crfm-helm 0.4.0__py3-none-any.whl → 0.5.10__py3-none-any.whl

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.

Potentially problematic release.

This version of crfm-helm might be problematic. Click here for more details.

Files changed (1033) hide show
  1. crfm_helm-0.5.10.dist-info/METADATA +369 -0
  2. crfm_helm-0.5.10.dist-info/RECORD +1008 -0
  3. {crfm_helm-0.4.0.dist-info → crfm_helm-0.5.10.dist-info}/WHEEL +1 -1
  4. helm/benchmark/adaptation/adapter_spec.py +80 -29
  5. helm/benchmark/adaptation/adapters/adapter.py +2 -2
  6. helm/benchmark/adaptation/adapters/adapter_factory.py +39 -28
  7. helm/benchmark/adaptation/adapters/binary_ranking_adapter.py +1 -1
  8. helm/benchmark/adaptation/adapters/chat_adapter.py +49 -0
  9. helm/benchmark/adaptation/adapters/ehr_instruction_adapter.py +108 -0
  10. helm/benchmark/adaptation/adapters/generation_adapter.py +2 -1
  11. helm/benchmark/adaptation/adapters/in_context_learning_adapter.py +24 -8
  12. helm/benchmark/adaptation/adapters/language_modeling_adapter.py +3 -4
  13. helm/benchmark/adaptation/adapters/multimodal/generation_multimodal_adapter.py +4 -2
  14. helm/benchmark/adaptation/adapters/multimodal/in_context_learning_multimodal_adapter.py +2 -1
  15. helm/benchmark/adaptation/adapters/multimodal/multimodal_prompt.py +7 -0
  16. helm/benchmark/adaptation/adapters/multimodal/multiple_choice_joint_multimodal_adapter.py +112 -0
  17. helm/benchmark/adaptation/adapters/multimodal/test_in_context_learning_multimodal_adapter.py +6 -3
  18. helm/benchmark/adaptation/adapters/multimodal/test_multimodal_prompt.py +3 -1
  19. helm/benchmark/adaptation/adapters/multiple_choice_calibrated_adapter.py +1 -1
  20. helm/benchmark/adaptation/adapters/multiple_choice_joint_adapter.py +18 -8
  21. helm/benchmark/adaptation/adapters/multiple_choice_joint_chain_of_thought_adapter.py +87 -0
  22. helm/benchmark/adaptation/adapters/multiple_choice_separate_adapter.py +1 -1
  23. helm/benchmark/adaptation/adapters/test_adapter.py +5 -4
  24. helm/benchmark/adaptation/adapters/test_generation_adapter.py +46 -22
  25. helm/benchmark/adaptation/adapters/test_language_modeling_adapter.py +17 -29
  26. helm/benchmark/adaptation/adapters/test_multiple_choice_joint_adapter.py +138 -16
  27. helm/benchmark/adaptation/common_adapter_specs.py +443 -0
  28. helm/benchmark/adaptation/prompt.py +1 -1
  29. helm/benchmark/adaptation/request_state.py +6 -1
  30. helm/benchmark/adaptation/scenario_state.py +6 -2
  31. helm/benchmark/annotation/aci_bench_annotator.py +84 -0
  32. helm/benchmark/annotation/air_bench_annotator.py +79 -0
  33. helm/benchmark/annotation/alrage_annotator.py +90 -0
  34. helm/benchmark/annotation/annotator.py +48 -0
  35. helm/benchmark/annotation/annotator_factory.py +50 -0
  36. helm/benchmark/annotation/anthropic_red_team_annotator.py +57 -0
  37. helm/benchmark/annotation/autobencher_capabilities_annotator.py +107 -0
  38. helm/benchmark/annotation/autobencher_safety_annotator.py +98 -0
  39. helm/benchmark/annotation/bigcodebench_annotator.py +108 -0
  40. helm/benchmark/annotation/bird_sql_annotator.py +58 -0
  41. helm/benchmark/annotation/call_center_annotator.py +258 -0
  42. helm/benchmark/annotation/chw_care_plan_annotator.py +82 -0
  43. helm/benchmark/annotation/czech_bank_qa_annotator.py +78 -0
  44. helm/benchmark/annotation/dischargeme_annotator.py +96 -0
  45. helm/benchmark/annotation/ehr_sql_annotator.py +87 -0
  46. helm/benchmark/annotation/financebench_annotator.py +79 -0
  47. helm/benchmark/annotation/harm_bench_annotator.py +55 -0
  48. helm/benchmark/annotation/helpdesk_call_summarization_annotator.py +131 -0
  49. helm/benchmark/annotation/image2struct/image_compiler_annotator.py +93 -0
  50. helm/benchmark/annotation/image2struct/latex_compiler_annotator.py +59 -0
  51. helm/benchmark/annotation/image2struct/lilypond_compiler_annotator.py +86 -0
  52. helm/benchmark/annotation/image2struct/webpage_compiler_annotator.py +132 -0
  53. helm/benchmark/annotation/live_qa_annotator.py +76 -0
  54. helm/benchmark/annotation/med_dialog_annotator.py +88 -0
  55. helm/benchmark/annotation/medalign_annotator.py +89 -0
  56. helm/benchmark/annotation/medi_qa_annotator.py +87 -0
  57. helm/benchmark/annotation/medication_qa_annotator.py +86 -0
  58. helm/benchmark/annotation/mental_health_annotator.py +87 -0
  59. helm/benchmark/annotation/mimic_bhc_annotator.py +89 -0
  60. helm/benchmark/annotation/mimic_rrs_annotator.py +89 -0
  61. helm/benchmark/annotation/model_as_judge.py +309 -0
  62. helm/benchmark/annotation/mtsamples_procedures_annotator.py +87 -0
  63. helm/benchmark/annotation/mtsamples_replicate_annotator.py +90 -0
  64. helm/benchmark/annotation/omni_math/gpt_evaluation_template.txt +152 -0
  65. helm/benchmark/annotation/omni_math/gpt_evaluation_zero_shot_template.txt +36 -0
  66. helm/benchmark/annotation/omni_math_annotator.py +131 -0
  67. helm/benchmark/annotation/simple_safety_tests_annotator.py +50 -0
  68. helm/benchmark/annotation/spider_annotator.py +18 -0
  69. helm/benchmark/annotation/starr_patient_instructions_annotator.py +87 -0
  70. helm/benchmark/annotation/test_annotator_factory.py +26 -0
  71. helm/benchmark/annotation/test_dummy_annotator.py +44 -0
  72. helm/benchmark/annotation/wildbench/eval_template.pairwise.v2.md +75 -0
  73. helm/benchmark/annotation/wildbench/eval_template.score.v2.md +66 -0
  74. helm/benchmark/annotation/wildbench_annotator.py +119 -0
  75. helm/benchmark/annotation/xstest_annotator.py +100 -0
  76. helm/benchmark/annotation_executor.py +144 -0
  77. helm/benchmark/augmentations/cleva_perturbation.py +9 -8
  78. helm/benchmark/augmentations/contraction_expansion_perturbation.py +2 -2
  79. helm/benchmark/augmentations/contrast_sets_perturbation.py +2 -2
  80. helm/benchmark/augmentations/data_augmenter.py +0 -2
  81. helm/benchmark/augmentations/dialect_perturbation.py +4 -5
  82. helm/benchmark/augmentations/extra_space_perturbation.py +2 -2
  83. helm/benchmark/augmentations/filler_words_perturbation.py +2 -2
  84. helm/benchmark/augmentations/gender_perturbation.py +3 -3
  85. helm/benchmark/augmentations/lowercase_perturbation.py +2 -2
  86. helm/benchmark/augmentations/mild_mix_perturbation.py +6 -6
  87. helm/benchmark/augmentations/misspelling_perturbation.py +2 -2
  88. helm/benchmark/augmentations/person_name_perturbation.py +4 -5
  89. helm/benchmark/augmentations/perturbation.py +26 -4
  90. helm/benchmark/augmentations/perturbation_description.py +1 -1
  91. helm/benchmark/augmentations/space_perturbation.py +2 -2
  92. helm/benchmark/augmentations/suffix_perturbation.py +29 -0
  93. helm/benchmark/augmentations/synonym_perturbation.py +4 -3
  94. helm/benchmark/augmentations/test_perturbation.py +56 -19
  95. helm/benchmark/augmentations/translate_perturbation.py +31 -0
  96. helm/benchmark/augmentations/typos_perturbation.py +2 -2
  97. helm/benchmark/config_registry.py +7 -1
  98. helm/benchmark/data_preprocessor.py +2 -2
  99. helm/benchmark/executor.py +54 -25
  100. helm/benchmark/huggingface_registration.py +28 -10
  101. helm/benchmark/metrics/air_bench_metrics.py +3212 -0
  102. helm/benchmark/metrics/alrage_metric.py +35 -0
  103. helm/benchmark/metrics/annotation_metrics.py +108 -0
  104. helm/benchmark/metrics/basic_metrics.py +437 -667
  105. helm/benchmark/metrics/bbq_metrics.py +17 -6
  106. helm/benchmark/metrics/bias_metrics.py +18 -9
  107. helm/benchmark/metrics/bias_word_lists.py +1 -1
  108. helm/benchmark/metrics/bigcodebench_metrics.py +25 -0
  109. helm/benchmark/metrics/bird_sql_metrics.py +28 -0
  110. helm/benchmark/metrics/classification_metrics.py +107 -22
  111. helm/benchmark/metrics/cleva_accuracy_metrics.py +8 -5
  112. helm/benchmark/metrics/cleva_harms_metrics.py +12 -11
  113. helm/benchmark/metrics/code_metrics.py +5 -5
  114. helm/benchmark/metrics/code_metrics_helper.py +11 -3
  115. helm/benchmark/metrics/codeinsights_code_efficiency_metrics.py +186 -0
  116. helm/benchmark/metrics/codeinsights_code_evaluation_metrics.py +477 -0
  117. helm/benchmark/metrics/codeinsights_correct_code_metrics.py +366 -0
  118. helm/benchmark/metrics/codeinsights_edge_case_metrics.py +92 -0
  119. helm/benchmark/metrics/codeinsights_metric_specs.py +51 -0
  120. helm/benchmark/metrics/comet_metric.py +125 -0
  121. helm/benchmark/metrics/common_metric_specs.py +174 -0
  122. helm/benchmark/metrics/conv_fin_qa_calc_metrics.py +83 -0
  123. helm/benchmark/metrics/copyright_metrics.py +5 -5
  124. helm/benchmark/metrics/czech_bank_qa_metrics.py +29 -0
  125. helm/benchmark/metrics/decodingtrust_fairness_metrics.py +72 -0
  126. helm/benchmark/metrics/decodingtrust_ood_knowledge_metrics.py +66 -0
  127. helm/benchmark/metrics/decodingtrust_privacy_metrics.py +101 -0
  128. helm/benchmark/metrics/decodingtrust_stereotype_bias_metrics.py +202 -0
  129. helm/benchmark/metrics/disinformation_metrics.py +8 -114
  130. helm/benchmark/metrics/dry_run_metrics.py +35 -6
  131. helm/benchmark/metrics/efficiency_metrics.py +287 -0
  132. helm/benchmark/metrics/ehr_sql_metrics.py +159 -0
  133. helm/benchmark/metrics/evaluate_instances_metric.py +59 -0
  134. helm/benchmark/metrics/evaluate_reference_metrics.py +831 -0
  135. helm/benchmark/metrics/fin_qa_metrics.py +60 -0
  136. helm/benchmark/metrics/fin_qa_metrics_helper.py +398 -0
  137. helm/benchmark/metrics/gpqa_chain_of_thought_metric.py +115 -0
  138. helm/benchmark/metrics/gpt4_audio_critique_metrics.py +167 -0
  139. helm/benchmark/metrics/gpt4_audio_refusal_metrics.py +145 -0
  140. helm/benchmark/metrics/gpt4v_originality_critique_metrics.py +126 -0
  141. helm/benchmark/metrics/helpdesk_call_summarization_metrics.py +48 -0
  142. helm/benchmark/metrics/ifeval/instructions.py +1574 -0
  143. helm/benchmark/metrics/ifeval/instructions_registry.py +182 -0
  144. helm/benchmark/metrics/ifeval/instructions_registry.pyi +3 -0
  145. helm/benchmark/metrics/ifeval/instructions_util.py +153 -0
  146. helm/benchmark/metrics/ifeval_metrics.py +67 -0
  147. helm/benchmark/metrics/image_generation/aesthetics_metrics.py +54 -0
  148. helm/benchmark/metrics/image_generation/aesthetics_scorer.py +66 -0
  149. helm/benchmark/metrics/image_generation/clip_score_metrics.py +84 -0
  150. helm/benchmark/metrics/image_generation/denoised_runtime_metric.py +42 -0
  151. helm/benchmark/metrics/image_generation/detection_metrics.py +57 -0
  152. helm/benchmark/metrics/image_generation/detectors/base_detector.py +8 -0
  153. helm/benchmark/metrics/image_generation/detectors/vitdet.py +178 -0
  154. helm/benchmark/metrics/image_generation/efficiency_metrics.py +41 -0
  155. helm/benchmark/metrics/image_generation/fidelity_metrics.py +168 -0
  156. helm/benchmark/metrics/image_generation/fractal_dimension/__init__.py +0 -0
  157. helm/benchmark/metrics/image_generation/fractal_dimension/fractal_dimension_util.py +63 -0
  158. helm/benchmark/metrics/image_generation/fractal_dimension/test_fractal_dimension_util.py +33 -0
  159. helm/benchmark/metrics/image_generation/fractal_dimension_metric.py +50 -0
  160. helm/benchmark/metrics/image_generation/gender_metrics.py +58 -0
  161. helm/benchmark/metrics/image_generation/image_critique_metrics.py +284 -0
  162. helm/benchmark/metrics/image_generation/lpips_metrics.py +82 -0
  163. helm/benchmark/metrics/image_generation/multi_scale_ssim_metrics.py +82 -0
  164. helm/benchmark/metrics/image_generation/nsfw_detector.py +96 -0
  165. helm/benchmark/metrics/image_generation/nsfw_metrics.py +103 -0
  166. helm/benchmark/metrics/image_generation/nudity_metrics.py +38 -0
  167. helm/benchmark/metrics/image_generation/photorealism_critique_metrics.py +153 -0
  168. helm/benchmark/metrics/image_generation/psnr_metrics.py +78 -0
  169. helm/benchmark/metrics/image_generation/q16/__init__.py +0 -0
  170. helm/benchmark/metrics/image_generation/q16/q16_toxicity_detector.py +90 -0
  171. helm/benchmark/metrics/image_generation/q16/test_q16.py +20 -0
  172. helm/benchmark/metrics/image_generation/q16_toxicity_metrics.py +48 -0
  173. helm/benchmark/metrics/image_generation/skin_tone_metrics.py +164 -0
  174. helm/benchmark/metrics/image_generation/uiqi_metrics.py +92 -0
  175. helm/benchmark/metrics/image_generation/watermark/__init__.py +0 -0
  176. helm/benchmark/metrics/image_generation/watermark/test_watermark_detector.py +16 -0
  177. helm/benchmark/metrics/image_generation/watermark/watermark_detector.py +87 -0
  178. helm/benchmark/metrics/image_generation/watermark_metrics.py +48 -0
  179. helm/benchmark/metrics/instruction_following_critique_metrics.py +48 -5
  180. helm/benchmark/metrics/kpi_edgar_metrics.py +142 -0
  181. helm/benchmark/metrics/language_modeling_metrics.py +111 -0
  182. helm/benchmark/metrics/live_qa_metrics.py +35 -0
  183. helm/benchmark/metrics/llm_jury_metrics.py +58 -0
  184. helm/benchmark/metrics/lmkt_metric_specs.py +12 -0
  185. helm/benchmark/metrics/lmkt_metrics.py +47 -0
  186. helm/benchmark/metrics/machine_translation_metrics.py +89 -0
  187. helm/benchmark/metrics/medcalc_bench_metrics.py +137 -0
  188. helm/benchmark/metrics/medec_metrics.py +124 -0
  189. helm/benchmark/metrics/melt_bias_metric.py +234 -0
  190. helm/benchmark/metrics/melt_bias_word_lists.py +1367 -0
  191. helm/benchmark/metrics/melt_metric_specs.py +43 -0
  192. helm/benchmark/metrics/melt_toxicity_metric.py +107 -0
  193. helm/benchmark/metrics/metric.py +121 -175
  194. helm/benchmark/metrics/metric_name.py +0 -1
  195. helm/benchmark/metrics/metric_service.py +23 -7
  196. helm/benchmark/metrics/mimiciv_billing_code_metrics.py +127 -0
  197. helm/benchmark/metrics/nltk_helper.py +32 -0
  198. helm/benchmark/metrics/omni_math_metrics.py +44 -0
  199. helm/benchmark/metrics/openai_mrcr_metrics.py +52 -0
  200. helm/benchmark/metrics/output_processing_metric.py +60 -0
  201. helm/benchmark/metrics/output_processors.py +15 -0
  202. helm/benchmark/metrics/paraphrase_generation_metrics.py +5 -6
  203. helm/benchmark/metrics/prometheus_vision_critique_metrics.py +185 -0
  204. helm/benchmark/metrics/ranking_metrics.py +5 -5
  205. helm/benchmark/metrics/reference_metric.py +148 -0
  206. helm/benchmark/metrics/reka_vibe_critique_metrics.py +158 -0
  207. helm/benchmark/metrics/ruler_qa_metrics.py +34 -0
  208. helm/benchmark/metrics/safety_metrics.py +91 -0
  209. helm/benchmark/metrics/seahelm_metrics.py +201 -0
  210. helm/benchmark/metrics/seahelm_metrics_specs.py +10 -0
  211. helm/benchmark/metrics/spider_metrics.py +7 -0
  212. helm/benchmark/metrics/statistic.py +1 -1
  213. helm/benchmark/metrics/summac/model_summac.py +8 -11
  214. helm/benchmark/metrics/summarization_critique_metrics.py +4 -4
  215. helm/benchmark/metrics/summarization_metrics.py +150 -11
  216. helm/benchmark/metrics/test_bias_metrics.py +5 -1
  217. helm/benchmark/metrics/test_classification_metrics.py +145 -70
  218. helm/benchmark/metrics/test_disinformation_metrics.py +78 -0
  219. helm/benchmark/metrics/{test_basic_metrics.py → test_evaluate_reference_metrics.py} +20 -1
  220. helm/benchmark/metrics/test_metric.py +3 -3
  221. helm/benchmark/metrics/test_statistic.py +2 -2
  222. helm/benchmark/metrics/tokens/ai21_token_cost_estimator.py +1 -1
  223. helm/benchmark/metrics/tokens/auto_token_cost_estimator.py +6 -6
  224. helm/benchmark/metrics/tokens/cohere_token_cost_estimator.py +1 -1
  225. helm/benchmark/metrics/tokens/free_token_cost_estimator.py +1 -1
  226. helm/benchmark/metrics/tokens/gooseai_token_cost_estimator.py +11 -3
  227. helm/benchmark/metrics/tokens/openai_token_cost_estimator.py +1 -1
  228. helm/benchmark/metrics/tokens/test_ai21_token_cost_estimator.py +3 -3
  229. helm/benchmark/metrics/tokens/test_openai_token_cost_estimator.py +7 -7
  230. helm/benchmark/metrics/toxicity_metrics.py +37 -7
  231. helm/benchmark/metrics/toxicity_utils.py +23 -0
  232. helm/benchmark/metrics/ultra_suite_asr_classification_metrics.py +52 -0
  233. helm/benchmark/metrics/unitxt_metrics.py +107 -0
  234. helm/benchmark/metrics/vision_language/__init__.py +0 -0
  235. helm/benchmark/metrics/vision_language/emd_utils.py +347 -0
  236. helm/benchmark/metrics/vision_language/image_metrics.py +537 -0
  237. helm/benchmark/metrics/vision_language/image_utils.py +100 -0
  238. helm/benchmark/metrics/wildbench_metrics.py +54 -0
  239. helm/benchmark/model_deployment_registry.py +69 -5
  240. helm/benchmark/model_metadata_registry.py +58 -2
  241. helm/benchmark/multi_gpu_runner.py +133 -0
  242. helm/benchmark/presentation/contamination.py +3 -3
  243. helm/benchmark/presentation/create_plots.py +51 -20
  244. helm/benchmark/presentation/run_display.py +51 -12
  245. helm/benchmark/presentation/run_entry.py +2 -2
  246. helm/benchmark/presentation/schema.py +83 -66
  247. helm/benchmark/presentation/summarize.py +483 -388
  248. helm/benchmark/presentation/table.py +8 -8
  249. helm/benchmark/presentation/taxonomy_info.py +20 -0
  250. helm/benchmark/presentation/test_contamination.py +2 -2
  251. helm/benchmark/presentation/test_create_plots.py +4 -1
  252. helm/benchmark/presentation/test_run_entry.py +2 -2
