@huggingface/tasks 0.19.35 → 0.19.36

This diff represents the content of publicly available package versions that have been released to one of the supported registries. The information contained in this diff is provided for informational purposes only and reflects changes between package versions as they appear in their respective public registries.
Files changed (62) hide show
  1. package/dist/commonjs/local-apps.d.ts +7 -0
  2. package/dist/commonjs/local-apps.d.ts.map +1 -1
  3. package/dist/commonjs/local-apps.js +31 -0
  4. package/dist/commonjs/tasks/any-to-any/data.js +3 -3
  5. package/dist/commonjs/tasks/image-segmentation/data.js +2 -2
  6. package/dist/commonjs/tasks/image-text-to-text/data.d.ts.map +1 -1
  7. package/dist/commonjs/tasks/image-text-to-text/data.js +10 -38
  8. package/dist/commonjs/tasks/image-to-3d/data.js +2 -2
  9. package/dist/commonjs/tasks/image-to-image/data.d.ts.map +1 -1
  10. package/dist/commonjs/tasks/image-to-image/data.js +13 -25
  11. package/dist/commonjs/tasks/image-to-text/data.d.ts.map +1 -1
  12. package/dist/commonjs/tasks/image-to-text/data.js +10 -30
  13. package/dist/commonjs/tasks/keypoint-detection/data.js +5 -5
  14. package/dist/commonjs/tasks/object-detection/data.js +2 -2
  15. package/dist/commonjs/tasks/text-generation/data.d.ts.map +1 -1
  16. package/dist/commonjs/tasks/text-generation/data.js +14 -10
  17. package/dist/commonjs/tasks/text-to-image/data.js +6 -6
  18. package/dist/commonjs/tasks/text-to-speech/data.d.ts.map +1 -1
  19. package/dist/commonjs/tasks/text-to-speech/data.js +10 -14
  20. package/dist/commonjs/tasks/text-to-video/data.js +3 -3
  21. package/dist/commonjs/tasks/visual-document-retrieval/data.d.ts.map +1 -1
  22. package/dist/commonjs/tasks/visual-document-retrieval/data.js +6 -2
  23. package/dist/commonjs/tasks/zero-shot-object-detection/data.js +4 -4
  24. package/dist/esm/local-apps.d.ts +7 -0
  25. package/dist/esm/local-apps.d.ts.map +1 -1
  26. package/dist/esm/local-apps.js +31 -0
  27. package/dist/esm/tasks/any-to-any/data.js +3 -3
  28. package/dist/esm/tasks/image-segmentation/data.js +2 -2
  29. package/dist/esm/tasks/image-text-to-text/data.d.ts.map +1 -1
  30. package/dist/esm/tasks/image-text-to-text/data.js +10 -38
  31. package/dist/esm/tasks/image-to-3d/data.js +2 -2
  32. package/dist/esm/tasks/image-to-image/data.d.ts.map +1 -1
  33. package/dist/esm/tasks/image-to-image/data.js +13 -25
  34. package/dist/esm/tasks/image-to-text/data.d.ts.map +1 -1
  35. package/dist/esm/tasks/image-to-text/data.js +10 -30
  36. package/dist/esm/tasks/keypoint-detection/data.js +5 -5
  37. package/dist/esm/tasks/object-detection/data.js +2 -2
  38. package/dist/esm/tasks/text-generation/data.d.ts.map +1 -1
  39. package/dist/esm/tasks/text-generation/data.js +14 -10
  40. package/dist/esm/tasks/text-to-image/data.js +6 -6
  41. package/dist/esm/tasks/text-to-speech/data.d.ts.map +1 -1
  42. package/dist/esm/tasks/text-to-speech/data.js +10 -14
  43. package/dist/esm/tasks/text-to-video/data.js +3 -3
  44. package/dist/esm/tasks/visual-document-retrieval/data.d.ts.map +1 -1
  45. package/dist/esm/tasks/visual-document-retrieval/data.js +6 -2
  46. package/dist/esm/tasks/zero-shot-object-detection/data.js +4 -4
  47. package/package.json +1 -1
  48. package/src/local-apps.ts +32 -0
  49. package/src/tasks/any-to-any/data.ts +3 -3
  50. package/src/tasks/image-segmentation/data.ts +2 -2
  51. package/src/tasks/image-text-to-text/data.ts +10 -38
  52. package/src/tasks/image-to-3d/data.ts +2 -2
  53. package/src/tasks/image-to-image/data.ts +13 -25
  54. package/src/tasks/image-to-text/data.ts +10 -30
  55. package/src/tasks/keypoint-detection/data.ts +5 -5
  56. package/src/tasks/object-detection/data.ts +2 -2
  57. package/src/tasks/text-generation/data.ts +14 -10
  58. package/src/tasks/text-to-image/data.ts +6 -6
  59. package/src/tasks/text-to-speech/data.ts +10 -14
  60. package/src/tasks/text-to-video/data.ts +3 -3
  61. package/src/tasks/visual-document-retrieval/data.ts +7 -2
  62. package/src/tasks/zero-shot-object-detection/data.ts +4 -4
@@ -203,6 +203,13 @@ export declare const LOCAL_APPS: {
203
