nohup: ignoring input
INFO:baidu_bos_manager:✅ BOS连接测试成功
INFO:baidu_bos_manager:✅ BOS客户端初始化成功: dmtyz-demo
INFO:mongodb_manager:✅ MongoDB连接成功: mmeb
INFO:mongodb_manager:✅ MongoDB索引创建完成
INFO:__main__:初始化多模态检索系统...
INFO:multimodal_retrieval_local:使用GPU: [0, 1]
INFO:multimodal_retrieval_local:加载本地模型和处理器: /root/models/Ops-MM-embedding-v1-7B
The image processor of type `Qwen2VLImageProcessor` is now loaded as a fast processor by default, even if the model checkpoint was saved with a slow processor. This is a breaking change and may produce slightly different outputs. To continue using the slow processor, instantiate this class with `use_fast=False`. Note that this behavior will be extended to all models in a future release.
You have video processor config saved in `preprocessor.json` file which is deprecated. Video processor configs should be saved in their own `video_preprocessor.json` file. You can rename the file or load and save the processor back which renames it automatically. Loading from `preprocessor.json` will be removed in v5.0.
INFO:multimodal_retrieval_local:Processor类型: <class 'transformers.models.qwen2_vl.processing_qwen2_vl.Qwen2VLProcessor'>
INFO:multimodal_retrieval_local:Processor方法: ['__annotations__', '__call__', '__class__', '__delattr__', '__dict__', '__dir__', '__doc__', '__eq__', '__format__', '__ge__', '__getattribute__', '__gt__', '__hash__', '__init__', '__init_subclass__', '__le__', '__lt__', '__module__', '__ne__', '__new__', '__reduce__', '__reduce_ex__', '__repr__', '__setattr__', '__sizeof__', '__str__', '__subclasshook__', '__weakref__', '_auto_class', '_check_special_mm_tokens', '_create_repo', '_get_arguments_from_pretrained', '_get_files_timestamps', '_get_num_multimodal_tokens', '_merge_kwargs', '_upload_modified_files', 'apply_chat_template', 'attributes', 'audio_tokenizer', 'batch_decode', 'chat_template', 'check_argument_for_proper_class', 'decode', 'feature_extractor_class', 'from_args_and_dict', 'from_pretrained', 'get_possibly_dynamic_module', 'get_processor_dict', 'image_processor', 'image_processor_class', 'image_token', 'image_token_id', 'model_input_names', 'optional_attributes', 'optional_call_args', 'post_process_image_text_to_text', 'push_to_hub', 'register_for_auto_class', 'save_pretrained', 'to_dict', 'to_json_file', 'to_json_string', 'tokenizer', 'tokenizer_class', 'validate_init_kwargs', 'video_processor', 'video_processor_class', 'video_token', 'video_token_id']
INFO:multimodal_retrieval_local:Image processor类型: <class 'transformers.models.qwen2_vl.image_processing_qwen2_vl_fast.Qwen2VLImageProcessorFast'>
INFO:multimodal_retrieval_local:Image processor方法: ['__backends', '__call__', '__class__', '__delattr__', '__dict__', '__dir__', '__doc__', '__eq__', '__format__', '__ge__', '__getattribute__', '__gt__', '__hash__', '__init__', '__init_subclass__', '__le__', '__lt__', '__module__', '__ne__', '__new__', '__reduce__', '__reduce_ex__', '__repr__', '__setattr__', '__sizeof__', '__slotnames__', '__str__', '__subclasshook__', '__weakref__', '_auto_class', '_create_repo', '_further_process_kwargs', '_fuse_mean_std_and_rescale_factor', '_get_files_timestamps', '_prepare_image_like_inputs', '_prepare_images_structure', '_preprocess', '_preprocess_image_like_inputs', '_process_image', '_processor_class', '_set_processor_class', '_upload_modified_files', '_valid_kwargs_names', '_validate_preprocess_kwargs', 