| { | |
| "_attn_implementation_autoset": true, | |
| "ae_mode": "token", | |
| "attn_implementation": "flash_attention_2", | |
| "auto_map": { | |
| "AutoConfig": "modelling_pisco.COCOMConfig", | |
| "AutoModel": "modelling_pisco.COCOM" | |
| }, | |
| "compr_base_model_name": "/root/autodl-fs/models/Mistral-7B-Instruct-v0.2", | |
| "compr_every_n_layer": null, | |
| "compr_linear_type": "concat", | |
| "compr_mlp_hidden_dim": 8096, | |
| "compr_model_name": null, | |
| "compr_n_layers": null, | |
| "compr_rate": 16, | |
| "compr_rms_norm": false, | |
| "compr_use_mlp": true, | |
| "decoder_model_name": "/root/autodl-fs/models/Mistral-7B-Instruct-v0.2", | |
| "device_map": null, | |
| "different_mem_tokens": true, | |
| "doc_max_length": 128, | |
| "generation_top_k": 30, | |
| "kbtc_training": false, | |
| "load_adapters": true, | |
| "lora": true, | |
| "lora_compressor": false, | |
| "lora_r": 16, | |
| "lora_r_compressor": 16, | |
| "max_new_tokens": 128, | |
| "model_type": "COCOM", | |
| "optimize_mem_tokens": false, | |
| "quantization": "no", | |
| "sep": true, | |
| "training_form": "both_separately", | |
| "transformers_version": "4.57.6" | |
| } | |