203
  displayOnModelPage: typeof isLlamaCppGgufModel;
204
204
  snippet: (model: ModelData, filepath?: string) => string;
205
205
  };
206
+ lemonade: {
207
+ prettyLabel: string;
208
+ docsUrl: string;
209
+ mainTask: "text-generation";
210
+ displayOnModelPage: typeof isLlamaCppGgufModel;
211
+ snippet: (model: ModelData, filepath?: string) => LocalAppSnippet[];
212
+ };
206
213
  };
207
214
  export type LocalAppKey = keyof typeof LOCAL_APPS;
208
215
  export {};
@@ -1 +1 @@
1
- {"version":3,"file":"local-apps.d.ts","sourceRoot":"","sources":["../../src/local-apps.ts"],"names":[],"mappings":"AACA,OAAO,KAAK,EAAE,SAAS,EAAE,MAAM,iBAAiB,CAAC;AACjD,OAAO,KAAK,EAAE,YAAY,EAAE,MAAM,gBAAgB,CAAC;AAKnD,MAAM,WAAW,eAAe;IAC/B;;OAEG;IACH,KAAK,EAAE,MAAM,CAAC;IACd;;OAEG;IACH,KAAK,CAAC,EAAE,MAAM,CAAC;IACf;;OAEG;IACH,OAAO,EAAE,MAAM,GAAG,MAAM,EAAE,CAAC;CAC3B;AAED;;GAEG;AACH,MAAM,MAAM,QAAQ,GAAG;IACtB;;OAEG;IACH,WAAW,EAAE,MAAM,CAAC;IACpB;;OAEG;IACH,OAAO,EAAE,MAAM,CAAC;IAChB;;OAEG;IACH,QAAQ,EAAE,YAAY,CAAC;IACvB;;OAEG;IACH,SAAS,CAAC,EAAE,OAAO,CAAC;IAEpB,UAAU,CAAC,EAAE,OAAO,CAAC;IACrB;;OAEG;IACH,kBAAkB,EAAE,CAAC,KAAK,EAAE,SAAS,KAAK,OAAO,CAAC;CAClD,GAAG,CACD;IACA;;OAEG;IACH,QAAQ,EAAE,CAAC,KAAK,EAAE,SAAS,EAAE,QAAQ,CAAC,EAAE,MAAM,KAAK,GAAG,CAAC;CACtD,GACD;IACA;;;;OAIG;IACH,OAAO,EAAE,CAAC,KAAK,EAAE,SAAS,EAAE,QAAQ,CAAC,EAAE,MAAM,KAAK,MAAM,GAAG,MAAM,EAAE,GAAG,eAAe,GAAG,eAAe,EAAE,CAAC;CACzG,CACH,CAAC;AAqBF,iBAAS,UAAU,CAAC,KAAK,EAAE,SAAS,GAAG,OAAO,CAE7C;AAED,iBAAS,mBAAmB,CAAC,KAAK,EAAE,SAAS,WAE5C;AAmOD;;;;;;;;;;GAUG;AACH,eAAO,MAAM,UAAU;;;;;;yBA7NS,SAAS,aAAa,MAAM,KAAG,eAAe,EAAE;;;;;;;yBAsCzC,SAAS,aAAa,MAAM,KAAG,eAAe,EAAE;;;;;;oCA0MzD,SAAS;yBAjKX,SAAS,KAAG,eAAe,EAAE;;;;;;;yBA4F5B,SAAS,KAAG,eAAe,EAAE;;;;;;;yBA7B/B,SAAS,KAAG,eAAe,EAAE;;;;;;;;;;;;;;yBAtFzB,SAAS,aAAa,MAAM,KAAG,eAAe,EAAE;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;yBAJjD,SAAS,aAAa,MAAM,KAAG,MAAM;;;;;;;yBA0J1B,SAAS,aAAa,MAAM,KAAG,MAAM;;CAqL1C,CAAC;AAErC,MAAM,MAAM,WAAW,GAAG,MAAM,OAAO,UAAU,CAAC"}
1
+ {"version":3,"file":"local-apps.d.ts","sourceRoot":"","sources":["../../src/local-apps.ts"],"names":[],"mappings":"AACA,OAAO,KAAK,EAAE,SAAS,EAAE,MAAM,iBAAiB,CAAC;AACjD,OAAO,KAAK,EAAE,YAAY,EAAE,MAAM,gBAAgB,CAAC;AAKnD,MAAM,WAAW,eAAe;IAC/B;;OAEG;IACH,KAAK,EAAE,MAAM,CAAC;IACd;;OAEG;IACH,KAAK,CAAC,EAAE,MAAM,CAAC;IACf;;OAEG;IACH,OAAO,EAAE,MAAM,GAAG,MAAM,EAAE,CAAC;CAC3B;AAED;;GAEG;AACH,MAAM,MAAM,QAAQ,GAAG;IACtB;;OAEG;IACH,WAAW,EAAE,MAAM,CAAC;IACpB;;OAEG;IACH,OAAO,EAAE,MAAM,CAAC;IAChB;;OAEG;IACH,QAAQ,EAAE,YAAY,CAAC;IACvB;;OAEG;IACH,SAAS,CAAC,EAAE,OAAO,CAAC;IAEpB,UAAU,CAAC,EAAE,OAAO,CAAC;IACrB;;OAEG;IACH,kBAAkB,EAAE,CAAC,KAAK,EAAE,SAAS,KAAK,OAAO,CAAC;CAClD,GAAG,CACD;IACA;;OAEG;IACH,QAAQ,EAAE,CAAC,KAAK,EAAE,SAAS,EAAE,QAAQ,CAAC,EAAE,MAAM,KAAK,GAAG,CAAC;CACtD,GACD;IACA;;;;OAIG;IACH,OAAO,EAAE,CAAC,KAAK,EAAE,SAAS,EAAE,QAAQ,CAAC,EAAE,MAAM,KAAK,MAAM,GAAG,MAAM,EAAE,GAAG,eAAe,GAAG,eAAe,EAAE,CAAC;CACzG,CACH,CAAC;AAqBF,iBAAS,UAAU,CAAC,KAAK,EAAE,SAAS,GAAG,OAAO,CAE7C;AAED,iBAAS,mBAAmB,CAAC,KAAK,EAAE,SAAS,WAE5C;AA4PD;;;;;;;;;;GAUG;AACH,eAAO,MAAM,UAAU;;;;;;yBAtPS,SAAS,aAAa,MAAM,KAAG,eAAe,EAAE;;;;;;;yBAsCzC,SAAS,aAAa,MAAM,KAAG,eAAe,EAAE;;;;;;oCAmOzD,SAAS;yBA1LX,SAAS,KAAG,eAAe,EAAE;;;;;;;yBA4F5B,SAAS,KAAG,eAAe,EAAE;;;;;;;yBA7B/B,SAAS,KAAG,eAAe,EAAE;;;;;;;;;;;;;;yBAtFzB,SAAS,aAAa,MAAM,KAAG,eAAe,EAAE;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;yBAJjD,SAAS,aAAa,MAAM,KAAG,MAAM;;;;;;;yBA0J1B,SAAS,aAAa,MAAM,KAAG,MAAM;;;;;;;yBAI9C,SAAS,aAAa,MAAM,KAAG,eAAe,EAAE;;CAiN5C,CAAC;AAErC,MAAM,MAAM,WAAW,GAAG,MAAM,OAAO,UAAU,CAAC"}
@@ -233,6 +233,30 @@ const snippetMlxLm = (model) => {
233
233
  const snippetDockerModelRunner = (model, filepath) => {
234
234
  return `docker model run hf.co/${model.id}${getQuantTag(filepath)}`;
235
235
  };
236
+ const snippetLemonade = (model, filepath) => {
237
+ const tagName = getQuantTag(filepath);
238
+ const modelName = model.id.split("/")[1];
239
+ return [
240
+ {
241
+ title: "Pull the model",
242
+ setup: "# Download Lemonade from https://lemonade-server.ai/",
243
+ content: [
244
+ `lemonade-server pull user.${modelName} \\
245
+ --checkpoint ${model.id}${tagName} \\
246
+ --recipe llamacpp`,
247
+ "# Note: If you installed from source, use the lemonade-server-dev command instead.",
