| { | |
| "architectures": [ | |
| "LlamaForCausalLM" | |
| ], | |
| "attention_bias": false, | |
| "attention_dropout": 0.0, | |
| "bos_token_id": 100257, | |
| "eos_token_id": 100257, | |
| "hidden_act": "silu", | |
| "hidden_size": 6144, | |
| "initializer_range": 0.006, | |
| "intermediate_size": 19648, | |
| "max_position_embeddings": 32768, | |
| "model_type": "llama", | |
| "num_attention_heads": 64, | |
| "num_hidden_layers": 44, | |
| "num_key_value_heads": 8, | |
| "pretraining_tp": 1, | |
| "rms_norm_eps": 1e-05, | |
| "rope_scaling": null, | |
| "rope_theta": 8000000, | |
| "tie_word_embeddings": false, | |
| "torch_dtype": "bfloat16", | |
| "use_cache": true, | |
| "vocab_size": 100352, | |
| "quantization_config": { | |
| "quant_method": "exl3", | |
| "version": "0.0.1", | |
| "bits": 4.0, | |
| "head_bits": 6, | |
| "calibration": { | |
| "rows": 100, | |
| "cols": 2048 | |
| }, | |
| "out_scales": "auto" | |
| } | |
| } |