diff --git a/cv/detection/yolov5/pytorch/README.md b/cv/detection/yolov5/pytorch/README.md index 475fb9973d038502a820104710e926c3d84f3630..15c74620d5ca6f66cb6b5fba0839c01b3c195ee7 100644 --- a/cv/detection/yolov5/pytorch/README.md +++ b/cv/detection/yolov5/pytorch/README.md @@ -6,9 +6,8 @@ YOLOv5 🚀 is a family of object detection architectures and models pretrained ```shell # install libGL -yum install mesa-libGL - -pip3 install -r requirements.txt +yum install -y mesa-libGL +bash init.sh ``` ## Step 2: Preparing datasets @@ -38,10 +37,12 @@ coco2017 Modify the configuration file(data/coco.yaml) - $ vim data/coco.yaml - $ # path: the root of coco data - $ # train: the relative path of train images - $ # val: the relative path of valid images +```bash +$ vim data/coco.yaml +# path: the root of coco data +# train: the relative path of train images +# val: the relative path of valid images +``` ## Training the detector @@ -49,45 +50,65 @@ Train the yolov5 model as follows, the train log is saved in ./runs/train/exp ### On single GPU - $ cd yolov5 - $ python3 train.py --data ./data/coco.yaml --batch-size 32 --cfg ./models/yolov5s.yaml --weights '' +```bash +$ python3 train.py --data ./data/coco.yaml --batch-size 32 --cfg ./models/yolov5s.yaml --weights '' +``` ### On single GPU (AMP) - $ python3 train.py --data ./data/coco.yaml --batch-size 32 --cfg ./models/yolov5s.yaml --weights '' --amp - +```bash +$ python3 train.py --data ./data/coco.yaml --batch-size 32 --cfg ./models/yolov5s.yaml --weights '' --amp +``` ### Multiple GPUs on one machine - $ # eight cards - $ python3 -m torch.distributed.launch --nproc_per_node 8 train.py --data ./data/coco.yaml --batch-size 256 --cfg ./models/yolov5s.yaml --weights '' --device 0,1,2,3,4,5,6,7 +```bash +# eight cards +# YOLOv5s +$ python3 -m torch.distributed.launch --nproc_per_node 8 \ + train.py \ + --data ./data/coco.yaml \ + --batch-size 64 \ + --cfg ./models/yolov5s.yaml --weights '' \ + --device 0,1,2,3,4,5,6,7 + +# YOLOv5m +$ bash run.sh +``` ### Multiple GPUs on one machine (AMP) - $ # eight cards - $ python3 -m torch.distributed.launch --nproc_per_node 8 train.py --data ./data/coco.yaml --batch-size 256 --cfg ./models/yolov5s.yaml --weights '' --device 0,1,2,3,4,5,6,7 --amp - +```bash +# eight cards +$ python3 -m torch.distributed.launch --nproc_per_node 8 \ + train.py \ + --data ./data/coco.yaml \ + --batch-size 256 \ + --cfg ./models/yolov5s.yaml --weights '' \ + --device 0,1,2,3,4,5,6,7 --amp +``` ## Test the detector Test the yolov5 model as follows, the result is saved in ./runs/detect: - $ cd yolov5 - $ python3 detect.py --source ./data/images/bus.jpg --weights yolov5s.pt --img 640 - $ python3 detect.py --source ./data/images/zidane.jpg --weights yolov5s.pt --img 640 - +$ cd yolov5 +$ python3 detect.py --source ./data/images/bus.jpg --weights yolov5s.pt --img 640 +$ python3 detect.py --source ./data/images/zidane.jpg --weights yolov5s.pt --img 640 ## Results on BI-V100 + | GPUs | FP16 | Batch size | FPS | E2E | mAP@.5 | -|------|------|------------|-----|-----|--------| -| 1x1 | True | 64 | 81 | N/A | N/A | -| 1x8 | True | 64 | 598 | 24h | 0.632 | +| ------ | ------ | ------------ | ----- | ----- | -------- | +| 1x1 | True | 64 | 81 | N/A | N/A | +| 1x8 | True | 64 | 598 | 24h | 0.632 | -| Convergence criteria | Configuration (x denotes number of GPUs) | Performance | Accuracy | Power（W） | Scalability | Memory utilization（G） | Stability | -|----------------------|------------------------------------------|-------------|----------|------------|-------------|-------------------------|-----------| -| mAP:0.5 | SDK V2.2,bs:128,8x,AMP | 1228 | 0.56 | 140\*8 | 0.92 | 27.3\*8 | 1 | +| Convergence criteria | Configuration (x denotes number of GPUs) | Performance | Accuracy | Power（W） | Scalability | Memory utilization（G） | Stability | +| ---------------------- | ------------------------------------------ | ------------- | ---------- | ------------ | ------------- | ------------------------- | ----------- | +| mAP:0.5 | SDK V2.2, bs:128, 8x, AMP | 1228 | 0.56 | 140\*8 | 0.92 | 27.3\*8 | 1 | ## Reference + https://github.com/ultralytics/yolov5 diff --git a/cv/detection/yolov5/pytorch/init.sh b/cv/detection/yolov5/pytorch/init.sh new file mode 100644 index 0000000000000000000000000000000000000000..44510f81a271bef34200e15ac322c9e7735edcef --- /dev/null +++ b/cv/detection/yolov5/pytorch/init.sh @@ -0,0 +1,42 @@ +#!/bin/bash +# Copyright (c) 2023, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); you may +# not use this file except in compliance with the License. You may obtain +# a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT +# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the +# License for the specific language governing permissions and limitations +# under the License. + +if [[ "$(uname)" == "Linux" ]]; then + if command -v apt &> /dev/null; then + apt install -y numactl + elif command -v yum &> /dev/null; then + yum install -y numactl + else + echo "Unsupported package manager" + exit 1 + fi +else + echo "Unsupported operating system" + exit 1 +fi + +pip3 install -r requirements.txt +PY_VERSION=$(python3 -V 2>&1|awk '{print $2}'|awk -F '.' '{print $2}') +if [ "$PY_VERSION" == "10" ]; then + pip3 install matplotlib==3.8.2 + pip3 install numpy==1.22.4 + pip3 install Pillow==9.5 +else + echo "only for python3.10" +fi + +wandb disabled +pip3 install pycocotools diff --git a/cv/detection/yolov5/pytorch/models/common.py b/cv/detection/yolov5/pytorch/models/common.py index 2308ace43b1a4a928abab9705d17058153d8ca7c..7603e9e4719202e9143bcfd80984bf92128085d9 100644 --- a/cv/detection/yolov5/pytorch/models/common.py +++ b/cv/detection/yolov5/pytorch/models/common.py @@ -49,24 +49,6 @@ class Conv(nn.Module): return self.act(self.conv(x)) -class SPPF(nn.Module): - # Spatial Pyramid Pooling - Fast (SPPF) layer for YOLOv5 by Glenn Jocher - def __init__(self, c1, c2, k=5): # equivalent to SPP(k=(5, 9, 13)) - super().__init__() - c_ = c1 // 2 # hidden channels - self.cv1 = Conv(c1, c_, 1, 1) - self.cv2 = Conv(c_ * 4, c2, 1, 1) - self.m = nn.MaxPool2d(kernel_size=k, stride=1, padding=k // 2) - - def forward(self, x): - x = self.cv1(x) - with warnings.catch_warnings(): - warnings.simplefilter('ignore') # suppress torch 1.9.0 max_pool2d() warning - y1 = self.m(x) - y2 = self.m(y1) - return self.cv2(torch.cat([x, y1, y2, self.m(y2)], 1)) - - class TransformerLayer(nn.Module): # Transformer layer https://arxiv.org/abs/2010.11929 (LayerNorm layers removed for better performance) def __init__(self, c, num_heads): diff --git a/cv/detection/yolov5/pytorch/run.sh b/cv/detection/yolov5/pytorch/run.sh new file mode 100644 index 0000000000000000000000000000000000000000..ed5cda7d4b22ed2e4e09c1e0cd43208694c31cc6 --- /dev/null +++ b/cv/detection/yolov5/pytorch/run.sh @@ -0,0 +1,38 @@ +#!