hojas de lámina convoluta. Las espiguillas se hallan dispuestas en panojas generalmente laxas. Todas las Stipa se reconocen por tener unas aristas muy largas, que cuando son maduras en algunos se enrollan entre ellas quedando completamente enmarañadas.", "messages": null, "license": "CC BY-SA 4.0", "source": "wikipedia:es:Stipa", "greedy": false, "greedy_prompt": null}
+{"seq_id": 136, "group": "multilingual", "kind": "text", "lang": "es", "text": "La ética o filosofía moral es la rama de la filosofía que estudia la conducta humana, lo correcto y lo incorrecto, lo bueno y lo malo, la moral, el buen vivir, la virtud, la felicidad y el deber. La ética contemporánea se suele dividir en tres ramas o niveles: la metaética estudia el origen, naturaleza y significado de los conceptos éticos, la ética normativa busca normas o estándares para regular la conducta humana, y la ética aplicada examina controversias éticas específicas.\n\nÉtica y moral son conceptos muy relacionados que a veces se usan como sinónimos, pero tradicionalmente se diferencian en que la ética es la disciplina académica que estudia la moral. La ética no inventa los problemas morales, sino que reflexiona sobre ellos. Las acciones relevantes para la ética son las acciones morales, que son aquellas realizadas de manera libre, ya sean privadas, interpersonales o políticas. La ética no se limita a observar y describir esas acciones, sino que busca determinar si son buenas o malas, emitir juicio sobre ellas y así ayudar a encauzar la conducta humana.\n\nEl estudio de la ética se remonta a los orígenes mismos de la filosofía en la Antigua Grecia, y su desarrollo histórico ha sido amplio y variado. A lo largo de la historia ha habido diversas maneras de entender la ética y distintas propuestas morales orientadoras de la vida humana.\n\nAunque la ética siempre fue una rama de la filosofía, su amplio alcance la conecta con muchas otras disciplinas, incluyendo la antropología, biología, economía, historia, política, sociología y teología.\n\nDefiniciones\nLa palabra ética se deriva del griego antiguo ēthikós (), que significa \"relativa al carácter de uno\", que a su vez proviene de la palabra raíz êthos () que significa \"carácter, naturaleza moral\". Esta palabra fue transferida al latín como ethica y luego al francés como éthique, del cual fue transferido al español.", "messages": null, "license": "CC BY-SA 4.0", "source": "wikipedia:es:Ética", "greedy": false, "greedy_prompt": null}
+{"seq_id": 137, "group": "multilingual", "kind": "text", "lang": "es", "text": "El 12 de septiembre es el 255.º (ducentésimo quincuagésimo quinto) día del año —el 256.º (ducentésimo quincuagésimo sexto) en los años bisiestos— en el calendario gregoriano. Quedan 110 días para finalizar el año.\n\nAcontecimientos  \n\n : en Grecia se libra la batalla de Maratón, donde los griegos derrotaron a los invasores persas. Dos días después habría sucedido la fábula de Filípides (que corrió 42 km desde Maratón a Atenas para morir diciendo «Hemos vencido»).\n 1213: Simón de Montfort derrota a Pedro II de Aragón en la batalla de Muret (Cruzada albigense).\n 1229: Mallorca es conquistada por Jaime I de Aragón, apodado El Conquistador.\n 1297: en la provincia de Zamora se firma el Tratado de Alcañices, por el cual los pueblos de Aracena y Aroche (Huelva), en poder de Portugal, pasan definitivamente a la Corona de Castilla.\n 1502: a las costas de la actual Nicaragua llegan las naves de Cristóbal Colón, bautizando el Cabo Gracias a Dios por haber salido de una tormenta.