248
+ ].join("\n"),
249
+ },
250
+ {
251
+ title: "Run and chat with the model",
252
+ content: `lemonade-server run user.${modelName}`,
253
+ },
254
+ {
255
+ title: "List all available models",
256
+ content: "lemonade-server list",
257
+ },
258
+ ];
259
+ };
236
260
  /**
237
261
  * Add your new local app here.
238
262
  *
@@ -408,4 +432,11 @@ exports.LOCAL_APPS = {
408
432
  displayOnModelPage: isLlamaCppGgufModel,
409
433
  snippet: snippetDockerModelRunner,
410
434
  },
435
+ lemonade: {
436
+ prettyLabel: "Lemonade",
437
+ docsUrl: "https://lemonade-server.ai",
438
+ mainTask: "text-generation",
439
+ displayOnModelPage: isLlamaCppGgufModel,
440
+ snippet: snippetLemonade,
441
+ },
411
442
  };
@@ -39,7 +39,7 @@ const taskData = {
39
39
  },
40
40
  {
41
41
  description: "Robust model that can take in image and text and generate image and text.",
42
- id: "deepseek-ai/Janus-Pro-7B",
42
+ id: "OmniGen2/OmniGen2",
43
43
  },
44
44
  {
45
45
  description: "Any-to-any model with speech, video, audio, image and text understanding capabilities.",
@@ -47,13 +47,13 @@ const taskData = {
47
47
  },
48
48
  {
49
49
  description: "A model that can understand image and text and generate image and text.",
50
- id: "EPFL-VILAB/4M-21_XL",
50
+ id: "ByteDance-Seed/BAGEL-7B-MoT",
51
51
  },
52
52
  ],
53
53
  spaces: [
54
54
  {
55
55
  description: "An application to chat with an any-to-any (image & text) model.",
56
- id: "deepseek-ai/Janus-Pro-7B",
56
+ id: "OmniGen2/OmniGen2",
57
57
  },
58
58
  ],
59
59
  summary: "Any-to-any models can understand two or more modalities and output two or more modalities.",
@@ -42,8 +42,8 @@ const taskData = {
42
42
  models: [
43
43
  {
44
44
  // TO DO: write description
45
- description: "Solid semantic segmentation model trained on ADE20k.",
46
- id: "openmmlab/upernet-convnext-small",
45
+ description: "Solid panoptic segmentation model trained on COCO.",
46
+ id: "tue-mps/coco_panoptic_eomt_large_640",
47
47
  },
48
48
  {
49
49
  description: "Background removal model.",
@@ -1 +1 @@
1
- {"version":3,"file":"data.d.ts","sourceRoot":"","sources":["../../../../src/tasks/image-text-to-text/data.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,cAAc,EAAE,MAAM,aAAa,CAAC;AAElD,QAAA,MAAM,QAAQ,EAAE,cA6Gf,CAAC;AAEF,eAAe,QAAQ,CAAC"}
1
+ {"version":3,"file":"data.d.ts","sourceRoot":"","sources":["../../../../src/tasks/image-text-to-text/data.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,cAAc,EAAE,MAAM,aAAa,CAAC;AAElD,QAAA,MAAM,QAAQ,EAAE,cAiFf,CAAC;AAEF,eAAe,QAAQ,CAAC"}
@@ -46,33 +46,21 @@ const taskData = {
46
46
  id: "HuggingFaceTB/SmolVLM-Instruct",
47
47
  },
48
48
  {
49
- description: "A screenshot understanding model used to control computers.",
50
- id: "microsoft/OmniParser-v2.0",
49
+ description: "Cutting-edge reasoning vision language model.",
50
+ id: "zai-org/GLM-4.5V",
51
51
  },
52
52
  {
53
- description: "Cutting-edge vision language model.",
54
- id: "allenai/Molmo-7B-D-0924",
53
+ description: "Cutting-edge small vision language model to convert documents to text.",
54
+ id: "rednote-hilab/dots.ocr",
55
55
  },
56
56
  {
57
57
  description: "Small yet powerful model.",
58
- id: "vikhyatk/moondream2",
59
- },
60
- {
61
- description: "Strong image-text-to-text model.",
62
- id: "Qwen/Qwen2.5-VL-7B-Instruct",
58
+ id: "Qwen/Qwen2.5-VL-3B-Instruct",
63
59
  },
64
60
  {
65
61
  description: "Image-text-to-text model with agentic capabilities.",
66
62
  id: "microsoft/Magma-8B",
67
63
  },
68
- {
69
- description: "Strong image-text-to-text model focused on documents.",
70
- id: "allenai/olmOCR-7B-0225-preview",
71
- },
72
- {
73
- description: "Small yet strong image-text-to-text model.",
74
- id: "ibm-granite/granite-vision-3.2-2b",
75
- },
76
64
  ],
77
65
  spaces: [
78
66
  {
@@ -80,32 +68,16 @@ const taskData = {
80
68
  id: "opencompass/open_vlm_leaderboard",
81
69
  },
82
70
  {
83
- description: "Vision language models arena, where models are ranked by votes of users.",
84
- id: "WildVision/vision-arena",
85
- },
86
- {
87
- description: "Powerful vision-language model assistant.",
88
- id: "akhaliq/Molmo-7B-D-0924",
