chris-hughes10 / yolov7-training Goto Github PK

(Pdb) torch.load('yolov7-tiny.pt', map_location="cpu")
*** AttributeError: Can't get attribute 'Model' on <module 'models.yolo' (namespace)>

Transferred 555/566 items from https://github.com/Chris-hughes10/Yolov7-training/releases/download/0.1.0/yolov7_training_state_dict.pt

Starting training run

Starting epoch 1
  0%|                                                                                                   | 0/1013 [00:00<?, ?it/s]
torch.Size([9, 6]) 4.  <--- output of print(image_preds.shape, PredIdx.OBJ)
../aten/src/ATen/native/cuda/ScatterGatherKernel.cu:367: operator(): block: [0,0,0], thread: [0,0,0] Assertion `idx_dim >= 0 && idx_dim < index_size && "index out of bounds"` failed.
Traceback (most recent call last):
    main()
  File "/opt/conda/envs/yolov7/lib/python3.9/site-packages/func_to_script/core.py", line 108, in scripted_function
    return func(**args)
  File "/home/varshanth/yolov7/fine_tune.py", line 77, in main
    trainer.train(
  File "/opt/conda/envs/yolov7/lib/python3.9/site-packages/pytorch_accelerated/trainer.py", line 467, in train
    self._run_training()
  File "/opt/conda/envs/yolov7/lib/python3.9/site-packages/pytorch_accelerated/trainer.py", line 676, in _run_training
    self._run_train_epoch(self._train_dataloader)
  File "/opt/conda/envs/yolov7/lib/python3.9/site-packages/pytorch_accelerated/trainer.py", line 749, in _run_train_epoch
    self._perform_forward_and_backward_passes(batch)
  File "/opt/conda/envs/yolov7/lib/python3.9/site-packages/pytorch_accelerated/trainer.py", line 776, in _perform_forward_and_backward_passes
    batch_output = self.calculate_train_batch_loss(batch)
  File "/home/varshanth/yolov7/yolov7/trainer.py", line 107, in calculate_train_batch_loss
    loss, _ = self.loss_func(
  File "/home/varshanth/yolov7/yolov7/loss.py", line 204, in __call__
    box_loss, obj_loss, cls_loss = self._compute_losses(
  File "/home/varshanth/yolov7/yolov7/loss.py", line 237, in _compute_losses_for_train
    anchor_boxes_per_layer, targets_per_layer = self.simOTA_assignment(
  File "/home/varshanth/yolov7/yolov7/loss.py", line 619, in simOTA_assignment
    pred_objectness = image_preds[:, [PredIdx.OBJ]]
RuntimeError: CUDA error: device-side assert triggered
CUDA kernel errors might be asynchronously reported at some other API call,so the stacktrace below might be incorrect.
For debugging consider passing CUDA_LAUNCH_BLOCKING=1.
  0%|                                                                                                   | 0/1013 [00:03<?, ?it/s

 The truth value of an array with more than one element is ambiguous. Use a.any() or a.all()

    model = create_yolov7_model(
        architecture="yolov7-w6", num_classes=num_classes, pretrained=pretrained
    )

image_size: int = 640
num_classes = 1
pretrained=False
DATA_PATH = Path("Yolov7-trainer/datasets/data")
ckpt_path = "./lightning_logs/checkpoints/epoch=4-step=205.ckpt"
batch_size: int = 8

data_path = Path(DATA_PATH)
images_path = data_path / "training_images"
annotations_file_path = data_path / "annotations.csv"


model = create_yolov7_model(
        architecture="yolov7", num_classes=num_classes, pretrained=True
    )
loss_func = create_yolov7_loss(model, image_size=image_size)

optimizer = torch.optim.SGD(
    model.parameters(), lr=0.01, momentum=0.9, nesterov=True
)
loaded_model = Yolov7Trainer.load_from_checkpoint(ckpt_path, model=model, optimizer=optimizer, loss_func=loss_func)
loaded_model.eval()

#single image prediction
img = np.array(Image.open(f'{images_path}/vid_4_2160.jpg').convert("RGB"))
original_image_sizes = torch.tensor([img.shape[:2]])

transforms = create_yolov7_transforms()
img = transforms(image=img, bboxes=[], labels = [])

with torch.no_grad():
    img = torch.from_numpy(img["image"]).permute(2,0,1).unsqueeze(0).type('torch.FloatTensor')
    y_hat = loaded_model(img.cuda())
    preds = model.postprocess(y_hat, conf_thres=0.8) #threshold
    preds = filter_eval_predictions(preds)  #nms

img = np.array(Image.open(f'{images_path}/vid_4_2160.jpg').convert("RGB"))
img = transforms(image=img, bboxes=[], labels = [])
fig, axe = plt.subplots()
axe.imshow(img["image"])
bboxes = preds[0][:,:4]
for box in bboxes.cpu():
    box_cvt = torchvision.ops.box_convert(box, in_fmt="xyxy", out_fmt="xywh")
    rect = patches.Rectangle((box_cvt[0],box_cvt[1]), 
                             box_cvt[2], 
                             box_cvt[3],
                             edgecolor="b",
                             fill=False)
    axe.add_patch(rect)

pred_class_probs = (
    (pred_class_scores.sigmoid_() * pred_objectness.sigmoid_())
    .unsqueeze(dim=0)
    .repeat(num_image_targets, 1, 1)
)
y = pred_class_probs.sqrt_()
pair_wise_cls_loss = F.binary_cross_entropy_with_logits(
    torch.log(y/(1-y)), target_class_probs, reduction="none"
).sum(-1)

   if RESUME_LOCAL_PATH is not None:
        ckpoint = torch.load(RESUME_LOCAL_PATH)
        optimizer.state_dict().update(ckpoint['optimizer_state_dict'])
   
   ......

   state_dict = torch.load(RESUME_LOCAL_PATH )
   state_dict = intersect_dicts(
                state_dict,
                model.state_dict(),
                exclude=["anchor"],
            )

   model.load_state_dict(state_dict, strict=False)

Namespace(weights=['best_model.pt'], source='/content/gdrive/MyDrive/CookiePy/Script/yolov7/data/TestSet/images', img_size=640, conf_thres=0.5, iou_thres=0.45, device='', view_img=False, save_txt=False, save_conf=False, nosave=False, classes=None, agnostic_nms=False, augment=False, update=False, project='runs/detect', name='exp', exist_ok=False, no_trace=True)
YOLOR 🚀 55b90e1 torch 2.0.1+cu118 CPU

Traceback (most recent call last):
  File "/content/gdrive/MyDrive/CookiePy/Script/yolov7/detect.py", line 195, in <module>
    detect()
  File "/content/gdrive/MyDrive/CookiePy/Script/yolov7/detect.py", line 34, in detect
    model = attempt_load(weights, map_location=device)  # load FP32 model
  File "/content/gdrive/MyDrive/CookiePy/Script/yolov7/models/experimental.py", line 253, in attempt_load
    model.append(ckpt['ema' if ckpt.get('ema') else 'model'].float().fuse().eval())  # FP32 model
KeyError: 'model'