\n 1504: Cristóbal Colón parte hacia España luego de concluir su cuarto viaje a América.\n 1532: en Angrogna (norte de Italia) se inicia el Sínodo de Chanforan, en el cual los valdenses de Francia e Italia deciden adherirse a la Reforma.\n 1537: Carlos V autoriza que cuando en el Río de la Plata fallezca la primera autoridad real, la gente elija libremente su sucesor.\n 1581: Felipe II es proclamado rey de Portugal.\n 1631: frente a las costas brasileñas de Pernambuco, la escuadra española derrota a la neerlandesa en un combate naval.\n 1683: en la Batalla de Kahlenberg, las tropas cristianas ―mandadas por Juan III Sobieski― vencen a los otomanos de Kara Mustafá.\n 1827: en Valparaíso (Chile) se funda el periódico El Mercurio, el más antiguo en lengua española.\n 1843: en la provincia de Corrientes (Argentina) se funda la aldea de Paso de los Libres (fronteriza con la localidad brasileña de Uruguayana).\n 1847: cerca de la ciudad de México, en el marco de la intervención estadounidense en México, comienza la batalla de Chapultepec.\n 1848: Suiza se convierte en un estado federal.", "messages": null, "license": "CC BY-SA 4.0", "source": "wikipedia:es:12 de septiembre", "greedy": false, "greedy_prompt": null}
+{"seq_id": 138, "group": "multilingual", "kind": "text", "lang": "es", "text": "El Zilog Z80 (Z80) es un microprocesador de 8 bits cuya arquitectura se encuentra a medio camino entre la organización de acumulador y de registros generales. Si consideramos al Z80 como procesador de arquitectura de registros generales, se sitúa dentro del tipo de registro-memoria.\n\nFue lanzado al mercado en marzo de 1976 por la compañía Zilog, y se popularizó en los años 1980 a través de ordenadores como el Sinclair ZX Spectrum, Amstrad CPC o los ordenadores de sistema MSX. Es uno de los procesadores de más éxito del mercado, del cual se han producido infinidad de versiones clónicas, y sigue siendo usado de forma extensiva en la actualidad en multitud de sistemas embebidos.\n\nHistoria\n\nIntroducción \nEl Z80 fue diseñado principalmente por Federico Faggin, que estuvo trabajando en Intel como diseñador jefe del Intel 4004 y del Intel 8080. Cuando se terminó de producir, en 1974, Federico Faggin dejó Intel, fundó Zilog y comenzó a trabajar en el diseño de Z80 basándose en la experiencia adquirida creando el Intel 4004, 8080 y basándose en la estructura de este último y algunos circuitos integrados más. Masatoshi Shima, el principal diseñador de niveles de lógica y transistores del 4004 y el 8080 bajo la supervisión de Faggin, se unió al equipo de Zilog. Dos años después estaba a la venta el Z80.\n\nEl Z80 estaba diseñado para ser compatible a nivel de código con el Intel 8080, de forma que la mayoría de los programas para el 8080 pudieran funcionar en él, especialmente el sistema operativo CP/M.\n\nEl Z80 tenía ocho mejoras fundamentales respecto al Intel 8080:\n\n Un conjunto de instrucciones mejorado, incluyendo los nuevos registros índice IX e IY y las instrucciones necesarias para manejarlos.\n Dos bancos de registros que podían ser cambiados de forma rápida para acelerar la respuesta a interrupciones.\n Instrucciones de movimiento de bloques, E/S de bloques y búsqueda de bytes.\n Instrucciones de manipulación de bits.\n Número de transistores de 6500.\n Un contador de direcciones para el refresco de la DRAM integrado, que en el 8080 tenía que ser proporcionado por el conjunto de circuitos de soporte.\n Alimentación única de 5 voltios.", "messages": null, "license": "CC BY-SA 4.0", "source": "wikipedia:es:Zilog Z80", "greedy": false, "greedy_prompt": null}
