crfm-helm 0.4.0__py3-none-any.whl → 0.5.0__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.
- {crfm_helm-0.4.0.dist-info → crfm_helm-0.5.0.dist-info}/METADATA +134 -31
- crfm_helm-0.5.0.dist-info/RECORD +642 -0
- {crfm_helm-0.4.0.dist-info → crfm_helm-0.5.0.dist-info}/WHEEL +1 -1
- helm/benchmark/adaptation/adapter_spec.py +31 -3
- helm/benchmark/adaptation/adapters/adapter.py +2 -2
- helm/benchmark/adaptation/adapters/adapter_factory.py +24 -27
- helm/benchmark/adaptation/adapters/generation_adapter.py +1 -0
- helm/benchmark/adaptation/adapters/in_context_learning_adapter.py +20 -4
- helm/benchmark/adaptation/adapters/language_modeling_adapter.py +2 -3
- helm/benchmark/adaptation/adapters/multimodal/multiple_choice_joint_multimodal_adapter.py +104 -0
- helm/benchmark/adaptation/adapters/multimodal/test_in_context_learning_multimodal_adapter.py +2 -1
- helm/benchmark/adaptation/adapters/test_adapter.py +2 -1
- helm/benchmark/adaptation/adapters/test_generation_adapter.py +32 -8
- helm/benchmark/adaptation/adapters/test_language_modeling_adapter.py +7 -19
- helm/benchmark/adaptation/adapters/test_multiple_choice_joint_adapter.py +60 -6
- helm/benchmark/adaptation/common_adapter_specs.py +376 -0
- helm/benchmark/adaptation/request_state.py +6 -1
- helm/benchmark/adaptation/scenario_state.py +6 -2
- helm/benchmark/annotation/annotator.py +43 -0
- helm/benchmark/annotation/annotator_factory.py +61 -0
- helm/benchmark/annotation/image2structure/image_compiler_annotator.py +88 -0
- helm/benchmark/annotation/image2structure/latex_compiler_annotator.py +59 -0
- helm/benchmark/annotation/image2structure/lilypond_compiler_annotator.py +84 -0
- helm/benchmark/annotation/image2structure/webpage_compiler_annotator.py +132 -0
- helm/benchmark/annotation/test_annotator_factory.py +26 -0
- helm/benchmark/annotation/test_dummy_annotator.py +44 -0
- helm/benchmark/annotation_executor.py +124 -0
- helm/benchmark/augmentations/data_augmenter.py +0 -2
- helm/benchmark/augmentations/gender_perturbation.py +1 -1
- helm/benchmark/augmentations/perturbation.py +8 -2
- helm/benchmark/augmentations/perturbation_description.py +1 -1
- helm/benchmark/augmentations/suffix_perturbation.py +29 -0
- helm/benchmark/augmentations/test_perturbation.py +11 -7
- helm/benchmark/augmentations/translate_perturbation.py +30 -0
- helm/benchmark/config_registry.py +7 -1
- helm/benchmark/executor.py +46 -16
- helm/benchmark/huggingface_registration.py +20 -7
- helm/benchmark/metrics/basic_metrics.py +169 -664
- helm/benchmark/metrics/bbq_metrics.py +3 -4
- helm/benchmark/metrics/bias_metrics.py +6 -6
- helm/benchmark/metrics/classification_metrics.py +11 -8
- helm/benchmark/metrics/cleva_accuracy_metrics.py +8 -5
- helm/benchmark/metrics/cleva_harms_metrics.py +2 -2
- helm/benchmark/metrics/code_metrics_helper.py +0 -2
- helm/benchmark/metrics/common_metric_specs.py +167 -0
- helm/benchmark/metrics/decodingtrust_fairness_metrics.py +72 -0
- helm/benchmark/metrics/decodingtrust_ood_knowledge_metrics.py +66 -0
- helm/benchmark/metrics/decodingtrust_privacy_metrics.py +101 -0
- helm/benchmark/metrics/decodingtrust_stereotype_bias_metrics.py +202 -0
- helm/benchmark/metrics/disinformation_metrics.py +4 -110
- helm/benchmark/metrics/dry_run_metrics.py +2 -2
- helm/benchmark/metrics/efficiency_metrics.py +206 -0
- helm/benchmark/metrics/evaluate_instances_metric.py +59 -0
- helm/benchmark/metrics/evaluate_reference_metrics.py +376 -0
- helm/benchmark/metrics/image_generation/aesthetics_metrics.py +54 -0
- helm/benchmark/metrics/image_generation/aesthetics_scorer.py +66 -0
- helm/benchmark/metrics/image_generation/clip_score_metrics.py +73 -0
- helm/benchmark/metrics/image_generation/denoised_runtime_metric.py +42 -0
- helm/benchmark/metrics/image_generation/detection_metrics.py +57 -0
- helm/benchmark/metrics/image_generation/detectors/base_detector.py +8 -0
- helm/benchmark/metrics/image_generation/detectors/vitdet.py +178 -0
- helm/benchmark/metrics/image_generation/efficiency_metrics.py +41 -0
- helm/benchmark/metrics/image_generation/fidelity_metrics.py +168 -0
- helm/benchmark/metrics/image_generation/fractal_dimension/__init__.py +0 -0
- helm/benchmark/metrics/image_generation/fractal_dimension/fractal_dimension_util.py +63 -0
- helm/benchmark/metrics/image_generation/fractal_dimension/test_fractal_dimension_util.py +33 -0
- helm/benchmark/metrics/image_generation/fractal_dimension_metric.py +50 -0
- helm/benchmark/metrics/image_generation/gender_metrics.py +58 -0
- helm/benchmark/metrics/image_generation/image_critique_metrics.py +284 -0
- helm/benchmark/metrics/image_generation/lpips_metrics.py +82 -0
- helm/benchmark/metrics/image_generation/multi_scale_ssim_metrics.py +82 -0
