| { | |
| "model_type": "oculus", | |
| "architectures": ["OculusForConditionalGeneration"], | |
| "architecture_name": "Oceanir-Oculus OO1", | |
| "version": "0.1", | |
| "vision_config": { | |
| "hidden_size": 1024, | |
| "num_layers": 24, | |
| "num_heads": 16, | |
| "image_size": 224, | |
| "patch_size": 16 | |
| }, | |
| "language_config": { | |
| "hidden_size": 1536, | |
| "num_layers": 16, | |
| "num_heads": 24, | |
| "vocab_size": 131072, | |
| "max_position_embeddings": 32768 | |
| }, | |
| "trained_components": { | |
| "projector": "trained_components/projector.npz", | |
| "heads": "trained_components/heads.pth" | |
| }, | |
| "projector_config": { | |
| "fused_vision_dim": 2176, | |
| "hidden_dim": 4352, | |
| "output_dim": 1536, | |
| "num_tokens": 64 | |
| }, | |
| "task_heads": { | |
| "segmentation_classes": 150, | |
| "detection_classes": 80, | |
| "ui_element_classes": 50 | |
| }, | |
| "reasoning": { | |
| "enabled": true, | |
| "thinking_token": "<think>", | |
| "thinking_end_token": "</think>", | |
| "max_thinking_tokens": 256, | |
| "style": "structured" | |
| }, | |
| "focus_system": { | |
| "enabled": true, | |
| "focus_token": "<focus>", | |
| "focus_end_token": "</focus>", | |
| "max_regions": 4, | |
| "min_crop_size": 64, | |
| "auto_threshold": 0.7 | |
| }, | |
| "structured_output": { | |
| "enabled": true, | |
| "json_token": "<json>", | |
| "json_end_token": "</json>", | |
| "box_token": "<box>", | |
| "box_end_token": "</box>", | |
| "point_token": "<point>", | |
| "point_end_token": "</point>" | |
| }, | |
| "ocr": { | |
| "enabled": true, | |
| "languages": ["en"], | |
| "confidence_threshold": 0.5 | |
| }, | |
| "ui_understanding": { | |
| "enabled": true, | |
| "element_classes": 50 | |
| }, | |
| "generation": { | |
| "max_new_tokens": 512, | |
| "temperature": 0.7, | |
| "top_p": 0.95 | |
| } | |
| } | |