+{"seq_id": 139, "group": "multilingual", "kind": "text", "lang": "es", "text": "Un videojuego de disparos en primera persona o FPS (del inglés, first-person shooter) es un género de videojuegos que simula el uso de armas de fuego desde una perspectiva de primera persona. El género comparte rasgos comunes con otros juegos de disparos, que a su vez hacen que caiga bajo el título de juego de acción. Desde el inicio del género, los gráficos avanzados en 3D y pseudo-3D han desafiado el desarrollo de hardware, y los juegos multijugador han sido integrales.\n\nEl género de disparos en primera persona se remonta a Maze War, cuyo desarrollo comenzó en 1973, y en 1974 a Spasim. Más tarde, y después de títulos más lúdicos como MIDI Maze en 1987, el género se convirtió en una forma más violenta con Wolfenstein 3D de 1992, a la que se le atribuye la creación del arquetipo básico del género en el que se basan los títulos posteriores. Uno de esos títulos, y el progenitor de la mayor aceptación y popularidad general del género, fue Doom, uno de los juegos más influyentes en este género; durante algunos años, el término «clon de Doom» se usó para designar este género debido a la influencia de este juego. «De tiros en corredor» (corridor shooter) fue otro nombre común para el género en sus primeros años, ya que las limitaciones de procesamiento del hardware de la época significaban que la mayor parte de la acción en los juegos tenía que llevarse a cabo en áreas cerradas.\n\nHalf-Life de 1998 —junto con su secuela de 2004, Half-Life 2— mejoró la narrativa y los elementos del rompecabezas. En 1999, el mod de Half-Life Counter-Strike fue lanzado y junto con Doom es quizás uno de los de tiros en primera persona más influyentes y paradigmáticos. Tras este éxito llegaron famosas sagas como Quake, Battlefield, Medal of Honor, o Unreal.", "messages": null, "license": "CC BY-SA 4.0", "source": "wikipedia:es:Videojuego de disparos en primera persona", "greedy": false, "greedy_prompt": null}
+{"seq_id": 140, "group": "multilingual", "kind": "text", "lang": "ar", "text": "شكيم هو الاسم القديم لمدينة نابلس أو لمدينة كانت تقع في الطرف الشرقي لمدينة نابلس الحالية، قام الكنعانيون بتأسيسها في أواسط الألفية الثالثة قبل الميلاد، وتعني كلمة شكيم المنكب أو الأرض المرتفعة، وموقع المدينة في الغالب شرق مدينة نابلس الحالية إلى الشمال قليلا من القرية المعروفة اليوم باسم بلاطة البلد، وكان الرحالة الألماني هيرمان تيرش قد استرعى اهتمامه تل كبير عليه قطع حجرية تعود إلى بقايا جدار يمتد إلى مسافة 8 أمتار على الجانب الغربي من التل، وتوقع أن تكون هذه الآثار هي آثار مدينة شكيم بحكم موقعها المميز المعترض للممر بين جبلي عيبال وجرزيم وتحكمه بسهل عسكر.\n\nأظهرت الحفريات التي قامت بها بعثة أمريكية مشتركة من جامعة درو ومعهد مكورميك اللاهوتي والمدرسة الأمريكية للأبحاث الشرقية بين عامي 1956-1967 أن مدينة شكيم كنعانية قد بنيت في أواسط الألف الثالث قبل الميلاد فوق مستوطن أقدم يعود إلى الألف الرابع قبل الميلاد؛ حيث اكتشفت بعض الفخاريات التي تعود إلى العصر الحجري-المعدني (الكالكوليثي). كما أظهرت الحفريات أن المدينة قد تميزت مع بداية الألف الثاني قبل الميلاد بتحصيناتها الكنعانية القوية.\n\nالتسمية \nشكيم تسمية كنعانية اللفظ، تعني المنكب أو الكتف أو المكان المرتفع، وجاء في التوراة أن يعقوب حينما أتى إلى شكيم وأذن له بالنزول بجوارها، كان حاكم المدينة حمور الحوى الكنعاني وذكرت أن اسم ابنه كان شكيم أيضاً.\n\nالتاريخ", "messages": null, "license": "CC BY-SA 4.0", "source": "wikipedia:ar:شكيم", "greedy": false, "greedy_prompt": null}
