tarekziade HF Staff
Publish MIC usage, slim fixture and reviewed Transformers expectations
09d714a verified | { | |
| "mic": "0.3", | |
| "name": "OWL-ViT open-vocabulary object detection", | |
| "usage": "OwlViTProcessor + OwlViTForObjectDetection, batched text queries and images, explicit detection loss for training", | |
| "architecture": "OwlViTForObjectDetection", | |
| "checkpoints": { | |
| "google/owlvit-base-patch32": { | |
| "fixture": "huggingface-mic/owlvit-base-patch32-mic" | |
| } | |
| }, | |
| "fixtures": { | |
| "huggingface-mic/owlvit-base-patch32-mic": { | |
| "built_from": "google/owlvit-base-patch32", | |
| "covers": [ | |
| "google/owlvit-base-patch32" | |
| ], | |
| "build": { | |
| "seed": 17, | |
| "model_class": "OwlViTForObjectDetection", | |
| "accept_insensitive": { | |
| "attention_mask:all_ones": "OWL-ViT's text tower is causal and pools at the EOS token, so padding after EOS cannot move any output" | |
| } | |
| } | |
| } | |
| }, | |
| "integrations": { | |
| "transformers": { | |
| "entry_point": "integrations/transformers/usage.md", | |
| "format": "runnable-markdown", | |
| "requires": [ | |
| "transformers>=5", | |
| "torch", | |
| "torchvision" | |
| ], | |
| "targets": [ | |
| "fixtures", | |
| "checkpoints" | |
| ], | |
| "variants": { | |
| "default": {}, | |
| "eager": { | |
| "attn_implementation": "eager" | |
| }, | |
| "bfloat16": { | |
| "dtype": "bfloat16" | |
| } | |
| }, | |
| "training": true, | |
| "resources": { | |
| "fixture": { | |
| "memory_gb": 2, | |
| "download_mb": 20 | |
| }, | |
| "google/owlvit-base-patch32": { | |
| "memory_gb": 4, | |
| "download_mb": 620 | |
| } | |
| } | |
| } | |
| }, | |
| "outputs": { | |
| "runtime": { | |
| "compare": "exact" | |
| }, | |
| "detections.boxes": { | |
| "compare": "allclose", | |
| "atol": 0.0001, | |
| "rtol": 0.0001 | |
| }, | |
| "detections.scores": { | |
| "compare": "allclose", | |
| "atol": 0.0001, | |
| "rtol": 0.0001 | |
| }, | |
| "detections.labels": { | |
| "compare": "exact" | |
| }, | |
| "training.loss": { | |
| "compare": "allclose", | |
| "atol": 0.0001, | |
| "rtol": 0.0001 | |
| }, | |
| "training.gradients": { | |
| "compare": "all_true" | |
| }, | |
| "training.updated": { | |
| "compare": "true" | |
| } | |
| } | |
| } | |