'center_crop', 'compile_friendly_resize', 'convert_to_rgb', 'crop_size', 'data_format', 'default_to_square', 'device', 'disable_grouping', 'do_center_crop', 'do_convert_rgb', 'do_normalize', 'do_rescale', 'do_resize', 'fetch_images', 'filter_out_unused_kwargs', 'from_dict', 'from_json_file', 'from_pretrained', 'get_image_processor_dict', 'get_number_of_image_patches', 'image_mean', 'image_processor_type', 'image_std', 'input_data_format', 'max_pixels', 'merge_size', 'min_pixels', 'model_input_names', 'normalize', 'patch_size', 'preprocess', 'push_to_hub', 'register_for_auto_class', 'resample', 'rescale', 'rescale_and_normalize', 'rescale_factor', 'resize', 'return_tensors', 'save_pretrained', 'size', 'temporal_patch_size', 'to_dict', 'to_json_file', 'to_json_string', 'unused_kwargs', 'valid_kwargs']

Loading checkpoint shards:   0%|          | 0/4 [00:00<?, ?it/s]
Loading checkpoint shards:  25%|██▌       | 1/4 [03:33<10:40, 213.50s/it]
Loading checkpoint shards:  50%|█████     | 2/4 [05:26<05:08, 154.38s/it]
Loading checkpoint shards:  75%|███████▌  | 3/4 [07:17<02:14, 134.57s/it]
Loading checkpoint shards: 100%|██████████| 4/4 [07:30<00:00, 86.66s/it] 
Loading checkpoint shards: 100%|██████████| 4/4 [07:30<00:00, 112.67s/it]
INFO:multimodal_retrieval_local:向量维度: 3584
INFO:multimodal_retrieval_local:模型和处理器加载成功
INFO:multimodal_retrieval_local:加载现有索引: /root/mmeb/local_faiss_index.index
INFO:multimodal_retrieval_local:索引加载成功，包含0个向量
INFO:multimodal_retrieval_local:元数据加载成功，包含0条记录
INFO:multimodal_retrieval_local:多模态检索系统初始化完成，使用本地模型: /root/models/Ops-MM-embedding-v1-7B
INFO:multimodal_retrieval_local:向量存储路径: /root/mmeb/local_faiss_index
INFO:__main__:多模态检索系统初始化完成
 * Serving Flask app 'web_app_local'
 * Debug mode: off
INFO:werkzeug:WARNING: This is a development server. Do not use it in a production deployment. Use a production WSGI server instead.
 * Running on all addresses (0.0.0.0)
 * Running on http://127.0.0.1:5000
 * Running on http://192.168.48.82:5000
INFO:werkzeug:Press CTRL+C to quit
INFO:werkzeug:127.0.0.1 - - [22/Sep/2025 05:52:11] "GET / HTTP/1.1" 200 -
INFO:werkzeug:127.0.0.1 - - [22/Sep/2025 05:52:13] "GET /api/system_info HTTP/1.1" 200 -
INFO:werkzeug:127.0.0.1 - - [22/Sep/2025 05:52:13] "GET /api/system_info HTTP/1.1" 200 -
INFO:werkzeug:127.0.0.1 - - [22/Sep/2025 05:52:14] "GET /favicon.ico HTTP/1.1" 404 -
INFO:__main__:处理图像: 微信图片_20250910164839_1_13.jpg (99396 字节)
INFO:__main__:成功加载图像: 20250910164839_1_13.jpg, 格式: JPEG, 模式: RGB, 大小: (939, 940)
INFO:multimodal_retrieval_local:add_images: 开始添加图像，数量: 1
INFO:multimodal_retrieval_local:add_images: 编码图像
INFO:multimodal_retrieval_local:encode_image: 开始编码图像，类型: <class 'list'>
INFO:multimodal_retrieval_local:encode_image: 图像列表，长度: 1
INFO:multimodal_retrieval_local:encode_image: 处理图像输入
INFO:multimodal_retrieval_local:encode_image: 图像 0 格式: JPEG, 模式: RGB, 大小: (939, 940)
INFO:multimodal_retrieval_local:encode_image: 使用image_processor处理图像
INFO:werkzeug:127.0.0.1 - - [22/Sep/2025 05:52:40] "GET / HTTP/1.1" 200 -
INFO:werkzeug:127.0.0.1 - - [22/Sep/2025 05:52:41] "GET /api/system_info HTTP/1.1" 200 -
INFO:werkzeug:127.0.0.1 - - [22/Sep/2025 05:52:41] "GET /api/system_info HTTP/1.1" 200 -
INFO:werkzeug:127.0.0.1 - - [22/Sep/2025 05:52:42] "GET /favicon.ico HTTP/1.1" 404 -
INFO:multimodal_retrieval_local:encode_image: 处理后的输入键: ['pixel_values']