+{"seq_id": 141, "group": "multilingual", "kind": "text", "lang": "ar", "text": "جُبْران خليل جُبْران (6 يناير 1883 – 10 أبريل 1931 م) شاعر، كاتب، فيلسوف، عالم روحانيات، رسّام، فنان تشكيلي، نحّات عُثماني من أدباء وشعراء المهجر، ولد في بلدة بشري في شمال لبنان زمن متصرفية جبل لبنان، في سوريا العثمانية ونشأ فقيرًا، كانت والدته كاميلا في الثلاثين من عمرها عندما ولدته وكان والده خليل هو زوجها الثالث. ولم يتلق جُبران التعليم الرسمي خلال شبابه في متصرفية جبل لبنان. هاجر صبيًا مع عائلته إلى الولايات المتحدة الأمريكية، ليدرُس الأدب وليبدأ مسيرتهُ الأدبية، والكتابة باللغتين العربية والإنجليزية، امتاز أسلوبه بالرومانسية ويعتبر من رموز عصر نهضة الأدب العربي الحديث، وخاصةً في الشعر النثري.\n\nكان جبران عضوًا في رابطة القلم في نيويورك، المعروفة حينها بشُعراء المهجر جنبًا إلى جنب مؤلفين لبنانيين مثل الأمين الريحاني وميخائيل نعيمة وإيليا أبو ماضي. اشتهر في المهجر بكتابه النبي الذي صدر في عام 1923، وهو مثال مُبكر على «الخيال الملهم» بما في ذلك سلسلة من المقالات الفلسفية المكتوبة في النثر الشعري باللغة الإنجليزية، وحصل الكتاب على مبيعات جيدة على الرغم من الإستقبال الناقد والرائع. عُرفَ أيضاً بالشاعر الأكثر مبيعًا بعد شكسبير ولاوزي،  وقد تُرجمَ كتاب النبي إلى ما يصل إلى 110 لغة  منها الصينية.", "messages": null, "license": "CC BY-SA 4.0", "source": "wikipedia:ar:جبران خليل جبران", "greedy": false, "greedy_prompt": null}
+{"seq_id": 142, "group": "multilingual", "kind": "text", "lang": "ar", "text": "فدوى طوقان (1917 - 2003) من أهم شاعرات فلسطين في القرن العشرين من مدينة نابلس وهي من عائلة فلسطينية معروفة، ولقبت بشاعرة فلسطين، حيث مثّل شعرها أساسًا قويًا للتجارب الأنثوية في الحب والثورة واحتجاج المرأة على المجتمع.\n\nالسيرة الشخصية \nولدت فدوى طوقان في مدينة نابلس، وتلقت تعليمها حتى المرحلة الابتدائية، حيث اعتبرت عائلتها مشاركة المرأة في الحياة العامة أمرًا غير مقبول، فتركت مقاعد الدراسة واستمرت في تثقيف نفسها بنفسها، ثم درست على يد أخيها شاعر فلسطين الكبير إبراهيم طوقان، الذي نمى مواهبها ووجهها نحو كتابة الشعر، كما شجعها على نشره في العديد من الصحف العربية، وأسماها «أم تمّام». ثم أسماها محمود درويش لاحقاً «أم الشعر الفلسطيني».\nومع أنها وقّعت قصائدها الأولى باسم «دنانير»، وهو اسم جارية، إلا أن أحب أسمائها المستعارة إلى قلبها كان «المطوّقة» لأنه يتضمن إشارة مزدوجة، بل تورية فصيحة إلى حال الشاعرة بالتحديد. فالمطوقة تعني انتسابها إلى عائلة طوقان المعروفة، وترمز، في الوقت نفسه، إلى أحوالها في مجتمع تقليدي غير رحيم.\nوقد توالت النكبات في حياة فدوى طوقان بعدما توفي والدها ثم أخوها ومعلمها إبراهيم، وأعقب ذلك احتلال فلسطين إبان نكبة 1948، وقد تركت تلك المآسي المتلاحقة أثرها الواضح في نفسية فدوى طوقان كما يتبين من شعرها في ديوانها الأول وحدي مع الأيام. وفي الوقت نفسه فان ذلك دفع فدوى طوقان إلى المشاركة في الحياة السياسية خلال الخمسينيات من القرن الماضي.", "messages": null, "license": "CC BY-SA 4.0", "source": "wikipedia:ar:فدوى طوقان", "greedy": false, "greedy_prompt": null}
+{"seq_id": 143, "group": "multilingual", "kind": "text", "lang": "ar", "text": "أَبُو تَمَّامْ حَبِيبْ بْنْ أَوْسْ بْنْ اَلْحَارِثْ اَلطَّائِيّ (188 - 231 هـ / 803-845م)، أحد أمراء البيان، ولد بمدينة جاسم (من قرى حوران بسورية) ورحل إلى مصر واستقدمه المعتصم إلى بغداد فأجازه وقدمه على شعراء وقته فأقام في العراق ثم ولي بريد الموصل فلم يتم سنتين حتى توفي بها.\n\nكان أسمر، طويلاً، فصيحاً، حلو الكلام، فيه تمتمة يسيرة، يحفظ أربعة عشر ألف أرجوزة من أراجيز العرب غير القصائد والمقاطيع. وفي أخبار أبي تمام للصولي: .\n\nفي شعره قوة وجزالة، واختلف في التفضيل بينه وبين المتنبي والبحتري، له تصانيف، منها فحول الشعراء، وديوان الحماسة، ومختار أشعار القبائل، ونقائض جرير والأخطل، نُسِبَ إليه ولعله للأصمعي كما يرى الميمني.\n\nنشأته \nولد في جاسم من قرى حوران (في سوريا حالياً) أيام الرشيد، وكان أولا حدثا يسقي الماء بمصر، ثم جالس الأدباء، وأخذ عنهم وكان يتوقد ذكاء. وسحت قريحته بالنظم البديع. فسمع به المعتصم، فطلبه، وقدمه على الشعراء، وله فيه قصائد. وكان يوصف بطيب الأخلاق والظرف والسماحة.\nوقيل: قدم في زي الأعراب، فجلس إلى حلقة من الشعراء، وطلب منهم أن يسمعوا من نظمه، فشاع وذاع وخضعوا له. وصار من أمره ما صار.", "messages": null, "license": "CC BY-SA 4.0", "source": "wikipedia:ar:أبو تمام", "greedy": false, "greedy_prompt": null}