89
- },
90
- {
91
- description: "Powerful vision language assistant that can understand multiple images.",
92
- id: "HuggingFaceTB/SmolVLM2",
93
- },
94
- {
95
- description: "An application for chatting with an image-text-to-text model.",
96
- id: "GanymedeNil/Qwen2-VL-7B",
97
- },
98
- {
99
- description: "An application that parses screenshots into actions.",
100
- id: "showlab/ShowUI",
71
+ description: "An application that compares object detection capabilities of different vision language models.",
72
+ id: "sergiopaniego/vlm_object_understanding",
101
73
  },
102
74
  {
103
- description: "An application that detects gaze.",
104
- id: "moondream/gaze-demo",
75
+ description: "An application to compare different OCR models.",
76
+ id: "prithivMLmods/Multimodal-OCR",
105
77
  },
106
78
  ],
107
79
  summary: "Image-text-to-text models take in an image and text prompt and output text. These models are also called vision-language models, or VLMs. The difference from image-to-text models is that these models take an additional text input, not restricting the model to certain use cases like image captioning, and may also be trained to accept a conversation as input.",
108
- widgetModels: ["Qwen/Qwen2-VL-7B-Instruct"],
80
+ widgetModels: ["zai-org/GLM-4.5V"],
109
81
  youtubeId: "IoGaGfU1CIg",
110
82
  };
111
83
  exports.default = taskData;
@@ -33,8 +33,8 @@ const taskData = {
33
33
  id: "TencentARC/InstantMesh",
34
34
  },
35
35
  {
36
- description: "Fast image-to-3D mesh model by StabilityAI",
37
- id: "stabilityai/TripoSR",
36
+ description: "3D world generation model.",
37
+ id: "tencent/HunyuanWorld-1",
38
38
  },
39
39
  {
40
40
  description: "A scaled up image-to-3D mesh model derived from TripoSR.",
@@ -1 +1 @@
1
- {"version":3,"file":"data.d.ts","sourceRoot":"","sources":["../../../../src/tasks/image-to-image/data.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,cAAc,EAAE,MAAM,aAAa,CAAC;AAElD,QAAA,MAAM,QAAQ,EAAE,cAuGf,CAAC;AAEF,eAAe,QAAQ,CAAC"}
1
+ {"version":3,"file":"data.d.ts","sourceRoot":"","sources":["../../../../src/tasks/image-to-image/data.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,cAAc,EAAE,MAAM,aAAa,CAAC;AAElD,QAAA,MAAM,QAAQ,EAAE,cA2Ff,CAAC;AAEF,eAAe,QAAQ,CAAC"}
@@ -50,16 +50,16 @@ const taskData = {
50
50
  id: "fal/AuraSR-v2",
51
51
  },
52
52
  {
53
- description: "A model that increases the resolution of an image.",
54
- id: "keras-io/super-resolution",
53
+ description: "Powerful image editing model.",
54
+ id: "black-forest-labs/FLUX.1-Kontext-dev",
55
55
  },
56
56
  {
57
- description: "A model for applying edits to images through image controls.",
58
- id: "Yuanshi/OminiControl",
57
+ description: "Virtual try-on model.",
58
+ id: "yisol/IDM-VTON",
59
59
  },
60
60
  {
61
- description: "A model that generates images based on segments in the input image and the text prompt.",
62
- id: "mfidabel/controlnet-segment-anything",
61
+ description: "Image re-lighting model.",
62
+ id: "kontext-community/relighting-kontext-dev-lora-v3",
63
63
  },
64
64
  {
65
65
  description: "Strong model for inpainting and outpainting.",
@@ -72,32 +72,20 @@ const taskData = {
72
72
  ],
73
73
  spaces: [
74
74
  {
75
- description: "Image enhancer application for low light.",
76
- id: "keras-io/low-light-image-enhancement",
75
+ description: "Image editing application.",
76
+ id: "black-forest-labs/FLUX.1-Kontext-Dev",
77
77
  },
78
78
  {
79
- description: "Style transfer application.",
80
- id: "keras-io/neural-style-transfer",
79
+ description: "Image relighting application.",
80
+ id: "lllyasviel/iclight-v2-vary",
81
81
  },
82
82
  {
83
- description: "An application that generates images based on segment control.",
84
- id: "mfidabel/controlnet-segment-anything",
85
- },
86
- {
87
- description: "Image generation application that takes image control and text prompt.",
88
- id: "hysts/ControlNet",
89
- },
90
- {
91
- description: "Colorize any image using this app.",
92
- id: "ioclab/brightness-controlnet",
93
- },
94
- {
95
- description: "Edit images with instructions.",
96
- id: "timbrooks/instruct-pix2pix",
83
+ description: "An application for image upscaling.",