INFO:__main__:处理图像: 微信图片_20250910164839_1_13.jpg (99396 字节)
INFO:__main__:成功加载图像: 20250910164839_1_13.jpg, 格式: JPEG, 模式: RGB, 大小: (939, 940)
INFO:multimodal_retrieval_local:add_images: 开始添加图像，数量: 1
INFO:multimodal_retrieval_local:add_images: 编码图像
INFO:multimodal_retrieval_local:encode_image: 开始编码图像，类型: <class 'list'>
INFO:multimodal_retrieval_local:encode_image: 图像列表，长度: 1
INFO:multimodal_retrieval_local:encode_image: 处理图像输入
INFO:multimodal_retrieval_local:encode_image: 图像 0 格式: JPEG, 模式: RGB, 大小: (939, 940)
INFO:multimodal_retrieval_local:encode_image: 使用image_processor处理图像
INFO:multimodal_retrieval_local:encode_image: 运行模型推理
INFO:multimodal_retrieval_local:Model类型: <class 'transformers.models.qwen2_vl.modeling_qwen2_vl.Qwen2VLModel'>
INFO:multimodal_retrieval_local:Model属性: ['T_destination', '__annotations__', '__call__', '__class__', '__delattr__', '__dict__', '__dir__', '__doc__', '__eq__', '__format__', '__ge__', '__getattr__', '__getattribute__', '__getstate__', '__gt__', '__hash__', '__init__', '__init_subclass__', '__le__', '__lt__', '__module__', '__ne__', '__new__', '__reduce__', '__reduce_ex__', '__repr__', '__setattr__', '__setstate__', '__sizeof__', '__str__', '__subclasshook__', '__weakref__', '_apply', '_auto_class', '_backward_compatibility_gradient_checkpointing', '_backward_hooks', '_backward_pre_hooks', '_buffers', '_call_impl', '_can_compile_fullgraph', '_can_record_outputs', '_can_set_attn_implementation', '_check_and_adjust_attn_implementation', '_checkpoint_conversion_mapping', '_compiled_call_impl', '_convert_head_mask_to_5d', '_copy_lm_head_original_to_resized', '_create_repo', '_dispatch_accelerate_model', '_fix_state_dict_key_on_load', '_fix_state_dict_key_on_save', '_fix_state_dict_keys_on_save', '_flash_attn_2_can_dispatch', '_flash_attn_3_can_dispatch', '_flex_attn_can_dispatch', '_forward_hooks', '_forward_hooks_always_called', '_forward_hooks_with_kwargs', '_forward_pre_hooks', '_forward_pre_hooks_with_kwargs', '_from_config', '_get_backward_hooks', '_get_backward_pre_hooks', '_get_files_timestamps', '_get_key_renaming_mapping', '_get_name', '_get_no_split_modules', '_get_resized_embeddings', '_get_resized_lm_head', '_hf_hook', '_hf_peft_config_loaded', '_hook_rss_memory_post_forward', '_hook_rss_memory_pre_forward', '_init_added_embeddings_weights_with_mean', '_init_added_lm_head_bias_with_mean', '_init_added_lm_head_weights_with_mean', '_init_weights', '_initialize_missing_keys', '_initialize_weights', '_input_embed_layer', '_is_full_backward_hook', '_is_hf_initialized', '_is_stateful', '_keep_in_fp32_modules', '_keep_in_fp32_modules', '_keep_in_fp32_modules_strict', '_keep_in_fp32_modules_strict', '_keys_to_ignore_on_load_missing', '_keys_to_ignore_on_load_unexpected', '_keys_to_ignore_on_save', '_load_from_flax', '_load_from_state_dict', '_load_from_tf', '_load_pretrained_model', '_load_state_dict_post_hooks', '_load_state_dict_pre_hooks', '_maybe_warn_non_full_backward_hook', '_modules', '_move_missing_keys_from_meta_to_cpu', '_named_members', '_no_split_modules', '_no_split_modules', '_non_persistent_buffers_set', '_old_forward', '_parameters', '_pp_plan', '_pp_plan', '_register_load_state_dict_pre_hook', '_register_state_dict_hook', '_replicate_for_data_parallel', '_resize_token_embeddings', '_save_to_state_dict', '_sdpa_can_dispatch', '_set_default_torch_dtype', '_set_gradient_checkpointing', '_skip_keys_device_placement', '_slow_forward', '_state_dict_hooks', '_state_dict_pre_hooks', '_supports_attention_backend', '_supports_flash_attn', '_supports_flex_attn', '_supports_sdpa', '_tie_encoder_decoder_weights', '_tie_or_clone_weights', '_tied_weights_keys', '_tp_plan', '_tp_size', '_upload_modified_files', '_version', '_wrapped_call_impl', 