Peaceuai/GLM_model4
05
1{2 "add_bias_linear": false,3 "add_qkv_bias": true,4 "apply_query_key_layer_scaling": true,5 "apply_residual_connection_post_layernorm": false,6 "architectures": [7 "CharacterGLMForConditionalGeneration"8 ],9 "attention_dropout": 0.0,10 "attention_softmax_in_fp32": true,11 "auto_map": {12 "AutoConfig": "configuration_chatglm.ChatGLMConfig",13 "AutoModel": "modeling_characterglm.CharacterGLMForConditionalGeneration",14 "AutoModelForCausalLM": "thu-coai/CharacterGLM-6B--modeling_characterglm.CharacterGLMForConditionalGeneration",15 "AutoModelForSeq2SeqLM": "thu-coai/CharacterGLM-6B--modeling_characterglm.CharacterGLMForConditionalGeneration",16 "AutoModelForSequenceClassification": "thu-coai/CharacterGLM-6B--modeling_chatglm.ChatGLMForSequenceClassification"17 },18 "bias_dropout_fusion": true,19 "classifier_dropout": null,20 "eos_token_id": 2,21 "ffn_hidden_size": 13696,22 "fp32_residual_connection": false,23 "hidden_dropout": 0.0,24 "hidden_size": 4096,25 "kv_channels": 128,26 "layernorm_epsilon": 1e-05,27 "model_type": "chatglm",28 "multi_query_attention": true,29 "multi_query_group_num": 2,30 "num_attention_heads": 32,31 "num_layers": 28,32 "original_rope": true,33 "pad_token_id": 0,34 "padded_vocab_size": 65024,35 "post_layer_norm": true,36 "pre_seq_len": null,37 "prefix_projection": false,38 "quantization_bit": 0,39 "rmsnorm": true,40 "seq_length": 32768,41 "tie_word_embeddings": false,42 "torch_dtype": "float16",43 "transformers_version": "4.50.0",44 "use_cache": true,45 "vocab_size": 6502446}47 