{
    "pipeline_class": "GlmPipeline",
    "img_context_token_id": 151343,
    "image_encoder": {
        "vision_encoder": {
            "model_id": "glm4v_vit",
            "num_layers": 24,
            "pipeline_num_layers": [24,0,0,0],
            "hidden_size": 1536,
            "attention_bias": false,
            "out_hidden_size": 4096,
            "ffn_hidden_size": 4096,
            "llm_hidden_size": 1280,
            "gated_linear_unit": true,
            "add_bias_linear": false,
            "num_attention_heads": 12,
            "hidden_dropout": 0.0,
            "attention_dropout": 0.0,
            "in_channels": 3,
            "patch_size": 14,
            "spatial_merge_size": 2,
            "temporal_patch_size": 2,
            "layernorm_epsilon": 1e-06,
            "normalization": "RMSNorm",
            "image_size": 336,
            "fp16": false,
            "bf16": true,
            "params_dtype": "bf16",
            "activation_func": "silu",
            "hidden_act": "silu",
            "freeze": true,
            "use_fused_rotary_pos_emb": true,
            "post_layer_norm": false,
            "tokens_per_second": 2,
            "rms_norm_eps": 1e-05,
            "window_attn_size": 112
        },
        "vision_projector": {
            "model_id": "GlmMLP",
            "num_layers": 1,
            "gated_linear_unit": false,
            "bias_activation_fusion": false,
            "add_bias_linear": true,
            "input_size": 1280,
            "hidden_size": 2048,
            "ffn_hidden_size": 5120,
            "bf16": true,
            "params_dtype": "bf16",
            "freeze": true,
            "layernorm_epsilon": 1e-06,
            "normalization": "RMSNorm",
            "out_hidden_size": 4096,
            "intermediate_size": 13696,
            "hidden_act": "silu"
        }
    },
    "text_decoder": {
        "model_id": "glm4v_lm",
        "num_layers": 40,
        "pipeline_num_layers": [7, 11, 11, 11],
        "hidden_size": 4096,
        "ffn_hidden_size": 13696,
        "num_attention_heads": 32,
        "max_position_embeddings": 128000,
        "vocab_size": 151552,
        "rope_theta": 10000.0,
        "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": "mrope",
        "normalization": "RMSNorm",
        "activation_func": "silu",
        "use_fused_rotary_pos_emb": true,
        "partial_rotary_factor": 0.5,
        "attention_softmax_in_fp32": true,
        "params_dtype": "bf16",
        "hidden_act": "silu",
        "bf16": true,
        "parallel_output": true,
        "group_query_attention": true,
        "num_query_groups": 2,
        "num_key_value_heads": 2,
        "mrope_section": [8, 12, 12],
        "rope_scaling": null,
        "gated_linear_unit": true,
        "layernorm_epsilon": 1e-06,
        "add_bias_linear":false,
        "add_qkv_bias": true,
        "use_remove_padding": false,
        "rms_norm_eps": 1e-05,
        "is_encoder_decoder": false,
        "use_infer_fa": true
    },
    "text_encoder": null,
    "video_encoder": null,
    "dtype": "bf16",
    "device": "npu",
    "tokenizer": {
        "hub_backend": "hf",
        "autotokenizer_name": "AutoTokenizer",
        "from_pretrained": "ckpt/hf_path/GLM4.1V-9B-Instruct"
    },
    "generation_config": {
        "bos_token_id": 151643,
        "do_sample": true,
        "kv_cache":true,
        "output_attentions": false,
        "output_hidden_states": false,
        "max_length": 20,
        "min_length": 0,
        "min_new_tokens": null,
        "constraints": null,
        "prompt_lookup_num_tokens": null,
        "guidance_scale": null,
        "bad_words_ids": null,
        "begin_suppress_tokens": null,
        "diversity_penalty": 0.0,
        "early_stopping": false,
        "encoder_no_repeat_ngram_size": 0,
        "encoder_repetition_penalty": 1.0,
        "epsilon_cutoff": 0.0,
        "eta_cutoff": 0.0,
        "exponential_decay_length_penalty": null,
        "forced_bos_token_id": null,
        "forced_decoder_ids": null,
        "forced_eos_token_id": null,
        "length_penalty": 1.0,
        "low_memory": null,
        "max_time": null,
        "no_repeat_ngram_size": 0,
        "num_assistant_tokens": 5,
        "num_assistant_tokens_schedule": "heuristic",
        "num_beam_groups": 1,
        "num_return_groups": 1,
        "num_return_sequences": 1,
        "output_scores": false,
        "output_logits": null,
        "penalty_alpha": null,
        "remove_invalid_values": false,
        "repetition_penalty": 1.1,
        "return_dict_in_generate": false,
        "sequence_bias": null,
        "spatial_merge_size": 2,
        "suppress_tokens": null,
        "typical_p": 1.0,
        "force_words_ids": null,
        "num_beams": 1,
        "renormalize_logits": false,
        "use_cache": true,
        "eos_token_id": [
            151329,
            151336,
            151338,
            151348
        ],
        "max_new_tokens": 8192,
        "pad_token_id": 151329,
        "vision_start_token_id": 151341,
        "vision_end_token_id": 151342,
        "image_token_id": 151343,
        "video_token_id": 151656,
        "temperature": 0,
        "top_k": 2,
        "top_p": 0,
        "dola_layers": null,
        "cache_implementation": null,
        "cache_config": null,
        "return_legacy_cache": null,
        "min_p": null,
        "token_healing": false,
        "watermarking_config": null,
        "decoder_start_token_id": null,
        "max_matching_ngram_size": null,
        "stop_strings": null
    },
    "patch":{
        "infer_fa": true
    },
    "min_pixels": 1003520,
    "max_pixels": 12845056,
    "image_path": "examples/qwen2vl/demo.jpeg",
    "prompts": "Describe this image and keep it within 100 words.",
    "return_ids": true
}