'active_adapter', 'active_adapters', 'add_adapter', 'add_memory_hooks', 'add_model_tags', 'add_module', 'apply', 'base_model', 'base_model_prefix', 'bfloat16', 'buffers', 'call_super_init', 'can_generate', 'can_record_outputs', 'children', 'compile', 'config', 'config_class', 'cpu', 'create_extended_attention_mask_for_decoder', 'cuda', 'cuda', 'delete_adapter', 'dequantize', 'device', 'disable_adapters', 'disable_input_require_grads', 'double', 'dtype', 'dummy_inputs', 'dump_patches', 'enable_adapters', 'enable_input_require_grads', 'estimate_tokens', 'eval', 'extra_repr', 'float', 'floating_point_ops', 'forward', 'forward', 'framework', 'from_pretrained', 'generation_config', 'get_adapter_state_dict', 'get_buffer', 'get_compiled_call', 'get_correct_attn_implementation', 'get_decoder', 'get_extended_attention_mask', 'get_extra_state', 'get_head_mask', 'get_image_features', 'get_init_context', 'get_input_embeddings', 'get_memory_footprint', 'get_output_embeddings', 'get_parameter', 'get_parameter_or_buffer', 'get_placeholder_mask', 'get_position_embeddings', 'get_rope_index', 'get_submodule', 'get_video_features', 'gradient_checkpointing_disable', 'gradient_checkpointing_enable', 'half', 'hf_device_map', 'init_weights', 'initialize_weights', 'invert_attention_mask', 'ipu', 'is_backend_compatible', 'is_gradient_checkpointing', 'is_parallelizable', 'language_model', 'load_adapter', 'load_state_dict', 'loss_function', 'loss_type', 'main_input_name', 'model_tags', 'modules', 'mtia', 'name_or_path', 'named_buffers', 'named_children', 'named_modules', 'named_parameters', 'num_parameters', 'parameters', 'post_init', 'prune_heads', 'push_to_hub', 'register_backward_hook', 'register_buffer', 'register_for_auto_class', 'register_forward_hook', 'register_forward_pre_hook', 'register_full_backward_hook', 'register_full_backward_pre_hook', 'register_load_state_dict_post_hook', 'register_load_state_dict_pre_hook', 'register_module', 'register_parameter', 'register_state_dict_post_hook', 'register_state_dict_pre_hook', 'requires_grad_', 'reset_memory_hooks_state', 'resize_position_embeddings', 'resize_token_embeddings', 'retrieve_modules_from_names', 'reverse_bettertransformer', 'rope_deltas', 'save_pretrained', 'set_adapter', 'set_attn_implementation', 'set_decoder', 'set_extra_state', 'set_input_embeddings', 'set_output_embeddings', 'set_submodule', 'share_memory', 'smart_apply', 'state_dict', 'supports_gradient_checkpointing', 'supports_pp_plan', 'supports_tp_plan', 'tie_weights', 'to', 'to', 'to_bettertransformer', 'to_empty', 'tp_size', 'train', 'training', 'type', 'visual', 'warn_if_padding_and_no_attention_mask', 'warnings_issued', 'xpu', 'zero_grad']
ERROR:multimodal_retrieval_local:encode_image: 处理图像时出错: embedding(): argument 'indices' (position 2) must be Tensor, not NoneType
ERROR:multimodal_retrieval_local:add_images: 图像编码失败，返回空数组
INFO:multimodal_retrieval_local:索引保存成功: /root/mmeb/local_faiss_index.index
INFO:multimodal_retrieval_local:元数据保存成功: /root/mmeb/local_faiss_index_metadata.json