  253. helm/benchmark/presentation/test_schema.py +11 -0
  254. helm/benchmark/presentation/test_summarize.py +148 -6
  255. helm/benchmark/presentation/torr_robustness_summarizer.py +178 -0
  256. helm/benchmark/reeval_run.py +202 -0
  257. helm/benchmark/reeval_runner.py +355 -0
  258. helm/benchmark/run.py +151 -87
  259. helm/benchmark/run_expander.py +418 -33
  260. helm/benchmark/run_spec.py +93 -0
  261. helm/benchmark/run_spec_factory.py +180 -0
  262. helm/benchmark/run_specs/__init__.py +0 -0
  263. helm/benchmark/run_specs/air_bench_run_specs.py +58 -0
  264. helm/benchmark/run_specs/arabic_run_specs.py +197 -0
  265. helm/benchmark/run_specs/audio_run_specs.py +657 -0
  266. helm/benchmark/run_specs/bluex_run_specs.py +40 -0
  267. helm/benchmark/run_specs/call_center_run_specs.py +201 -0
  268. helm/benchmark/run_specs/capabilities_run_specs.py +308 -0
  269. helm/benchmark/run_specs/classic_run_specs.py +1393 -0
  270. helm/benchmark/run_specs/cleva_run_specs.py +277 -0
  271. helm/benchmark/run_specs/codeinsights_run_specs.py +192 -0
  272. helm/benchmark/run_specs/decodingtrust_run_specs.py +316 -0
  273. helm/benchmark/run_specs/enem_challenge_specs.py +31 -0
  274. helm/benchmark/run_specs/enterprise_run_specs.py +280 -0
  275. helm/benchmark/run_specs/experimental_run_specs.py +224 -0
  276. helm/benchmark/run_specs/finance_run_specs.py +114 -0
  277. helm/benchmark/run_specs/healthqa_br_run_specs.py +40 -0
  278. helm/benchmark/run_specs/heim_run_specs.py +625 -0
  279. helm/benchmark/run_specs/imdb_ptbr_run_specs.py +30 -0
  280. helm/benchmark/run_specs/instruction_following_run_specs.py +129 -0
  281. helm/benchmark/run_specs/lite_run_specs.py +307 -0
  282. helm/benchmark/run_specs/lmkt_run_specs.py +144 -0
  283. helm/benchmark/run_specs/long_context_run_specs.py +188 -0
  284. helm/benchmark/run_specs/medhelm/__init__.py +0 -0
  285. helm/benchmark/run_specs/medhelm/benchmark_config.py +219 -0
  286. helm/benchmark/run_specs/medhelm_run_specs.py +1570 -0
  287. helm/benchmark/run_specs/melt_run_specs.py +783 -0
  288. helm/benchmark/run_specs/mmlu_clinical_afr_run_specs.py +49 -0
  289. helm/benchmark/run_specs/multilingual_run_specs.py +50 -0
  290. helm/benchmark/run_specs/oab_exams_specs.py +32 -0
  291. helm/benchmark/run_specs/safety_run_specs.py +191 -0
  292. helm/benchmark/run_specs/seahelm_run_specs.py +652 -0
  293. helm/benchmark/run_specs/simple_run_specs.py +104 -0
  294. helm/benchmark/run_specs/speech_disorder_audio_run_specs.py +167 -0
  295. helm/benchmark/run_specs/sql_run_specs.py +54 -0
  296. helm/benchmark/run_specs/tweetsentbr_run_specs.py +32 -0
  297. helm/benchmark/run_specs/unitxt_run_specs.py +51 -0
  298. helm/benchmark/run_specs/vlm_run_specs.py +1057 -0
  299. helm/benchmark/run_specs/winogrande_afr_run_specs.py +47 -0
  300. helm/benchmark/runner.py +63 -62
  301. helm/benchmark/runner_config_registry.py +21 -0
  302. helm/benchmark/scenarios/aci_bench_scenario.py +149 -0
  303. helm/benchmark/scenarios/air_bench_scenario.py +76 -0
  304. helm/benchmark/scenarios/alghafa_scenario.py +126 -0
  305. helm/benchmark/scenarios/alrage_scenario.py +54 -0
  306. helm/benchmark/scenarios/anthropic_hh_rlhf_scenario.py +27 -3
  307. helm/benchmark/scenarios/anthropic_red_team_scenario.py +82 -0
  308. helm/benchmark/scenarios/arabic_exams_scenario.py +114 -0
  309. helm/benchmark/scenarios/arabic_mmlu_scenario.py +82 -0
  310. helm/benchmark/scenarios/aratrust_scenario.py +95 -0
  311. helm/benchmark/scenarios/audio_language/__init__.py +0 -0
  312. helm/benchmark/scenarios/audio_language/air_bench_chat_scenario.py +130 -0
  313. helm/benchmark/scenarios/audio_language/air_bench_foundation_scenario.py +154 -0
  314. helm/benchmark/scenarios/audio_language/ami_scenario.py +96 -0
  315. helm/benchmark/scenarios/audio_language/audio_mnist_scenario.py +62 -0
  316. helm/benchmark/scenarios/audio_language/audio_pairs_scenario.py +62 -0
  317. helm/benchmark/scenarios/audio_language/audiocaps_scenario.py +59 -0
  318. helm/benchmark/scenarios/audio_language/casual_conversations2_scenario.py +152 -0
  319. helm/benchmark/scenarios/audio_language/common_voice_15_scenario.py +99 -0
  320. helm/benchmark/scenarios/audio_language/corebench_scenario.py +77 -0
  321. helm/benchmark/scenarios/audio_language/covost2_scenario.py +163 -0
  322. helm/benchmark/scenarios/audio_language/fleurs_fairness_scenario.py +83 -0
  323. helm/benchmark/scenarios/audio_language/fleurs_scenario.py +312 -0
  324. helm/benchmark/scenarios/audio_language/iemocap_audio_scenario.py +83 -0
  325. helm/benchmark/scenarios/audio_language/librispeech_fairness_scenario.py +96 -0
  326. helm/benchmark/scenarios/audio_language/librispeech_scenario.py +80 -0
  327. helm/benchmark/scenarios/audio_language/meld_audio_scenario.py +113 -0
  328. helm/benchmark/scenarios/audio_language/multilingual_librispeech_scenario.py +80 -0
  329. helm/benchmark/scenarios/audio_language/mustard_scenario.py +142 -0
  330. helm/benchmark/scenarios/audio_language/mutox_scenario.py +254 -0
  331. helm/benchmark/scenarios/audio_language/parade_scenario.py +97 -0
  332. helm/benchmark/scenarios/audio_language/speech_robust_bench_scenario.py +124 -0
  333. helm/benchmark/scenarios/audio_language/ultra_suite_asr_classification_scenario.py +74 -0
  334. helm/benchmark/scenarios/audio_language/ultra_suite_asr_transcription_scenario.py +70 -0
  335. helm/benchmark/scenarios/audio_language/ultra_suite_classification_scenario.py +79 -0
  336. helm/benchmark/scenarios/audio_language/ultra_suite_disorder_breakdown_scenario.py +78 -0
  337. helm/benchmark/scenarios/audio_language/ultra_suite_disorder_symptoms_scenario.py +78 -0
  338. helm/benchmark/scenarios/audio_language/vocal_sound_scenario.py +83 -0
  339. helm/benchmark/scenarios/audio_language/voice_jailbreak_attacks_scenario.py +87 -0
  340. helm/benchmark/scenarios/audio_language/voxceleb2_scenario.py +105 -0
  341. helm/benchmark/scenarios/autobencher_capabilities_scenario.py +68 -0
  342. helm/benchmark/scenarios/autobencher_safety_scenario.py +51 -0
  343. helm/benchmark/scenarios/babi_qa_scenario.py +16 -1
  344. helm/benchmark/scenarios/banking77_scenario.py +77 -0
  345. helm/benchmark/scenarios/bbq_scenario.py +17 -2
  346. helm/benchmark/scenarios/best_chatgpt_prompts.yaml +473 -0
  347. helm/benchmark/scenarios/big_bench_scenario.py +11 -1
  348. helm/benchmark/scenarios/bigcodebench_scenario.py +58 -0
  349. helm/benchmark/scenarios/bird_sql_scenario.py +112 -0
  350. helm/benchmark/scenarios/bird_sql_scenario_helper.py +118 -0
  351. helm/benchmark/scenarios/blimp_scenario.py +1 -1
  352. helm/benchmark/scenarios/bluex_scenario.py +70 -0
  353. helm/benchmark/scenarios/bold_scenario.py +18 -3
  354. helm/benchmark/scenarios/boolq_scenario.py +21 -1
  355. helm/benchmark/scenarios/call_center_scenario.py +84 -0
  356. helm/benchmark/scenarios/casehold_scenario.py +79 -0
  357. helm/benchmark/scenarios/chw_care_plan_scenario.py +129 -0
  358. helm/benchmark/scenarios/ci_mcqa_scenario.py +80 -0
  359. helm/benchmark/scenarios/civil_comments_scenario.py +14 -1
  360. helm/benchmark/scenarios/clear_scenario.py +180 -0
  361. helm/benchmark/scenarios/cleva_scenario.py +482 -3
  362. helm/benchmark/scenarios/code_scenario.py +46 -4
  363. helm/benchmark/scenarios/codeinsights_code_efficiency_scenario.py +197 -0
  364. helm/benchmark/scenarios/codeinsights_correct_code_scenario.py +78 -0
  365. helm/benchmark/scenarios/codeinsights_edge_case_scenario.py +192 -0
  366. helm/benchmark/scenarios/codeinsights_student_coding_scenario.py +162 -0
  367. helm/benchmark/scenarios/codeinsights_student_mistake_scenario.py +188 -0
  368. helm/benchmark/scenarios/commonsense_scenario.py +33 -1
  369. helm/benchmark/scenarios/compositional_instructions.yaml +70 -0
  370. helm/benchmark/scenarios/conv_fin_qa_calc_scenario.py +118 -0
  371. helm/benchmark/scenarios/copyright_scenario.py +35 -1
  372. helm/benchmark/scenarios/covid_dialog_scenario.py +10 -1
  373. helm/benchmark/scenarios/cti_to_mitre_scenario.py +261 -0
  374. helm/benchmark/scenarios/custom_mcqa_scenario.py +1 -1
  375. helm/benchmark/scenarios/czech_bank_qa_scenario.py +148 -0
  376. helm/benchmark/scenarios/decodingtrust_adv_demonstration_scenario.py +190 -0
  377. helm/benchmark/scenarios/decodingtrust_adv_robustness_scenario.py +143 -0
  378. helm/benchmark/scenarios/decodingtrust_fairness_scenario.py +98 -0
  379. helm/benchmark/scenarios/decodingtrust_machine_ethics_scenario.py +344 -0
  380. helm/benchmark/scenarios/decodingtrust_ood_robustness_scenario.py +217 -0
  381. helm/benchmark/scenarios/decodingtrust_privacy_scenario.py +571 -0
  382. helm/benchmark/scenarios/decodingtrust_stereotype_bias_scenario.py +80 -0
  383. helm/benchmark/scenarios/decodingtrust_toxicity_prompts_scenario.py +90 -0
  384. helm/benchmark/scenarios/dialogue_scenarios.py +13 -3
  385. helm/benchmark/scenarios/dischargeme_scenario.py +196 -0
  386. helm/benchmark/scenarios/disinformation_scenario.py +32 -1
  387. helm/benchmark/scenarios/dyck_language_scenario.py +25 -1
  388. helm/benchmark/scenarios/echr_judgment_classification_scenario.py +113 -0
  389. helm/benchmark/scenarios/ehr_sql_scenario.py +137 -0
  390. helm/benchmark/scenarios/ehrshot_scenario.py +1541 -0
  391. helm/benchmark/scenarios/enem_challenge_scenario.py +77 -0
  392. helm/benchmark/scenarios/entity_data_imputation_scenario.py +33 -3
  393. helm/benchmark/scenarios/entity_matching_scenario.py +26 -2
  394. helm/benchmark/scenarios/ewok_scenario.py +116 -0
  395. helm/benchmark/scenarios/exams_multilingual_scenario.py +115 -0
  396. helm/benchmark/scenarios/fin_qa_scenario.py +139 -0
  397. helm/benchmark/scenarios/financebench_scenario.py +74 -0
  398. helm/benchmark/scenarios/financial_phrasebank_scenario.py +115 -0
  399. helm/benchmark/scenarios/gold_commodity_news_scenario.py +145 -0
  400. helm/benchmark/scenarios/gpqa_scenario.py +98 -0
  401. helm/benchmark/scenarios/grammar.py +2 -2
  402. helm/benchmark/scenarios/grammar_scenario.py +21 -2
  403. helm/benchmark/scenarios/gsm_scenario.py +31 -1
  404. helm/benchmark/scenarios/harm_bench_gcg_transfer_scenario.py +61 -0
  405. helm/benchmark/scenarios/harm_bench_scenario.py +70 -0
  406. helm/benchmark/scenarios/headqa_scenario.py +158 -0
  407. helm/benchmark/scenarios/healthqa_br_scenario.py +80 -0
  408. helm/benchmark/scenarios/helpdesk_call_summarization_scenario.py +50 -0
  409. helm/benchmark/scenarios/ice_scenario.py +28 -4
  410. helm/benchmark/scenarios/ifeval_scenario.py +71 -0
  411. helm/benchmark/scenarios/image_generation/__init__.py +0 -0
  412. helm/benchmark/scenarios/image_generation/common_syntactic_processes_scenario.py +105 -0
  413. helm/benchmark/scenarios/image_generation/cub200_scenario.py +95 -0
  414. helm/benchmark/scenarios/image_generation/daily_dalle_scenario.py +124 -0
  415. helm/benchmark/scenarios/image_generation/demographic_stereotypes_scenario.py +82 -0
  416. helm/benchmark/scenarios/image_generation/detection_scenario.py +83 -0
  417. helm/benchmark/scenarios/image_generation/draw_bench_scenario.py +74 -0
  418. helm/benchmark/scenarios/image_generation/i2p_scenario.py +57 -0
  419. helm/benchmark/scenarios/image_generation/landing_page_scenario.py +46 -0
  420. helm/benchmark/scenarios/image_generation/logos_scenario.py +223 -0
  421. helm/benchmark/scenarios/image_generation/magazine_cover_scenario.py +91 -0
  422. helm/benchmark/scenarios/image_generation/mental_disorders_scenario.py +46 -0
  423. helm/benchmark/scenarios/image_generation/mscoco_scenario.py +91 -0
  424. helm/benchmark/scenarios/image_generation/paint_skills_scenario.py +72 -0
  425. helm/benchmark/scenarios/image_generation/parti_prompts_scenario.py +94 -0
  426. helm/benchmark/scenarios/image_generation/radiology_scenario.py +42 -0
  427. helm/benchmark/scenarios/image_generation/relational_understanding_scenario.py +52 -0
  428. helm/benchmark/scenarios/image_generation/time_most_significant_historical_figures_scenario.py +124 -0
  429. helm/benchmark/scenarios/image_generation/winoground_scenario.py +62 -0
  430. helm/benchmark/scenarios/imdb_ptbr_scenario.py +60 -0
  431. helm/benchmark/scenarios/imdb_scenario.py +26 -3
  432. helm/benchmark/scenarios/infinite_bench_en_mc_scenario.py +111 -0
  433. helm/benchmark/scenarios/infinite_bench_en_qa_scenario.py +85 -0
  434. helm/benchmark/scenarios/infinite_bench_en_sum_scenario.py +98 -0
  435. helm/benchmark/scenarios/interactive_qa_mmlu_scenario.py +2 -2
  436. helm/benchmark/scenarios/koala_scenario.py +21 -1
  437. helm/benchmark/scenarios/kpi_edgar_scenario.py +172 -0
  438. helm/benchmark/scenarios/legal_contract_summarization_scenario.py +149 -0
  439. helm/benchmark/scenarios/legal_opinion_sentiment_classification_scenario.py +77 -0
  440. helm/benchmark/scenarios/legal_summarization_scenario.py +61 -1
  441. helm/benchmark/scenarios/legal_support_scenario.py +24 -1
  442. helm/benchmark/scenarios/legalbench_scenario.py +45 -3
  443. helm/benchmark/scenarios/lex_glue_scenario.py +23 -2
  444. helm/benchmark/scenarios/lextreme_scenario.py +22 -1
  445. helm/benchmark/scenarios/live_qa_scenario.py +94 -0
  446. helm/benchmark/scenarios/lm_entry_scenario.py +185 -0
  447. helm/benchmark/scenarios/lmkt_scenarios.py +288 -0
  448. helm/benchmark/scenarios/lsat_qa_scenario.py +15 -1
  449. helm/benchmark/scenarios/madinah_qa_scenario.py +73 -0
  450. helm/benchmark/scenarios/math_scenario.py +81 -22
  451. helm/benchmark/scenarios/mbzuai_human_translated_arabic_mmlu.py +68 -0
  452. helm/benchmark/scenarios/me_q_sum_scenario.py +10 -1
  453. helm/benchmark/scenarios/med_dialog_scenario.py +56 -22
  454. helm/benchmark/scenarios/med_mcqa_scenario.py +24 -1
  455. helm/benchmark/scenarios/med_paragraph_simplification_scenario.py +10 -1
  456. helm/benchmark/scenarios/med_qa_scenario.py +30 -1
  457. helm/benchmark/scenarios/medalign_scenario.py +117 -0
  458. helm/benchmark/scenarios/medalign_scenario_helper.py +326 -0
  459. helm/benchmark/scenarios/medbullets_scenario.py +167 -0
  460. helm/benchmark/scenarios/medcalc_bench_scenario.py +149 -0
  461. helm/benchmark/scenarios/medec_scenario.py +148 -0
  462. helm/benchmark/scenarios/medhallu_scenario.py +95 -0
  463. helm/benchmark/scenarios/medhelm/__init__.py +0 -0
  464. helm/benchmark/scenarios/medhelm/judges.yaml +14 -0
  465. helm/benchmark/scenarios/medhelm_configurable_scenario.py +101 -0
  466. helm/benchmark/scenarios/medi_qa_scenario.py +134 -0
  467. helm/benchmark/scenarios/medication_qa_scenario.py +96 -0
  468. helm/benchmark/scenarios/melt_ir_scenario.py +171 -0
  469. helm/benchmark/scenarios/melt_knowledge_scenario.py +246 -0
  470. helm/benchmark/scenarios/melt_lm_scenarios.py +252 -0
  471. helm/benchmark/scenarios/melt_scenarios.py +793 -0
  472. helm/benchmark/scenarios/melt_srn_scenario.py +342 -0
  473. helm/benchmark/scenarios/melt_synthetic_reasoning_scenario.py +222 -0
  474. helm/benchmark/scenarios/melt_translation_scenario.py +152 -0
  475. helm/benchmark/scenarios/mental_health_scenario.py +146 -0
  476. helm/benchmark/scenarios/mimic_bhc_scenario.py +127 -0
  477. helm/benchmark/scenarios/mimic_rrs_scenario.py +121 -0
  478. helm/benchmark/scenarios/mimiciv_billing_code_scenario.py +99 -0
  479. helm/benchmark/scenarios/mmlu_clinical_afr_scenario.py +74 -0
  480. helm/benchmark/scenarios/mmlu_pro_scenario.py +113 -0
  481. helm/benchmark/scenarios/mmlu_scenario.py +32 -1
  482. helm/benchmark/scenarios/mmmlu_scenario.py +85 -0
  483. helm/benchmark/scenarios/msmarco_scenario.py +31 -1
  484. helm/benchmark/scenarios/mtsamples_procedures_scenario.py +166 -0
  485. helm/benchmark/scenarios/mtsamples_replicate_scenario.py +164 -0
  486. helm/benchmark/scenarios/n2c2_ct_matching_scenario.py +297 -0
  487. helm/benchmark/scenarios/narrativeqa_scenario.py +20 -1
  488. helm/benchmark/scenarios/natural_qa_scenario.py +33 -1
  489. helm/benchmark/scenarios/newsqa_scenario.py +1 -1
  490. helm/benchmark/scenarios/oab_exams_scenario.py +57 -0
  491. helm/benchmark/scenarios/omni_math_scenario.py +71 -0
  492. helm/benchmark/scenarios/open_assistant_scenario.py +33 -2
  493. helm/benchmark/scenarios/openai_mrcr_scenario.py +94 -0
  494. helm/benchmark/scenarios/opinions_qa_scenario.py +1 -5
  495. helm/benchmark/scenarios/pubmed_qa_scenario.py +81 -43
  496. helm/benchmark/scenarios/quac_scenario.py +24 -1
  497. helm/benchmark/scenarios/race_based_med_scenario.py +175 -0
  498. helm/benchmark/scenarios/raft_scenario.py +33 -3
  499. helm/benchmark/scenarios/real_toxicity_prompts_scenario.py +14 -1
  500. helm/benchmark/scenarios/ruler_qa_scenario_helper.py +171 -0
  501. helm/benchmark/scenarios/ruler_qa_scenarios.py +128 -0
  502. helm/benchmark/scenarios/scenario.py +44 -1
  503. helm/benchmark/scenarios/seahelm_scenario.py +2295 -0
  504. helm/benchmark/scenarios/self_instruct_scenario.py +29 -1
  505. helm/benchmark/scenarios/shc_bmt_scenario.py +97 -0
  506. helm/benchmark/scenarios/shc_cdi_scenario.py +95 -0
  507. helm/benchmark/scenarios/shc_conf_scenario.py +99 -0