+{"seq_id": 144, "group": "multilingual", "kind": "text", "lang": "ar", "text": "أبو الوليد عبد الملك بن مروان بن الحكم بن أبي العاص بن أمية القرشي (26 هـ - 86 هـ / 646 - 705م) الخليفة الخامس من خلفاء بني أمية والمؤسس الثاني للدولة الأموية. ولد في المدينة وتفقه فيها علوم الدين، وكان قبل توليه الخلافة ممن اشتهر بالعلم والفقه والعبادة، وكان أحد فقهاء المدينة الأربعة، قال الأعمش عن أبي الزناد: «كان فقهاء المدينة أربعة: سعيد بن المسيب، وعروة بن الزبير، وقبيصة بن ذؤيب، وعبد الملك بن مروان». استلم الحكم بعد أبيه مروان بن الحكم سنة 65 هـ الموافق 684م، وحكم دولة الخلافة الإسلامية واحدًا وعشرين عامًا.\n\nتسلم عبد الملك بن مروان حكم الدولة الأموية في وقت كانت الفتن والاضطرابات والانقسامات تعصف بها. في الكوفة كان أنصار الحسين يشعرون بالتقصير والذنب بعد معركة كربلاء، وعندما عم الاضطراب أنحاء بلاد العالم الإسلامي بعد موت يزيد بن معاوية، خرجت من الكوفة ثورة التوابين في 5 ربيع الثاني 65 هـ وفي منطقة عين الوردة استطاع عبد الملك بن مروان القضاء على ثورتهم. ما إن أُخمدت ثورة التوابين حتى خرج المختار الثقفي بشعار يا لثارات الحسين، واصطدم مع الأمويين في أكثر من وقعة انتصر فيها، لكن مصعب بن الزبير كفى عبد الملك مواصلة قتاله واستطاع قتله وإخضاع الكوفة تحت سلطة ابن الزبير.", "messages": null, "license": "CC BY-SA 4.0", "source": "wikipedia:ar:عبد الملك بن مروان", "greedy": false, "greedy_prompt": null}
+{"seq_id": 145, "group": "multilingual", "kind": "text", "lang": "ar", "text": "محافظة المفرق هي إحدى المحافظات الإثني عشر المكونة للمملكة الأردنية الهاشمية وهي ثاني أكبر محافظة من حيث المساحة بعد محافظة معان، تشكل منطقة البادية الشمالية الغالبية العظمى من مساحتها. تقـع المحافظة في الشمـال الشرقي للأردن، يحدها العـراق من الشرق وسوريـا من الشمال والسعودية من الجنوب والشرق. وتبلغ مساحتها (26,552) كم2 أي ما يعادل 29.6% من مساحة الأردن وهي من الشرق ذات طبيعة صحراوية غنيّة بالمياه الجوفية تمدّ الأردن بمحاصيل الخضروات والثروة الحيوانية. أمّا من الغرب فهي ذات طبيعة خصبة منتجة لزيت الزيتون.\n\nشهدت المحافظة خلال السنوات القليلة الماضية نشاطا واضحا في المجالين الصناعي والزراعي، كما أصبحت منطقة البادية الشمالية المصدر الرئيسي في الإنتاج الزراعي والحيواني حيث تحولت البادية إلى حقول خضراء نتيجة انتشار الآبار الارتوازية في تلك المنطقة، وتعتبر المحافظة المصدر الثاني بعد الأغوار من حيث الإنتاج الزراعي.\n\nمدينة المفرق هي مركز المحافظة التي تحتوي على العديد من المعالم الأثرية أبرزها مدينة أم الجمال إحدى مدن حلف الديكابولس الروماني. كما أنشئ فيها مخيم الزعتري لللاجئين السوريين عام 2012. وتضم أيضا مقبرة الشهداء العراقيين التي تعود إلى عام  1948 المعروف بعام النكبة والذي خاضت فيه دول عربية حرباً شرسة مع العدو الصهيوني في نفس العام. كما تضم المحافظة جامعة حكومية واحدة وهي جامعة آل البيت.\n\nالتسمية", "messages": null, "license": "CC BY-SA 4.0", "source": "wikipedia:ar:المفرق (محافظة)", "greedy": false, "greedy_prompt": null}