INFO:werkzeug:127.0.0.1 - - [22/Sep/2025 05:52:46] "POST /api/add_image HTTP/1.1" 200 -
INFO:multimodal_retrieval_local:encode_image: 处理后的输入键: ['pixel_values']
INFO:multimodal_retrieval_local:encode_image: 运行模型推理
INFO:multimodal_retrieval_local:Model类型: <class 'transformers.models.qwen2_vl.modeling_qwen2_vl.Qwen2VLModel'>
INFO:multimodal_retrieval_local:Model属性: ['T_destination', '__annotations__', '__call__', '__class__', '__delattr__', '__dict__', '__dir__', '__doc__', '__eq__', '__format__', '__ge__', '__getattr__', '__getattribute__', '__getstate__', '__gt__', '__hash__', '__init__', '__init_subclass__', '__le__', '__lt__', '__module__', '__ne__', '__new__', '__reduce__', '__reduce_ex__', '__repr__', '__setattr__', '__setstate__', '__sizeof__', '__str__', '__subclasshook__', '__weakref__', '_apply', '_auto_class', '_backward_compatibility_gradient_checkpointing', '_backward_hooks', '_backward_pre_hooks', '_buffers', '_call_impl', '_can_compile_fullgraph', '_can_record_outputs', '_can_set_attn_implementation', '_check_and_adjust_attn_implementation', '_checkpoint_conversion_mapping', '_compiled_call_impl', '_convert_head_mask_to_5d', '_copy_lm_head_original_to_resized', '_create_repo', '_dispatch_accelerate_model', '_fix_state_dict_key_on_load', '_fix_state_dict_key_on_save', '_fix_state_dict_keys_on_save', '_flash_attn_2_can_dispatch', '_flash_attn_3_can_dispatch', '_flex_attn_can_dispatch', '_forward_hooks', '_forward_hooks_always_called', '_forward_hooks_with_kwargs', '_forward_pre_hooks', '_forward_pre_hooks_with_kwargs', '_from_config', '_get_backward_hooks', '_get_backward_pre_hooks', '_get_files_timestamps', '_get_key_renaming_mapping', '_get_name', '_get_no_split_modules', '_get_resized_embeddings', '_get_resized_lm_head', '_hf_hook', '_hf_peft_config_loaded', '_hook_rss_memory_post_forward', '_hook_rss_memory_pre_forward', '_init_added_embeddings_weights_with_mean', '_init_added_lm_head_bias_with_mean', '_init_added_lm_head_weights_with_mean', '_init_weights', '_initialize_missing_keys', '_initialize_weights', '_input_embed_layer', '_is_full_backward_hook', '_is_hf_initialized', '_is_stateful', '_keep_in_fp32_modules', '_keep_in_fp32_modules', '_keep_in_fp32_modules_strict', '_keep_in_fp32_modules_strict', '_keys_to_ignore_on_load_missing', '_keys_to_ignore_on_load_unexpected', '_keys_to_ignore_on_save', '_load_from_flax', '_load_from_state_dict', '_load_from_tf', '_load_pretrained_model', '_load_state_dict_post_hooks', '_load_state_dict_pre_hooks', '_maybe_warn_non_full_backward_hook', '_modules', '_move_missing_keys_from_meta_to_cpu', '_named_members', '_no_split_modules', '_no_split_modules', '_non_persistent_buffers_set', '_old_forward', '_parameters', '_pp_plan', '_pp_plan', '_register_load_state_dict_pre_hook', '_register_state_dict_hook', '_replicate_for_data_parallel', '_resize_token_embeddings', '_save_to_state_dict', '_sdpa_can_dispatch', '_set_default_torch_dtype', '_set_gradient_checkpointing', '_skip_keys_device_placement', '_slow_forward', '_state_dict_hooks', '_state_dict_pre_hooks', '_supports_attention_backend', '_supports_flash_attn', '_supports_flex_attn', '_supports_sdpa', '_tie_encoder_decoder_weights', '_tie_or_clone_weights', '_tied_weights_keys', '_tp_plan', '_tp_size', '_upload_modified_files', '_version', '_wrapped_call_impl', 'active_adapter', 'active_adapters', 'add_adapter', 'add_memory_hooks', 