  508. helm/benchmark/scenarios/shc_ent_scenario.py +98 -0
  509. helm/benchmark/scenarios/shc_gip_scenario.py +94 -0
  510. helm/benchmark/scenarios/shc_privacy_scenario.py +100 -0
  511. helm/benchmark/scenarios/shc_proxy_scenario.py +98 -0
  512. helm/benchmark/scenarios/shc_ptbm_scenario.py +104 -0
  513. helm/benchmark/scenarios/shc_sei_scenario.py +94 -0
  514. helm/benchmark/scenarios/shc_sequoia_scenario.py +98 -0
  515. helm/benchmark/scenarios/simple_safety_tests_scenario.py +44 -0
  516. helm/benchmark/scenarios/simple_scenarios.py +122 -1
  517. helm/benchmark/scenarios/situation_prompts.yaml +49 -0
  518. helm/benchmark/scenarios/spider_scenario.py +109 -0
  519. helm/benchmark/scenarios/starr_patient_instructions_scenario.py +119 -0
  520. helm/benchmark/scenarios/summarization_scenario.py +48 -1
  521. helm/benchmark/scenarios/sumosum_scenario.py +157 -0
  522. helm/benchmark/scenarios/synthetic_efficiency_scenario.py +22 -1
  523. helm/benchmark/scenarios/synthetic_reasoning_natural_scenario.py +24 -1
  524. helm/benchmark/scenarios/synthetic_reasoning_scenario.py +11 -1
  525. helm/benchmark/scenarios/test_air_bench_scenario.py +27 -0
  526. helm/benchmark/scenarios/test_alghafa_scenario.py +29 -0
  527. helm/benchmark/scenarios/test_alrage_scenario.py +23 -0
  528. helm/benchmark/scenarios/test_arabic_exams_scenario.py +21 -0
  529. helm/benchmark/scenarios/test_aratrust_scenario.py +21 -0
  530. helm/benchmark/scenarios/test_bigcodebench_scenario.py +26 -0
  531. helm/benchmark/scenarios/test_bluex_scenario.py +59 -0
  532. helm/benchmark/scenarios/test_commonsense_scenario.py +21 -0
  533. helm/benchmark/scenarios/test_czech_bank_qa_scenario.py +18 -0
  534. helm/benchmark/scenarios/test_enem_challenge_scenario.py +53 -0
  535. helm/benchmark/scenarios/test_ewok_scenario.py +29 -0
  536. helm/benchmark/scenarios/test_exams_multilingual_scenario.py +29 -0
  537. helm/benchmark/scenarios/test_financebench_scenario.py +26 -0
  538. helm/benchmark/scenarios/test_gold_commodity_news_scenario.py +18 -0
  539. helm/benchmark/scenarios/test_gpqa_scenario.py +44 -0
  540. helm/benchmark/scenarios/test_gsm_scenario.py +31 -0
  541. helm/benchmark/scenarios/test_healtha_br_scenario.py +57 -0
  542. helm/benchmark/scenarios/test_ifeval_scenario.py +36 -0
  543. helm/benchmark/scenarios/test_imdb_ptbr_scenario.py +27 -0
  544. helm/benchmark/scenarios/test_infinite_bench_en_qa_scenario.py +18 -0
  545. helm/benchmark/scenarios/test_infinite_bench_en_sum_scenario.py +31 -0
  546. helm/benchmark/scenarios/test_legalbench_scenario.py +30 -0
  547. helm/benchmark/scenarios/test_math_scenario.py +4 -3
  548. helm/benchmark/scenarios/test_med_qa_scenario.py +30 -0
  549. helm/benchmark/scenarios/test_mmlu_clinical_afr_scenario.py +21 -0
  550. helm/benchmark/scenarios/test_mmlu_pro_scenario.py +53 -0
  551. helm/benchmark/scenarios/test_mmlu_scenario.py +33 -0
  552. helm/benchmark/scenarios/test_narrativeqa_scenario.py +73 -0
  553. helm/benchmark/scenarios/test_oab_exams_scenario.py +51 -0
  554. helm/benchmark/scenarios/test_omni_math_scenario.py +27 -0
  555. helm/benchmark/scenarios/test_scenario.py +6 -3
  556. helm/benchmark/scenarios/test_simple_scenarios.py +50 -0
  557. helm/benchmark/scenarios/test_tweetsentbr_scenario.py +24 -0
  558. helm/benchmark/scenarios/test_wildbench_scenario.py +15 -0
  559. helm/benchmark/scenarios/test_winogrande_afr_scenario.py +19 -0
  560. helm/benchmark/scenarios/thai_exam_scenario.py +239 -0
  561. helm/benchmark/scenarios/the_pile_scenario.py +13 -1
  562. helm/benchmark/scenarios/truthful_qa_scenario.py +26 -2
  563. helm/benchmark/scenarios/tweetsentbr_scenario.py +66 -0
  564. helm/benchmark/scenarios/twitter_aae_scenario.py +20 -1
  565. helm/benchmark/scenarios/unitxt_scenario.py +62 -0
  566. helm/benchmark/scenarios/verifiability_judgment_scenario.py +4 -2
  567. helm/benchmark/scenarios/vicuna_scenario.py +22 -2
  568. helm/benchmark/scenarios/vision_language/a_okvqa_scenario.py +83 -0
  569. helm/benchmark/scenarios/vision_language/bingo_scenario.py +103 -0
  570. helm/benchmark/scenarios/vision_language/blink_scenario.py +140 -0
  571. helm/benchmark/scenarios/vision_language/crossmodal_3600_scenario.py +135 -0
  572. helm/benchmark/scenarios/vision_language/exams_v_scenario.py +104 -0
  573. helm/benchmark/scenarios/vision_language/fair_face_scenario.py +136 -0
  574. helm/benchmark/scenarios/vision_language/flickr30k_scenario.py +74 -0
  575. helm/benchmark/scenarios/vision_language/gqa_scenario.py +91 -0
  576. helm/benchmark/scenarios/vision_language/hateful_memes_scenario.py +94 -0
  577. helm/benchmark/scenarios/vision_language/heim_human_eval_scenario.py +113 -0
  578. helm/benchmark/scenarios/vision_language/image2struct/__init__.py +0 -0
  579. helm/benchmark/scenarios/vision_language/image2struct/chart2csv_scenario.py +55 -0
  580. helm/benchmark/scenarios/vision_language/image2struct/image2struct_scenario.py +225 -0
  581. helm/benchmark/scenarios/vision_language/image2struct/latex_scenario.py +21 -0
  582. helm/benchmark/scenarios/vision_language/image2struct/musicsheet_scenario.py +16 -0
  583. helm/benchmark/scenarios/vision_language/image2struct/utils_latex.py +339 -0
  584. helm/benchmark/scenarios/vision_language/image2struct/webpage/__init__.py +0 -0
  585. helm/benchmark/scenarios/vision_language/image2struct/webpage/driver.py +84 -0
  586. helm/benchmark/scenarios/vision_language/image2struct/webpage/jekyll_server.py +182 -0
  587. helm/benchmark/scenarios/vision_language/image2struct/webpage/utils.py +31 -0
  588. helm/benchmark/scenarios/vision_language/image2struct/webpage_scenario.py +256 -0
  589. helm/benchmark/scenarios/vision_language/math_vista_scenario.py +117 -0
  590. helm/benchmark/scenarios/vision_language/mementos_scenario.py +124 -0
  591. helm/benchmark/scenarios/vision_language/mm_safety_bench_scenario.py +103 -0
  592. helm/benchmark/scenarios/vision_language/mm_star_scenario.py +95 -0
  593. helm/benchmark/scenarios/vision_language/mme_scenario.py +148 -0
  594. helm/benchmark/scenarios/vision_language/mmmu_scenario.py +187 -0
  595. helm/benchmark/scenarios/vision_language/mscoco_captioning_scenario.py +92 -0
  596. helm/benchmark/scenarios/vision_language/mscoco_categorization_scenario.py +117 -0
  597. helm/benchmark/scenarios/vision_language/msr_vtt_scenario.py +75 -0
  598. helm/benchmark/scenarios/vision_language/multipanelvqa_scenario.py +169 -0
  599. helm/benchmark/scenarios/vision_language/originality_scenario.py +35 -0
  600. helm/benchmark/scenarios/vision_language/pairs_scenario.py +247 -0
  601. helm/benchmark/scenarios/vision_language/pope_scenario.py +105 -0
  602. helm/benchmark/scenarios/vision_language/real_world_qa_scenario.py +57 -0
  603. helm/benchmark/scenarios/vision_language/seed_bench_scenario.py +131 -0
  604. helm/benchmark/scenarios/vision_language/unicorn_scenario.py +108 -0
  605. helm/benchmark/scenarios/vision_language/vibe_eval_scenario.py +98 -0
  606. helm/benchmark/scenarios/vision_language/viz_wiz_scenario.py +4 -5
  607. helm/benchmark/scenarios/vision_language/vqa_rad_scenario.py +88 -0
  608. helm/benchmark/scenarios/vision_language/vqa_scenario.py +8 -4
  609. helm/benchmark/scenarios/wikifact_scenario.py +31 -1
  610. helm/benchmark/scenarios/wikitext_103_scenario.py +1 -1
  611. helm/benchmark/scenarios/wildbench_scenario.py +101 -0
  612. helm/benchmark/scenarios/winogrande_afr_scenario.py +78 -0
  613. helm/benchmark/scenarios/wmt_14_scenario.py +33 -2
  614. helm/benchmark/scenarios/xstest_scenario.py +35 -0
  615. helm/benchmark/server.py +32 -2
  616. helm/benchmark/slurm_jobs.py +1 -2
  617. helm/benchmark/slurm_runner.py +78 -50
  618. helm/benchmark/static/schema_air_bench.yaml +3149 -0
  619. helm/benchmark/static/schema_arabic.yaml +271 -0
  620. helm/benchmark/static/schema_audio.yaml +763 -0
  621. helm/benchmark/static/schema_autobencher.yaml +150 -0
  622. helm/benchmark/static/schema_call_center.yaml +269 -0
  623. helm/benchmark/static/schema_capabilities.yaml +254 -0
  624. helm/benchmark/static/schema_classic.yaml +259 -1140
  625. helm/benchmark/static/schema_cleva.yaml +768 -0
  626. helm/benchmark/static/schema_czech_bank.yaml +148 -0
  627. helm/benchmark/static/schema_decodingtrust.yaml +444 -0
  628. helm/benchmark/static/schema_enem_challenge.yaml +146 -0
  629. helm/benchmark/static/schema_enterprise.yaml +319 -0
  630. helm/benchmark/static/schema_ewok.yaml +367 -0
  631. helm/benchmark/static/schema_finance.yaml +191 -0
  632. helm/benchmark/static/schema_heim.yaml +1389 -0
  633. helm/benchmark/static/schema_image2struct.yaml +588 -0
  634. helm/benchmark/static/schema_instruction_following.yaml +161 -0
  635. helm/benchmark/static/schema_legal.yaml +566 -0
  636. helm/benchmark/static/schema_lite.yaml +3 -286
  637. helm/benchmark/static/schema_long_context.yaml +282 -0
  638. helm/benchmark/static/schema_medhelm.yaml +1176 -0
  639. helm/benchmark/static/schema_melt.yaml +1257 -0
  640. helm/benchmark/static/schema_mmlu.yaml +1449 -0
  641. helm/benchmark/static/schema_mmlu_winogrande_afr.yaml +1045 -0
  642. helm/benchmark/static/schema_safety.yaml +283 -0
  643. helm/benchmark/static/schema_seahelm.yaml +723 -0
  644. helm/benchmark/static/schema_slp.yaml +219 -0
  645. helm/benchmark/static/schema_slphelm.yaml +162 -0
  646. helm/benchmark/static/schema_social_audio.yaml +224 -0
  647. helm/benchmark/static/schema_sql.yaml +171 -0
  648. helm/benchmark/static/schema_thai.yaml +244 -0
  649. helm/benchmark/static/schema_torr.yaml +474 -0
  650. helm/benchmark/static/schema_tweetsentbr.yaml +146 -0
  651. helm/benchmark/static/schema_unitxt.yaml +370 -0
  652. helm/benchmark/static/schema_vhelm.yaml +933 -0
  653. helm/benchmark/static/schema_vhelm_lite.yaml +109 -0
  654. helm/benchmark/static/schema_video.yaml +219 -0
  655. helm/benchmark/static_build/assets/air-overview-DpBbyagA.png +0 -0
  656. helm/benchmark/static_build/assets/audio-table-Dn5NMMeJ.png +0 -0
  657. helm/benchmark/static_build/assets/heim-logo-BJtQlEbV.png +0 -0
  658. helm/benchmark/static_build/assets/helm-safety-COfndXuS.png +0 -0
  659. helm/benchmark/static_build/assets/helmhero-D9TvmJsp.png +0 -0
  660. helm/benchmark/static_build/assets/index-oIeiQW2g.css +1 -0
  661. helm/benchmark/static_build/assets/index-qOFpOyHb.js +10 -0
  662. helm/benchmark/static_build/assets/medhelm-overview-CND0EIsy.png +0 -0
  663. helm/benchmark/static_build/assets/medhelm-v1-overview-Cu2tphBB.png +0 -0
  664. helm/benchmark/static_build/assets/overview-BwypNWnk.png +0 -0
  665. helm/benchmark/static_build/assets/process-flow-DWDJC733.png +0 -0
  666. helm/benchmark/static_build/assets/react-BteFIppM.js +85 -0
  667. helm/benchmark/static_build/assets/recharts-DxuQtTOs.js +97 -0
  668. helm/benchmark/static_build/assets/tremor-DR4fE7ko.js +10 -0
  669. helm/benchmark/static_build/assets/vhelm-aspects-NiDQofvP.png +0 -0
  670. helm/benchmark/static_build/assets/vhelm-framework-NxJE4fdA.png +0 -0
  671. helm/benchmark/static_build/assets/vhelm-model-ypCL5Yvq.png +0 -0
  672. helm/benchmark/static_build/config.js +4 -0
  673. helm/benchmark/static_build/index.html +19 -0
  674. helm/benchmark/test_data_preprocessor.py +3 -3
  675. helm/benchmark/test_run_expander.py +1 -1
  676. helm/benchmark/window_services/default_window_service.py +3 -45
  677. helm/benchmark/window_services/encoder_decoder_window_service.py +4 -15
  678. helm/benchmark/window_services/ice_window_service.py +1 -35
  679. helm/benchmark/window_services/image_generation/__init__.py +0 -0
  680. helm/benchmark/window_services/image_generation/clip_window_service.py +13 -0
  681. helm/benchmark/window_services/image_generation/lexica_search_window_service.py +9 -0
  682. helm/benchmark/window_services/image_generation/openai_dalle_window_service.py +9 -0
  683. helm/benchmark/window_services/image_generation/test_clip_window_service.py +29 -0
  684. helm/benchmark/window_services/image_generation/test_openai_dalle_window_service.py +30 -0
  685. helm/benchmark/window_services/local_window_service.py +22 -5
  686. helm/benchmark/window_services/test_anthropic_window_service.py +5 -4
  687. helm/benchmark/window_services/test_bloom_window_service.py +5 -4
  688. helm/benchmark/window_services/test_flan_t5_window_service.py +2 -1
  689. helm/benchmark/window_services/test_gpt2_window_service.py +9 -4
  690. helm/benchmark/window_services/test_gpt4_window_service.py +10 -4
  691. helm/benchmark/window_services/test_gptj_window_service.py +11 -5
  692. helm/benchmark/window_services/test_gptneox_window_service.py +6 -5
  693. helm/benchmark/window_services/test_openai_window_service.py +18 -12
  694. helm/benchmark/window_services/test_opt_window_service.py +6 -5
  695. helm/benchmark/window_services/test_palmyra_window_service.py +5 -4
  696. helm/benchmark/window_services/test_t0pp_window_service.py +5 -4
  697. helm/benchmark/window_services/test_t511b_window_service.py +5 -4
  698. helm/benchmark/window_services/test_ul2_window_service.py +5 -4
  699. helm/benchmark/window_services/test_utils.py +6 -6
  700. helm/benchmark/window_services/test_yalm_window_service.py +5 -4
  701. helm/benchmark/window_services/tokenizer_service.py +7 -13
  702. helm/benchmark/window_services/window_service.py +42 -0
  703. helm/benchmark/window_services/window_service_factory.py +4 -1
  704. helm/benchmark/window_services/yalm_window_service.py +1 -28
  705. helm/clients/__init__.py +0 -0
  706. helm/{proxy/clients → clients}/ai21_client.py +78 -12
  707. helm/clients/aleph_alpha_client.py +114 -0
  708. helm/{proxy/clients → clients}/anthropic_client.py +304 -21
  709. helm/clients/audio_language/__init__.py +0 -0
  710. helm/clients/audio_language/diva_llama_client.py +122 -0
  711. helm/clients/audio_language/llama_omni/arguments.py +61 -0
  712. helm/clients/audio_language/llama_omni/constants.py +9 -0
  713. helm/clients/audio_language/llama_omni/conversation.py +213 -0
  714. helm/clients/audio_language/llama_omni/model/__init__.py +0 -0
  715. helm/clients/audio_language/llama_omni/model/builder.py +88 -0
  716. helm/clients/audio_language/llama_omni/model/language_model/omni_speech2s_llama.py +190 -0
  717. helm/clients/audio_language/llama_omni/model/language_model/omni_speech_llama.py +118 -0
  718. helm/clients/audio_language/llama_omni/model/omni_speech_arch.py +249 -0
  719. helm/clients/audio_language/llama_omni/model/speech_encoder/builder.py +9 -0
  720. helm/clients/audio_language/llama_omni/model/speech_encoder/speech_encoder.py +27 -0
  721. helm/clients/audio_language/llama_omni/model/speech_generator/builder.py +9 -0
  722. helm/clients/audio_language/llama_omni/model/speech_generator/generation.py +622 -0
  723. helm/clients/audio_language/llama_omni/model/speech_generator/speech_generator.py +104 -0
  724. helm/clients/audio_language/llama_omni/model/speech_projector/builder.py +9 -0
  725. helm/clients/audio_language/llama_omni/model/speech_projector/speech_projector.py +27 -0
  726. helm/clients/audio_language/llama_omni/preprocess.py +295 -0
  727. helm/clients/audio_language/llama_omni/utils.py +202 -0
  728. helm/clients/audio_language/llama_omni_client.py +199 -0
  729. helm/clients/audio_language/qwen2_5_omni_client.py +210 -0
  730. helm/clients/audio_language/qwen2_audiolm_client.py +191 -0
  731. helm/clients/audio_language/qwen_audiolm_client.py +153 -0
  732. helm/clients/audio_language/qwen_omni/configuration_qwen2_5_omni.py +519 -0
  733. helm/clients/audio_language/qwen_omni/modeling_qwen2_5_omni.py +4308 -0
  734. helm/clients/audio_language/qwen_omni/processing_qwen2_5_omni.py +270 -0
  735. helm/clients/audio_language/qwen_omni/qwen2_5_omni_utils/__init__.py +0 -0
  736. helm/clients/audio_language/qwen_omni/qwen2_5_omni_utils/v2_5/__init__.py +8 -0
  737. helm/clients/audio_language/qwen_omni/qwen2_5_omni_utils/v2_5/audio_process.py +56 -0
  738. helm/clients/audio_language/qwen_omni/qwen2_5_omni_utils/v2_5/vision_process.py +380 -0
  739. helm/clients/audio_language/test.py +62 -0
  740. helm/{proxy/clients → clients}/auto_client.py +72 -31
  741. helm/clients/azure_openai_client.py +55 -0
  742. helm/clients/bedrock_client.py +381 -0
  743. helm/clients/bedrock_utils.py +105 -0
  744. helm/{proxy/clients → clients}/client.py +92 -17
  745. helm/clients/clip_score_client.py +49 -0
  746. helm/clients/clip_scorers/__init__.py +0 -0
  747. helm/clients/clip_scorers/base_clip_scorer.py +18 -0
  748. helm/clients/clip_scorers/clip_scorer.py +50 -0
  749. helm/clients/clip_scorers/multilingual_clip_scorer.py +50 -0
  750. helm/{proxy/clients → clients}/cohere_client.py +105 -14
  751. helm/clients/dspy_client.py +135 -0
  752. helm/clients/gcs_client.py +82 -0
  753. helm/{proxy/clients → clients}/google_client.py +8 -6
  754. helm/clients/google_translate_client.py +35 -0
  755. helm/clients/grok_client.py +36 -0
  756. helm/{proxy/clients → clients}/http_model_client.py +8 -8
  757. helm/{proxy/clients → clients}/huggingface_client.py +157 -86
  758. helm/clients/huggingface_pipeline_client.py +138 -0
