{
"model_id": "InternVL2.5",
"pre_process": true,
"post_process": true,
"add_text_encoder": false,
"img_embedding_idx": 1,
"downsample_ratio": 0.5,
"select_layer": -1,
"ps_version": "v2",
"add_rmsnorm_offset": false,
"img_context_token_id": 151667,
"text_decoder": {
"model_id": "internllm",
"num_layers": 36,
"pipeline_num_layers": [36],
"hidden_size": 2048,
"ffn_hidden_size": 11008,
"num_attention_heads": 16,
"num_key_value_heads": 2,
"max_position_embeddings": 32768,
"vocab_size": 151674,
"rotary_percent": 1.0,
"rotary_base": 1000000,
"untie_embeddings_and_output_weights": true,
"disable_bias_linear": true,
"attention_dropout": 0.0,
"init_method_std": 0.01,
"hidden_dropout": 0.0,
"position_embedding_type": "rope",
"normalization": "RMSNorm",
"activation_func": "silu",
"attention_softmax_in_fp32": true,
"use_fused_rotary_pos_emb": false,
"params_dtype": "bf16",
"bf16": true,
"fp16_lm_cross_entropy": false,
"parallel_output": true,
"group_query_attention": true,
"num_query_groups": 2,
"rope_scaling": null,
"gated_linear_unit": true,
"layernorm_epsilon": 1e-06,
"add_bias_linear":false,
"add_qkv_bias": true
},
"image_encoder": {
"vision_encoder": {
"model_id": "InternViT",
"num_layers": 24,
"pipeline_num_layers": [24],
"hidden_size": 1024,
"ffn_hidden_size": 4096,
"num_attention_heads": 16,
"num_channels": 3,
"patch_size": 14,
"image_size": 448,
"add_qkv_bias": true,
"qk_layernorm": false,
"activation_func": "gelu",
"normalization": "LayerNorm",
"layernorm_epsilon": 1e-6,
"hidden_dropout": 0.0,
"drop_path_rate": 0.0,
"attention_dropout": 0.0,
"init_method_std": 0.02,
"initializer_factor": 1.0,
"output_hidden_states": false,
"use_return_dict": false,
"params_dtype": "bf16",
"post_layer_norm": false,
"downsample_ratio": 0.5,
"fp16": false,
"bf16": true,
"attention_softmax_in_fp32": false,
"select_layer": -1,
"ps_version": "v2",
"pre_tockens": 2147483647,
"next_tockens": 2147483647,
"freeze": true
},
"vision_projector": {
"model_id": "InternVLMLP",
"downsample_ratio": 0.5,
"vit_hidden_size": 1024,
"llm_hidden_size": 2048,
"params_dtype":"bf16",
"num_layers":1
}
},
"text_encoder": null,
"video_encoder": null
}