/bin/bash +# Copyright (c) 2023, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); you may +# not use this file except in compliance with the License. You may obtain +# a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT +# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the +# License for the specific language governing permissions and limitations +# under the License. + +start_time=$(date +%s) +unset no_proxy use_proxy https_proxy http_proxy +EXIT_STATUS=0 +check_status() { + if ((${PIPESTATUS[0]} != 0)); then + EXIT_STATUS=1 + fi +} + +python3 -m torch.distributed.launch --nproc_per_node=16 \ + train.py --batch-size 32 \ + --data ./data/coco.yaml --weights "" \ + --cfg models/yolov5m.yaml --workers 16 \ + --epochs 3 --linear-lr "$@" +check_status + +wait + +end_time=$(date +%s) +e2e_time=$(($end_time - $start_time)) +echo "end to end time: $e2e_time" >>total_time.log +exit ${EXIT_STATUS} diff --git a/cv/detection/yolov5/pytorch/run_inference.sh b/cv/detection/yolov5/pytorch/run_inference.sh index 23a092df8f0d2705b8a5eff76d586d683f466463..3f4226fbc9ee681d9f1d0ea0fc1207d5a16beabb 100644 --- a/cv/detection/yolov5/pytorch/run_inference.sh +++ b/cv/detection/yolov5/pytorch/run_inference.sh @@ -13,4 +13,14 @@ # WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the # License for the specific language governing permissions and limitations # under the License. -python3 test.py --task val --data data/coco.yaml --weights weights/yolov5s.pt 2>&1 | tee inferencelog.log; + +EXIT_STATUS=0 +check_status() { + if ((${PIPESTATUS[0]} != 0)); then + EXIT_STATUS=1 + fi +} + +python3 test.py --task val --data data/coco128.yaml --weights weights/yolov5s.pt 2>&1 | tee inferencelog.log +check_status +exit ${EXIT_STATUS} diff --git a/cv/detection/yolov5/pytorch/start_scripts/train_yolov5s_coco128_amp_torch.sh b/cv/detection/yolov5/pytorch/start_scripts/train_yolov5s_coco128_amp_torch.sh new file mode 100644 index 0000000000000000000000000000000000000000..b84888011df7f1ee71b7b90a6014655560baea16 --- /dev/null +++ b/cv/detection/yolov5/pytorch/start_scripts/train_yolov5s_coco128_amp_torch.sh @@ -0,0 +1,21 @@ +#!/bin/bash +# Copyright (c) 2023, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); you may +# not use this file except in compliance with the License. You may obtain +# a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT +# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the +# License for the specific language governing permissions and limitations +# under the License. + +export PYTORCH_DISABLE_VEC_KERNEL=1 +export PT_USE_CUDNN_BATCHNORM_SPATIAL_PERSISTENT=1 +cd .. +bash run_training.sh --data ./data/coco128.yaml --amp "$@" +cd - diff --git a/cv/detection/yolov5/pytorch/start_scripts/train_yolov5s_coco128_dist_torch.sh b/cv/detection/yolov5/pytorch/start_scripts/train_yolov5s_coco128_dist_torch.sh new file mode 100644 index 0000000000000000000000000000000000000000..0a37d6c7a3d5650cb5cd1f8d575d2cf60f029cd5 --- /dev/null +++ b/cv/detection/yolov5/pytorch/start_scripts/train_yolov5s_coco128_dist_torch.sh @@ -0,0 +1,18 @@ +#!/bin/bash +# Copyright (c) 2023, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); you may +# not use this file except in compliance with the License. You may obtain +# a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT +# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the +# License for the specific language governing permissions and limitations +# under the License. +cd .. +bash run_dist_training.sh --data ./data/coco128.yaml "$@" +cd - diff --git a/cv/detection/yolov5/pytorch/start_scripts/train_yolov5s_coco128_torch.sh b/cv/detection/yolov5/pytorch/start_scripts/train_yolov5s_coco128_torch.sh new file mode 100644 index 0000000000000000000000000000000000000000..c206974245492b3568c5eb85f0ded2bf0f809236 --- /dev/null +++ b/cv/detection/yolov5/pytorch/start_scripts/train_yolov5s_coco128_torch.sh @@ -0,0 +1,18 @@ +#!