| { | |
| "model": "internlm2", | |
| "bias": false, | |
| "bos_token_id": 1, | |
| "eos_token_id": 2, | |
| "pad_token_id": 2, | |
| "hidden_act": "silu", | |
| "hidden_size": 2048, | |
| "initializer_range": 0.02, | |
| "intermediate_size": 8192, | |
| "max_position_embeddings": 32768, | |
| "num_attention_heads": 16, | |
| "num_hidden_layers": 24, | |
| "num_key_value_heads": 8, | |
| "layer_norm_eps": 1e-05, | |
| "rope_scaling": { | |
| "type": "dynamic", | |
| "factor": 2.0 | |
| }, | |
| "rope_theta": 1000000, | |
| "tie_word_embeddings": false, | |
| "segment_vocab_size": 0, | |
| "skip_init": true, | |
| "rope_rank": "updown", | |
| "torch_dtype": "float16", | |
| "vocab_size": 92544, | |
| "generation_config": {"eos_token_id": [2, 92542], "max_length": 32768} | |
| } |