- helm/benchmark/metrics/image_generation/nsfw_detector.py +96 -0
- helm/benchmark/metrics/image_generation/nsfw_metrics.py +103 -0
- helm/benchmark/metrics/image_generation/nudity_metrics.py +38 -0
- helm/benchmark/metrics/image_generation/photorealism_critique_metrics.py +153 -0
- helm/benchmark/metrics/image_generation/psnr_metrics.py +78 -0
- helm/benchmark/metrics/image_generation/q16/__init__.py +0 -0
- helm/benchmark/metrics/image_generation/q16/q16_toxicity_detector.py +90 -0
- helm/benchmark/metrics/image_generation/q16/test_q16.py +18 -0
- helm/benchmark/metrics/image_generation/q16_toxicity_metrics.py +48 -0
- helm/benchmark/metrics/image_generation/skin_tone_metrics.py +164 -0
- helm/benchmark/metrics/image_generation/uiqi_metrics.py +92 -0
- helm/benchmark/metrics/image_generation/watermark/__init__.py +0 -0
- helm/benchmark/metrics/image_generation/watermark/test_watermark_detector.py +16 -0
- helm/benchmark/metrics/image_generation/watermark/watermark_detector.py +87 -0
- helm/benchmark/metrics/image_generation/watermark_metrics.py +48 -0
- helm/benchmark/metrics/instruction_following_critique_metrics.py +3 -1
- helm/benchmark/metrics/language_modeling_metrics.py +99 -0
- helm/benchmark/metrics/machine_translation_metrics.py +89 -0
- helm/benchmark/metrics/metric.py +93 -172
- helm/benchmark/metrics/metric_name.py +0 -1
- helm/benchmark/metrics/metric_service.py +16 -0
- helm/benchmark/metrics/paraphrase_generation_metrics.py +3 -4
- helm/benchmark/metrics/ranking_metrics.py +2 -2
- helm/benchmark/metrics/reference_metric.py +148 -0
- helm/benchmark/metrics/summac/model_summac.py +0 -2
- helm/benchmark/metrics/summarization_metrics.py +2 -2
- helm/benchmark/metrics/test_classification_metrics.py +8 -5
- helm/benchmark/metrics/test_disinformation_metrics.py +78 -0
- helm/benchmark/metrics/{test_basic_metrics.py → test_evaluate_reference_metrics.py} +5 -1
- helm/benchmark/metrics/test_metric.py +2 -2
- helm/benchmark/metrics/tokens/gooseai_token_cost_estimator.py +10 -2
- helm/benchmark/metrics/toxicity_metrics.py +1 -1
- helm/benchmark/metrics/toxicity_utils.py +23 -0
- helm/benchmark/metrics/unitxt_metrics.py +81 -0
- helm/benchmark/metrics/vision_language/__init__.py +0 -0
- helm/benchmark/metrics/vision_language/emd_utils.py +341 -0
- helm/benchmark/metrics/vision_language/image_metrics.py +450 -0
- helm/benchmark/metrics/vision_language/image_utils.py +100 -0
- helm/benchmark/model_deployment_registry.py +74 -0
- helm/benchmark/model_metadata_registry.py +36 -0
- helm/benchmark/multi_gpu_runner.py +133 -0
- helm/benchmark/presentation/create_plots.py +8 -7
- helm/benchmark/presentation/run_display.py +26 -10
- helm/benchmark/presentation/schema.py +15 -40
- helm/benchmark/presentation/summarize.py +119 -79
- helm/benchmark/presentation/table.py +8 -8
- helm/benchmark/presentation/test_contamination.py +2 -2
- helm/benchmark/presentation/test_run_entry.py +1 -2
- helm/benchmark/presentation/test_summarize.py +3 -3
- helm/benchmark/run.py +54 -26
- helm/benchmark/run_expander.py +214 -16
- helm/benchmark/run_spec.py +93 -0
- helm/benchmark/run_spec_factory.py +162 -0
- helm/benchmark/run_specs/__init__.py +0 -0
- helm/benchmark/run_specs/classic_run_specs.py +1510 -0
- helm/benchmark/run_specs/cleva_run_specs.py +277 -0
- helm/benchmark/run_specs/decodingtrust_run_specs.py +314 -0
- helm/benchmark/run_specs/heim_run_specs.py +623 -0
- helm/benchmark/run_specs/instruction_following_run_specs.py +129 -0
- helm/benchmark/run_specs/lite_run_specs.py +307 -0
- helm/benchmark/run_specs/simple_run_specs.py +104 -0
- helm/benchmark/run_specs/unitxt_run_specs.py +42 -0
- helm/benchmark/run_specs/vlm_run_specs.py +501 -0
- helm/benchmark/runner.py +51 -57
- helm/benchmark/runner_config_registry.py +21 -0
- helm/benchmark/scenarios/bbq_scenario.py +1 -1
- helm/benchmark/scenarios/bold_scenario.py +2 -2
- helm/benchmark/scenarios/code_scenario.py +1 -0
- helm/benchmark/scenarios/decodingtrust_adv_demonstration_scenario.py +169 -0
- helm/benchmark/scenarios/decodingtrust_adv_robustness_scenario.py +121 -0
- helm/benchmark/scenarios/decodingtrust_fairness_scenario.py +77 -0
- helm/benchmark/scenarios/decodingtrust_machine_ethics_scenario.py +324 -0
- helm/benchmark/scenarios/decodingtrust_ood_robustness_scenario.py +204 -0
- helm/benchmark/scenarios/decodingtrust_privacy_scenario.py +559 -0
- helm/benchmark/scenarios/decodingtrust_stereotype_bias_scenario.py +67 -0
- helm/benchmark/scenarios/decodingtrust_toxicity_prompts_scenario.py +78 -0
- helm/benchmark/scenarios/dialogue_scenarios.py +0 -1
- helm/benchmark/scenarios/image_generation/__init__.py +0 -0
- helm/benchmark/scenarios/image_generation/common_syntactic_processes_scenario.py +105 -0