  759. helm/clients/ibm_client.py +269 -0
  760. helm/clients/image_generation/__init__.py +0 -0
  761. helm/clients/image_generation/adobe_vision_client.py +80 -0
  762. helm/clients/image_generation/aleph_alpha_image_generation_client.py +100 -0
  763. helm/clients/image_generation/cogview2/__init__.py +0 -0
  764. helm/clients/image_generation/cogview2/coglm_strategy.py +96 -0
  765. helm/clients/image_generation/cogview2/coglm_utils.py +82 -0
  766. helm/clients/image_generation/cogview2/sr_pipeline/__init__.py +15 -0
  767. helm/clients/image_generation/cogview2/sr_pipeline/direct_sr.py +99 -0
  768. helm/clients/image_generation/cogview2/sr_pipeline/dsr_model.py +254 -0
  769. helm/clients/image_generation/cogview2/sr_pipeline/dsr_sampling.py +190 -0
  770. helm/clients/image_generation/cogview2/sr_pipeline/iterative_sr.py +144 -0
  771. helm/clients/image_generation/cogview2/sr_pipeline/itersr_model.py +269 -0
  772. helm/clients/image_generation/cogview2/sr_pipeline/itersr_sampling.py +120 -0
  773. helm/clients/image_generation/cogview2/sr_pipeline/sr_group.py +42 -0
  774. helm/clients/image_generation/cogview2_client.py +192 -0
  775. helm/clients/image_generation/dalle2_client.py +194 -0
  776. helm/clients/image_generation/dalle3_client.py +108 -0
  777. helm/clients/image_generation/dalle_mini/__init__.py +3 -0
  778. helm/clients/image_generation/dalle_mini/data.py +442 -0
  779. helm/clients/image_generation/dalle_mini/model/__init__.py +5 -0
  780. helm/clients/image_generation/dalle_mini/model/configuration.py +175 -0
  781. helm/clients/image_generation/dalle_mini/model/modeling.py +1834 -0
  782. helm/clients/image_generation/dalle_mini/model/partitions.py +84 -0
  783. helm/clients/image_generation/dalle_mini/model/processor.py +63 -0
  784. helm/clients/image_generation/dalle_mini/model/text.py +251 -0
  785. helm/clients/image_generation/dalle_mini/model/tokenizer.py +9 -0
  786. helm/clients/image_generation/dalle_mini/model/utils.py +29 -0
  787. helm/clients/image_generation/dalle_mini/vqgan_jax/__init__.py +1 -0
  788. helm/clients/image_generation/dalle_mini/vqgan_jax/configuration_vqgan.py +40 -0
  789. helm/clients/image_generation/dalle_mini/vqgan_jax/convert_pt_model_to_jax.py +107 -0
  790. helm/clients/image_generation/dalle_mini/vqgan_jax/modeling_flax_vqgan.py +610 -0
  791. helm/clients/image_generation/dalle_mini_client.py +191 -0
  792. helm/clients/image_generation/deep_floyd_client.py +80 -0
  793. helm/clients/image_generation/huggingface_diffusers_client.py +250 -0
  794. helm/clients/image_generation/image_generation_client_utils.py +9 -0
  795. helm/clients/image_generation/lexica_client.py +88 -0
  796. helm/clients/image_generation/mindalle/__init__.py +0 -0
  797. helm/clients/image_generation/mindalle/models/__init__.py +216 -0
  798. helm/clients/image_generation/mindalle/models/stage1/__init__.py +0 -0
  799. helm/clients/image_generation/mindalle/models/stage1/layers.py +312 -0
  800. helm/clients/image_generation/mindalle/models/stage1/vqgan.py +103 -0
  801. helm/clients/image_generation/mindalle/models/stage2/__init__.py +0 -0
  802. helm/clients/image_generation/mindalle/models/stage2/layers.py +144 -0
  803. helm/clients/image_generation/mindalle/models/stage2/transformer.py +268 -0
  804. helm/clients/image_generation/mindalle/models/tokenizer.py +30 -0
  805. helm/clients/image_generation/mindalle/utils/__init__.py +3 -0
  806. helm/clients/image_generation/mindalle/utils/config.py +129 -0
  807. helm/clients/image_generation/mindalle/utils/sampling.py +149 -0
  808. helm/clients/image_generation/mindalle/utils/utils.py +89 -0
  809. helm/clients/image_generation/mindalle_client.py +116 -0
  810. helm/clients/image_generation/nudity_check_client.py +64 -0
  811. helm/clients/image_generation/together_image_generation_client.py +113 -0
  812. helm/{proxy/clients → clients}/lit_gpt_client.py +6 -6
  813. helm/{proxy/clients → clients}/megatron_client.py +7 -5
  814. helm/clients/mistral_client.py +180 -0
  815. helm/clients/moderation_api_client.py +111 -0
  816. helm/clients/nvidia_nim_client.py +32 -0
  817. helm/clients/open_lm_client.py +43 -0
  818. helm/clients/openai_client.py +604 -0
  819. helm/clients/openai_responses_client.py +200 -0
  820. helm/clients/openrouter_client.py +31 -0
  821. helm/{proxy/clients → clients}/palmyra_client.py +31 -14
  822. helm/{proxy/clients → clients}/perspective_api_client.py +18 -14
  823. helm/clients/reka_client.py +190 -0
  824. helm/clients/simple_client.py +64 -0
  825. helm/clients/stanfordhealthcare_azure_openai_client.py +58 -0
  826. helm/clients/stanfordhealthcare_claude_client.py +31 -0
  827. helm/clients/stanfordhealthcare_google_client.py +43 -0
  828. helm/clients/stanfordhealthcare_http_model_client.py +95 -0
  829. helm/clients/stanfordhealthcare_openai_client.py +62 -0
  830. helm/clients/stanfordhealthcare_shc_openai_client.py +42 -0
  831. helm/{proxy/clients → clients}/test_auto_client.py +13 -15
  832. helm/clients/test_client.py +98 -0
  833. helm/{proxy/clients → clients}/test_huggingface_client.py +31 -16
  834. helm/clients/test_openrouter_client.py +69 -0
  835. helm/clients/test_simple_client.py +19 -0
  836. helm/clients/test_together_client.py +184 -0
  837. helm/clients/together_client.py +599 -0
  838. helm/clients/upstage_client.py +23 -0
  839. helm/clients/vertexai_client.py +488 -0
  840. helm/clients/vision_language/__init__.py +0 -0
  841. helm/clients/vision_language/huggingface_vision2seq_client.py +148 -0
  842. helm/clients/vision_language/huggingface_vlm_client.py +114 -0
  843. helm/{proxy/clients → clients}/vision_language/idefics_client.py +61 -51
  844. helm/clients/vision_language/open_flamingo/__init__.py +2 -0
  845. helm/clients/vision_language/open_flamingo/src/__init__.py +0 -0
  846. helm/clients/vision_language/open_flamingo/src/factory.py +147 -0
  847. helm/clients/vision_language/open_flamingo/src/flamingo.py +337 -0
  848. helm/clients/vision_language/open_flamingo/src/flamingo_lm.py +155 -0
  849. helm/clients/vision_language/open_flamingo/src/helpers.py +267 -0
  850. helm/clients/vision_language/open_flamingo/src/utils.py +47 -0
  851. helm/clients/vision_language/open_flamingo_client.py +155 -0
  852. helm/clients/vision_language/paligemma_client.py +147 -0
  853. helm/clients/vision_language/palmyra_vision_client.py +101 -0
  854. helm/clients/vision_language/qwen2_vlm_client.py +189 -0
  855. helm/clients/vision_language/qwen_vlm_client.py +174 -0
  856. helm/clients/vllm_client.py +80 -0
  857. helm/clients/vllm_granite_thinking_client.py +56 -0
  858. helm/clients/writer_client.py +105 -0
  859. helm/clients/yi_client.py +28 -0
  860. helm/common/audio_utils.py +111 -0
  861. helm/common/cache.py +23 -33
  862. helm/common/cache_backend_config.py +47 -0
  863. helm/common/clip_score_request.py +41 -0
  864. helm/common/context.py +80 -0
  865. helm/common/credentials_utils.py +5 -5
  866. helm/common/critique_request.py +10 -2
  867. helm/common/file_caches/__init__.py +0 -0
  868. helm/common/file_caches/file_cache.py +16 -0
  869. helm/common/file_caches/local_file_cache.py +61 -0
  870. helm/common/file_caches/test_local_file_cache.py +25 -0
  871. helm/common/file_upload_request.py +27 -0
  872. helm/common/general.py +10 -3
  873. helm/common/hierarchical_logger.py +124 -12
  874. helm/common/image_generation_parameters.py +25 -0
  875. helm/common/images_utils.py +60 -5
  876. helm/common/key_value_store.py +41 -10
  877. helm/common/local_context.py +140 -0
  878. helm/common/media_object.py +14 -1
  879. helm/common/moderations_api_request.py +71 -0
  880. helm/common/mongo_key_value_store.py +8 -7
  881. helm/common/multimodal_request_utils.py +57 -0
  882. helm/common/nudity_check_request.py +29 -0
  883. helm/common/object_spec.py +23 -8
  884. helm/common/optional_dependencies.py +1 -1
  885. helm/common/reeval_parameters.py +12 -0
  886. helm/common/remote_context.py +61 -0
  887. helm/common/request.py +45 -19
  888. helm/common/response_format.py +18 -0
  889. helm/common/test_cache.py +1 -48
  890. helm/common/test_general.py +10 -0
  891. helm/common/test_logging.py +94 -0
  892. helm/common/test_media_object.py +1 -1
  893. helm/common/tokenization_request.py +1 -10
  894. helm/config/model_deployments.yaml +4713 -1005
  895. helm/config/model_metadata.yaml +4045 -255
  896. helm/config/tokenizer_configs.yaml +1091 -50
  897. helm/proxy/accounts.py +31 -4
  898. helm/proxy/cli.py +6 -4
  899. helm/proxy/critique/mechanical_turk_critique_importer.py +3 -0
  900. helm/proxy/critique/mechanical_turk_utils.py +1 -1
  901. helm/proxy/critique/model_critique_client.py +40 -10
  902. helm/proxy/example_queries.py +33 -28
  903. helm/proxy/retry.py +5 -0
  904. helm/proxy/server.py +82 -18
  905. helm/proxy/services/remote_service.py +32 -7
  906. helm/proxy/services/server_service.py +71 -69
  907. helm/proxy/services/service.py +30 -6
  908. helm/proxy/services/test_remote_service.py +6 -5
  909. helm/proxy/services/test_service.py +1 -13
  910. helm/proxy/static/help.html +99 -0
  911. helm/proxy/static/index.css +61 -0
  912. helm/proxy/static/index.html +40 -0
  913. helm/proxy/static/index.js +462 -0
  914. helm/proxy/test_accounts.py +32 -0
  915. helm/proxy/test_retry.py +1 -1
  916. helm/proxy/token_counters/auto_token_counter.py +37 -37
  917. helm/proxy/token_counters/test_auto_token_counter.py +164 -0
  918. helm/proxy/token_counters/token_counter.py +3 -5
  919. helm/tokenizers/__init__.py +0 -0
  920. helm/tokenizers/ai21_tokenizer.py +52 -0
  921. helm/{proxy/tokenizers → tokenizers}/aleph_alpha_tokenizer.py +1 -1
  922. helm/{proxy/tokenizers → tokenizers}/auto_tokenizer.py +9 -12
  923. helm/{proxy/tokenizers → tokenizers}/caching_tokenizer.py +2 -30
  924. helm/tokenizers/cohere_tokenizer.py +50 -0
  925. helm/tokenizers/grok_tokenizer.py +55 -0
  926. helm/{proxy/tokenizers → tokenizers}/http_model_tokenizer.py +4 -4
  927. helm/{proxy/tokenizers → tokenizers}/huggingface_tokenizer.py +44 -41
  928. helm/{proxy/tokenizers → tokenizers}/lit_gpt_tokenizer.py +1 -1
  929. helm/tokenizers/simple_tokenizer.py +33 -0
  930. helm/tokenizers/test_ai21_tokenizer.py +48 -0
  931. helm/{proxy/tokenizers → tokenizers}/test_anthropic_tokenizer.py +6 -2
  932. helm/tokenizers/test_cohere_tokenizer.py +39 -0
  933. helm/tokenizers/test_grok_tokenizer.py +33 -0
  934. helm/{proxy/tokenizers → tokenizers}/test_huggingface_tokenizer.py +9 -2
  935. helm/tokenizers/test_simple_tokenizer.py +33 -0
  936. helm/{proxy/tokenizers → tokenizers}/test_yalm_tokenizer.py +1 -1
  937. helm/{proxy/tokenizers → tokenizers}/tiktoken_tokenizer.py +1 -1
  938. helm/{proxy/tokenizers → tokenizers}/tokenizer.py +3 -1
  939. helm/{proxy/tokenizers → tokenizers}/vertexai_tokenizer.py +1 -1
  940. helm/{proxy/tokenizers → tokenizers}/yalm_tokenizer.py +8 -6
  941. helm/tokenizers/yalm_tokenizer_data/__init__.py +0 -0
  942. helm/{proxy/tokenizers → tokenizers}/yalm_tokenizer_data/test_yalm_tokenizer.py +1 -1
  943. helm/tokenizers/yalm_tokenizer_data/voc_100b.sp +0 -0
  944. helm/{proxy/tokenizers → tokenizers}/yalm_tokenizer_data/yalm_tokenizer.py +1 -1
  945. crfm_helm-0.4.0.dist-info/METADATA +0 -264
  946. crfm_helm-0.4.0.dist-info/RECORD +0 -397
  947. helm/benchmark/data_overlap/data_overlap_spec.py +0 -86
  948. helm/benchmark/data_overlap/export_scenario_text.py +0 -119
  949. helm/benchmark/data_overlap/light_scenario.py +0 -60
  950. helm/benchmark/metrics/numeracy_metrics.py +0 -72
  951. helm/benchmark/metrics/test_numeracy_metrics.py +0 -95
  952. helm/benchmark/run_specs.py +0 -2762
  953. helm/benchmark/scenarios/numeracy_scenario.py +0 -784
  954. helm/benchmark/static/benchmarking.css +0 -156
  955. helm/benchmark/static/benchmarking.js +0 -1705
  956. helm/benchmark/static/config.js +0 -3
  957. helm/benchmark/static/images/helm-logo.png +0 -0
  958. helm/benchmark/static/images/language-model-helm.png +0 -0
  959. helm/benchmark/static/images/organizations/ai21.png +0 -0
  960. helm/benchmark/static/images/organizations/anthropic.png +0 -0
  961. helm/benchmark/static/images/organizations/bigscience.png +0 -0
  962. helm/benchmark/static/images/organizations/cohere.png +0 -0
  963. helm/benchmark/static/images/organizations/eleutherai.png +0 -0
  964. helm/benchmark/static/images/organizations/google.png +0 -0
  965. helm/benchmark/static/images/organizations/meta.png +0 -0
  966. helm/benchmark/static/images/organizations/microsoft.png +0 -0
  967. helm/benchmark/static/images/organizations/nvidia.png +0 -0
  968. helm/benchmark/static/images/organizations/openai.png +0 -0
  969. helm/benchmark/static/images/organizations/together.png +0 -0
  970. helm/benchmark/static/images/organizations/tsinghua-keg.png +0 -0
  971. helm/benchmark/static/images/organizations/yandex.png +0 -0
  972. helm/benchmark/static/images/scenarios-by-metrics.png +0 -0
  973. helm/benchmark/static/images/taxonomy-scenarios.png +0 -0
  974. helm/benchmark/static/index.html +0 -68
  975. helm/benchmark/static/json-urls.js +0 -69
  976. helm/benchmark/static/plot-captions.js +0 -27
  977. helm/benchmark/static/utils.js +0 -285
  978. helm/benchmark/test_model_deployment_definition.py +0 -92
  979. helm/benchmark/test_model_properties.py +0 -1570
  980. helm/benchmark/vlm_run_specs.py +0 -97
  981. helm/benchmark/window_services/ai21_window_service.py +0 -258
  982. helm/benchmark/window_services/cohere_window_service.py +0 -163
  983. helm/benchmark/window_services/flan_t5_window_service.py +0 -29
  984. helm/benchmark/window_services/gpt2_window_service.py +0 -32
  985. helm/benchmark/window_services/huggingface_window_service.py +0 -60
  986. helm/benchmark/window_services/t0pp_window_service.py +0 -35
  987. helm/benchmark/window_services/t511b_window_service.py +0 -30
  988. helm/benchmark/window_services/test_ai21_window_service.py +0 -163
  989. helm/benchmark/window_services/test_cohere_window_service.py +0 -74
  990. helm/benchmark/window_services/test_cohere_window_service_utils.py +0 -8328
  991. helm/benchmark/window_services/test_ice_window_service.py +0 -326
  992. helm/benchmark/window_services/test_mt_nlg_window_service.py +0 -48
  993. helm/benchmark/window_services/ul2_window_service.py +0 -30
  994. helm/benchmark/window_services/wider_ai21_window_service.py +0 -24
  995. helm/common/cache_utils.py +0 -14
  996. helm/proxy/clients/aleph_alpha_client.py +0 -95
  997. helm/proxy/clients/goose_ai_client.py +0 -99
  998. helm/proxy/clients/microsoft_client.py +0 -180
  999. helm/proxy/clients/openai_client.py +0 -206
  1000. helm/proxy/clients/simple_client.py +0 -60
  1001. helm/proxy/clients/test_client.py +0 -49
  1002. helm/proxy/clients/test_together_client.py +0 -97
  1003. helm/proxy/clients/together_client.py +0 -334
  1004. helm/proxy/clients/vertexai_client.py +0 -115
  1005. helm/proxy/token_counters/ai21_token_counter.py +0 -20
  1006. helm/proxy/token_counters/cohere_token_counter.py +0 -13
  1007. helm/proxy/token_counters/free_token_counter.py +0 -12
  1008. helm/proxy/token_counters/gooseai_token_counter.py +0 -24
  1009. helm/proxy/token_counters/openai_token_counter.py +0 -22
  1010. helm/proxy/token_counters/test_ai21_token_counter.py +0 -88
  1011. helm/proxy/token_counters/test_openai_token_counter.py +0 -81
  1012. helm/proxy/tokenizers/ai21_tokenizer.py +0 -60
  1013. helm/proxy/tokenizers/anthropic_tokenizer.py +0 -52
  1014. helm/proxy/tokenizers/cohere_tokenizer.py +0 -83
  1015. helm/proxy/tokenizers/ice_tokenizer.py +0 -30
  1016. helm/proxy/tokenizers/simple_tokenizer.py +0 -32
  1017. helm/proxy/tokenizers/test_ice_tokenizer.py +0 -57
  1018. {crfm_helm-0.4.0.dist-info → crfm_helm-0.5.10.dist-info}/entry_points.txt +0 -0
  1019. {crfm_helm-0.4.0.dist-info → crfm_helm-0.5.10.dist-info/licenses}/LICENSE +0 -0
  1020. {crfm_helm-0.4.0.dist-info → crfm_helm-0.5.10.dist-info}/top_level.txt +0 -0
  1021. /helm/benchmark/{data_overlap → annotation}/__init__.py +0 -0
  1022. /helm/{proxy/clients → benchmark/annotation/image2struct}/__init__.py +0 -0
  1023. /helm/{proxy/clients/vision_language → benchmark/metrics/ifeval}/__init__.py +0 -0
  1024. /helm/{proxy/tokenizers → benchmark/metrics/image_generation}/__init__.py +0 -0
  1025. /helm/{proxy/tokenizers/yalm_tokenizer_data → benchmark/metrics/image_generation/detectors}/__init__.py +0 -0
  1026. /helm/benchmark/{static/images/crfm-logo.png → static_build/assets/crfm-logo-Du4T1uWZ.png} +0 -0
  1027. /helm/benchmark/{static/images/helm-logo-simple.png → static_build/assets/helm-logo-simple-DzOhNN41.png} +0 -0
  1028. /helm/{proxy/clients → clients}/ai21_utils.py +0 -0
  1029. /helm/{proxy/clients → clients}/cohere_utils.py +0 -0
  1030. /helm/{proxy/clients → clients}/lit_gpt_generate.py +0 -0
  1031. /helm/{proxy/clients → clients}/toxicity_classifier_client.py +0 -0
  1032. /helm/{benchmark → proxy}/static/general.js +0 -0
  1033. /helm/{benchmark → proxy}/static/info-icon.png +0 -0
helm/config/tokenizer_configs.yaml CHANGED
@@ -10,129 +10,273 @@
10
10
 
