edbeeching's picture
edbeeching HF staff
Upload eval_results/HuggingFaceH4/mistral-7b-ift/v41.19/arc/results_2024-03-10T10-19-03.380911.json with huggingface_hub
cbb62c4 verified
raw
history blame
2.31 kB
{
"config_general": {
"lighteval_sha": "?",
"num_fewshot_seeds": 1,
"override_batch_size": 1,
"max_samples": null,
"job_id": "",
"start_time": 90298.766914565,
"end_time": 90418.615718823,
"total_evaluation_time_secondes": "119.84880425799929",
"model_name": "HuggingFaceH4/mistral-7b-ift",
"model_sha": "f0d1762aee3d7ab24ac13b2521b76c9b32c8bf25",
"model_dtype": "torch.bfloat16",
"model_size": "13.99 GB",
"config": null
},
"results": {
"lighteval|arc:challenge|25": {
"acc": 0.4931740614334471,
"acc_stderr": 0.014610029151379813,
"acc_norm": 0.5221843003412969,
"acc_norm_stderr": 0.01459700192707613
}
},
"versions": {
"lighteval|arc:challenge|25": 0
},
"config_tasks": {
"lighteval|arc:challenge": {
"name": "arc:challenge",
"prompt_function": "arc",
"hf_repo": "ai2_arc",
"hf_subset": "ARC-Challenge",
"metric": [
"loglikelihood_acc",
"loglikelihood_acc_norm_nospace"
],
"hf_avail_splits": [
"train",
"test"
],
"evaluation_splits": [
"test"
],
"few_shots_split": null,
"few_shots_select": "random_sampling_from_train",
"generation_size": 1,
"stop_sequence": [
"\n"
],
"output_regex": null,
"frozen": false,
"suite": [
"lighteval",
"arc"
],
"original_num_docs": 1172,
"effective_num_docs": 1172
}
},
"summary_tasks": {
"lighteval|arc:challenge|25": {
"hashes": {
"hash_examples": "17b0cae357c0259e",
"hash_full_prompts": "f42cdfcbcc9b6fa4",
"hash_input_tokens": "bedd26d7d7ad2546",
"hash_cont_tokens": "8b6f018bc80a3417"
},
"truncated": 0,
"non_truncated": 1172,
"padded": 4660,
"non_padded": 27,
"effective_few_shots": 25.0,
"num_truncated_few_shots": 0
}
},
"summary_general": {
"hashes": {
"hash_examples": "aaa6929c6d3771fb",
"hash_full_prompts": "099f2d4c459a2c1d",
"hash_input_tokens": "7eee0cf412c3eaf0",
"hash_cont_tokens": "6857c42b72c1a62a"
},
"truncated": 0,
"non_truncated": 1172,
"padded": 4660,
"non_padded": 27,
"num_truncated_few_shots": 0
}
}