- helm/benchmark/scenarios/image_generation/cub200_scenario.py +95 -0
- helm/benchmark/scenarios/image_generation/daily_dalle_scenario.py +124 -0
- helm/benchmark/scenarios/image_generation/demographic_stereotypes_scenario.py +82 -0
- helm/benchmark/scenarios/image_generation/detection_scenario.py +83 -0
- helm/benchmark/scenarios/image_generation/draw_bench_scenario.py +74 -0
- helm/benchmark/scenarios/image_generation/i2p_scenario.py +57 -0
- helm/benchmark/scenarios/image_generation/landing_page_scenario.py +46 -0
- helm/benchmark/scenarios/image_generation/logos_scenario.py +223 -0
- helm/benchmark/scenarios/image_generation/magazine_cover_scenario.py +91 -0
- helm/benchmark/scenarios/image_generation/mental_disorders_scenario.py +46 -0
- helm/benchmark/scenarios/image_generation/mscoco_scenario.py +91 -0
- helm/benchmark/scenarios/image_generation/paint_skills_scenario.py +72 -0
- helm/benchmark/scenarios/image_generation/parti_prompts_scenario.py +94 -0
- helm/benchmark/scenarios/image_generation/radiology_scenario.py +42 -0
- helm/benchmark/scenarios/image_generation/relational_understanding_scenario.py +52 -0
- helm/benchmark/scenarios/image_generation/time_most_significant_historical_figures_scenario.py +124 -0
- helm/benchmark/scenarios/image_generation/winoground_scenario.py +62 -0
- helm/benchmark/scenarios/imdb_scenario.py +0 -1
- helm/benchmark/scenarios/live_qa_scenario.py +94 -0
- helm/benchmark/scenarios/lm_entry_scenario.py +185 -0
- helm/benchmark/scenarios/math_scenario.py +19 -2
- helm/benchmark/scenarios/medication_qa_scenario.py +60 -0
- helm/benchmark/scenarios/numeracy_scenario.py +1 -1
- helm/benchmark/scenarios/opinions_qa_scenario.py +0 -4
- helm/benchmark/scenarios/scenario.py +4 -0
- helm/benchmark/scenarios/simple_scenarios.py +122 -1
- helm/benchmark/scenarios/test_math_scenario.py +6 -0
- helm/benchmark/scenarios/test_scenario.py +6 -3
- helm/benchmark/scenarios/test_simple_scenarios.py +50 -0
- helm/benchmark/scenarios/thai_exam_scenario.py +135 -0
- helm/benchmark/scenarios/unitxt_scenario.py +56 -0
- helm/benchmark/scenarios/verifiability_judgment_scenario.py +3 -1
- helm/benchmark/scenarios/vicuna_scenario.py +1 -1
- helm/benchmark/scenarios/vision_language/bingo_scenario.py +103 -0
- helm/benchmark/scenarios/vision_language/hateful_memes_scenario.py +92 -0
- helm/benchmark/scenarios/vision_language/heim_human_eval_scenario.py +113 -0
- helm/benchmark/scenarios/vision_language/image2structure/__init__.py +0 -0
- helm/benchmark/scenarios/vision_language/image2structure/chart2csv_scenario.py +55 -0
- helm/benchmark/scenarios/vision_language/image2structure/image2structure_scenario.py +214 -0
- helm/benchmark/scenarios/vision_language/image2structure/latex_scenario.py +25 -0
- helm/benchmark/scenarios/vision_language/image2structure/musicsheet_scenario.py +20 -0
- helm/benchmark/scenarios/vision_language/image2structure/utils_latex.py +347 -0
- helm/benchmark/scenarios/vision_language/image2structure/webpage/__init__.py +0 -0
- helm/benchmark/scenarios/vision_language/image2structure/webpage/driver.py +84 -0
- helm/benchmark/scenarios/vision_language/image2structure/webpage/jekyll_server.py +182 -0
- helm/benchmark/scenarios/vision_language/image2structure/webpage/utils.py +31 -0
- helm/benchmark/scenarios/vision_language/image2structure/webpage_scenario.py +225 -0
- helm/benchmark/scenarios/vision_language/mementos_scenario.py +124 -0
- helm/benchmark/scenarios/vision_language/mme_scenario.py +145 -0
- helm/benchmark/scenarios/vision_language/mmmu_scenario.py +187 -0
- helm/benchmark/scenarios/vision_language/multipanelvqa_scenario.py +169 -0
- helm/benchmark/scenarios/vision_language/pope_scenario.py +104 -0
- helm/benchmark/scenarios/vision_language/seed_bench_scenario.py +129 -0
- helm/benchmark/scenarios/vision_language/unicorn_scenario.py +108 -0
- helm/benchmark/scenarios/vision_language/viz_wiz_scenario.py +1 -2
- helm/benchmark/scenarios/vision_language/vqa_scenario.py +1 -1
- helm/benchmark/scenarios/wmt_14_scenario.py +1 -1
- helm/benchmark/server.py +24 -1
- helm/benchmark/slurm_runner.py +70 -49
- helm/benchmark/static/benchmarking.js +1 -1
- helm/benchmark/static/schema_classic.yaml +258 -1066
- helm/benchmark/static/schema_instruction_following.yaml +210 -0
- helm/benchmark/static/schema_lite.yaml +2 -227
- helm/benchmark/static/schema_mmlu.yaml +1507 -0
- helm/benchmark/static/schema_unitxt.yaml +428 -0
- helm/benchmark/static/schema_vlm.yaml +576 -0
- helm/benchmark/static_build/assets/01-694cb9b7.png +0 -0
- helm/benchmark/static_build/assets/ai21-0eb91ec3.png +0 -0
- helm/benchmark/static_build/assets/aleph-alpha-7ce10034.png +0 -0