/bin/bash +# Copyright (c) 2023, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); you may +# not use this file except in compliance with the License. You may obtain +# a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT +# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the +# License for the specific language governing permissions and limitations +# under the License. +cd .. +bash run_training.sh --data ./data/coco128.yaml "$@" +cd - diff --git a/cv/detection/yolov5/pytorch/start_scripts/train_yolov5s_coco_amp_torch.sh b/cv/detection/yolov5/pytorch/start_scripts/train_yolov5s_coco_amp_torch.sh new file mode 100644 index 0000000000000000000000000000000000000000..fe09a71af5a7f123c4fff99f2aa8892c50dff3f9 --- /dev/null +++ b/cv/detection/yolov5/pytorch/start_scripts/train_yolov5s_coco_amp_torch.sh @@ -0,0 +1,21 @@ +#!/bin/bash +# Copyright (c) 2023, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); you may +# not use this file except in compliance with the License. You may obtain +# a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT +# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the +# License for the specific language governing permissions and limitations +# under the License. + +export PYTORCH_DISABLE_VEC_KERNEL=1 +export PT_USE_CUDNN_BATCHNORM_SPATIAL_PERSISTENT=1 +cd .. +bash run_training.sh --data ./data/coco.yaml --amp "$@" +cd - diff --git a/cv/detection/yolov5/pytorch/start_scripts/train_yolov5s_coco_dist_torch.sh b/cv/detection/yolov5/pytorch/start_scripts/train_yolov5s_coco_dist_torch.sh new file mode 100644 index 0000000000000000000000000000000000000000..3434fee60cc38489585c7b96c785980f149701fb --- /dev/null +++ b/cv/detection/yolov5/pytorch/start_scripts/train_yolov5s_coco_dist_torch.sh @@ -0,0 +1,18 @@ +#!/bin/bash +# Copyright (c) 2023, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); you may +# not use this file except in compliance with the License. You may obtain +# a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT +# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the +# License for the specific language governing permissions and limitations +# under the License. + +cd .. +bash run_dist_training.sh --data ./data/coco.yaml "$@" diff --git a/cv/detection/yolov5/pytorch/start_scripts/train_yolov5s_coco_torch.sh b/cv/detection/yolov5/pytorch/start_scripts/train_yolov5s_coco_torch.sh new file mode 100644 index 0000000000000000000000000000000000000000..94da449ae65fb8dbceccd0a0234685def4ba53ed --- /dev/null +++ b/cv/detection/yolov5/pytorch/start_scripts/train_yolov5s_coco_torch.sh @@ -0,0 +1,19 @@ +#!/bin/bash +# Copyright (c) 2023, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. +# +# Licensed under the Apache License, Version 2.0 (the "License"); you may +# not use this file except in compliance with the License. You may obtain +# a copy of the License at +# +# http://www.apache.org/licenses/LICENSE-2.0 +# +# Unless required by applicable law or agreed to in writing, software +# distributed under the License is distributed on an "AS IS" BASIS, WITHOUT +# WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the +# License for the specific language governing permissions and limitations +# under the License. + +cd .. +bash run_training.sh --data ./data/coco.yaml "$@" +cd - \ No newline at end of file diff --git a/cv/detection/yolov5/pytorch/train.py b/cv/detection/yolov5/pytorch/train.py index 2bc68e9815ce4679668fd47eec8eb55ce98e629a..f8f0f95a22846e7f2f76f649f68e88357cf30bfb 100644 --- a/cv/detection/yolov5/pytorch/train.py +++ b/cv/detection/yolov5/pytorch/train.py @@ -335,7 +335,7 @@ def train(hyp, # path/to/hyp.yaml or hyp dictionary logger.info('Using SyncBatchNorm()') # Trainloader - dataloader, dataset = create_dataloader(train_path, imgsz, batch_size // WORLD_SIZE, gs, single_cls, + dataloader, dataset = create_dataloader(train_path, imgsz, batch_size, gs, single_cls, hyp=hyp, augment=False, cache=opt.cache_images, rect=opt.rect, rank=RANK, workers=workers, image_weights=opt.image_weights, quad=opt.quad, prefix=colorstr('train: ')) @@ -345,7 +345,7 @@ def train(hyp, # path/to/hyp.yaml or hyp dictionary # Process 0 if RANK in [-1, 0]: - testloader = create_dataloader(test_path, imgsz_test, batch_size // WORLD_SIZE * 2, gs, single_cls, + testloader = create_dataloader(test_path, imgsz_test, batch_size, gs, single_cls, hyp=hyp, cache=opt.cache_images and not notest, rect=True, rank=-1, workers=workers, pad=0.5, prefix=colorstr('val: '))[0] @@ -397,10 +397,6 @@ def train(hyp, # path/to/hyp.yaml or hyp dictionary f'Using {dataloader.num_workers} dataloader workers\n' f'Logging results to {save_dir}\n' f'Starting training for {epochs} epochs...') - - run_steps = 0 - time_step = [] - time_step.append(time.time()) for epoch in range(start_epoch, epochs): # epoch ------------------------------------------------------------------ model.train() @@ -426,11 +422,13 @@ def train(hyp, # path/to/hyp.yaml or hyp dictionary if RANK != -1: dataloader.sampler.set_epoch(epoch) pbar = enumerate(dataloader) - logger.info(('\n' + '%10s' * 8) % ('Epoch', 'gpu_mem', 'box', 'obj', 'cls', 'total', 'labels', 'img_size')) + logger.info(('\n' + '%10s' * 8) % ('Epoch', 'gpu_mem', 'box', 'obj', 'cls', 'total', 'img_size', "total_fps")) if RANK in [-1, 0]: pbar = tqdm(pbar, total=nb) # progress bar optimizer.zero_grad() for i, (imgs, targets, paths, _) in pbar: # batch ------------------------------------------------------------- + step_start_time = time.time() + ni = i + nb * epoch # number integrated batches (since train start) imgs = imgs.to(device, non_blocking=True).float() / 255.0 # uint8 to float32, 0-255 to 0.0-1.0 @@ -470,6 +468,10 @@ def train(hyp, # path/to/hyp.yaml or hyp dictionary if opt.quad: loss *= 4. + if not math.isfinite(loss[0]): + print("Loss is {}, stopping training".format(loss[0])) + sys.exit(1) + # Backward if opt.amp: scaler.scale(loss).backward() @@ -491,18 +493,18 @@ def train(hyp, # path/to/hyp.yaml or hyp dictionary ema.update(model) last_opt_step = ni + step_end_time = time.time() + fps = len(imgs) / (step_end_time - step_start_time) + if torch.distributed.is_initialized(): + fps = fps * torch.distributed.get_world_size() + # Print if RANK in [-1, 0]: - time_step.append(time.time()) mloss = (mloss * i + loss_items) / (i + 1) # update mean losses mem = '%.3gG' % (torch.cuda.memory_reserved() / 1E9 if torch.cuda.is_available() else 0) # (GB) s = ('%10s' * 2 + '%10.4g' * 6) % ( - f'{epoch}/{epochs - 1}', mem, *mloss, targets.shape[0], imgs.shape[-1]) - time_iter = time_step[i+1] - time_step[i] - fps = opt.batch_size / time_iter - performance = " timer: %.6f sec. || fps: %.3f" % (time_iter, fps) - - pbar.set_description(s + performance) + f'{epoch}/{epochs - 