84
+ id: "jasperai/Flux.1-dev-Controlnet-Upscaler",
97
85
  },
98
86
  ],
99
87
  summary: "Image-to-image is the task of transforming an input image through a variety of possible manipulations and enhancements, such as super-resolution, image inpainting, colorization, and more.",
100
- widgetModels: ["stabilityai/stable-diffusion-2-inpainting"],
88
+ widgetModels: ["Qwen/Qwen-Image"],
101
89
  youtubeId: "",
102
90
  };
103
91
  exports.default = taskData;
@@ -1 +1 @@
1
- {"version":3,"file":"data.d.ts","sourceRoot":"","sources":["../../../../src/tasks/image-to-text/data.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,cAAc,EAAE,MAAM,aAAa,CAAC;AAElD,QAAA,MAAM,QAAQ,EAAE,cA6Ef,CAAC;AAEF,eAAe,QAAQ,CAAC"}
1
+ {"version":3,"file":"data.d.ts","sourceRoot":"","sources":["../../../../src/tasks/image-to-text/data.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,cAAc,EAAE,MAAM,aAAa,CAAC;AAElD,QAAA,MAAM,QAAQ,EAAE,cAyDf,CAAC;AAEF,eAAe,QAAQ,CAAC"}
@@ -31,46 +31,26 @@ const taskData = {
31
31
  metrics: [],
32
32
  models: [
33
33
  {
34
- description: "A robust image captioning model.",
35
- id: "Salesforce/blip2-opt-2.7b",
34
+ description: "Strong OCR model.",
35
+ id: "allenai/olmOCR-7B-0725",
36
36
  },
37
37
  {
38
- description: "A powerful and accurate image-to-text model that can also localize concepts in images.",
39
- id: "microsoft/kosmos-2-patch14-224",
40
- },
41
- {
42
- description: "A strong optical character recognition model.",
43
- id: "facebook/nougat-base",
44
- },
45
- {
46
- description: "A powerful model that lets you have a conversation with the image.",
47
- id: "llava-hf/llava-1.5-7b-hf",
38
+ description: "Powerful image captioning model.",
39
+ id: "fancyfeast/llama-joycaption-beta-one-hf-llava",
48
40
  },
49
41
  ],
50
42
  spaces: [
51
43
  {
52
- description: "An application that compares various image captioning models.",
53
- id: "nielsr/comparing-captioning-models",
54
- },
55
- {
56
- description: "A robust image captioning application.",
57
- id: "flax-community/image-captioning",
58
- },
59
- {
60
- description: "An application that transcribes handwritings into text.",
61
- id: "nielsr/TrOCR-handwritten",
62
- },
63
- {
64
- description: "An application that can caption images and answer questions about a given image.",
65
- id: "Salesforce/BLIP",
44
+ description: "SVG generator app from images.",
45
+ id: "multimodalart/OmniSVG-3B",
66
46
  },
67
47
  {
68
- description: "An application that can caption images and answer questions with a conversational agent.",
69
- id: "Salesforce/BLIP2",
48
+ description: "An application that converts documents to markdown.",
49
+ id: "numind/NuMarkdown-8B-Thinking",
70
50
  },
71
51
  {
72
- description: "An image captioning application that demonstrates the effect of noise on captions.",
73
- id: "johko/capdec-image-captioning",
52
+ description: "An application that can caption images.",
53
+ id: "fancyfeast/joy-caption-beta-one",
74
54
  },
75
55
  ],
76
56
  summary: "Image to text models output a text from a given image. Image captioning or optical character recognition can be considered as the most common applications of image to text.",
@@ -33,11 +33,11 @@ const taskData = {
33
33
  },
34
34
  {
35
35
  description: "Strong keypoint detection model used to detect human pose.",
36
- id: "facebook/sapiens-pose-1b",
36
+ id: "qualcomm/RTMPose-Body2d",
37
37
  },
38
38
  {
39
- description: "Powerful keypoint detection model used to detect human pose.",
40
- id: "usyd-community/vitpose-plus-base",
39
+ description: "Powerful keypoint matching model.",
40
+ id: "ETH-CVG/lightglue_disk",
41
41
  },
42
42
  ],
43
43
  spaces: [
@@ -46,8 +46,8 @@ const taskData = {
46
46
  id: "datasciencedojo/Hand-Keypoint-Detection-Realtime",
47
47
  },
48
48
  {
49
- description: "An application to try a universal keypoint detection model.",
50
- id: "merve/SuperPoint",
49
+ description: "An application for keypoint detection and matching.",
50
+ id: "ETH-CVG/LightGlue",
51
51
  },
52
52
  ],
53
53
  summary: "Keypoint detection is the task of identifying meaningful distinctive points or features in an image.",