- helm/benchmark/static_build/assets/anthropic-70d8bc39.png +0 -0
- helm/benchmark/static_build/assets/bigscience-7f0400c0.png +0 -0
- helm/benchmark/static_build/assets/cohere-3550c6cb.png +0 -0
- helm/benchmark/static_build/assets/crfm-logo-74391ab8.png +0 -0
- helm/benchmark/static_build/assets/eleutherai-b9451114.png +0 -0
- helm/benchmark/static_build/assets/google-06d997ad.png +0 -0
- helm/benchmark/static_build/assets/heim-logo-3e5e3aa4.png +0 -0
- helm/benchmark/static_build/assets/helm-logo-simple-2ed5400b.png +0 -0
- helm/benchmark/static_build/assets/helmhero-28e90f4d.png +0 -0
- helm/benchmark/static_build/assets/index-5088afcb.css +1 -0
- helm/benchmark/static_build/assets/index-d839df55.js +9 -0
- helm/benchmark/static_build/assets/meta-5580e9f1.png +0 -0
- helm/benchmark/static_build/assets/microsoft-f5ee5016.png +0 -0
- helm/benchmark/static_build/assets/mistral-18e1be23.png +0 -0
- helm/benchmark/static_build/assets/nvidia-86fa75c1.png +0 -0
- helm/benchmark/static_build/assets/openai-3f8653e4.png +0 -0
- helm/benchmark/static_build/assets/react-d4a0b69b.js +85 -0
- helm/benchmark/static_build/assets/recharts-6d337683.js +97 -0
- helm/benchmark/static_build/assets/tii-24de195c.png +0 -0
- helm/benchmark/static_build/assets/together-a665a35b.png +0 -0
- helm/benchmark/static_build/assets/tremor-54a99cc4.js +10 -0
- helm/benchmark/static_build/assets/tsinghua-keg-97d4b395.png +0 -0
- helm/benchmark/static_build/assets/vhelm-framework-cde7618a.png +0 -0
- helm/benchmark/static_build/assets/vhelm-model-6d812526.png +0 -0
- helm/benchmark/static_build/assets/yandex-38e09d70.png +0 -0
- helm/benchmark/static_build/config.js +4 -0
- helm/benchmark/static_build/index.html +20 -0
- helm/benchmark/test_data_preprocessor.py +3 -3
- helm/benchmark/test_model_deployment_definition.py +14 -16
- helm/benchmark/test_run_expander.py +1 -1
- helm/benchmark/window_services/ai21_window_service.py +22 -33
- helm/benchmark/window_services/cohere_window_service.py +1 -63
- helm/benchmark/window_services/default_window_service.py +2 -44
- helm/benchmark/window_services/encoder_decoder_window_service.py +0 -11
- helm/benchmark/window_services/ice_window_service.py +0 -34
- helm/benchmark/window_services/image_generation/__init__.py +0 -0
- helm/benchmark/window_services/image_generation/clip_window_service.py +15 -0
- helm/benchmark/window_services/image_generation/lexica_search_window_service.py +9 -0
- helm/benchmark/window_services/image_generation/openai_dalle_window_service.py +9 -0
- helm/benchmark/window_services/image_generation/test_clip_window_service.py +29 -0
- helm/benchmark/window_services/image_generation/test_openai_dalle_window_service.py +30 -0
- helm/benchmark/window_services/local_window_service.py +21 -4
- helm/benchmark/window_services/test_anthropic_window_service.py +2 -1
- helm/benchmark/window_services/test_bloom_window_service.py +2 -1
- helm/benchmark/window_services/test_cohere_window_service.py +2 -1
- helm/benchmark/window_services/test_flan_t5_window_service.py +2 -1
- helm/benchmark/window_services/test_gpt2_window_service.py +2 -2
- helm/benchmark/window_services/test_gpt4_window_service.py +2 -1
- helm/benchmark/window_services/test_gptj_window_service.py +3 -2
- helm/benchmark/window_services/test_gptneox_window_service.py +3 -2
- helm/benchmark/window_services/test_ice_window_service.py +2 -1
- helm/benchmark/window_services/test_openai_window_service.py +2 -1
- helm/benchmark/window_services/test_opt_window_service.py +3 -2
- helm/benchmark/window_services/test_palmyra_window_service.py +2 -1
- helm/benchmark/window_services/test_t0pp_window_service.py +2 -1
- helm/benchmark/window_services/test_t511b_window_service.py +2 -1
- helm/benchmark/window_services/test_ul2_window_service.py +2 -1
- helm/benchmark/window_services/test_utils.py +3 -2
- helm/benchmark/window_services/test_yalm_window_service.py +2 -1
- helm/benchmark/window_services/window_service.py +42 -0
- helm/benchmark/window_services/window_service_factory.py +4 -1
- helm/benchmark/window_services/yalm_window_service.py +0 -27
- helm/clients/__init__.py +0 -0
- helm/{proxy/clients → clients}/ai21_client.py +3 -9
- helm/clients/aleph_alpha_client.py +112 -0
- helm/{proxy/clients → clients}/anthropic_client.py +203 -18
- helm/{proxy/clients → clients}/auto_client.py +59 -31
- helm/clients/bedrock_client.py +128 -0
- helm/clients/bedrock_utils.py +72 -0
- helm/{proxy/clients → clients}/client.py +65 -7
- helm/clients/clip_score_client.py +49 -0
- helm/clients/clip_scorers/__init__.py +0 -0
- helm/clients/clip_scorers/base_clip_scorer.py +18 -0
- helm/clients/clip_scorers/clip_scorer.py +50 -0
- helm/clients/clip_scorers/multilingual_clip_scorer.py +50 -0