1}', mem, *mloss, imgs.shape[-1], fps) + pbar.set_description(s) if nb > 1000: log_freq = 100 @@ -677,7 +679,7 @@ def parse_opt(known=False): parser.add_argument('--bbox_interval', type=int, default=-1, help='Set bounding-box image logging interval for W&B') parser.add_argument('--save_period', type=int, default=-1, help='Log model after every "save_period" epoch') parser.add_argument('--artifact_alias', type=str, default="latest", help='version of dataset artifact to be used') - parser.add_argument('--local_rank', type=int, default=-1, help='DDP parameter, do not modify') + parser.add_argument('--local_rank', '--local-rank', type=int, default=-1, help='DDP parameter, do not modify') parser.add_argument('--amp', action='store_true', default=False, help='use amp to train and test') opt = parser.parse_known_args()[0] if known else parser.parse_args() return opt @@ -709,6 +711,12 @@ def main(opt): print("Global setting:", LOCAL_RANK, RANK, WORLD_SIZE) + try: + from dltest import show_training_arguments + show_training_arguments(opt) + except: + pass + # DDP mode device = select_device(opt.device, batch_size=opt.batch_size) if LOCAL_RANK != -1: @@ -724,7 +732,7 @@ def main(opt): dist_backend = os.environ[DIST_BACKEND_ENV] dist.init_process_group(backend=dist_backend, rank=RANK, world_size=WORLD_SIZE) - assert opt.batch_size % WORLD_SIZE == 0, '--batch-size must be multiple of CUDA device count' + # assert opt.batch_size % WORLD_SIZE == 0, '--batch-size must be multiple of CUDA device count' assert not opt.image_weights, '--image-weights argument is not compatible with DDP training' # Train diff --git a/cv/detection/yolov5/pytorch/utils/loss.py b/cv/detection/yolov5/pytorch/utils/loss.py index 932f8df15956046f322933bd3d8bbbb2627a6f70..370f7323e54e8cd01e8fd7cef9708fc192bce3c2 100644 --- a/cv/detection/yolov5/pytorch/utils/loss.py +++ b/cv/detection/yolov5/pytorch/utils/loss.py @@ -183,8 +183,9 @@ class ComputeLoss: ], device=targets.device).float() * g # offsets for i in range(self.nl): + # anchors = self.anchors[i] anchors, shape = self.anchors[i], p[i].shape - gain[2:6] = torch.tensor(shape)[[3, 2, 3, 2]] # xyxy gain + gain[2:6] = torch.tensor(p[i].shape)[[3, 2, 3, 2]] # xyxy gain # Match targets to anchors t = targets * gain @@ -216,7 +217,8 @@ class ComputeLoss: # Append a = t[:, 6].long() # anchor indices - indices.append((b, a, gj.clamp_(0, shape[2] - 1), gi.clamp_(0, shape[3] - 1))) # image, anchor, grid + # indices.append((b, a, gj.clamp_(0, gain[3] - 1), gi.clamp_(0, gain[2] - 1))) # image, anchor, grid indices + indices.append((b,a,gj.clamp_(0,shape[2] - 1),gi.clamp_(0,shape[3] - 1))) # image, anchor, grid indices tbox.append(torch.cat((gxy - gij, gwh), 1)) # box anch.append(anchors[a]) # anchors tcls.append(c) # class diff --git a/cv/detection/yolov5/pytorch/utils/wandb_logging/wandb_utils.py b/cv/detection/yolov5/pytorch/utils/wandb_logging/wandb_utils.py index f031a819b977c8a606a99a7b8f31fd86d4a2c110..45aa088fa5e05538ecbab2362457cb38d369f4cf 100644 --- a/cv/detection/yolov5/pytorch/utils/wandb_logging/wandb_utils.py +++ b/cv/detection/yolov5/pytorch/utils/wandb_logging/wandb_utils.py @@ -1,3 +1,6 @@ +# Copyright (c) 2024, Shanghai Iluvatar CoreX Semiconductor Co., Ltd. +# All Rights Reserved. + """Utilities and tools for tracking runs with Weights & Biases.""" import logging import os @@ -18,7 +21,7 @@ try: from wandb import init, finish except ImportError: wandb = None - +wandb = None RANK = int(os.getenv('RANK', -1)) WANDB_ARTIFACT_PREFIX = 'wandb-artifact://'