config.json: drop the float quantization_config.bits (the Hub requires an int); 3.5 stays in quantization_config.json
6f4bc66 verified | { | |
| "architectures": [ | |
| "LlamaForCausalLM" | |
| ], | |
| "attention_bias": false, | |
| "attention_dropout": 0.0, | |
| "bos_token_id": 166100, | |
| "embd_pdrop": 0.0, | |
| "eos_token_id": 166101, | |
| "ffn_hidden_size": 22016, | |
| "head_dim": 128, | |
| "hidden_act": "silu", | |
| "hidden_size": 7168, | |
| "initializer_range": 0.02, | |
| "intermediate_size": 22016, | |
| "kv_channels": 128, | |
| "max_position_embeddings": 524288, | |
| "mlp_bias": false, | |
| "model_type": "llama", | |
| "num_attention_heads": 56, | |
| "num_hidden_layers": 52, | |
| "num_key_value_heads": 8, | |
| "num_query_groups": 8, | |
| "pad_token_id": 0, | |
| "pretraining_tp": 1, | |
| "resid_pdrop": 0.0, | |
| "rms_norm_eps": 1e-05, | |
| "rope_scaling": null, | |
| "rope_theta": 120000000, | |
| "rotary_base": 120000000, | |
| "seq_length": 524288, | |
| "tie_word_embeddings": false, | |
| "torch_dtype": "bfloat16", | |
| "transformers_version": "4.51.1", | |
| "use_cache": false, | |
| "vocab_size": 166912, | |
| "quantization_config": { | |
| "quant_method": "exl3", | |
| "version": "1.6.5", | |
| "head_bits": 6, | |
| "calibration": { | |
| "rows": 250, | |
| "cols": 2048 | |
| }, | |
| "out_scales": "auto", | |
| "codebook": "mul1" | |
| } | |
| } |