| dataset_path: lmms-lab/LMMs-Eval-Lite | |
| task: "ai2d_lite" | |
| dataset_kwargs: | |
| token: True | |
| dataset_name: ai2d | |
| test_split: lite | |
| output_type: generate_until | |
| doc_to_visual: !function utils.ai2d_doc_to_visual | |
| doc_to_text: !function utils.ai2d_doc_to_text | |
| doc_to_target: !function utils.ai2d_doc_to_target | |
| lmms_eval_specific_kwargs: | |
| default: | |
| prompt_format: mcq | |
| pre_prompt: "" | |
| post_prompt: "\nAnswer with the option's letter from the given choices directly." | |
| gpt4v: | |
| prompt_format: mcq | |
| pre_prompt: "" | |
| post_prompt: "\nAbove choices are given in {option}. {content} format.\nPlease answer with the option letter from the given choices directly." | |
| qwen_vl: | |
| prompt_format: qa | |
| pre_prompt: "" | |
| post_prompt: " Answer:" | |
| xcomposer2_4khd: | |
| prompt_format: mcq_xcomposer | |
| pre_prompt: "[UNUSED_TOKEN_146]user\nQuestion: " | |
| post_prompt: "[UNUSED_TOKEN_145]\n[UNUSED_TOKEN_146]assistant\nThe answer is" | |
| model_specific_target_kwargs: | |
| default: "mcq" | |
| qwen_vl: "qa" | |
| generation_kwargs: | |
| max_new_tokens: 16 | |
| temperature: 0 | |
| do_sample: False | |
| filter_list: | |
| - name: "flexible-extract" | |
| filter: | |
| - function: !function utils.MultiChoiceRegexFilter | |
| group_select: 0 | |
| ignore_case: true | |
| ignore_punctuation: true | |
| regex_pattern: "([A-Z])\\." | |
| metric_list: | |
| - metric: exact_match | |
| aggregation: mean | |
| higher_is_better: true | |
| ignore_case: true | |
| ignore_punctuation: true | |
| metadata: | |
| - version: 0.0 |