11
11
  tokenizer_configs:
12
12
 
13
- - name: simple/model1
13
+ - name: simple/tokenizer1
14
14
  tokenizer_spec:
15
- class_name: "helm.proxy.tokenizers.simple_tokenizer.SimpleTokenizer"
15
+ class_name: "helm.tokenizers.simple_tokenizer.SimpleTokenizer"
16
16
  end_of_text_token: "</s>"
17
17
  prefix_token: "<s>"
18
18
 
19
19
  # AI21
20
- - name: ai21/j1
20
+ - name: ai21/j2-tokenizer
21
21
  tokenizer_spec:
22
- class_name: "helm.proxy.tokenizers.ai21_tokenizer.AI21Tokenizer"
23
- end_of_text_token: " "
24
- prefix_token: ""
22
+ class_name: "helm.tokenizers.ai21_tokenizer.AI21LocalTokenizer"
23
+ end_of_text_token: "<|endoftext|>"
24
+ prefix_token: "<|startoftext|>"
25
+ - name: ai21/jamba-tokenizer
26
+ tokenizer_spec:
27
+ class_name: "helm.tokenizers.ai21_tokenizer.AI21LocalTokenizer"
28
+ end_of_text_token: "<|endoftext|>"
29
+ prefix_token: "<|startoftext|>"
30
+ - name: ai21/jamba-instruct-tokenizer
31
+ tokenizer_spec:
32
+ class_name: "helm.tokenizers.ai21_tokenizer.AI21LocalTokenizer"
33
+ end_of_text_token: "<|endoftext|>"
34
+ prefix_token: "<|startoftext|>"
35
+ - name: ai21/jamba-1.5-mini-tokenizer
36
+ tokenizer_spec:
37
+ class_name: "helm.tokenizers.ai21_tokenizer.AI21LocalTokenizer"
38
+ end_of_text_token: "<|endoftext|>"
39
+ prefix_token: "<|startoftext|>"
40
+ - name: ai21/jamba-1.5-large-tokenizer
41
+ tokenizer_spec:
42
+ class_name: "helm.tokenizers.ai21_tokenizer.AI21LocalTokenizer"
43
+ end_of_text_token: "<|endoftext|>"
44
+ prefix_token: "<|startoftext|>"
25
45
 
