Create processor_config.json

processor_config.json

@@ -0,0 +1,265 @@
+{
+  "auto_map": {
+    "AutoProcessor": "processing_llava.LlavaProcessor"
+  },
+  "image_processor_type": "SiglipImageProcessor",
+  "processor_class": "LlavaProcessor",
+  "tokenizer_class": "LlamaTokenizer",
+  
+  "image_processor": {
+    "do_resize": true,
+    "size": {
+      "shortest_edge": 384,
+      "longest_edge": 384
+    },
+    "resample": 3,
+    "do_center_crop": false,
+    "crop_size": {
+      "height": 384,
+      "width": 384
+    },
+    "do_rescale": true,
+    "rescale_factor": 0.00392156862745098,
+    "do_normalize": true,
+    "image_mean": [0.5, 0.5, 0.5],
+    "image_std": [0.5, 0.5, 0.5],
+    "do_convert_rgb": true,
+    "image_grid_pinpoints": [
+      [336, 672],
+      [672, 336],
+      [672, 672],
+      [1008, 336],
+      [336, 1008]
+    ]
+  },
+  
+  "tokenizer": {
+    "add_bos_token": true,
+    "add_eos_token": false,
+    "bos_token": "<s>",
+    "eos_token": "</s>",
+    "unk_token": "<unk>",
+    "pad_token": null,
+    "model_max_length": 200000,
+    "clean_up_tokenization_spaces": false,
+    "tokenizer_class": "LlamaTokenizer",
+    "legacy": true,
+    "use_default_system_prompt": false
+  },
+  
+  "chat_template": "{% if messages[0]['role'] == 'system' %}{% set system_message = messages[0]['content'] %}{% set messages = messages[1:] %}{% else %}{% set system_message = 'You are Helion, a helpful, respectful, and honest AI assistant with vision and tool use capabilities.' %}{% endif %}{{ bos_token }}{% if system_message %}<|system|>\n{{ system_message }}\n</s>{% endif %}{% for message in messages %}{% if message['role'] == 'user' %}<|user|>\n{% if message.get('images') %}{% for image in message['images'] %}<image>{% endfor %}{% endif %}{{ message['content'] }}\n</s>{% elif message['role'] == 'assistant' %}<|assistant|>\n{{ message['content'] }}\n</s>{% elif message['role'] == 'tool' %}<|tool|>\n{{ message['content'] }}\n</s>{% endif %}{% endfor %}{% if add_generation_prompt %}<|assistant|>\n{% endif %}",
+  
+  "vision_feature_layer": -2,
+  "vision_feature_select_strategy": "default",
+  "image_token_index": 32000,
+  "vision_aspect_ratio": "anyres",
+  
+  "patch_size": 14,
+  "vision_config": {
+    "hidden_size": 1152,
+    "image_size": 384,
+    "intermediate_size": 4304,
+    "model_type": "siglip_vision_model",
+    "num_attention_heads": 16,
+    "num_hidden_layers": 27,
+    "patch_size": 14,
+    "projection_dim": 768
+  },
+  
+  "text_config": {
+    "model_type": "llama",
+    "vocab_size": 32000,
+    "hidden_size": 4096,
+    "intermediate_size": 14336,
+    "num_hidden_layers": 48,
+    "num_attention_heads": 32,
+    "num_key_value_heads": 8,
+    "max_position_embeddings": 200000,
+    "rope_theta": 500000.0,
+    "rope_scaling": {
+      "type": "linear",
+      "factor": 8.0
+    }
+  },
+  
+  "multimodal_config": {
+    "max_images_per_sample": 10,
+    "image_seq_length": 729,
+    "num_image_tokens": 1,
+    "image_aspect_ratio": "anyres",
+    "image_grid_pinpoints": [
+      [336, 672],
+      [672, 336],
+      [672, 672],
+      [1008, 336],
+      [336, 1008]
+    ],
+    "use_image_start_end": false,
+    "vision_feature_layer": -2,
+    "vision_feature_select_strategy": "default",
+    "mm_vision_select_feature": "patch"
+  },
+  
+  "generation_config": {
+    "do_sample": true,
+    "temperature": 0.7,
+    "top_p": 0.9,
+    "top_k": 50,
+    "max_new_tokens": 2048,
+    "repetition_penalty": 1.1,
+    "length_penalty": 1.0,
+    "no_repeat_ngram_size": 0,
+    "num_beams": 1,
+    "early_stopping": false,
+    "use_cache": true,
+    "pad_token_id": 0,
+    "bos_token_id": 1,
+    "eos_token_id": 2
+  },
+  
+  "preprocessing": {
+    "image_processing": {