- helm/{proxy/clients → clients}/cohere_client.py +4 -11
- helm/clients/gcs_client.py +82 -0
- helm/{proxy/clients → clients}/google_client.py +5 -5
- helm/clients/google_translate_client.py +35 -0
- helm/{proxy/clients → clients}/http_model_client.py +5 -7
- helm/{proxy/clients → clients}/huggingface_client.py +43 -64
- helm/clients/image_generation/__init__.py +0 -0
- helm/clients/image_generation/adobe_vision_client.py +78 -0
- helm/clients/image_generation/aleph_alpha_image_generation_client.py +98 -0
- helm/clients/image_generation/cogview2/__init__.py +0 -0
- helm/clients/image_generation/cogview2/coglm_strategy.py +96 -0
- helm/clients/image_generation/cogview2/coglm_utils.py +82 -0
- helm/clients/image_generation/cogview2/sr_pipeline/__init__.py +15 -0
- helm/clients/image_generation/cogview2/sr_pipeline/direct_sr.py +96 -0
- helm/clients/image_generation/cogview2/sr_pipeline/dsr_model.py +254 -0
- helm/clients/image_generation/cogview2/sr_pipeline/dsr_sampling.py +190 -0
- helm/clients/image_generation/cogview2/sr_pipeline/iterative_sr.py +141 -0
- helm/clients/image_generation/cogview2/sr_pipeline/itersr_model.py +269 -0
- helm/clients/image_generation/cogview2/sr_pipeline/itersr_sampling.py +120 -0
- helm/clients/image_generation/cogview2/sr_pipeline/sr_group.py +42 -0
- helm/clients/image_generation/cogview2_client.py +191 -0
- helm/clients/image_generation/dalle2_client.py +192 -0
- helm/clients/image_generation/dalle3_client.py +108 -0
- helm/clients/image_generation/dalle_mini/__init__.py +3 -0
- helm/clients/image_generation/dalle_mini/data.py +442 -0
- helm/clients/image_generation/dalle_mini/model/__init__.py +5 -0
- helm/clients/image_generation/dalle_mini/model/configuration.py +175 -0
- helm/clients/image_generation/dalle_mini/model/modeling.py +1834 -0
- helm/clients/image_generation/dalle_mini/model/partitions.py +84 -0
- helm/clients/image_generation/dalle_mini/model/processor.py +63 -0
- helm/clients/image_generation/dalle_mini/model/text.py +251 -0
- helm/clients/image_generation/dalle_mini/model/tokenizer.py +9 -0
- helm/clients/image_generation/dalle_mini/model/utils.py +29 -0
- helm/clients/image_generation/dalle_mini/vqgan_jax/__init__.py +1 -0
- helm/clients/image_generation/dalle_mini/vqgan_jax/configuration_vqgan.py +40 -0
- helm/clients/image_generation/dalle_mini/vqgan_jax/convert_pt_model_to_jax.py +107 -0
- helm/clients/image_generation/dalle_mini/vqgan_jax/modeling_flax_vqgan.py +610 -0
- helm/clients/image_generation/dalle_mini_client.py +190 -0
- helm/clients/image_generation/deep_floyd_client.py +78 -0
- helm/clients/image_generation/huggingface_diffusers_client.py +249 -0
- helm/clients/image_generation/image_generation_client_utils.py +9 -0
- helm/clients/image_generation/lexica_client.py +86 -0
- helm/clients/image_generation/mindalle/__init__.py +0 -0
- helm/clients/image_generation/mindalle/models/__init__.py +216 -0
- helm/clients/image_generation/mindalle/models/stage1/__init__.py +0 -0
- helm/clients/image_generation/mindalle/models/stage1/layers.py +312 -0
- helm/clients/image_generation/mindalle/models/stage1/vqgan.py +103 -0
- helm/clients/image_generation/mindalle/models/stage2/__init__.py +0 -0
- helm/clients/image_generation/mindalle/models/stage2/layers.py +144 -0
- helm/clients/image_generation/mindalle/models/stage2/transformer.py +268 -0
- helm/clients/image_generation/mindalle/models/tokenizer.py +30 -0
- helm/clients/image_generation/mindalle/utils/__init__.py +3 -0
- helm/clients/image_generation/mindalle/utils/config.py +129 -0
- helm/clients/image_generation/mindalle/utils/sampling.py +149 -0
- helm/clients/image_generation/mindalle/utils/utils.py +89 -0
- helm/clients/image_generation/mindalle_client.py +115 -0
- helm/clients/image_generation/nudity_check_client.py +64 -0
- helm/clients/image_generation/together_image_generation_client.py +111 -0
- helm/{proxy/clients → clients}/lit_gpt_client.py +4 -4
- helm/{proxy/clients → clients}/megatron_client.py +5 -5
- helm/clients/mistral_client.py +134 -0
- helm/clients/moderation_api_client.py +109 -0
- helm/clients/open_lm_client.py +43 -0
- helm/clients/openai_client.py +302 -0
- helm/{proxy/clients → clients}/palmyra_client.py +6 -8
- helm/{proxy/clients → clients}/perspective_api_client.py +7 -8
- helm/clients/simple_client.py +64 -0
- helm/{proxy/clients → clients}/test_auto_client.py +13 -15
- helm/clients/test_client.py +100 -0
- helm/{proxy/clients → clients}/test_huggingface_client.py +15 -16
- helm/clients/test_simple_client.py +19 -0
- helm/{proxy/clients → clients}/test_together_client.py +20 -8
- helm/{proxy/clients → clients}/together_client.py +12 -72
- helm/clients/vertexai_client.py +391 -0