26
46
  # AlephAlpha
27
47
  - name: AlephAlpha/luminous-base
28
48
  tokenizer_spec:
29
- class_name: "helm.proxy.tokenizers.aleph_alpha_tokenizer.AlephAlphaTokenizer"
49
+ class_name: "helm.tokenizers.aleph_alpha_tokenizer.AlephAlphaTokenizer"
30
50
  end_of_text_token: ""
31
51
  prefix_token: ""
32
52
  - name: AlephAlpha/luminous-extended
33
53
  tokenizer_spec:
34
- class_name: "helm.proxy.tokenizers.aleph_alpha_tokenizer.AlephAlphaTokenizer"
54
+ class_name: "helm.tokenizers.aleph_alpha_tokenizer.AlephAlphaTokenizer"
35
55
  end_of_text_token: ""
36
56
  prefix_token: ""
37
57
  - name: AlephAlpha/luminous-supreme
38
58
  tokenizer_spec:
39
- class_name: "helm.proxy.tokenizers.aleph_alpha_tokenizer.AlephAlphaTokenizer"
59
+ class_name: "helm.tokenizers.aleph_alpha_tokenizer.AlephAlphaTokenizer"
40
60
  end_of_text_token: ""
41
61
  prefix_token: ""
42
62
  - name: AlephAlpha/luminous-world
43
63
  tokenizer_spec:
44
- class_name: "helm.proxy.tokenizers.aleph_alpha_tokenizer.AlephAlphaTokenizer"
64
+ class_name: "helm.tokenizers.aleph_alpha_tokenizer.AlephAlphaTokenizer"
45
65
  end_of_text_token: ""
46
66
  prefix_token: ""
47
67
 
68
+ # Alibaba DAMO Academy
69
+
70
+ - name: damo/seallm-7b-v2
71
+ tokenizer_spec:
72
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
73
+ args:
74
+ pretrained_model_name_or_path: SeaLLMs/SeaLLM-7B-v2
75
+ end_of_text_token: "</s>"
76
+ prefix_token: "<s>"
77
+
78
+ - name: damo/seallm-7b-v2.5
79
+ tokenizer_spec:
80
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
81
+ args:
82
+ pretrained_model_name_or_path: SeaLLMs/SeaLLM-7B-v2.5
83
+ end_of_text_token: "<eos>"
84
+ prefix_token: "<bos>"
85
+
48
86
  # Anthropic
49
87
  - name: anthropic/claude
50
88
  tokenizer_spec:
51
- class_name: "helm.proxy.tokenizers.anthropic_tokenizer.AnthropicTokenizer"
89
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
90
+ args:
91
+ pretrained_model_name_or_path: Xenova/claude-tokenizer
52
92
  end_of_text_token: "<|endoftext|>"
53
93
  prefix_token: "<|endoftext|>"
54
94
 
55
95
  # Bigcode
56
96
  - name: bigcode/santacoder
57
97
  tokenizer_spec:
58
- class_name: "helm.proxy.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
98
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
59
99
  end_of_text_token: "<|endoftext|>"
60
100
  prefix_token: "<|endoftext|>"
61
101
  - name: bigcode/starcoder
62
102
  tokenizer_spec:
63
- class_name: "helm.proxy.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
103
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
64
104
  end_of_text_token: "<|endoftext|>"
65
105
  prefix_token: "<|endoftext|>"
66
106
 
67
107
  # Bigscience
68
108
  - name: bigscience/bloom
69
109
  tokenizer_spec:
70
- class_name: "helm.proxy.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
110
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
71
111
  end_of_text_token: "</s>"
72
- prefix_token: "</s>"
112
+ prefix_token: "<s>"
73
113
  - name: bigscience/T0pp
74
114
  tokenizer_spec:
75
- class_name: "helm.proxy.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
115
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
76
116
  end_of_text_token: "</s>"
77
117
  prefix_token: ""
78
118
 
79
119
  # Cohere
80
- - name: cohere/cohere
120
+ - name: cohere/command
81
121
  tokenizer_spec:
82
- class_name: "helm.proxy.tokenizers.cohere_tokenizer.CohereTokenizer"
83
- end_of_text_token: ""
84
- prefix_token: ":"
122
+ class_name: "helm.tokenizers.cohere_tokenizer.CohereLocalTokenizer"
123
+ end_of_text_token: "<EOS_TOKEN>"
124
+ prefix_token: "<BOS_TOKEN>"
125
+
126
+ - name: cohere/command-light
127
+ tokenizer_spec:
128
+ class_name: "helm.tokenizers.cohere_tokenizer.CohereLocalTokenizer"
129
+ end_of_text_token: "<EOS_TOKEN>"
130
+ prefix_token: "<BOS_TOKEN>"
131
+
132
+ - name: cohere/command-r
133
+ tokenizer_spec:
134
+ class_name: "helm.tokenizers.cohere_tokenizer.CohereLocalTokenizer"
135
+ end_of_text_token: "<EOS_TOKEN>"
136
+ prefix_token: "<BOS_TOKEN>"
137
+
138
+ - name: cohere/command-r-plus
139
+ tokenizer_spec:
140
+ class_name: "helm.tokenizers.cohere_tokenizer.CohereLocalTokenizer"
141
+ end_of_text_token: "<EOS_TOKEN>"
142
+ prefix_token: "<BOS_TOKEN>"
143
+
144
+ - name: cohere/c4ai-command-r-v01
145
+ tokenizer_spec:
146
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
147
+ args:
148
+ pretrained_model_name_or_path: CohereForAI/c4ai-command-r-v01
149
+ end_of_text_token: "<EOS_TOKEN>"
150
+ prefix_token: "<BOS_TOKEN>"
151
+
152
+ - name: cohere/c4ai-command-r-plus
153
+ tokenizer_spec:
154
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
155
+ args:
156
+ pretrained_model_name_or_path: CohereForAI/c4ai-command-r-plus
157
+ end_of_text_token: "<EOS_TOKEN>"
158
+ prefix_token: "<BOS_TOKEN>"
159
+
160
+ # Databricks
161
+ - name: databricks/dbrx-instruct
162
+ tokenizer_spec:
163
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
164
+ end_of_text_token: "<|endoftext|>"
165
+ prefix_token: "<|endoftext|>"
166
+
167
+ # DeepSeek
168
+ - name: deepseek-ai/deepseek-llm-67b-chat
169
+ tokenizer_spec:
170
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
171
+ end_of_text_token: "<｜end▁of▁sentence｜>"
172
+ prefix_token: "<｜begin▁of▁sentence｜>"
173
+
174
+ - name: deepseek-ai/deepseek-v3
175
+ tokenizer_spec:
176
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
177
+ end_of_text_token: "<｜end▁of▁sentence｜>"
178
+ prefix_token: "<｜begin▁of▁sentence｜>"
179
+
180
+ - name: deepseek-ai/deepseek-r1
181
+ tokenizer_spec:
182
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
183
+ end_of_text_token: "<｜end▁of▁sentence｜>"
184
+ prefix_token: "<｜begin▁of▁sentence｜>"
85
185
 
86
186
  # EleutherAI
87
187
  - name: EleutherAI/gpt-j-6B
88
188
  tokenizer_spec:
89
- class_name: "helm.proxy.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
189
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
90
190
  end_of_text_token: "<|endoftext|>"
91
191
  prefix_token: "<|endoftext|>"
192
+
92
193
  - name: EleutherAI/gpt-neox-20b
93
194
  tokenizer_spec:
94
- class_name: "helm.proxy.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
195
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
95
196
  end_of_text_token: "<|endoftext|>"
96
197
  prefix_token: "<|endoftext|>"
97
198
 
98
199
  # Facebook
99
200
  - name: facebook/opt-66b
100
201
  tokenizer_spec:
101
- class_name: "helm.proxy.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
202
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
102
203
  end_of_text_token: "</s>"
103
204
  prefix_token: "</s>"
104
205
 
105
206
  # Google
106
207
  - name: google/t5-11b
107
208
  tokenizer_spec:
108
- class_name: "helm.proxy.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
209
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
210
+ args:
211
+ pretrained_model_name_or_path: google-t5/t5-11b
109
212
  end_of_text_token: "</s>"
110
213
  prefix_token: ""
111
214
  - name: google/flan-t5-xxl
112
215
  tokenizer_spec:
113
- class_name: "helm.proxy.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
216
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
114
217
  end_of_text_token: "</s>"
115
218
  prefix_token: ""
116
219
  - name: google/ul2
117
220
  tokenizer_spec:
118
- class_name: "helm.proxy.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
221
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
119
222
  end_of_text_token: "</s>"
120
223
  prefix_token: ""
121
224
  - name: google/mt5-base
122
225
  tokenizer_spec:
123
- class_name: "helm.proxy.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
226
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
124
227
  end_of_text_token: "</s>"
125
228
  prefix_token: ""
126
229
  - name: google/text-bison@001
127
230
  tokenizer_spec:
128
- class_name: "helm.proxy.tokenizers.vertexai_tokenizer.VertexAITokenizer"
231
+ class_name: "helm.tokenizers.vertexai_tokenizer.VertexAITokenizer"
232
+ end_of_text_token: "</s>"
233
+ prefix_token: ""
234
+ - name: google/text-bison@002
235
+ tokenizer_spec:
236
+ class_name: "helm.tokenizers.vertexai_tokenizer.VertexAITokenizer"
129
237
  end_of_text_token: "</s>"
130
238
  prefix_token: ""
131
239
  - name: google/text-unicorn@001
132
240
  tokenizer_spec:
133
- class_name: "helm.proxy.tokenizers.vertexai_tokenizer.VertexAITokenizer"
241
+ class_name: "helm.tokenizers.vertexai_tokenizer.VertexAITokenizer"
134
242
  end_of_text_token: "</s>"
135
243
  prefix_token: ""
244
+ - name: google/gemma-2b
245
+ tokenizer_spec:
246
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
247
+ end_of_text_token: "<eos>"
248
+ prefix_token: "<bos>"
249
+ - name: google/gemma-2-9b
250
+ tokenizer_spec:
251
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
252
+ end_of_text_token: "<eos>"
253
+ prefix_token: "<bos>"
254
+ - name: google/medgemma-4b-it
255
+ tokenizer_spec:
256
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
257
+ args:
258
+ trust_remote_code: true
259
+ end_of_text_token: "<eos>"
260
+ prefix_token: "<bos>"
261
+
262
+ # Grok
263
+ - name: xai/grok-3-beta
264
+ tokenizer_spec:
265
+ class_name: "helm.tokenizers.grok_tokenizer.GrokAPITokenizer"
266
+ end_of_text_token: ""
267
+ prefix_token: ""
268
+
269
+ - name: xai/grok-3-mini-beta
270
+ tokenizer_spec:
271
+ class_name: "helm.tokenizers.grok_tokenizer.GrokAPITokenizer"
272
+ end_of_text_token: ""
273
+ prefix_token: ""
274
+
275
+ - name: xai/grok-4-0709
276
+ tokenizer_spec:
277
+ class_name: "helm.tokenizers.grok_tokenizer.GrokAPITokenizer"
278
+ end_of_text_token: ""
279
+ prefix_token: ""
136
280
 
137
281
  # Hf-internal-testing
138
282
 
@@ -140,43 +284,67 @@ tokenizer_configs:
140
284
  # https://huggingface.co/docs/transformers/main/en/model_doc/llama#transformers.LlamaTokenizerFast.example
141
285
  - name: hf-internal-testing/llama-tokenizer
142
286
  tokenizer_spec:
143
- class_name: "helm.proxy.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
287
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
144
288
  end_of_text_token: "</s>"
145
289
  prefix_token: "<s>"
146
290
 
147
291
  # HuggingFaceM4
148
292
  - name: HuggingFaceM4/idefics-9b
149
293
  tokenizer_spec:
150
- class_name: "helm.proxy.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
294
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
151
295
  end_of_text_token: "</s>"
152
296
  prefix_token: "<s>"
153
297
  - name: HuggingFaceM4/idefics-9b-instruct
154
298
  tokenizer_spec:
155
- class_name: "helm.proxy.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
299
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
156
300
  end_of_text_token: "</s>"
157
301
  prefix_token: "<s>"
158
302
  - name: HuggingFaceM4/idefics-80b
159
303
  tokenizer_spec:
160
- class_name: "helm.proxy.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
304
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
161
305
  end_of_text_token: "</s>"
162
306
  prefix_token: "<s>"
163
307
  - name: HuggingFaceM4/idefics-80b-instruct
164
308
  tokenizer_spec:
165
- class_name: "helm.proxy.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
309
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
166
310
  end_of_text_token: "</s>"
167
311
  prefix_token: "<s>"
312
+
313
+ - name: anas-awadalla/mpt-7b
314
+ tokenizer_spec:
315
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
316
+ end_of_text_token: "<|endoftext|>"
317
+ prefix_token: ""
168
318
 
169
319
  # Huggingface
170
320
  - name: huggingface/gpt2
171
321
  tokenizer_spec:
172
- class_name: "helm.proxy.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
322
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
323
+ args:
324
+ pretrained_model_name_or_path: openai-community/gpt2
173
325
  end_of_text_token: "<|endoftext|>"
174
326
  prefix_token: "<|endoftext|>"
175
327
 
328
+ - name: huggingface/smollm2-135m
329
+ tokenizer_spec:
330
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
331
+ args:
332
+ pretrained_model_name_or_path: HuggingFaceTB/SmolLM2-135M
333
+ end_of_text_token: "<|endoftext|>"
334
+ prefix_token: "<|endoftext|>"
335
+
336
+ - name: huggingface/smollm2-135m-instruct
337
+ tokenizer_spec:
338
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
339
+ args:
340
+ pretrained_model_name_or_path: HuggingFaceTB/SmolLM2-135M-Instruct
341
+ end_of_text_token: "<|endoftext|>"
342
+ prefix_token: "<|im_end|>"
343
+
176
344
  # Lighting AI
177
345
  - name: lightningai/lit-gpt
178
346
  tokenizer_spec:
179
- class_name: "helm.proxy.tokenizers.lit_gpt_tokenizer.LitGPTTokenizer"
347
+ class_name: "helm.tokenizers.lit_gpt_tokenizer.LitGPTTokenizer"
180
348
  end_of_text_token: "<|endoftext|>"
181
349
  prefix_token: "<|endoftext|>"
182
350
 
@@ -194,70 +362,943 @@ tokenizer_configs:
194
362
  # 'https://huggingface.co/models'
195
363
  - name: meta-llama/Llama-2-7b-hf
196
364
  tokenizer_spec:
197
- class_name: "helm.proxy.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
365
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
198
366
  end_of_text_token: "</s>"
199
367
  prefix_token: "<s>"
200
368
 
369
+ - name: meta/llama-3-8b
370
+ tokenizer_spec:
371
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
372
+ args:
373
+ pretrained_model_name_or_path: meta-llama/Meta-Llama-3-8B
374
+ prefix_token: "<|begin_of_text|>"
375
+ end_of_text_token: "<|end_of_text|>"
376
+
377
+ - name: meta/llama-3-8b-instruct
378
+ tokenizer_spec:
379
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
380
+ args:
381
+ pretrained_model_name_or_path: meta-llama/Meta-Llama-3.1-8B-Instruct
382
+ prefix_token: "<|begin_of_text|>"
383
+ end_of_text_token: "<|eot_id|>"
384
+
385
+ - name: meta/llama-3.1-8b
386
+ tokenizer_spec:
387
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
388
+ args:
389
+ pretrained_model_name_or_path: meta-llama/Meta-Llama-3.1-8B-Instruct
390
+ prefix_token: "<|begin_of_text|>"
391
+ end_of_text_token: "<|end_of_text|>"
392
+
393
+ - name: meta/llama-3.1-8b-instruct
394
+ tokenizer_spec:
395
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
396
+ args:
397
+ pretrained_model_name_or_path: meta-llama/Meta-Llama-3.1-8B-Instruct
398
+ prefix_token: "<|begin_of_text|>"
399
+ end_of_text_token: "<|eot_id|>"
400
+
401
+ - name: meta/llama-3.2-3b-instruct
402
+ tokenizer_spec:
403
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
404
+ args:
405
+ pretrained_model_name_or_path: meta-llama/Llama-3.2-3B-Instruct
406
+ prefix_token: "<|begin_of_text|>"
407
+ end_of_text_token: "<|eot_id|>"
408
+
409
+ - name: meta/llama-3.2-1b-instruct
410
+ tokenizer_spec:
411
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
412
+ args:
413
+ pretrained_model_name_or_path: meta-llama/Llama-3.2-1B-Instruct
414
+ prefix_token: "<|begin_of_text|>"
415
+ end_of_text_token: "<|eot_id|>"
416
+
417
+ - name: meta/llama-3.1-8b-instruct
418
+ tokenizer_spec:
419
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
420
+ args:
421
+ pretrained_model_name_or_path: meta-llama/Llama-3.1-8B-Instruct
422
+ prefix_token: "<|begin_of_text|>"
423
+ end_of_text_token: "<|eot_id|>"
424
+
425
+ - name: meta/llama-3.2-11b-vision-instruct
426
+ tokenizer_spec:
427
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
428
+ args:
429
+ pretrained_model_name_or_path: meta-llama/Llama-3.2-11B-Vision-Instruct
430
+ prefix_token: "<|begin_of_text|>"
431
+ end_of_text_token: "<|eot_id|>"
432
+
433
+ - name: meta/llama-3.3-70b-instruct
434
+ tokenizer_spec:
435
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
436
+ args:
437
+ pretrained_model_name_or_path: meta-llama/Llama-3.3-70B-Instruct
438
+ prefix_token: "<|begin_of_text|>"
439
+ end_of_text_token: "<|eot_id|>"
440
+
441
+ - name: meta/llama-4-scout-17b-16e-instruct
442
+ tokenizer_spec:
443
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
444
+ args:
445
+ pretrained_model_name_or_path: meta-llama/Llama-4-Scout-17B-16E-Instruct
446
+ prefix_token: "<|begin_of_text|>"
447
+ end_of_text_token: "<|end_of_text|>"
201
448
 
