{ "config_general": { "model_name": "Lazycuber/L2-7b-Guanaco-Uncensored", "model_sha": "0dd5d093a206aa7930c3e13b0cf4e5a4c7945b98", "model_size": "12.61 GB", "model_dtype": "torch.float16", "lighteval_sha": "0f318ecf002208468154899217b3ba7c6ae09374", "num_few_shot_default": 0, "num_fewshot_seeds": 1, "override_batch_size": 1, "max_samples": null, "job_id": "" }, "results": { "harness|drop|3": { "em": 0.0014681208053691276, "em_stderr": 0.00039210421902985767, "f1": 0.056828859060402796, "f1_stderr": 0.0013179206618636607 }, "harness|gsm8k|5": { "acc": 0.07960576194086429, "acc_stderr": 0.007455924338676278 }, "harness|winogrande|5": { "acc": 0.7537490134175217, "acc_stderr": 0.012108365307437523 }, "all": { "em": 0.0014681208053691276, "em_stderr": 0.00039210421902985767, "f1": 0.056828859060402796, "f1_stderr": 0.0013179206618636607, "acc": 0.416677387679193, "acc_stderr": 0.0097821448230569 } }, "versions": { "harness|drop|3": 1, "harness|gsm8k|5": 0, "harness|winogrande|5": 0, "all": 0 }, "config_tasks": { "harness|drop": "LM Harness task", "harness|gsm8k": "LM Harness task", "harness|winogrande": "LM Harness task" }, "summary_tasks": { "harness|drop|3": { "hashes": { "hash_examples": "1d27416e8324e9a3", "hash_full_prompts": "a5513ff9a741b385", "hash_input_tokens": "42076f0efbb50aa6", "hash_cont_tokens": "6c4c0a15a547258d" }, "truncated": 3, "non-truncated": 9533, "padded": 0, "non-padded": 9536, "effective_few_shots": 3.0, "num_truncated_few_shots": 0 }, "harness|gsm8k|5": { "hashes": { "hash_examples": "4c0843a5d99bcfdc", "hash_full_prompts": "41d55e83abc0e02d", "hash_input_tokens": "bda342e47b5099b2", "hash_cont_tokens": "dde6454510196e7c" }, "truncated": 0, "non-truncated": 1319, "padded": 0, "non-padded": 1319, "effective_few_shots": 5.0, "num_truncated_few_shots": 0 }, "harness|winogrande|5": { "hashes": { "hash_examples": "aada0a176fd81218", "hash_full_prompts": "c8655cbd12de8409", "hash_input_tokens": "c0bedf98cb040854", "hash_cont_tokens": "f08975ad6f2d5864" }, "truncated": 0, "non-truncated": 2534, "padded": 2432, "non-padded": 102, "effective_few_shots": 5.0, "num_truncated_few_shots": 0 } }, "summary_general": { "hashes": { "hash_examples": "9b4d8993161e637d", "hash_full_prompts": "08215e527b7e60a5", "hash_input_tokens": "a12f3e3c934bd78b", "hash_cont_tokens": "21dd8d5546892d1c" }, "total_evaluation_time_secondes": "10531.30201625824", "truncated": 3, "non-truncated": 13386, "padded": 2432, "non-padded": 10957, "num_truncated_few_shots": 0 } }