- helm/clients/vision_language/__init__.py +0 -0
- helm/clients/vision_language/huggingface_vlm_client.py +104 -0
- helm/{proxy/clients → clients}/vision_language/idefics_client.py +53 -48
- helm/clients/vision_language/open_flamingo/__init__.py +2 -0
- helm/clients/vision_language/open_flamingo/src/__init__.py +0 -0
- helm/clients/vision_language/open_flamingo/src/factory.py +147 -0
- helm/clients/vision_language/open_flamingo/src/flamingo.py +337 -0
- helm/clients/vision_language/open_flamingo/src/flamingo_lm.py +155 -0
- helm/clients/vision_language/open_flamingo/src/helpers.py +267 -0
- helm/clients/vision_language/open_flamingo/src/utils.py +47 -0
- helm/clients/vision_language/open_flamingo_client.py +155 -0
- helm/clients/vision_language/qwen_vlm_client.py +171 -0
- helm/clients/vllm_client.py +46 -0
- helm/common/cache.py +16 -4
- helm/common/cache_backend_config.py +47 -0
- helm/common/clip_score_request.py +41 -0
- helm/common/file_caches/__init__.py +0 -0
- helm/common/file_caches/file_cache.py +16 -0
- helm/common/file_caches/local_file_cache.py +61 -0
- helm/common/file_caches/test_local_file_cache.py +25 -0
- helm/common/file_upload_request.py +27 -0
- helm/common/general.py +1 -1
- helm/common/image_generation_parameters.py +25 -0
- helm/common/images_utils.py +24 -1
- helm/common/key_value_store.py +35 -4
- helm/common/media_object.py +13 -0
- helm/common/moderations_api_request.py +71 -0
- helm/common/mongo_key_value_store.py +3 -3
- helm/common/multimodal_request_utils.py +31 -0
- helm/common/nudity_check_request.py +29 -0
- helm/common/request.py +15 -17
- helm/common/test_general.py +6 -0
- helm/common/tokenization_request.py +1 -1
- helm/config/model_deployments.yaml +1069 -546
- helm/config/model_metadata.yaml +753 -31
- helm/config/tokenizer_configs.yaml +142 -43
- helm/proxy/accounts.py +31 -4
- helm/proxy/critique/mechanical_turk_critique_importer.py +3 -0
- helm/proxy/critique/model_critique_client.py +8 -6
- helm/proxy/example_queries.py +29 -17
- helm/proxy/server.py +70 -5
- helm/proxy/services/remote_service.py +31 -0
- helm/proxy/services/server_service.py +96 -16
- helm/proxy/services/service.py +30 -0
- helm/proxy/services/test_remote_service.py +4 -3
- helm/proxy/services/test_service.py +0 -12
- helm/proxy/test_accounts.py +32 -0
- helm/proxy/token_counters/auto_token_counter.py +37 -37
- helm/proxy/token_counters/test_auto_token_counter.py +164 -0
- helm/proxy/token_counters/token_counter.py +3 -5
- helm/tokenizers/__init__.py +0 -0
- helm/{proxy/tokenizers → tokenizers}/ai21_tokenizer.py +3 -3
- helm/{proxy/tokenizers → tokenizers}/anthropic_tokenizer.py +1 -1
- helm/{proxy/tokenizers → tokenizers}/auto_tokenizer.py +6 -9
- helm/{proxy/tokenizers → tokenizers}/cohere_tokenizer.py +1 -1
- helm/{proxy/tokenizers → tokenizers}/http_model_tokenizer.py +3 -3
- helm/{proxy/tokenizers → tokenizers}/huggingface_tokenizer.py +7 -26
- helm/tokenizers/simple_tokenizer.py +33 -0
- helm/{proxy/tokenizers → tokenizers}/test_anthropic_tokenizer.py +1 -1
- helm/{proxy/tokenizers → tokenizers}/test_huggingface_tokenizer.py +3 -0
- helm/tokenizers/test_simple_tokenizer.py +33 -0
- helm/{proxy/tokenizers → tokenizers}/vertexai_tokenizer.py +1 -1
- helm/{proxy/tokenizers → tokenizers}/yalm_tokenizer.py +5 -3
- helm/tokenizers/yalm_tokenizer_data/__init__.py +0 -0
- helm/tokenizers/yalm_tokenizer_data/voc_100b.sp +0 -0
- helm/{proxy/tokenizers → tokenizers}/yalm_tokenizer_data/yalm_tokenizer.py +1 -1
- crfm_helm-0.4.0.dist-info/RECORD +0 -397
- helm/benchmark/run_specs.py +0 -2762
- helm/benchmark/test_model_properties.py +0 -1570
- helm/benchmark/vlm_run_specs.py +0 -97
- helm/benchmark/window_services/flan_t5_window_service.py +0 -29
- helm/benchmark/window_services/gpt2_window_service.py +0 -32
- helm/benchmark/window_services/huggingface_window_service.py +0 -60
- helm/benchmark/window_services/t0pp_window_service.py +0 -35
- helm/benchmark/window_services/t511b_window_service.py +0 -30
- helm/benchmark/window_services/test_mt_nlg_window_service.py +0 -48
- helm/benchmark/window_services/ul2_window_service.py +0 -30
- helm/benchmark/window_services/wider_ai21_window_service.py +0 -24
- helm/common/cache_utils.py +0 -14
- helm/proxy/clients/aleph_alpha_client.py +0 -95
- helm/proxy/clients/goose_ai_client.py +0 -99
- helm/proxy/clients/microsoft_client.py +0 -180
- helm/proxy/clients/openai_client.py +0 -206
- helm/proxy/clients/simple_client.py +0 -60
- helm/proxy/clients/test_client.py +0 -49
- helm/proxy/clients/vertexai_client.py +0 -115
- helm/proxy/token_counters/ai21_token_counter.py +0 -20
- helm/proxy/token_counters/cohere_token_counter.py +0 -13
- helm/proxy/token_counters/free_token_counter.py +0 -12