202
449
  # 01-ai
203
450
  - name: 01-ai/Yi-6B
204
451
  tokenizer_spec:
205
- class_name: "helm.proxy.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
452
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
206
453
  end_of_text_token: "</s>"
207
454
  prefix_token: "<s>"
208
455
 
209
- # Microsoft
210
- - name: microsoft/gpt2
456
+ # AI Singapore
457
+ - name: aisingapore/sea-lion-7b
458
+ tokenizer_spec:
459
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
460
+ args:
461
+ trust_remote_code: true
462
+ use_fast: false
463
+ end_of_text_token: "<|endoftext|>"
464
+ prefix_token: ""
465
+
466
+
467
+
468
+ # Allen Institute for AI
469
+ # The allenai/olmo-7b requires Python 3.9 or newer.
470
+ # To use the allenai/olmo-7b tokenizer, run `pip install "crfm-helm[allenai]"` first.
471
+ - name: allenai/olmo-7b
472
+ tokenizer_spec:
473
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
474
+ args:
475
+ trust_remote_code: true
476
+ end_of_text_token: "<|endoftext|>"
477
+ prefix_token: ""
478
+
479
+ - name: allenai/OLMo-1.7-7B-hf
480
+ tokenizer_spec:
481
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
482
+ end_of_text_token: "<|endoftext|>"
483
+ prefix_token: ""
484
+
485
+ - name: allenai/olmo-2-1124-7b-instruct
486
+ tokenizer_spec:
487
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
488
+ end_of_text_token: "<|endoftext|>"
489
+ prefix_token: "<|endoftext|>"
490
+
491
+ - name: allenai/olmo-2-0325-32b-instruct
492
+ tokenizer_spec:
493
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
494
+ end_of_text_token: "<|endoftext|>"
495
+ prefix_token: "<|endoftext|>"
496
+
497
+ - name: allenai/olmoe-1b-7b-0125-instruct
211
498
  tokenizer_spec:
212
- class_name: "helm.proxy.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
499
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
500
+ end_of_text_token: "|||IP_ADDRESS|||"
501
+ prefix_token: "|||IP_ADDRESS|||"
502
+
503
+ # Marin Community
504
+ - name: marin-community/marin-8b-instruct
505
+ tokenizer_spec:
506
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
507
+ end_of_text_token: "<|eot_id|>"
508
+ prefix_token: "<|begin_of_text|>"
509
+
510
+ # Microsoft
511
+ - name: microsoft/phi-2
512
+ tokenizer_spec:
513
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
514
+ end_of_text_token: "<|endoftext|>"
515
+ prefix_token: "<|endoftext|>"
516
+
517
+ - name: microsoft/phi-3-small-8k-instruct
518
+ tokenizer_spec:
519
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
520
+ args:
521
+ trust_remote_code: true
213
522
  end_of_text_token: "<|endoftext|>"
214
- prefix_token: "<<"
523
+ prefix_token: "<|endoftext|>"
524
+
525
+ - name: microsoft/phi-3-medium-4k-instruct
526
+ tokenizer_spec:
527
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
528
+ end_of_text_token: "<|endoftext|>"
529
+ prefix_token: "<s>"
530
+
531
+ - name: microsoft/phi-3.5-mini-instruct
532
+ tokenizer_spec:
533
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
534
+ end_of_text_token: "<|endoftext|>"
535
+ prefix_token: "<s>"
536
+
537
+ - name: microsoft/phi-3.5-mini-instruct
538
+ tokenizer_spec:
539
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
540
+ end_of_text_token: "<|endoftext|>"
541
+ prefix_token: "<s>"
215
542
 
216
543
  # Mistralai
217
544
  - name: mistralai/Mistral-7B-v0.1
218
545
  tokenizer_spec:
219
- class_name: "helm.proxy.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
546
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
547
+ end_of_text_token: "</s>"
548
+ prefix_token: "<s>"
549
+
550
+ - name: mistralai/Mistral-7B-Instruct-v0.1
551
+ tokenizer_spec:
552
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
553
+ end_of_text_token: "</s>"
554
+ prefix_token: "<s>"
555
+
556
+ - name: mistralai/Mistral-7B-Instruct-v0.2
557
+ tokenizer_spec:
558
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
559
+ end_of_text_token: "</s>"
560
+ prefix_token: "<s>"
561
+
562
+ - name: mistralai/Mistral-7B-Instruct-v0.3
563
+ tokenizer_spec:
564
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
565
+ end_of_text_token: "</s>"
566
+ prefix_token: "<s>"
567
+
568
+ - name: mistralai/Mistral-Nemo-Base-2407
569
+ tokenizer_spec:
570
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
220
571
  end_of_text_token: "</s>"
221
572
  prefix_token: "<s>"
222
573
 
574
+ - name: mistralai/Mistral-Large-Instruct-2407
575
+ tokenizer_spec:
576
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
577
+ end_of_text_token: "</s>"
578
+ prefix_token: "<s>"
579
+
580
+ - name: mistralai/Mistral-Large-Instruct-2411
581
+ tokenizer_spec:
582
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
583
+ end_of_text_token: "</s>"
584
+ prefix_token: "<s>"
585
+
586
+ - name: mistralai/Ministral-8B-Instruct-2410
587
+ tokenizer_spec:
588
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
589
+ end_of_text_token: "</s>"
590
+ prefix_token: "<s>"
591
+
592
+ - name: mistralai/Mistral-Small-24B-Instruct-2501
593
+ tokenizer_spec:
594
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
595
+ end_of_text_token: "</s>"
596
+ prefix_token: "<s>"
597
+
598
+ # Moonshot AI
599
+ - name: moonshotai/kimi-k2-instruct
600
+ tokenizer_spec:
601
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
602
+ args:
603
+ pretrained_model_name_or_path: moonshotai/Kimi-K2-Instruct
604
+ trust_remote_code: true
605
+ revision: 4f239503ad9d1a042f0a4bacac457931ab972cfc
606
+ end_of_text_token: "[EOS]"
607
+ prefix_token: "[BOS]"
608
+
609
+ # Nectec
610
+ - name: nectec/OpenThaiLLM-Prebuilt-7B
611
+ tokenizer_spec:
612
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
613
+ end_of_text_token: "<|im_end|>"
614
+ prefix_token: ""
615
+
616
+ - name: nectec/Pathumma-llm-text-1.0.0
617
+ tokenizer_spec:
618
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
619
+ end_of_text_token: "<|im_end|>"
620
+ prefix_token: "<|im_start|>"
621
+
223
622
  # Neurips
224
623
  - name: neurips/local
225
624
  tokenizer_spec:
226
- class_name: "helm.proxy.tokenizers.http_model_tokenizer.HTTPModelTokenizer"
625
+ class_name: "helm.tokenizers.http_model_tokenizer.HTTPModelTokenizer"
227
626
  end_of_text_token: "<|endoftext|>"
228
627
  prefix_token: "<|endoftext|>"
229
628
 
230
- # Openai
629
+ # NVIDIA
630
+ - name: nvidia/nemotron-4-340b-instruct
631
+ tokenizer_spec:
632
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
633
+ args:
634
+ pretrained_model_name_or_path: Xenova/Nemotron-4-340B-Instruct-Tokenizer
635
+ revision: b7aa0de92cda9f9e722d58d6ca90f46ae17d4701
636
+ end_of_text_token: "<|endoftext|>"
637
+ prefix_token: "<|endoftext|>"
638
+
639
+ - name: nvidia/llama-3.1-nemotron-70b-instruct
640
+ tokenizer_spec:
641
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
642
+ args:
643
+ pretrained_model_name_or_path: nvidia/Llama-3.1-Nemotron-70B-Instruct-HF
644
+ end_of_text_token: "<|eot_id|>"
645
+ prefix_token: "<|begin_of_text|>"
646
+
647
+ # OpenAI
231
648
  - name: openai/cl100k_base
232
649
  tokenizer_spec:
233
- class_name: "helm.proxy.tokenizers.tiktoken_tokenizer.TiktokenTokenizer"
650
+ class_name: "helm.tokenizers.tiktoken_tokenizer.TiktokenTokenizer"
234
651
  end_of_text_token: "<|endoftext|>"
235
652
  prefix_token: "<|endoftext|>"
236
653
 
654
+ - name: openai/o200k_base
655
+ tokenizer_spec:
656
+ class_name: "helm.tokenizers.tiktoken_tokenizer.TiktokenTokenizer"
657
+ end_of_text_token: "<|endoftext|>"
658
+ prefix_token: "<|endoftext|>"
659
+
660
+ - name: openai/o200k_harmony
661
+ tokenizer_spec:
662
+ class_name: "helm.tokenizers.tiktoken_tokenizer.TiktokenTokenizer"
663
+ end_of_text_token: "<|endoftext|>"
664
+ prefix_token: "<|startoftext|>"
665
+
666
+ - name: openai/clip-vit-large-patch14
667
+ tokenizer_spec:
668
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
669
+ end_of_text_token: ""
670
+ prefix_token: ""
671
+
672
+ # OpenThaiGPT
673
+ - name: openthaigpt/openthaigpt-1.0.0-7b-chat
674
+ tokenizer_spec:
675
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
676
+ end_of_text_token: "</s>"
677
+ prefix_token: "<s>"
678
+
679
+ # Qwen
680
+ - name: qwen/qwen-7b
681
+ tokenizer_spec:
682
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
683
+ args:
684
+ pretrained_model_name_or_path: Qwen/Qwen-7B
685
+ trust_remote_code: true
686
+ end_of_text_token: "<|endoftext|>"
687
+ prefix_token: ""
688
+
689
+ - name: qwen/qwen1.5-7b
690
+ tokenizer_spec:
691
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
692
+ args:
693
+ pretrained_model_name_or_path: Qwen/Qwen1.5-7B
694
+ end_of_text_token: "<|endoftext|>"
695
+ prefix_token: ""
696
+
697
+ - name: qwen/qwen2-72b-instruct
698
+ tokenizer_spec:
699
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
700
+ args:
701
+ pretrained_model_name_or_path: Qwen/Qwen2-72B-Instruct
702
+ end_of_text_token: "<|im_end|>"
703
+ prefix_token: "<|im_start|>"
704
+
705
+ - name: qwen/qwen2.5-7b-instruct
706
+ tokenizer_spec:
707
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
708
+ args:
709
+ pretrained_model_name_or_path: Qwen/Qwen2.5-7B-Instruct
710
+ end_of_text_token: "<|im_end|>"
711
+ prefix_token: "<|im_start|>"
712
+
713
+ - name: qwen/qwen3-235b-a22b
714
+ tokenizer_spec:
715
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
716
+ args:
717
+ pretrained_model_name_or_path: Qwen/Qwen3-235B-A22B
718
+ end_of_text_token: "<|im_end|>"
719
+ prefix_token: "<|im_start|>"
720
+
721
+ - name: qwen/qwen3-235b-a22b-instruct-2507-fp8
722
+ tokenizer_spec:
723
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
724
+ end_of_text_token: "<|im_end|>"
725
+ prefix_token: ""
726
+
727
+ - name: qwen/qwen3-next-80b-a3b-thinking
728
+ tokenizer_spec:
729
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
730
+ end_of_text_token: "<|im_end|>"
731
+ prefix_token: ""
732
+
733
+ - name: qwen/qwq-32b-preview
734
+ tokenizer_spec:
735
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
736
+ end_of_text_token: "<|im_end|>"
737
+ prefix_token: ""
738
+
739
+ - name: qwen/qwen-vl
740
+ tokenizer_spec:
741
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
742
+ args:
743
+ pretrained_model_name_or_path: Qwen/Qwen-VL
744
+ trust_remote_code: true
745
+ # Source: https://github.com/QwenLM/Qwen-VL
746
+ end_of_text_token: "<|endoftext|>"
747
+ prefix_token: ""
748
+
749
+ - name: qwen/qwen-vl-chat
750
+ tokenizer_spec:
751
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
752
+ args:
753
+ pretrained_model_name_or_path: Qwen/Qwen-VL-Chat
754
+ trust_remote_code: true
755
+ # Source: https://github.com/QwenLM/Qwen-VL
756
+ end_of_text_token: "<|endoftext|>"
757
+ prefix_token: ""
758
+
759
+ - name: qwen/qwen-audio-chat
760
+ tokenizer_spec:
761
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
762
+ args:
763
+ pretrained_model_name_or_path: Qwen/Qwen-Audio-Chat
764
+ trust_remote_code: true
765
+ # Source: https://github.com/QwenLM/Qwen-Audio
766
+ end_of_text_token: "<|endoftext|>"
767
+ prefix_token: ""
768
+
769
+ - name: qwen/qwen2-audio-instruct
770
+ tokenizer_spec:
771
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
772
+ args:
773
+ pretrained_model_name_or_path: Qwen/Qwen2-Audio-7B-Instruct
774
+ trust_remote_code: false
775
+ end_of_text_token: "<|endoftext|>"
776
+ prefix_token: ""
777
+
778
+ - name: qwen/qwen2.5-omni-7b
779
+ tokenizer_spec:
780
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
781
+ args:
782
+ pretrained_model_name_or_path: Qwen/Qwen2.5-Omni-7B
783
+ trust_remote_code: false
784
+ end_of_text_token: "<|endoftext|>"
785
+ prefix_token: ""
786
+
787
+ # SambaLingo
788
+ - name: sambanova/sambalingo-thai-base
789
+ tokenizer_spec:
790
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
791
+ args:
792
+ pretrained_model_name_or_path: sambanovasystems/SambaLingo-Thai-Base
793
+ end_of_text_token: "</s>"
794
+ prefix_token: "<s>"
795
+
796
+ # Snowflake
797
+ - name: snowflake/snowflake-arctic-instruct
798
+ tokenizer_spec:
799
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
800
+ args:
801
+ pretrained_model_name_or_path: Snowflake/snowflake-arctic-instruct
802
+ trust_remote_code: true
803
+ end_of_text_token: "<|im_end|>"
804
+ prefix_token: "<|im_start|>"
805
+
237
806
  # Tiiuae
238
807
  - name: tiiuae/falcon-7b
239
808
  tokenizer_spec:
240
- class_name: "helm.proxy.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
809
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
810
+ end_of_text_token: "<|endoftext|>"
811
+ prefix_token: ""
812
+
813
+ - name: tiiuae/falcon3-1b-instruct
814
+ tokenizer_spec:
815
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
241
816
  end_of_text_token: "<|endoftext|>"
242
817
  prefix_token: ""
243
818
 
244
819
  # TsinghuaKEG
245
820
  - name: TsinghuaKEG/ice
246
821
  tokenizer_spec:
247
- class_name: "helm.proxy.tokenizers.ice_tokenizer.ICETokenizer"
822
+ class_name: "helm.tokenizers.ice_tokenizer.ICETokenizer"
248
823
  end_of_text_token: "</s>"
249
824
  prefix_token: ""
250
825
 
826
+ # Typhoon
827
+ - name: scb10x/typhoon-7b
828
+ tokenizer_spec:
829
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
830
+ end_of_text_token: "</s>"
831
+ prefix_token: "<s>"
832
+
833
+ # Upstage
834
+ - name: upstage/solar-pro-preview-instruct
835
+ tokenizer_spec:
836
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
837
+ args:
838
+ trust_remote_code: true
839
+ end_of_text_token: "<|im_end|>"
840
+ prefix_token: "<|startoftext|>"
841
+
251
842
  # Writer
252
843
  - name: writer/gpt2
253
844
  tokenizer_spec:
254
- class_name: "helm.proxy.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
845
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
846
+ args:
847
+ pretrained_model_name_or_path: openai-community/gpt2
255
848
  end_of_text_token: ""
256
849
  prefix_token: ""
257
850
 