+      "auto_augment": false,
+      "interpolation": "bilinear",
+      "fill_color": 0,
+      "do_flip": false,
+      "do_rotate": false
+    },
+    "text_processing": {
+      "lowercase": false,
+      "remove_accents": false,
+      "strip_accents": false,
+      "handle_chinese_chars": true
+    }
+  },
+  
+  "special_tokens": {
+    "image_token": "<image>",
+    "image_start_token": null,
+    "image_end_token": null,
+    "pad_token": null,
+    "additional_special_tokens": [
+      "<image>"
+    ]
+  },
+  
+  "model_metadata": {
+    "model_name": "Helion-V2.0-Thinking",
+    "model_type": "llava",
+    "model_version": "2.0",
+    "architecture": "LLaVA",
+    "base_model": "meta-llama/Llama-2-10b-hf",
+    "vision_encoder": "SigLIP-400M",
+    "context_length": 200000,
+    "total_parameters": "10.2B",
+    "license": "apache-2.0",
+    "created_by": "DeepXR"
+  },
+  
+  "performance_optimizations": {
+    "use_flash_attention_2": true,
+    "torch_dtype": "bfloat16",
+    "low_cpu_mem_usage": true,
+    "use_cache": true,
+    "gradient_checkpointing": false,
+    "offload_folder": null,
+    "device_map": "auto"
+  },
+  
+  "inference_optimization": {
+    "batch_size": 1,
+    "enable_torch_compile": false,
+    "torch_compile_mode": null,
+    "enable_xformers": false,
+    "enable_bettertransformer": false
+  },
+  
+  "safety_config": {
+    "enable_content_filtering": true,
+    "enable_pii_detection": true,
+    "enable_toxicity_detection": true,
+    "max_output_tokens": 4096,
+    "temperature_bounds": {
+      "min": 0.0,
+      "max": 2.0
+    },
+    "blocked_token_ids": []
+  },
+  
+  "prompt_engineering": {
+    "system_prompt_template": "You are Helion, an advanced AI assistant with vision and tool use capabilities. You are helpful, respectful, and honest.",
+    "user_prompt_prefix": "",
+    "user_prompt_suffix": "",
+    "assistant_prompt_prefix": "",
+    "stop_sequences": ["</s>", "<|end|>", "<|endoftext|>"],
+    "response_format": "text"
+  },
+  
+  "tool_use_config": {
+    "enable_tool_calling": true,
+    "tool_call_format": "json",
+    "max_tool_calls_per_turn": 5,
+    "tool_response_format": "structured",
+    "parallel_tool_calls": true,
+    "tool_choice": "auto"
+  },
+  
+  "vision_processing": {
+    "max_num_images": 10,
+    "image_token_length": 729,
+    "image_encoding": "base64",
+    "supported_formats": ["jpg", "jpeg", "png", "webp", "gif"],
+    "max_image_size_mb": 20,
+    "min_image_dimension": 224,
+    "max_image_dimension": 2048,
+    "preserve_aspect_ratio": true
+  },
+  
+  "context_management": {
+    "max_context_length": 200000,
+    "context_window_strategy": "sliding",
+    "truncation_strategy": "left",
+    "preserve_system_message": true,
+    "context_compression": false
+  },
+  
+  "output_formatting": {
+    "return_full_text": false,
+    "clean_up_tokenization_spaces": true,
+    "skip_special_tokens": true,
+    "return_tensors": false,
+    "return_token_type_ids": false,
+    "return_attention_mask": false
+  },
+  
+  "error_handling": {
+    "on_error": "raise",
+    "fallback_behavior": "default_response",
+    "retry_attempts": 0,
+    "timeout_seconds": 300
+  },
+  
+  "logging": {
+    "log_level": "WARNING",
+    "log_inputs": false,
+    "log_outputs": false,
+    "log_timings": false,
+    "log_memory_usage": false
+  },
+  
+  "compatibility": {
+    "transformers_version": "4.36.0",
+    "torch_version": "2.1.0",
+    "python_version": "3.10+",
+    "cuda_version": "12.1+",
+    "supports_quantization": true,
+    "supports_distributed": true
+  },
+  
+  "experimental_features": {
+    "enable_streaming": false,
+    "enable_batched_inference": false,
+    "enable_speculative_decoding": false,
+    "enable_kv_cache_quantization": false
+  }
+}