- helm/proxy/token_counters/gooseai_token_counter.py +0 -24
- helm/proxy/token_counters/openai_token_counter.py +0 -22
- helm/proxy/token_counters/test_ai21_token_counter.py +0 -88
- helm/proxy/token_counters/test_openai_token_counter.py +0 -81
- helm/proxy/tokenizers/simple_tokenizer.py +0 -32
- {crfm_helm-0.4.0.dist-info → crfm_helm-0.5.0.dist-info}/LICENSE +0 -0
- {crfm_helm-0.4.0.dist-info → crfm_helm-0.5.0.dist-info}/entry_points.txt +0 -0
- {crfm_helm-0.4.0.dist-info → crfm_helm-0.5.0.dist-info}/top_level.txt +0 -0
- /helm/{proxy/clients → benchmark/annotation}/__init__.py +0 -0
- /helm/{proxy/clients/vision_language → benchmark/annotation/image2structure}/__init__.py +0 -0
- /helm/{proxy/tokenizers → benchmark/metrics/image_generation}/__init__.py +0 -0
- /helm/{proxy/tokenizers/yalm_tokenizer_data → benchmark/metrics/image_generation/detectors}/__init__.py +0 -0
- /helm/{proxy/clients → clients}/ai21_utils.py +0 -0
- /helm/{proxy/clients → clients}/cohere_utils.py +0 -0
- /helm/{proxy/clients → clients}/lit_gpt_generate.py +0 -0
- /helm/{proxy/clients → clients}/toxicity_classifier_client.py +0 -0
- /helm/{proxy/tokenizers → tokenizers}/aleph_alpha_tokenizer.py +0 -0
- /helm/{proxy/tokenizers → tokenizers}/caching_tokenizer.py +0 -0
- /helm/{proxy/tokenizers → tokenizers}/ice_tokenizer.py +0 -0
- /helm/{proxy/tokenizers → tokenizers}/lit_gpt_tokenizer.py +0 -0
- /helm/{proxy/tokenizers → tokenizers}/test_ice_tokenizer.py +0 -0
- /helm/{proxy/tokenizers → tokenizers}/test_yalm_tokenizer.py +0 -0
- /helm/{proxy/tokenizers → tokenizers}/tiktoken_tokenizer.py +0 -0
- /helm/{proxy/tokenizers → tokenizers}/tokenizer.py +0 -0
- /helm/{proxy/tokenizers → tokenizers}/yalm_tokenizer_data/test_yalm_tokenizer.py +0 -0
helm/benchmark/slurm_runner.py
CHANGED
|
@@ -10,6 +10,10 @@ import sys
|
|
|
10
10
|
|
|
11
11
|
from helm.common.codec import from_json, to_json
|
|
12
12
|
from helm.common.general import write
|
|
13
|
+
from helm.benchmark.config_registry import (
|
|
14
|
+
register_configs_from_directory,
|
|
15
|
+
register_builtin_configs_from_helm_package,
|
|
16
|
+
)
|
|
13
17
|
from helm.benchmark.executor import ExecutionSpec
|
|
14
18
|
from helm.benchmark.runner import Runner, RunSpec, RunnerError
|
|
15
19
|
from helm.benchmark.slurm_jobs import (
|
|
@@ -24,10 +28,11 @@ from helm.benchmark.slurm_jobs import (
|
|
|
24
28
|
from helm.common.general import ensure_directory_exists
|
|
25
29
|
from helm.common.hierarchical_logger import hlog, htrack_block
|
|
26
30
|
|
|
31
|
+
from helm.benchmark.runner_config_registry import RUNNER_CONFIG
|
|
27
32
|
|
|
28
|
-
_DEFAULT_MAX_CONCURRENT_WORKER_SLURM_JOBS = 8
|
|
29
33
|
_MAX_CONCURRENT_WORKER_SLURM_JOBS_ENV_NAME = "HELM_MAX_CONCURRENT_WORKER_SLURM_JOBS"
|
|
30
34
|
_SLURM_NODE_NAMES_ENV_NAME = "HELM_SLURM_NODE_NAMES"
|
|
35
|
+
_DEFAULT_MAX_CONCURRENT_WORKER_SLURM = 8
|
|
31
36
|
|
|
32
37
|
|
|
33
38
|
@dataclass
|
|
@@ -89,12 +94,15 @@ class SlurmRunner(Runner):
|
|
|
89
94
|
self.slurm_runner_spec_path = os.path.join(self.slurm_base_dir, "slurm_runner_spec.json")
|
|
90
95
|
|
|
91
96
|
# Configure max concurrent worker Slurm jobs from the environment variable.
|
|
92
|
-
# TODO: Read from a configuration file instead
|
|
93
97
|
env_max_concurrent_worker_slurm_jobs = os.getenv(_MAX_CONCURRENT_WORKER_SLURM_JOBS_ENV_NAME)
|
|
94
98
|
self.max_concurrent_worker_slurm_jobs = (
|
|
95
99
|
int(env_max_concurrent_worker_slurm_jobs)
|
|
96
100
|
if env_max_concurrent_worker_slurm_jobs
|
|
97
|
-
else
|
|
101
|
+
else (
|
|
102
|
+
RUNNER_CONFIG.helm_max_concurrent_workers
|
|
103
|
+
if RUNNER_CONFIG.helm_max_concurrent_workers > 0
|
|
104
|
+
else _DEFAULT_MAX_CONCURRENT_WORKER_SLURM
|
|
105
|
+
)
|
|
98
106
|
)
|
|
99
107
|
|
|
100
108
|
def run_all(self, run_specs: List[RunSpec]):
|
|
@@ -222,8 +230,7 @@ class SlurmRunner(Runner):
|
|
|
222
230
|
break
|
|
223
231
|
|
|
224
232
|
# Refresh every minute
|
|
225
|
-
|
|
226
|
-
time.sleep(60)
|
|
233
|
+
time.sleep(RUNNER_CONFIG.slurm_monitor_interval)
|
|
227
234
|
finally:
|
|
228
235
|
# Cleanup by cancelling all jobs during program termination or if an exception is raised.
|
|
229
236
|
cancel_all_jobs()
|
|
@@ -261,34 +268,48 @@ class SlurmRunner(Runner):
|
|
|
261
268
|
run_spec_path,
|
|
262
269
|
]
|
|
263
270
|
)
|
|
264
|
-
|
|
265
|
-
|
|
266
|
-
|
|
267
|
-
|
|
268
|
-
|
|
269
|
-
|
|
270
|
-
|
|
271
|
-
|
|
272
|
-
|
|
273
|
-
|
|
274
|
-
|
|
275
|
-
|
|
276
|
-
|
|
277
|
-
|
|
278
|
-
|
|
279
|
-
|
|
280
|
-
|
|
281
|
-
|
|
282
|
-
|
|
283
|
-
|
|
284
|
-
|
|
285
|
-
|
|
286
|
-
|
|
287
|
-
|
|
288
|
-
|
|
289
|
-
|
|
290
|
-
|
|
291
|
-
|
|
271
|
+
if RUNNER_CONFIG.slurm_args is None:
