test_multigpu_config.json

{
    "ruadapt_base_model_name_or_path": "RefalMachine/ruadapt_qwen2.5_3B_ext_u48_full_lr5e4_peft_mlp_32_32_bs256",
    "raw_base_model_name_or_path": "Qwen/Qwen2.5-3B",
    "instruct_model_name_or_path": "Qwen/Qwen2.5-3B-Instruct",
    "custom_chat_template_path": "ruadapt/ushanka/custom_chat_templates/qwen_2.5_instruct_no_system.json",
    "output_dir": "/workdir/data/models/qwen/ruadapt_qwen2.5_3B_ext_u48_full_lr5e4_peft_mlp_32_32_bs256_as1.75_test_multigpu_short",
    "pipeline_config_path": "pipeline_configs/qwen_pipeline_config_3b_instruct_v3_transformers_test.json",
    "alpaca_eval_questions_path": "../saiga/llmarena_questions.json",
    "custom_bos_token": "<|endoftext|>",
    "custom_eos_token": null,
    "custom_pad_token": "<|endoftext|>",
    "skip_lep": false,
    "eval": false,
    "sample_rate": 1.0,
    "alpha_scale": 1.75,
    "not_scale_lm_head": false,
    "num_gpu": 2
}