258
851
  # Yandex
259
852
  - name: Yandex/yalm
260
853
  tokenizer_spec:
261
- class_name: "helm.proxy.tokenizers.yalm_tokenizer.YaLMTokenizer"
854
+ class_name: "helm.tokenizers.yalm_tokenizer.YaLMTokenizer"
262
855
  end_of_text_token: "</s>"
263
856
  prefix_token: "</s>"
857
+
858
+ # Diva Llama
859
+ - name: stanford/diva-llama
860
+ tokenizer_spec:
861
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
862
+ args:
863
+ pretrained_model_name_or_path: WillHeld/DiVA-llama-3-v0-8b
864
+ trust_remote_code: true
865
+ prefix_token: "<|begin_of_text|>"
866
+ end_of_text_token: "<|eot_id|>"
867
+
868
+ # LLaMA-Omni
869
+ - name: ictnlp/llama-3.1-8b-omni
870
+ tokenizer_spec:
871
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
872
+ args:
873
+ pretrained_model_name_or_path: ICTNLP/Llama-3.1-8B-Omni
874
+ trust_remote_code: false
875
+ end_of_text_token: "<|eot_id|>"
876
+ prefix_token: "<|begin_of_text|>"
877
+
878
+ # IBM - Granite 3.0
879
+ - name: ibm-granite/granite-3.0-2b-base
880
+ tokenizer_spec:
881
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
882
+ args:
883
+ pretrained_model_name_or_path: ibm-granite/granite-3.0-2b-base
884
+ end_of_text_token: ""
885
+ prefix_token: ""
886
+
887
+ - name: ibm-granite/granite-3.0-2b-instruct
888
+ tokenizer_spec:
889
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
890
+ args:
891
+ pretrained_model_name_or_path: ibm-granite/granite-3.0-2b-instruct
892
+ end_of_text_token: ""
893
+ prefix_token: ""
894
+
895
+ - name: ibm-granite/granite-3.0-8b-instruct
896
+ tokenizer_spec:
897
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
898
+ args:
899
+ pretrained_model_name_or_path: ibm-granite/granite-3.0-8b-instruct
900
+ end_of_text_token: ""
901
+ prefix_token: ""
902
+
903
+ - name: ibm-granite/granite-3.0-8b-base
904
+ tokenizer_spec:
905
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
906
+ args:
907
+ pretrained_model_name_or_path: ibm-granite/granite-3.0-8b-base
908
+ end_of_text_token: ""
909
+ prefix_token: ""
910
+
911
+ - name: ibm-granite/granite-3.0-3b-a800m-instruct
912
+ tokenizer_spec:
913
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
914
+ args:
915
+ pretrained_model_name_or_path: ibm-granite/granite-3.0-3b-a800m-instruct
916
+ end_of_text_token: ""
917
+ prefix_token: ""
918
+
919
+ - name: ibm-granite/granite-3.0-3b-a800m-base
920
+ tokenizer_spec:
921
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
922
+ args:
923
+ pretrained_model_name_or_path: ibm-granite/granite-3.0-3b-a800m-base
924
+ end_of_text_token: ""
925
+ prefix_token: ""
926
+
927
+ - name: ibm-granite/granite-3.0-1b-a400m-instruct
928
+ tokenizer_spec:
929
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
930
+ args:
931
+ pretrained_model_name_or_path: ibm-granite/granite-3.0-1b-a400m-instruct
932
+ end_of_text_token: ""
933
+ prefix_token: ""
934
+
935
+ - name: ibm-granite/granite-3.0-1b-a400m-base
936
+ tokenizer_spec:
937
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
938
+ args:
939
+ pretrained_model_name_or_path: ibm-granite/granite-3.0-1b-a400m-base
940
+ end_of_text_token: ""
941
+ prefix_token: ""
942
+
943
+ - name: ibm/granite-4.0-micro
944
+ tokenizer_spec:
945
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
946
+ args:
947
+ pretrained_model_name_or_path: ibm-granite/granite-4.0-micro
948
+ end_of_text_token: "<|end_of_text|>"
949
+ prefix_token: "<|end_of_text|>"
950
+
951
+ - name: ibm/granite-4.0-h-small
952
+ tokenizer_spec:
953
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
954
+ args:
955
+ pretrained_model_name_or_path: ibm-granite/granite-4.0-h-small
956
+ end_of_text_token: "<|end_of_text|>"
957
+ prefix_token: "<|end_of_text|>"
958
+
959
+ # Maritaca AI
960
+ - name: maritaca-ai/sabia-7b
961
+ tokenizer_spec:
962
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
963
+ args:
964
+ pretrained_model_name_or_path: maritaca-ai/sabia-7b
965
+ end_of_text_token: "</s>"
966
+ prefix_token: "<s>"
967
+
968
+ - name: maritaca-ai/sabia-2-tokenizer-medium
969
+ tokenizer_spec:
970
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
971
+ args:
972
+ pretrained_model_name_or_path: maritaca-ai/sabia-2-tokenizer-medium
973
+ end_of_text_token: "</s>"
974
+ prefix_token: "<s>"
975
+
976
+ # Granite-3.1-8b-base
977
+ - name: ibm-granite/granite-3.1-8b-base
978
+ tokenizer_spec:
979
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
980
+ args:
981
+ pretrained_model_name_or_path: ibm-granite/granite-3.1-8b-base
982
+ prefix_token: ""
983
+ end_of_text_token: "<|endoftext|>"
984
+
985
+ # Granite-3.1-8b-instruct
986
+ - name: ibm-granite/granite-3.1-8b-instruct
987
+ tokenizer_spec:
988
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
989
+ args:
990
+ pretrained_model_name_or_path: ibm-granite/granite-3.1-8b-instruct
991
+ prefix_token: ""
992
+ end_of_text_token: "<|endoftext|>"
993
+
994
+ # Granite-3.1-2b-instruct
995
+ - name: ibm-granite/granite-3.1-2b-instruct
996
+ tokenizer_spec:
997
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
998
+ args:
999
+ pretrained_model_name_or_path: ibm-granite/granite-3.1-2b-instruct
1000
+ prefix_token: ""
1001
+ end_of_text_token: ""
1002
+
1003
+ # Granite-3.1-2b-base
1004
+ - name: ibm-granite/granite-3.1-2b-base
1005
+ tokenizer_spec:
1006
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
1007
+ args:
1008
+ pretrained_model_name_or_path: ibm-granite/granite-3.1-2b-base
1009
+ prefix_token: ""
1010
+ end_of_text_token: ""
1011
+
1012
+ # Granite-3.1-3b-a800m-instruct
1013
+ - name: ibm-granite/granite-3.1-3b-a800m-instruct
1014
+ tokenizer_spec:
1015
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
1016
+ args:
1017
+ pretrained_model_name_or_path: ibm-granite/granite-3.1-3b-a800m-instruct
1018
+ prefix_token: ""
1019
+ end_of_text_token: ""
1020
+
1021
+ # Granite-3.1-3b-a800m-base
1022
+ - name: ibm-granite/granite-3.1-3b-a800m-base
1023
+ tokenizer_spec:
1024
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
1025
+ args:
1026
+ pretrained_model_name_or_path: ibm-granite/granite-3.1-3b-a800m-base
1027
+ prefix_token: ""
1028
+ end_of_text_token: ""
1029
+
1030
+ # Granite-3.1-1b-a400m-instruct
1031
+ - name: ibm-granite/granite-3.1-1b-a400m-instruct
1032
+ tokenizer_spec:
1033
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
1034
+ args:
1035
+ pretrained_model_name_or_path: ibm-granite/granite-3.1-1b-a400m-instruct
1036
+ prefix_token: ""
1037
+ end_of_text_token: ""
1038
+
1039
+ # Granite-3.1-1b-a400m-base
1040
+ - name: ibm-granite/granite-3.1-1b-a400m-base
1041
+ tokenizer_spec:
1042
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
1043
+ args:
1044
+ pretrained_model_name_or_path: ibm-granite/granite-3.1-1b-a400m-base
1045
+ prefix_token: ""
1046
+ end_of_text_token: ""
1047
+
1048
+ - name: ibm-granite/granite-20b-code-instruct-8k
1049
+ tokenizer_spec:
1050
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
1051
+ args:
1052
+ pretrained_model_name_or_path: ibm-granite/granite-20b-code-instruct-8k
1053
+ prefix_token: ""
1054
+ end_of_text_token: ""
1055
+
1056
+ - name: ibm-granite/granite-3b-code-instruct-128k
1057
+ tokenizer_spec:
1058
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
1059
+ args:
1060
+ pretrained_model_name_or_path: ibm-granite/granite-3b-code-instruct-128k
1061
+ prefix_token: ""
1062
+ end_of_text_token: ""
1063
+
1064
+ - name: ibm-granite/granite-34b-code-instruct-8k
1065
+ tokenizer_spec:
1066
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
1067
+ args:
1068
+ pretrained_model_name_or_path: ibm-granite/granite-34b-code-instruct-8k
1069
+ prefix_token: ""
1070
+ end_of_text_token: ""
1071
+
1072
+ - name: ibm-granite/granite-8b-code-instruct-128k
1073
+ tokenizer_spec:
1074
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
1075
+ args:
1076
+ pretrained_model_name_or_path: ibm-granite/granite-8b-code-instruct-128k
1077
+ prefix_token: ""
1078
+ end_of_text_token: ""
1079
+
1080
+
1081
+ - name: ibm-granite/granite-guardian-3.1-2b
1082
+ tokenizer_spec:
1083
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
1084
+ args:
1085
+ pretrained_model_name_or_path: ibm-granite/granite-guardian-3.1-2b
1086
+ prefix_token: ""
1087
+ end_of_text_token: ""
1088
+
1089
+ - name: ibm-granite/granite-guardian-3.1-8b
1090
+ tokenizer_spec:
1091
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
1092
+ args:
1093
+ pretrained_model_name_or_path: ibm-granite/granite-guardian-3.1-8b
1094
+ prefix_token: ""
1095
+ end_of_text_token: ""
1096
+
1097
+ # IBM Granite 3.3
1098
+ - name: ibm/granite-3.3-8b-instruct
1099
+ tokenizer_spec:
1100
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
1101
+ args:
1102
+ pretrained_model_name_or_path: ibm-granite/granite-3.3-8b-instruct
1103
+ end_of_text_token: "<|end_of_text|>"
1104
+ prefix_token: "<|end_of_text|>"
1105
+
1106
+ # Z.ai GLM-4.5-AIR-FP8
1107
+ - name: zai-org/glm-4.5-air-fp8
1108
+ tokenizer_spec:
1109
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
1110
+ end_of_text_token: "<|endoftext|>"
1111
+ prefix_token: ""
1112
+
1113
+ # DeepSeek-R1-Distill-Llama-3.1-8b
1114
+ - name: deepseek-ai/DeepSeek-R1-Distill-Llama-8B
1115
+ tokenizer_spec:
1116
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
1117
+ args:
1118
+ pretrained_model_name_or_path: deepseek-ai/DeepSeek-R1-Distill-Llama-8B
1119
+ end_of_text_token: "<｜end▁of▁sentence｜>"
1120
+ prefix_token: "<｜begin▁of▁sentence｜>"
1121
+
1122
+ # DeepSeek-R1-Distill-Llama-3.1-8b
1123
+ - name: deepseek-ai/deepseek-r1-distill-llama-70b
1124
+ tokenizer_spec:
1125
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
1126
+ end_of_text_token: "<｜end▁of▁sentence｜>"
1127
+ prefix_token: "<｜begin▁of▁sentence｜>"
1128
+
1129
+ # DeepSeek-R1-Distill-Qwen-14B
1130
+ - name: deepseek-ai/deepseek-r1-distill-qwen-14b
1131
+ tokenizer_spec:
1132
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
1133
+ end_of_text_token: "<｜end▁of▁sentence｜>"
1134
+ prefix_token: "<｜begin▁of▁sentence｜>"
1135
+
1136
+ # deepseek-ai/deepseek-coder-6.7b-instruct
1137
+ - name: deepseek-ai/deepseek-coder-6.7b-instruct
1138
+ tokenizer_spec:
1139
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
1140
+ args:
1141
+ pretrained_model_name_or_path: deepseek-ai/deepseek-coder-6.7b-instruct
1142
+ end_of_text_token: "<｜end▁of▁sentence｜>"
1143
+ prefix_token: "<｜begin▁of▁sentence｜>"
1144
+
1145
+ # vilm/vinallama-2.7b-chat
1146
+ - name: vilm/vinallama-2.7b-chat
1147
+ tokenizer_spec:
1148
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
1149
+ args:
1150
+ pretrained_model_name_or_path: vilm/vinallama-2.7b-chat
1151
+ end_of_text_token: "<im_end>"
1152
+ prefix_token: "<im_start>"
1153
+
1154
+ # vilm/vinallama-7b-chat
1155
+ - name: vilm/vinallama-7b-chat
1156
+ tokenizer_spec:
1157
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
1158
+ args:
1159
+ pretrained_model_name_or_path: vilm/vinallama-7b-chat
1160
+ end_of_text_token: "<im_end>"
1161
+ prefix_token: "<im_start>"
1162
+
1163
+ # vilm/vietcuna-7b-v3
1164
+ - name: vilm/vietcuna-7b-v3
1165
+ tokenizer_spec:
1166
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
1167
+ args:
1168
+ pretrained_model_name_or_path: vilm/vietcuna-7b-v3
1169
+ end_of_text_token: "</s>"
1170
+ prefix_token: "<s>"
1171
+
1172
+ # Viet-Mistral/Vistral-7B-Chat
1173
+ - name: Viet-Mistral/Vistral-7B-Chat
1174
+ tokenizer_spec:
1175
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
1176
+ args:
1177
+ pretrained_model_name_or_path: Viet-Mistral/Vistral-7B-Chat
1178
+ end_of_text_token: "</s>"
1179
+ prefix_token: "<s>"
1180
+
1181
+ # vinai/PhoGPT-7B5-Instruct
1182
+ - name: vinai/PhoGPT-7B5-Instruct
1183
+ tokenizer_spec:
1184
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
1185
+ args:
1186
+ pretrained_model_name_or_path: vinai/PhoGPT-7B5-Instruct
1187
+ end_of_text_token: "</s>"
1188
+ prefix_token: "<s>"
1189
+
1190
+ # vinai/PhoGPT-4B-Chat
1191
+ - name: vinai/PhoGPT-4B-Chat
1192
+ tokenizer_spec:
1193
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
1194
+ args:
1195
+ pretrained_model_name_or_path: vinai/PhoGPT-4B-Chat
1196
+ end_of_text_token: "</s>"
1197
+ prefix_token: "<s>"
1198
+
1199
+ # Gemma-3-Gaia-PT-BR-4b-it
1200
+ - name: CEIA-UFG/Gemma-3-Gaia-PT-BR-4b-it
1201
+ tokenizer_spec:
1202
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
1203
+ args:
1204
+ pretrained_model_name_or_path: CEIA-UFG/Gemma-3-Gaia-PT-BR-4b-it
1205
+ end_of_text_token: "<eos>"
1206
+ prefix_token: "<bos>"
1207
+
1208
+ # Bode 13B Alpaca PT-BR
1209
+ - name: recogna-nlp/bode-13b-alpaca-pt-br-no-peft
1210
+ tokenizer_spec:
1211
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
1212
+ args:
1213
+ pretrained_model_name_or_path: recogna-nlp/bode-13b-alpaca-pt-br-no-peft
1214
+ end_of_text_token: "</s>"
1215
+ prefix_token: "<s>"
1216
+
1217
+ # Cabrita 7B PT-BR tokenizer
1218
+ - name: 22h/cabrita_7b_pt_850000
1219
+ tokenizer_spec:
1220
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
1221
+ args:
1222
+ pretrained_model_name_or_path: 22h/cabrita_7b_pt_850000
1223
+ end_of_text_token: "</s>"
1224
+ prefix_token: "<s>"
1225
+
1226
+ # Gervásio 7B PT‑BR/PT‑PT tokenizer
1227
+ - name: PORTULAN/gervasio-7b-portuguese-ptbr-decoder
1228
+ tokenizer_spec:
1229
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
1230
+ args:
1231
+ pretrained_model_name_or_path: PORTULAN/gervasio-7b-portuguese-ptbr-decoder
1232
+ end_of_text_token: "</s>"
1233
+ prefix_token: "<s>"
1234
+
1235
+ # Tucano 2b4 PT-BR tokenizer
1236
+ - name: TucanoBR/Tucano-2b4
1237
+ tokenizer_spec:
1238
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
1239
+ args:
1240
+ pretrained_model_name_or_path: TucanoBR/Tucano-2b4
1241
+ end_of_text_token: "</s>"
1242
+ prefix_token: "<s>"
1243
+
1244
+ # TeenyTinyLlama 460M PT-BR tokenizer
1245
+ - name: nicholasKluge/TeenyTinyLlama-460m
1246
+ tokenizer_spec:
1247
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
1248
+ args:
1249
+ pretrained_model_name_or_path: nicholasKluge/TeenyTinyLlama-460m
1250
+ end_of_text_token: "</s>"
1251
+ prefix_token: "<s>"
1252
+
1253
+ # AceGPT-v2
1254
+ - name: freedomintelligence/acegpt-v2-8b-chat
1255
+ tokenizer_spec:
1256
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
1257
+ end_of_text_token: "<|end_of_text|>"
1258
+ prefix_token: "<|begin_of_text|>"
1259
+
1260
+ - name: freedomintelligence/acegpt-v2-32b-chat
1261
+ tokenizer_spec:
1262
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
1263
+ end_of_text_token: "<|endoftext|>"
1264
+ prefix_token: ""
1265
+
1266
+ - name: freedomintelligence/acegpt-v2-70b-chat
1267
+ tokenizer_spec:
1268
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
1269
+ end_of_text_token: "<|end_of_text|>"
1270
+ prefix_token: "<|begin_of_text|>"
1271
+
1272
+ # ALLaM
1273
+ - name: allam-ai/allam-7b-instruct-preview
1274
+ tokenizer_spec:
1275
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
1276
+ end_of_text_token: "</s>"
1277
+ prefix_token: "<s>"
1278
+
1279
+ # SILMA
1280
+ - name: silma-ai/silma-9b-instruct-v1.0
1281
+ tokenizer_spec:
1282
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
1283
+ end_of_text_token: "<eos>"
1284
+ prefix_token: "<bos>"
1285
+
1286
+ # Jais Family
1287
+ - name: inceptionai/jais-family-590m-chat
1288
+ tokenizer_spec:
1289
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
1290
+ end_of_text_token: "<|endoftext|>"
1291
+ prefix_token: "<|endoftext|>"
1292
+
1293
+ # Jais Adapted
1294
+ - name: inceptionai/jais-adapted-7b-chat
1295
+ tokenizer_spec:
1296
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
1297
+ end_of_text_token: "</s>"
1298
+ prefix_token: "<s>"
1299
+
1300
+ - name: inceptionai/jais-adapted-13b-chat
1301
+ tokenizer_spec:
1302
+ class_name: "helm.tokenizers.huggingface_tokenizer.HuggingFaceTokenizer"
1303
+ end_of_text_token: "</s>"
1304
+ prefix_token: "<s>"