|
|
272
|
+
raw_slurm_args: Dict[str, str] = {
|
|
273
|
+
"account": "nlp",
|
|
274
|
+
"cpus_per_task": "4",
|
|
275
|
+
"mem": "32G",
|
|
276
|
+
"gres": "gpu:0",
|
|
277
|
+
"open_mode": "append",
|
|
278
|
+
"partition": "john",
|
|
279
|
+
"time": "14-0", # Deadline of 14 days
|
|
280
|
+
"mail_type": "FAIL",
|
|
281
|
+
"job_name": run_name,
|
|
282
|
+
"output": log_path,
|
|
283
|
+
"chdir": os.getcwd(),
|
|
284
|
+
}
|
|
285
|
+
# TODO: Move resource requirements into RunSpec.
|
|
286
|
+
slurm_node_names = os.getenv(_SLURM_NODE_NAMES_ENV_NAME)
|
|
287
|
+
if run_spec.name.startswith("msmarco:"):
|
|
288
|
+
raw_slurm_args["mem"] = "64G"
|
|
289
|
+
if "device=cuda" in run_spec.name:
|
|
290
|
+
raw_slurm_args["gres"] = "gpu:1"
|
|
291
|
+
raw_slurm_args["partition"] = "jag-hi"
|
|
292
|
+
if "model=huggingface" in run_spec.name:
|
|
293
|
+
raw_slurm_args["gres"] = "gpu:1"
|
|
294
|
+
raw_slurm_args["partition"] = "sphinx"
|
|
295
|
+
if not slurm_node_names or "sphinx" not in slurm_node_names:
|
|
296
|
+
raise Exception(
|
|
297
|
+
f"Environment variable {_SLURM_NODE_NAMES_ENV_NAME} must be set to sphinx node names"
|
|
298
|
+
)
|
|
299
|
+
if slurm_node_names:
|
|
300
|
+
raw_slurm_args["nodelist"] = slurm_node_names
|
|
301
|
+
|
|
302
|
+
else:
|
|
303
|
+
raw_slurm_args = RUNNER_CONFIG.slurm_args
|
|
304
|
+
|
|
305
|
+
dynamic_slurm_args = {
|
|
306
|
+
"job_name": run_name,
|
|
307
|
+
"output": log_path,
|
|
308
|
+
"chdir": os.getcwd(),
|
|
309
|
+
}
|
|
310
|
+
|
|
311
|
+
# User should not set these manually, overwrite them if necessary
|
|
312
|
+
raw_slurm_args.update(dynamic_slurm_args)
|
|
292
313
|
|
|
293
314
|
slurm_args: Dict[str, str] = {key: shlex.quote(value) for key, value in raw_slurm_args.items()}
|
|
294
315
|
# Uncomment this to get notification emails from Slurm for Slurm worker jobs.
|
|
@@ -300,27 +321,15 @@ class SlurmRunner(Runner):
|
|
|
300
321
|
return slurm_job_id
|
|
301
322
|
|
|
302
323
|
|
|
303
|
-
def run_as_worker(slurm_runner_spec_path: str, run_spec_path: str):
|
|
304
|
-
"""Deserialize SlurmRunner and RunSpec from the given files, then run the RunSpec with the SlurmRunner.
|
|
305
|
-
|
|
306
|
-
Used by the worker Slurm jobs only."""
|
|
307
|
-
with open(slurm_runner_spec_path, "r") as f:
|
|
308
|
-
slurm_runner_spec = from_json(f.read(), SlurmRunnerSpec)
|
|
309
|
-
with open(run_spec_path, "r") as f:
|
|
310
|
-
run_spec = from_json(f.read(), RunSpec)
|
|
311
|
-
slurm_runner = SlurmRunner(**slurm_runner_spec.to_kwargs())
|
|
312
|
-
slurm_runner.run_one(run_spec)
|
|
313
|
-
|
|
314
|
-
|
|
315
324
|
def main():
|
|
316
325
|
"""Entry point for the SlurmRunner's worker Slurm jobs that run a single RunSpec.
|
|
317
326
|
|
|
318
327
|
This entry point should only be used by SlurmRunner. Users should use `helm-run` instead.
|
|
319
328
|
SlurmRunner has to use this entry point instead of helm-run because there is no way to
|
|
320
329
|
specify the worker Slurm job parameters through `helm-run`. In particular, there is no way
|
|
321
|
-
to run a specific `RunSpec` using the `--run-
|
|
322
|
-
`run-
|
|
323
|
-
convert a `RunSpec` into a `
|
|
330
|
+
to run a specific `RunSpec` using the `--run-entries` parameter of `helm-run`, because the
|
|
331
|
+
`run-entries` argument contains `RunEntry` description (not `RunSpec`s), and there is no way to
|
|
332
|
+
convert a `RunSpec` into a `RunEntry` description."""
|
|
324
333
|
parser = argparse.ArgumentParser()
|
|
325
334
|
parser.add_argument(
|
|
326
335
|
"--slurm-runner-spec-path",
|
|
@@ -335,7 +344,19 @@ def main():
|
|
|
335
344
|
required=True,
|
|
336
345
|
)
|
|
337
346
|
args = parser.parse_args()
|
|
338
|
-
|
|
347
|
+
|
|
348
|
+
# Deserialize SlurmRunner and RunSpec from the given files, then run the RunSpec with the SlurmRunner.
|
|
349
|
+
with open(args.slurm_runner_spec_path, "r") as f:
|
|
350
|
+
slurm_runner_spec = from_json(f.read(), SlurmRunnerSpec)
|
|
351
|
+
with open(args.run_spec_path, "r") as f:
|
|
352
|
+
run_spec = from_json(f.read(), RunSpec)
|
|
353
|
+
|
|
354
|
+
register_builtin_configs_from_helm_package()
|
|
355
|
+
if slurm_runner_spec.execution_spec.local_path is not None:
|
|
356
|
+
register_configs_from_directory(slurm_runner_spec.execution_spec.local_path)
|
|
357
|
+
|
|
358
|
+
slurm_runner = SlurmRunner(**slurm_runner_spec.to_kwargs())
|
|
359
|
+
slurm_runner.run_one(run_spec)
|
|
339
360
|
|
|
340
361
|
|
|
341
362
|
if __name__ == "__main__":
|