Wrong humaneval in the model description
Far from 50% as described.
Evaluating generations...
{
"humaneval": {
"pass@1": 0.4024390243902439
},
"config": {
"prefix": "",
"do_sample": true,
"temperature": 0.2,
"top_k": 0,
"top_p": 1.0,
"n_samples": 1,
"eos": "<|endoftext|>",
"seed": 0,
"model": "teknium/OpenHermes-2.5-Mistral-7B",
"modeltype": "causal",
"peft_model": null,
"revision": null,
"use_auth_token": false,
"trust_remote_code": false,
"tasks": null,
"instruction_tokens": null,
"precision": "fp32",
"load_in_8bit": false,
"load_in_4bit": false,
"left_padding": false,
"limit_start": 0,
"save_every_k_tasks": -1,
"postprocess": true,
"generation_only": false,
"load_generations_path": null,
"load_data_path": null,
"load_generations_intermediate_paths": null,
"prompt": "prompt",
"max_memory_per_gpu": null,
"check_references": false,
"batch_size": 1,
"max_length_generation": 512,
"allow_code_execution": true,
"save_generations": true,
"save_generations_path": "humaneval_generations.json",
"save_references": true,
"save_references_path": "humaneval_references.json",
"metric_output_path": "evaluation_results.json",
"limit": 200
}