+{"seq_id": 146, "group": "multilingual", "kind": "text", "lang": "ar", "text": "المدية، إحدى ولايات الجزائر ورمزها هو 26، يقدر عمر ولاية المدية بألف عام أو يزيد، فالمدية عاصمة  بايلك التيطري. يعود تاريخها إلى عهد قديم أي حوالي 350 هـ، بحيث قال أحد المؤرخين (المدية عتيقة قديمة، وأن المدية سبقت بني زيري وأنها أقدم من أشير...)، فقد تداولت عليها عدة حضارات وسكنتها الكثير من الشعوب وهي تقاطع مدينة تلمسان في عدة تقاليد لتشابه أنماط المعيشة لدى سكانها وطريقة عمرانهم، وأسلوب حياتهم ومن الشخصيات التي ولدت وترعرعت بالمدينة الشيخ ابن شنب وفضيل اسكندر، كما أنها كانت عاصمة الولاية الرابعة في حرب التحرير ضد المستعمر الفرنسي.\n\nوالمدية اليوم من بين ولايات الجزائر تتوفر على منتوج ثقافي، سياحي، وتاريخي، تضم 64 بلدية موزعة على 19 دائرة و5 مقاطعات إدارية هي: المدية، البرواقية، قصر البخاري، شلالة العذاورة، تابلاط.\n\nالمناطق السياحية في ولاية المدية \n المحمية الطبيعية بالحمدانية وبن شيكاو، وتيمزقيدة\n ضريح العقيد بوقرة\n الاثار الإسلامية (آشير) ببلدية الكاف الأخضر ناحية شلالة العذاورة على بعد  غربا على الطريق الوطني رقم 60.\n\nالموقع الجغرافي \nتقع المدية في الأطلس التلي على بعد  تقريبا جنوب الجزائر العاصمة، وتتربع على مساحة قدرها 8700 كلم2 وعلى ارتفاع  من سطح البحر. وتشترك ولاية المدية في الحدود مع العديد من ولايات الوسط:\n\n من الشمال: ولايات:البليدة، تيبازة، بومرداس،\n من الجنوب: ولايات: الجلفة، تيسمسيلت، المسيلة،\n من الشرق: البويرة،\n من الغرب: عين الدفلى،", "messages": null, "license": "CC BY-SA 4.0", "source": "wikipedia:ar:ولاية المدية", "greedy": false, "greedy_prompt": null}
+{"seq_id": 147, "group": "multilingual", "kind": "text", "lang": "ar", "text": "كاغوشيما  هي مدينة وميناء في اليابان. عاصمة محافظة كاغوشيما، تقع في جزيرة كيوشو، وتشرف على الخليج الذي يحمل نفس الاسم. مساحتها 547.06 كلم،² عدد سكانها 606386 (2011). يطلق على مدينة كاغوشيما اسم «نابولي الشرق» بسبب خليجها ومناخها الدافئ وإطلالة بركان ساكوراجيما.\n\nتاريخ \nكانت المدينة معقلا للزعماء الإقطاعيين الكبار (دايميو) بين سنوات 794-1185 م، كما اندلعت فيها أحداث انتفاضة ساتسوما عام 1877 م (قاد فيها سايغو تاكاموري مجموعة من الساموراي ضد النظام الجديد). تعرضت للخراب عام 1914 م، عندما اندلع البركان ساكوراجيما، وأخذ يلفظ حممه في خليج كاغوشيما.\n\nاقتصاد \n\nتعتبر مدينة كاغوشيما المركز الثقافي والتجاري لمنطقة جنوبي كيوشو. تشتهر بصناعة الخزف، والتي يطلق عليها اسم ساتسوما، الحرير، المواد القطنية، المواد الغذائية، وبعض أنواع السيجار.\n\nتعتمد المدينة أيضا على السياحة من خلال جزيرة ساكوراجيما التي تحتوي بركان ومتنزه جيولوجي، كما أن علم المدينة يحتوي صورة هذا البركان النشط الذي يعد أبرز معالمها.\n\nالتعليم\n\nالجامعي \nمن أشهر الجامعات في كاغوشيما هي جامعة كاغوشيما.\n\nمدن التوأمة \nمدينة كاغوشيما أقامت علاقات توأمة مع 6 مدن وهي:\n نابولي  3 أيار 1960.\n بيرث  24 نيسان 1974.\n تشانغشا  30 تشرين الأول 1982.\n ميامي (فلوريدا)  1 تشرين الثاني 1990.\n تسوروكا  7 تشرين الثاني 1969.\n\nمعرض الصور\n\nأعلام \n\n ريوجي فوجيياما\n تومويوكي هيراسي\n واتارو كوبو\n كوهي حتاتا\n كيسوكي إيواشيتا\n شوهي أكاساكي\n يوجي هيكوتيك", "messages": null, "license": "CC BY-SA 4.0", "source": "wikipedia:ar:كاغوشيما", "greedy": false, "greedy_prompt": null}
+{"seq_id": 148, "group": "json", "kind": "text", "lang": "en", "text": "Return a JSON object describing the person Asha who lives in Pune. Use the keys name, city, age and hobbies. Output only the JSON.\n{\n  \"name\": \"Asha\",\n  \"city\": \"Pune\",\n  \"age\": 39,\n  \"hobbies\": [\n    \"cycling\",\n    \"painting\"\n  ]\n}", "messages": null, "license": "CC0-1.0", "source": "generated:build.py", "greedy": true, "greedy_prompt": "Return a JSON object describing the person Asha who lives in Pune. Use the keys name, city, age and hobbies. Output only the JSON.\n"}
+{"seq_id": 149, "group": "json", "kind": "text", "lang": "en", "text": "Return a JSON object for an order of one kettle shipped to Lagos. Use the keys order_id, item, quantity, price and shipping. Output only the JSON.\n{\n  \"order_id\": \"A-1791\",\n  \"item\": \"kettle\",\n  \"quantity\": 1,\n  \"price\": 74.81,\n  \"shipping\": {\n    \"city\": \"Lagos\",\n    \"express\": false\n  }\n}", "messages": null, "license": "CC0-1.0", "source": "generated:build.py", "greedy": true, "greedy_prompt": "Return a JSON object for an order of one kettle shipped to Lagos. Use the keys order_id, item, quantity, price and shipping. Output only the JSON.\n"}