'add_model_tags', 'add_module', 'apply', 'base_model', 'base_model_prefix', 'bfloat16', 'buffers', 'call_super_init', 'can_generate', 'can_record_outputs', 'children', 'compile', 'config', 'config_class', 'cpu', 'create_extended_attention_mask_for_decoder', 'cuda', 'cuda', 'delete_adapter', 'dequantize', 'device', 'disable_adapters', 'disable_input_require_grads', 'double', 'dtype', 'dummy_inputs', 'dump_patches', 'enable_adapters', 'enable_input_require_grads', 'estimate_tokens', 'eval', 'extra_repr', 'float', 'floating_point_ops', 'forward', 'forward', 'framework', 'from_pretrained', 'generation_config', 'get_adapter_state_dict', 'get_buffer', 'get_compiled_call', 'get_correct_attn_implementation', 'get_decoder', 'get_extended_attention_mask', 'get_extra_state', 'get_head_mask', 'get_image_features', 'get_init_context', 'get_input_embeddings', 'get_memory_footprint', 'get_output_embeddings', 'get_parameter', 'get_parameter_or_buffer', 'get_placeholder_mask', 'get_position_embeddings', 'get_rope_index', 'get_submodule', 'get_video_features', 'gradient_checkpointing_disable', 'gradient_checkpointing_enable', 'half', 'hf_device_map', 'init_weights', 'initialize_weights', 'invert_attention_mask', 'ipu', 'is_backend_compatible', 'is_gradient_checkpointing', 'is_parallelizable', 'language_model', 'load_adapter', 'load_state_dict', 'loss_function', 'loss_type', 'main_input_name', 'model_tags', 'modules', 'mtia', 'name_or_path', 'named_buffers', 'named_children', 'named_modules', 'named_parameters', 'num_parameters', 'parameters', 'post_init', 'prune_heads', 'push_to_hub', 'register_backward_hook', 'register_buffer', 'register_for_auto_class', 'register_forward_hook', 'register_forward_pre_hook', 'register_full_backward_hook', 'register_full_backward_pre_hook', 'register_load_state_dict_post_hook', 'register_load_state_dict_pre_hook', 'register_module', 'register_parameter', 'register_state_dict_post_hook', 'register_state_dict_pre_hook', 'requires_grad_', 'reset_memory_hooks_state', 'resize_position_embeddings', 'resize_token_embeddings', 'retrieve_modules_from_names', 'reverse_bettertransformer', 'rope_deltas', 'save_pretrained', 'set_adapter', 'set_attn_implementation', 'set_decoder', 'set_extra_state', 'set_input_embeddings', 'set_output_embeddings', 'set_submodule', 'share_memory', 'smart_apply', 'state_dict', 'supports_gradient_checkpointing', 'supports_pp_plan', 'supports_tp_plan', 'tie_weights', 'to', 'to', 'to_bettertransformer', 'to_empty', 'tp_size', 'train', 'training', 'type', 'visual', 'warn_if_padding_and_no_attention_mask', 'warnings_issued', 'xpu', 'zero_grad']
ERROR:multimodal_retrieval_local:encode_image: 处理图像时出错: embedding(): argument 'indices' (position 2) must be Tensor, not NoneType
ERROR:multimodal_retrieval_local:add_images: 图像编码失败，返回空数组
INFO:multimodal_retrieval_local:索引保存成功: /root/mmeb/local_faiss_index.index
INFO:multimodal_retrieval_local:元数据保存成功: /root/mmeb/local_faiss_index_metadata.json
INFO:werkzeug:127.0.0.1 - - [22/Sep/2025 05:52:59] "POST /api/add_image HTTP/1.1" 200 -
INFO:multimodal_retrieval_local:索引保存成功: /root/mmeb/local_faiss_index.index
INFO:multimodal_retrieval_local:元数据保存成功: /root/mmeb/local_faiss_index_metadata.json
INFO:werkzeug:127.0.0.1 - - [22/Sep/2025 05:53:00] "POST /api/save_index HTTP/1.1" 200 -
INFO:werkzeug:127.0.0.1 - - [22/Sep/2025 05:53:01] "GET /api/system_info HTTP/1.1" 200 -