@@ -59,8 +59,8 @@ const taskData = {
59
59
  ],
60
60
  spaces: [
61
61
  {
62
- description: "Leaderboard to compare various object detection models across several metrics.",
63
- id: "hf-vision/object_detection_leaderboard",
62
+ description: "Real-time object detection demo.",
63
+ id: "Roboflow/RF-DETR",
64
64
  },
65
65
  {
66
66
  description: "An application that contains various object detection models to try from.",
@@ -1 +1 @@
1
- {"version":3,"file":"data.d.ts","sourceRoot":"","sources":["../../../../src/tasks/text-generation/data.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,cAAc,EAAE,MAAM,aAAa,CAAC;AAElD,QAAA,MAAM,QAAQ,EAAE,cAyHf,CAAC;AAEF,eAAe,QAAQ,CAAC"}
1
+ {"version":3,"file":"data.d.ts","sourceRoot":"","sources":["../../../../src/tasks/text-generation/data.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,cAAc,EAAE,MAAM,aAAa,CAAC;AAElD,QAAA,MAAM,QAAQ,EAAE,cA6Hf,CAAC;AAEF,eAAe,QAAQ,CAAC"}
@@ -60,20 +60,20 @@ const taskData = {
60
60
  models: [
61
61
  { description: "A text-generation model trained to follow instructions.", id: "google/gemma-2-2b-it" },
62
62
  {
63
- description: "Smaller variant of one of the most powerful models.",
64
- id: "deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B",
63
+ description: "Powerful text generation model for coding.",
64
+ id: "Qwen/Qwen3-Coder-480B-A35B-Instruct",
65
65
  },
66
66
  {
67
- description: "Very powerful text generation model trained to follow instructions.",
68
- id: "meta-llama/Meta-Llama-3.1-8B-Instruct",
67
+ description: "Great text generation model with top-notch tool calling capabilities.",
68
+ id: "openai/gpt-oss-120b",
69
69
  },
70
70
  {
71
- description: "Powerful text generation model by Microsoft.",
72
- id: "microsoft/phi-4",
71
+ description: "Powerful text generation model.",
72
+ id: "zai-org/GLM-4.5",
73
73
  },
74
74
  {
75
- description: "A very powerful model with reasoning capabilities.",
76
- id: "simplescaling/s1.1-32B",
75
+ description: "A powerful small model with reasoning capabilities.",
76
+ id: "Qwen/Qwen3-4B-Thinking-2507",
77
77
  },
78
78
  {
79
79
  description: "Strong conversational model that supports very long instructions.",
@@ -90,8 +90,12 @@ const taskData = {
90
90
  ],
91
91
  spaces: [
92
92
  {
93
- description: "A leaderboard to compare different open-source text generation models based on various benchmarks.",
94
- id: "open-llm-leaderboard/open_llm_leaderboard",
93
+ description: "An application that writes and executes code from text instructions and supports many models.",
94
+ id: "akhaliq/anycoder",
95
+ },
96
+ {
97
+ description: "An application that builds websites from natural language prompts.",
98
+ id: "enzostvs/deepsite",
95
99
  },
96
100
  {
97
101
  description: "A leaderboard for comparing chain-of-thought performance of models.",
@@ -47,19 +47,19 @@ const taskData = {
47
47
  models: [
48
48
  {
49
49
  description: "One of the most powerful image generation models that can generate realistic outputs.",
50
- id: "black-forest-labs/FLUX.1-dev",
50
+ id: "black-forest-labs/FLUX.1-Krea-dev",
51
51
  },
52
52
  {
53
- description: "A powerful yet fast image generation model.",
54
- id: "latent-consistency/lcm-lora-sdxl",
53
+ description: "A powerful image generation model.",
54
+ id: "Qwen/Qwen-Image",
55
55
  },
56
56
  {
57
- description: "Text-to-image model for photorealistic generation.",
58
- id: "Kwai-Kolors/Kolors",
57
+ description: "Powerful and fast image generation model.",
58
+ id: "ByteDance/SDXL-Lightning",
59
59
  },
60
60
  {
61
61
  description: "A powerful text-to-image model.",
62
- id: "stabilityai/stable-diffusion-3-medium-diffusers",
62
+ id: "ByteDance/Hyper-SD",
63
63
  },
64
64
  ],
65
65
  spaces: [
@@ -1 +1 @@
1
- {"version":3,"file":"data.d.ts","sourceRoot":"","sources":["../../../../src/tasks/text-to-speech/data.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,cAAc,EAAE,MAAM,aAAa,CAAC;AAElD,QAAA,MAAM,QAAQ,EAAE,cAqFf,CAAC;AAEF,eAAe,QAAQ,CAAC"}
1
+ {"version":3,"file":"data.d.ts","sourceRoot":"","sources":["../../../../src/tasks/text-to-speech/data.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,cAAc,EAAE,MAAM,aAAa,CAAC;AAElD,QAAA,MAAM,QAAQ,EAAE,cAiFf,CAAC;AAEF,eAAe,QAAQ,CAAC"}
