winglian/basilisk-4b

Basilisk 4B

Built on winglian/llama-2-4b, a 4B parameter Llama-2 model, this model is finetuned with open orca CoT data.

hf-causal-experimental (pretrained=winglian/basilisk-4b,use_accelerate=True,trust_remote_code=True), limit: None, provide_description: False, num_fewshot: 0, batch_size: None
|                      Task                      |Version|       Metric        |Value |   |Stderr|
|------------------------------------------------|------:|---------------------|-----:|---|-----:|
|agieval_aqua_rat                                |      0|acc                  |0.2362|_  |0.0267|
|                                                |       |acc_norm             |0.2283|_  |0.0264|
|agieval_logiqa_en                               |      0|acc                  |0.2688|_  |0.0174|
|                                                |       |acc_norm             |0.2811|_  |0.0176|
|agieval_lsat_ar                                 |      0|acc                  |0.2130|_  |0.0271|
|                                                |       |acc_norm             |0.1913|_  |0.0260|
|agieval_lsat_lr                                 |      0|acc                  |0.2255|_  |0.0185|
|                                                |       |acc_norm             |0.2745|_  |0.0198|
|agieval_lsat_rc                                 |      0|acc                  |0.2305|_  |0.0257|
|                                                |       |acc_norm             |0.2491|_  |0.0264|
|agieval_sat_en                                  |      0|acc                  |0.3641|_  |0.0336|
|                                                |       |acc_norm             |0.3495|_  |0.0333|
|agieval_sat_en_without_passage                  |      0|acc                  |0.2427|_  |0.0299|
|                                                |       |acc_norm             |0.2427|_  |0.0299|
|agieval_sat_math                                |      0|acc                  |0.2318|_  |0.0285|
|                                                |       |acc_norm             |0.2091|_  |0.0275|
|bigbench_causal_judgement                       |      0|multiple_choice_grade|0.5000|_  |0.0364|
|bigbench_date_understanding                     |      0|multiple_choice_grade|0.3930|_  |0.0255|
|bigbench_disambiguation_qa                      |      0|multiple_choice_grade|0.2674|_  |0.0276|
|bigbench_geometric_shapes                       |      0|multiple_choice_grade|0.1838|_  |0.0205|
|                                                |       |exact_str_match      |0.0279|_  |0.0087|
|bigbench_logical_deduction_five_objects         |      0|multiple_choice_grade|0.2380|_  |0.0191|
|bigbench_logical_deduction_seven_objects        |      0|multiple_choice_grade|0.1843|_  |0.0147|
|bigbench_logical_deduction_three_objects        |      0|multiple_choice_grade|0.3800|_  |0.0281|
|bigbench_movie_recommendation                   |      0|multiple_choice_grade|0.3480|_  |0.0213|
|bigbench_navigate                               |      0|multiple_choice_grade|0.5000|_  |0.0158|
|bigbench_reasoning_about_colored_objects        |      0|multiple_choice_grade|0.3680|_  |0.0108|
|bigbench_ruin_names                             |      0|multiple_choice_grade|0.2746|_  |0.0211|
|bigbench_salient_translation_error_detection    |      0|multiple_choice_grade|0.2806|_  |0.0142|
|bigbench_snarks                                 |      0|multiple_choice_grade|0.4972|_  |0.0373|
|bigbench_sports_understanding                   |      0|multiple_choice_grade|0.4939|_  |0.0159|
|bigbench_temporal_sequences                     |      0|multiple_choice_grade|0.2740|_  |0.0141|
|bigbench_tracking_shuffled_objects_five_objects |      0|multiple_choice_grade|0.1904|_  |0.0111|
|bigbench_tracking_shuffled_objects_seven_objects|      0|multiple_choice_grade|0.1394|_  |0.0083|
|bigbench_tracking_shuffled_objects_three_objects|      0|multiple_choice_grade|0.3800|_  |0.0281|

hf-causal-experimental (pretrained=winglian/basilisk-4b,use_accelerate=True,trust_remote_code=True), limit: None, provide_description: False, num_fewshot: 0, batch_size: 12
|    Task     |Version| Metric |Value |   |Stderr|
|-------------|------:|--------|-----:|---|-----:|
|arc_challenge|      0|acc     |0.3285|_  |0.0137|
|             |       |acc_norm|0.3532|_  |0.0140|
|arc_easy     |      0|acc     |0.6364|_  |0.0099|
|             |       |acc_norm|0.6035|_  |0.0100|
|boolq        |      1|acc     |0.7196|_  |0.0079|
|hellaswag    |      0|acc     |0.4239|_  |0.0049|
|             |       |acc_norm|0.5473|_  |0.0050|
|openbookqa   |      0|acc     |0.2220|_  |0.0186|
|             |       |acc_norm|0.3320|_  |0.0211|
|piqa         |      0|acc     |0.6937|_  |0.0108|
|             |       |acc_norm|0.6921|_  |0.0108|
|winogrande   |      0|acc     |0.5399|_  |0.0140|