tarekziade HF Staff
Publish MIC usage, slim fixture and reviewed Transformers expectations
851c4b7 verified | # Mask2Former: instance segmentation with Transformers | |
| This page is the Transformers contract of the Mask2Former MIC and also a usage example. [doc-builder](https://github.com/huggingface/doc-builder/blob/main/docs/runnable-code-blocks.md) renders the `runnable` blocks below as documentation, and its pytest plugin runs them as written against `facebook/mask2former-swin-tiny-coco-instance`. The MIC runner runs the same blocks against each slim fixture and each checkpoint in [`integration.json`](../../integration.json). Lines marked `# doc-builder: hide` only matter there: they swap in the pinned checkpoint or variant and print the result record. | |
| ## Predict instances on a batch of images | |
| ```py runnable:instance_segmentation | |
| import torch | |
| from transformers import AutoImageProcessor, Mask2FormerForUniversalSegmentation, set_seed | |
| import json # doc-builder: hide | |
| import os # doc-builder: hide | |
| model_id = "facebook/mask2former-swin-tiny-coco-instance" | |
| model_id = os.environ.get("MIC_MODEL", model_id) # doc-builder: hide | |
| variant = json.loads(os.environ.get("MIC_MODEL_KWARGS", "{}")) # doc-builder: hide | |
| device = "cuda" if torch.cuda.is_available() else "cpu" | |
| set_seed(0) # seeds torch, NumPy, and Python: loading and dropout draw random numbers | |
| processor = AutoImageProcessor.from_pretrained(model_id) | |
| model = Mask2FormerForUniversalSegmentation.from_pretrained(model_id).to(device) | |
| if variant: # doc-builder: hide | |
| del model # doc-builder: hide | |
| model = Mask2FormerForUniversalSegmentation.from_pretrained(model_id, **variant).to(device) # doc-builder: hide | |
| # Two small synthetic images so the batch perturbation is meaningful. | |
| images = [torch.randint(0, 256, (3, 64, 64), dtype=torch.uint8) for _ in range(2)] | |
| inputs = processor(images=images, return_tensors="pt") | |
| inputs = inputs.to(device, dtype=model.dtype) | |
| model.eval() | |
| with torch.no_grad(): | |
| outputs = model(**inputs) | |
| # class_queries_logits: (batch, num_queries + 1); masks_queries_logits: (batch, num_queries, h, w) | |
| class_logits = outputs.class_queries_logits | |
| mask_logits = outputs.masks_queries_logits | |
| print(class_logits.shape, mask_logits.shape) | |
| ``` | |
| ## One training step | |
| ```py runnable:instance_segmentation:2 | |
| # Mask2Former trains on per-image binary masks (at the pixel_values resolution) and | |
| # their int64 class ids, passed as mask_labels / class_labels lists. | |
| height, width = inputs["pixel_values"].shape[-2:] | |
| masks = torch.zeros(2, height, width, device=device) | |
| masks[0, height // 4 : 3 * height // 4, width // 4 : 3 * width // 4] = 1 | |
| masks[1, : height // 4, : width // 4] = 1 | |
| mask_labels = [masks.to(model.dtype) for _ in range(2)] | |
| class_labels = [torch.tensor([0, 2], dtype=torch.int64, device=device) for _ in range(2)] | |
| model.train() | |
| optimizer = torch.optim.SGD([p for p in model.parameters() if p.requires_grad], lr=0.01) | |
| with torch.autocast(device_type=model.device.type, dtype=model.dtype, enabled=model.dtype != torch.float32): | |
| loss = model(**inputs, mask_labels=mask_labels, class_labels=class_labels).loss | |
| loss.backward() | |
| before = {n: p.detach().cpu().clone() for n, p in model.named_parameters() if p.requires_grad} # doc-builder: hide | |
| optimizer.step() | |
| print(f"loss: {loss.item():.4f}") | |
| if os.environ.get("MIC_RUNNER"): # doc-builder: hide | |
| named = dict(model.named_parameters()) | |
| checked = ( | |
| "class_predictor.weight", | |
| "model.pixel_level_module.decoder.mask_projection.weight", | |
| "model.pixel_level_module.encoder.swin.encoder.layers.0.blocks.0.attention.q_proj.weight", | |
| ) | |
| result = { | |
| "runtime": { | |
| "device": model.device.type, | |
| "dtype": str(model.dtype).removeprefix("torch."), | |
| "attn_implementation": model.config._attn_implementation, | |
| }, | |
| "class_logits_probe": class_logits[0, :4].float().tolist(), | |
| "mask_logits_probe": mask_logits[0, 0, :2, :2].float().tolist(), | |
| "training": { | |
| "loss": loss.item(), | |
| "gradients": { | |
| n: bool(named[n].grad is not None and torch.count_nonzero(named[n].grad) > 0 and torch.isfinite(named[n].grad).all()) | |
| for n in checked | |
| }, | |
| "updated": any(not torch.equal(p.detach().cpu(), before[n]) for n, p in model.named_parameters() if n in before), | |
| }, | |
| } | |
| print(json.dumps(result)) | |
| ``` | |