21 lines
807 B
JSON
21 lines
807 B
JSON
{
|
|
"model_kind": "merged_lora",
|
|
"projector_type": "C",
|
|
"checkpoint_choice": "best_by_retrieval",
|
|
"checkpoint_path": "E:\\brain\\brainalign\\project\\outputs\\stage2_lastblock_C_deepgpu\\best_by_retrieval.pt",
|
|
"stage2_dir": "E:\\brain\\brainalign\\project\\outputs\\stage2_lastblock_C_deepgpu",
|
|
"qwen_model_path": "E:\\brain\\brainalign\\models\\Qwen2.5-1.5B-Instruct",
|
|
"base_repo_id": "Qwen/Qwen2.5-1.5B-Instruct",
|
|
"adapter_backend": "peft",
|
|
"stage1_head_name": "mean5_pca1024_contrastive_seed42",
|
|
"lora_r": 16,
|
|
"lora_alpha": 32,
|
|
"lora_dropout": 0.05,
|
|
"torch_dtype": "bfloat16",
|
|
"export_time": "2026-05-14T19:55:09",
|
|
"verification": {
|
|
"config_class": "Qwen2Config",
|
|
"tokenizer_class": "Qwen2TokenizerFast",
|
|
"model_class": "Qwen2ForCausalLM"
|
|
}
|
|
} |