{ "architectures": [ "GlmImageModel" ], "dtype": "float32", "image_end_token_id": 51, "image_start_token_id": 50, "image_token_id": 52, "model_type": "glm_image", "text_config": { "attention_bias": true, "attention_dropout": 0.0, "bos_token_id": 16384, "eos_token_id": 16385, "hidden_act": "silu", "hidden_size": 16, "initializer_range": 0.02, "intermediate_size": 22, "max_position_embeddings": 512, "model_type": "glm_image_text", "n_group": 1, "n_routed_experts": 8, "n_shared_experts": 1, "num_attention_heads": 2, "num_experts_per_tok": 8, "num_hidden_layers": 2, "num_key_value_heads": 1, "output_channels": 64, "pad_token_id": 16385, "rms_norm_eps": 1e-05, "rope_parameters": { "mrope_section": [ 2, 1, 1 ], "rope_theta": 10000, "rope_type": "default", "type": "default" }, "tie_word_embeddings": true, "topk_group": 1, "use_cache": true, "vision_vocab_size": 99, "vocab_size": 167864 }, "tie_word_embeddings": false, "transformers_version": "5.16.0.dev0", "vision_config": { "attention_bias": true, "attention_dropout": 0.0, "depth": 2, "hidden_act": "gelu", "hidden_size": 32, "image_size": 128, "in_channels": 3, "initializer_range": 0.02, "intermediate_size": 22, "layer_norm_eps": 1e-06, "model_type": "glm_image_vision", "num_heads": 16, "patch_size": 16, "spatial_merge_size": 1, "temporal_patch_size": 1 }, "vq_config": { "embed_dim": 48, "in_channels": 3, "initializer_range": 0.02, "latent_channels": 32, "model_type": "glm_image_vqmodel", "num_embeddings": 32 } }