@@ -39,24 +39,20 @@ const taskData = {
39
39
  ],
40
40
  models: [
41
41
  {
42
- description: "A prompt based, powerful TTS model.",
43
- id: "parler-tts/parler-tts-large-v1",
42
+ description: "Small yet powerful TTS model.",
43
+ id: "KittenML/kitten-tts-nano-0.1",
44
44
  },
45
45
  {
46
- description: "A powerful TTS model that supports English and Chinese.",
47
- id: "SWivid/F5-TTS",
46
+ description: "Bleeding edge TTS model.",
47
+ id: "ResembleAI/chatterbox",
48
48
  },
49
49
  {
50
50
  description: "A massively multi-lingual TTS model.",
51
51
  id: "fishaudio/fish-speech-1.5",
52
52
  },
53
53
  {
54
- description: "A powerful TTS model.",
55
- id: "OuteAI/OuteTTS-0.1-350M",
56
- },
57
- {
58
- description: "Small yet powerful TTS model.",
59
- id: "hexgrad/Kokoro-82M",
54
+ description: "A text-to-dialogue model.",
55
+ id: "nari-labs/Dia-1.6B-0626",
60
56
  },
61
57
  ],
62
58
  spaces: [
@@ -69,12 +65,12 @@ const taskData = {
69
65
  id: "fishaudio/fish-speech-1",
70
66
  },
71
67
  {
72
- description: "An application that generates speech in different styles in English and Chinese.",
73
- id: "mrfakename/E2-F5-TTS",
68
+ description: "Performant TTS application.",
69
+ id: "ResembleAI/Chatterbox",
74
70
  },
75
71
  {
76
- description: "An application that synthesizes emotional speech for diverse speaker prompts.",
77
- id: "parler-tts/parler-tts-expresso",
72
+ description: "An application to compare different TTS models.",
73
+ id: "TTS-AGI/TTS-Arena-V2",
78
74
  },
79
75
  {
80
76
  description: "An application that generates podcast episodes.",
@@ -74,8 +74,8 @@ const taskData = {
74
74
  id: "nvidia/Cosmos-1.0-Diffusion-7B-Text2World",
75
75
  },
76
76
  {
77
- description: "A robust model for video generation.",
78
- id: "Wan-AI/Wan2.1-T2V-1.3B",
77
+ description: "Very fast model for video generation.",
78
+ id: "Lightricks/LTX-Video-0.9.8-13B-distilled",
79
79
  },
80
80
  ],
81
81
  spaces: [
@@ -93,7 +93,7 @@ const taskData = {
93
93
  },
94
94
  ],
95
95
  summary: "Text-to-video models can be used in any application that requires generating consistent sequence of images from text. ",
96
- widgetModels: ["Wan-AI/Wan2.1-T2V-14B"],
96
+ widgetModels: ["Wan-AI/Wan2.2-TI2V-5B"],
97
97
  youtubeId: undefined,
98
98
  };
99
99
  exports.default = taskData;
@@ -1 +1 @@
1
- {"version":3,"file":"data.d.ts","sourceRoot":"","sources":["../../../../src/tasks/visual-document-retrieval/data.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,cAAc,EAAE,MAAM,aAAa,CAAC;AAElD,QAAA,MAAM,QAAQ,EAAE,cAkEf,CAAC;AAEF,eAAe,QAAQ,CAAC"}
1
+ {"version":3,"file":"data.d.ts","sourceRoot":"","sources":["../../../../src/tasks/visual-document-retrieval/data.ts"],"names":[],"mappings":"AAAA,OAAO,KAAK,EAAE,cAAc,EAAE,MAAM,aAAa,CAAC;AAElD,QAAA,MAAM,QAAQ,EAAE,cAuEf,CAAC;AAEF,eAAe,QAAQ,CAAC"}
@@ -52,8 +52,8 @@ const taskData = {
52
52
  id: "vidore/colqwen2-v1.0",
53
53
  },
54
54
  {
55
- description: "Very fast and efficient visual document retrieval model that works on five languages.",
56
- id: "marco/mcdse-2b-v1",
55
+ description: "Very fast and efficient visual document retrieval model that can also take in other modalities like audio.",
56
+ id: "Tevatron/OmniEmbed-v0.1",
57
57
  },
58
58
  ],
59
59
  spaces: [
@@ -61,6 +61,10 @@ const taskData = {
61
61
  description: "A leaderboard of visual document retrieval models.",
62
62
  id: "vidore/vidore-leaderboard",
63
63
  },
64
+ {
65
+ description: "Visual retrieval augmented generation demo based on ColQwen2 model.",
66
+ id: "vidore/visual-rag-tool",
67
+ },
64
68
  ],
65
69
  summary: "Visual document retrieval is the task of searching for relevant image-based documents, such as PDFs. These models take a text query and multiple documents as input and return the top-most relevant documents and relevancy scores as output.",
66
70
  widgetModels: [""],
@@ -38,17 +38,17 @@ const taskData = {
38
38
  models: [
39
39
  {
40
40
  description: "Solid zero-shot object detection model.",
41
- id: "IDEA-Research/grounding-dino-base",
41
+ id: "openmmlab-community/mm_grounding_dino_large_all",
42
42