+{"seq_id": 150, "group": "json", "kind": "text", "lang": "en", "text": "Return a JSON object with a list of three tasks for Li Wei. Each task has the keys title, done and priority. Output only the JSON.\n{\n  \"owner\": \"Li Wei\",\n  \"tasks\": [\n    {\n      \"title\": \"call plumber\",\n      \"done\": true,\n      \"priority\": 3\n    },\n    {\n      \"title\": \"fix bike\",\n      \"done\": false,\n      \"priority\": 1\n    },\n    {\n      \"title\": \"write report\",\n      \"done\": false,\n      \"priority\": 2\n    }\n  ]\n}", "messages": null, "license": "CC0-1.0", "source": "generated:build.py", "greedy": true, "greedy_prompt": "Return a JSON object with a list of three tasks for Li Wei. Each task has the keys title, done and priority. Output only the JSON.\n"}
+{"seq_id": 151, "group": "json", "kind": "text", "lang": "en", "text": "Return a JSON object for the weather in Lima with the keys city, temperature_c, conditions and wind. wind has speed_kmh and direction. Output only the JSON.\n{\n  \"city\": \"Lima\",\n  \"temperature_c\": 21,\n  \"conditions\": \"sunny\",\n  \"wind\": {\n    \"speed_kmh\": 15,\n    \"direction\": \"NE\"\n  }\n}", "messages": null, "license": "CC0-1.0", "source": "generated:build.py", "greedy": true, "greedy_prompt": "Return a JSON object for the weather in Lima with the keys city, temperature_c, conditions and wind. wind has speed_kmh and direction. Output only the JSON.\n"}
+{"seq_id": 152, "group": "json", "kind": "text", "lang": "en", "text": "Return a JSON object describing the person Noah who lives in Osaka. Use the keys name, city, age and hobbies. Output only the JSON.\n{\n  \"name\": \"Noah\",\n  \"city\": \"Osaka\",\n  \"age\": 54,\n  \"hobbies\": [\n    \"painting\",\n    \"chess\"\n  ]\n}", "messages": null, "license": "CC0-1.0", "source": "generated:build.py", "greedy": true, "greedy_prompt": "Return a JSON object describing the person Noah who lives in Osaka. Use the keys name, city, age and hobbies. Output only the JSON.\n"}
+{"seq_id": 153, "group": "json", "kind": "text", "lang": "en", "text": "Return a JSON object for an order of one water bottle shipped to Cairo. Use the keys order_id, item, quantity, price and shipping. Output only the JSON.\n{\n  \"order_id\": \"A-3028\",\n  \"item\": \"water bottle\",\n  \"quantity\": 2,\n  \"price\": 58.6,\n  \"shipping\": {\n    \"city\": \"Cairo\",\n    \"express\": true\n  }\n}", "messages": null, "license": "CC0-1.0", "source": "generated:build.py", "greedy": true, "greedy_prompt": "Return a JSON object for an order of one water bottle shipped to Cairo. Use the keys order_id, item, quantity, price and shipping. Output only the JSON.\n"}
+{"seq_id": 154, "group": "json", "kind": "text", "lang": "en", "text": "Return a JSON object with a list of three tasks for Lucas. Each task has the keys title, done and priority. Output only the JSON.\n{\n  \"owner\": \"Lucas\",\n  \"tasks\": [\n    {\n      \"title\": \"write report\",\n      \"done\": false,\n      \"priority\": 1\n    },\n    {\n      \"title\": \"fix bike\",\n      \"done\": false,\n      \"priority\": 3\n    },\n    {\n      \"title\": \"buy milk\",\n      \"done\": true,\n      \"priority\": 1\n    }\n  ]\n}", "messages": null, "license": "CC0-1.0", "source": "generated:build.py", "greedy": true, "greedy_prompt": "Return a JSON object with a list of three tasks for Lucas. Each task has the keys title, done and priority. Output only the JSON.\n"}