  },
43
43
  {
44
44
  description: "Cutting-edge zero-shot object detection model.",
45
- id: "google/owlv2-base-patch16-ensemble",
45
+ id: "fushh7/LLMDet",
46
46
  },
47
47
  ],
48
48
  spaces: [
49
49
  {
50
- description: "A demo to try the state-of-the-art zero-shot object detection model, OWLv2.",
51
- id: "merve/owlv2",
50
+ description: "A demo to compare different zero-shot object detection models per output and latency.",
51
+ id: "ariG23498/zero-shot-od",
52
52
  },
53
53
  {
54
54
  description: "A demo that combines a zero-shot object detection and mask generation model for zero-shot segmentation.",
@@ -203,6 +203,13 @@ export declare const LOCAL_APPS: {
203
203
  displayOnModelPage: typeof isLlamaCppGgufModel;
204
204
  snippet: (model: ModelData, filepath?: string) => string;
205
205
  };
206
+ lemonade: {
207
+ prettyLabel: string;
208
+ docsUrl: string;
209
+ mainTask: "text-generation";
210
+ displayOnModelPage: typeof isLlamaCppGgufModel;
211
+ snippet: (model: ModelData, filepath?: string) => LocalAppSnippet[];
212
+ };
206
213
  };
207
214
  export type LocalAppKey = keyof typeof LOCAL_APPS;
208
215
  export {};
@@ -1 +1 @@
1
- {"version":3,"file":"local-apps.d.ts","sourceRoot":"","sources":["../../src/local-apps.ts"],"names":[],"mappings":"AACA,OAAO,KAAK,EAAE,SAAS,EAAE,MAAM,iBAAiB,CAAC;AACjD,OAAO,KAAK,EAAE,YAAY,EAAE,MAAM,gBAAgB,CAAC;AAKnD,MAAM,WAAW,eAAe;IAC/B;;OAEG;IACH,KAAK,EAAE,MAAM,CAAC;IACd;;OAEG;IACH,KAAK,CAAC,EAAE,MAAM,CAAC;IACf;;OAEG;IACH,OAAO,EAAE,MAAM,GAAG,MAAM,EAAE,CAAC;CAC3B;AAED;;GAEG;AACH,MAAM,MAAM,QAAQ,GAAG;IACtB;;OAEG;IACH,WAAW,EAAE,MAAM,CAAC;IACpB;;OAEG;IACH,OAAO,EAAE,MAAM,CAAC;IAChB;;OAEG;IACH,QAAQ,EAAE,YAAY,CAAC;IACvB;;OAEG;IACH,SAAS,CAAC,EAAE,OAAO,CAAC;IAEpB,UAAU,CAAC,EAAE,OAAO,CAAC;IACrB;;OAEG;IACH,kBAAkB,EAAE,CAAC,KAAK,EAAE,SAAS,KAAK,OAAO,CAAC;CAClD,GAAG,CACD;IACA;;OAEG;IACH,QAAQ,EAAE,CAAC,KAAK,EAAE,SAAS,EAAE,QAAQ,CAAC,EAAE,MAAM,KAAK,GAAG,CAAC;CACtD,GACD;IACA;;;;OAIG;IACH,OAAO,EAAE,CAAC,KAAK,EAAE,SAAS,EAAE,QAAQ,CAAC,EAAE,MAAM,KAAK,MAAM,GAAG,MAAM,EAAE,GAAG,eAAe,GAAG,eAAe,EAAE,CAAC;CACzG,CACH,CAAC;AAqBF,iBAAS,UAAU,CAAC,KAAK,EAAE,SAAS,GAAG,OAAO,CAE7C;AAED,iBAAS,mBAAmB,CAAC,KAAK,EAAE,SAAS,WAE5C;AAmOD;;;;;;;;;;GAUG;AACH,eAAO,MAAM,UAAU;;;;;;yBA7NS,SAAS,aAAa,MAAM,KAAG,eAAe,EAAE;;;;;;;yBAsCzC,SAAS,aAAa,MAAM,KAAG,eAAe,EAAE;;;;;;oCA0MzD,SAAS;yBAjKX,SAAS,KAAG,eAAe,EAAE;;;;;;;yBA4F5B,SAAS,KAAG,eAAe,EAAE;;;;;;;yBA7B/B,SAAS,KAAG,eAAe,EAAE;;;;;;;;;;;;;;yBAtFzB,SAAS,aAAa,MAAM,KAAG,eAAe,EAAE;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;yBAJjD,SAAS,aAAa,MAAM,KAAG,MAAM;;;;;;;yBA0J1B,SAAS,aAAa,MAAM,KAAG,MAAM;;CAqL1C,CAAC;AAErC,MAAM,MAAM,WAAW,GAAG,MAAM,OAAO,UAAU,CAAC"}
1
+ {"version":3,"file":"local-apps.d.ts","sourceRoot":"","sources":["../../src/local-apps.ts"],"names":[],"mappings":"AACA,OAAO,KAAK,EAAE,SAAS,EAAE,MAAM,iBAAiB,CAAC;AACjD,OAAO,KAAK,EAAE,YAAY,EAAE,MAAM,gBAAgB,CAAC;AAKnD,MAAM,WAAW,eAAe;IAC/B;;OAEG;IACH,KAAK,EAAE,MAAM,CAAC;IACd;;OAEG;IACH,KAAK,CAAC,EAAE,MAAM,CAAC;IACf;;OAEG;IACH,OAAO,EAAE,MAAM,GAAG,MAAM,EAAE,CAAC;CAC3B;AAED;;GAEG;AACH,MAAM,MAAM,QAAQ,GAAG;IACtB;;OAEG;IACH,WAAW,EAAE,MAAM,CAAC;IACpB;;OAEG;IACH,OAAO,EAAE,MAAM,CAAC;IAChB;;OAEG;IACH,QAAQ,EAAE,YAAY,CAAC;IACvB;;OAEG;IACH,SAAS,CAAC,EAAE,OAAO,CAAC;IAEpB,UAAU,CAAC,EAAE,OAAO,CAAC;IACrB;;OAEG;IACH,kBAAkB,EAAE,CAAC,KAAK,EAAE,SAAS,KAAK,OAAO,CAAC;CAClD,GAAG,CACD;IACA;;OAEG;IACH,QAAQ,EAAE,CAAC,KAAK,EAAE,SAAS,EAAE,QAAQ,CAAC,EAAE,MAAM,KAAK,GAAG,CAAC;CACtD,GACD;IACA;;;;OAIG;IACH,OAAO,EAAE,CAAC,KAAK,EAAE,SAAS,EAAE,QAAQ,CAAC,EAAE,MAAM,KAAK,MAAM,GAAG,MAAM,EAAE,GAAG,eAAe,GAAG,eAAe,EAAE,CAAC;CACzG,CACH,CAAC;AAqBF,iBAAS,UAAU,CAAC,KAAK,EAAE,SAAS,GAAG,OAAO,CAE7C;AAED,iBAAS,mBAAmB,CAAC,KAAK,EAAE,SAAS,WAE5C;AA4PD;;;;;;;;;;GAUG;AACH,eAAO,MAAM,UAAU;;;;;;yBAtPS,SAAS,aAAa,MAAM,KAAG,eAAe,EAAE;;;;;;;yBAsCzC,SAAS,aAAa,MAAM,KAAG,eAAe,EAAE;;;;;;oCAmOzD,SAAS;yBA1LX,SAAS,KAAG,eAAe,EAAE;;;;;;;yBA4F5B,SAAS,KAAG,eAAe,EAAE;;;;;;;yBA7B/B,SAAS,KAAG,eAAe,EAAE;;;;;;;;;;;;;;yBAtFzB,SAAS,aAAa,MAAM,KAAG,eAAe,EAAE;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;;yBAJjD,SAAS,aAAa,MAAM,KAAG,MAAM;;;;;;;yBA0J1B,SAAS,aAAa,MAAM,KAAG,MAAM;;;;;;;yBAI9C,SAAS,aAAa,MAAM,KAAG,eAAe,EAAE;;CAiN5C,CAAC;AAErC,MAAM,MAAM,WAAW,GAAG,MAAM,OAAO,UAAU,CAAC"}