+{"seq_id": 155, "group": "json", "kind": "text", "lang": "en", "text": "Return a JSON object for the weather in Quito with the keys city, temperature_c, conditions and wind. wind has speed_kmh and direction. Output only the JSON.\n{\n  \"city\": \"Quito\",\n  \"temperature_c\": 13,\n  \"conditions\": \"fog\",\n  \"wind\": {\n    \"speed_kmh\": 9,\n    \"direction\": \"NE\"\n  }\n}", "messages": null, "license": "CC0-1.0", "source": "generated:build.py", "greedy": true, "greedy_prompt": "Return a JSON object for the weather in Quito with the keys city, temperature_c, conditions and wind. wind has speed_kmh and direction. Output only the JSON.\n"}
+{"seq_id": 156, "group": "json", "kind": "text", "lang": "en", "text": "Return a JSON object describing the person Hiro who lives in Hanoi. Use the keys name, city, age and hobbies. Output only the JSON.\n{\n  \"name\": \"Hiro\",\n  \"city\": \"Hanoi\",\n  \"age\": 55,\n  \"hobbies\": [\n    \"cooking\",\n    \"cycling\"\n  ]\n}", "messages": null, "license": "CC0-1.0", "source": "generated:build.py", "greedy": false, "greedy_prompt": "Return a JSON object describing the person Hiro who lives in Hanoi. Use the keys name, city, age and hobbies. Output only the JSON.\n"}
+{"seq_id": 157, "group": "json", "kind": "text", "lang": "en", "text": "Return a JSON object for an order of one camera strap shipped to Leeds. Use the keys order_id, item, quantity, price and shipping. Output only the JSON.\n{\n  \"order_id\": \"A-2688\",\n  \"item\": \"camera strap\",\n  \"quantity\": 2,\n  \"price\": 36.65,\n  \"shipping\": {\n    \"city\": \"Leeds\",\n    \"express\": true\n  }\n}", "messages": null, "license": "CC0-1.0", "source": "generated:build.py", "greedy": false, "greedy_prompt": "Return a JSON object for an order of one camera strap shipped to Leeds. Use the keys order_id, item, quantity, price and shipping. Output only the JSON.\n"}
+{"seq_id": 158, "group": "json", "kind": "text", "lang": "en", "text": "Return a JSON object with a list of three tasks for Omar. Each task has the keys title, done and priority. Output only the JSON.\n{\n  \"owner\": \"Omar\",\n  \"tasks\": [\n    {\n      \"title\": \"renew passport\",\n      \"done\": true,\n      \"priority\": 1\n    },\n    {\n      \"title\": \"write report\",\n      \"done\": false,\n      \"priority\": 3\n    },\n    {\n      \"title\": \"fix bike\",\n      \"done\": true,\n      \"priority\": 2\n    }\n  ]\n}", "messages": null, "license": "CC0-1.0", "source": "generated:build.py", "greedy": false, "greedy_prompt": "Return a JSON object with a list of three tasks for Omar. Each task has the keys title, done and priority. Output only the JSON.\n"}
+{"seq_id": 159, "group": "json", "kind": "text", "lang": "en", "text": "Return a JSON object for the weather in Turin with the keys city, temperature_c, conditions and wind. wind has speed_kmh and direction. Output only the JSON.\n{\n  \"city\": \"Turin\",\n  \"temperature_c\": 15,\n  \"conditions\": \"fog\",\n  \"wind\": {\n    \"speed_kmh\": 37,\n    \"direction\": \"NW\"\n  }\n}", "messages": null, "license": "CC0-1.0", "source": "generated:build.py", "greedy": false, "greedy_prompt": "Return a JSON object for the weather in Turin with the keys city, temperature_c, conditions and wind. wind has speed_kmh and direction. Output only the JSON.\n"}