katielink commited on
Commit
5fbf6fb
1 Parent(s): fd8c5bc

fix the wrong GPU index issue of multi-node

Browse files
configs/metadata.json CHANGED
@@ -1,7 +1,8 @@
1
  {
2
  "schema": "https://github.com/Project-MONAI/MONAI-extra-test-data/releases/download/0.8.1/meta_schema_20220324.json",
3
- "version": "0.1.3",
4
  "changelog": {
 
5
  "0.1.3": "remove error dollar symbol in readme",
6
  "0.1.2": "add RAM usage with CachDataset",
7
  "0.1.1": "deterministic retrain benchmark and add link",
@@ -16,7 +17,7 @@
16
  "0.0.2": "Update The Torch Vision Transform",
17
  "0.0.1": "initialize the model package structure"
18
  },
19
- "monai_version": "1.2.0rc6",
20
  "pytorch_version": "1.13.1",
21
  "numpy_version": "1.22.2",
22
  "optional_packages_version": {
 
1
  {
2
  "schema": "https://github.com/Project-MONAI/MONAI-extra-test-data/releases/download/0.8.1/meta_schema_20220324.json",
3
+ "version": "0.1.4",
4
  "changelog": {
5
+ "0.1.4": "fix the wrong GPU index issue of multi-node",
6
  "0.1.3": "remove error dollar symbol in readme",
7
  "0.1.2": "add RAM usage with CachDataset",
8
  "0.1.1": "deterministic retrain benchmark and add link",
 
17
  "0.0.2": "Update The Torch Vision Transform",
18
  "0.0.1": "initialize the model package structure"
19
  },
20
+ "monai_version": "1.2.0",
21
  "pytorch_version": "1.13.1",
22
  "numpy_version": "1.22.2",
23
  "optional_packages_version": {
configs/multi_gpu_evaluate.json CHANGED
@@ -1,5 +1,5 @@
1
  {
2
- "device": "$torch.device(f'cuda:{dist.get_rank()}')",
3
  "network": {
4
  "_target_": "torch.nn.parallel.DistributedDataParallel",
5
  "module": "$@network_def.to(@device)",
 
1
  {
2
+ "device": "$torch.device('cuda:' + os.environ['LOCAL_RANK'])",
3
  "network": {
4
  "_target_": "torch.nn.parallel.DistributedDataParallel",
5
  "module": "$@network_def.to(@device)",
configs/multi_gpu_train.json CHANGED
@@ -1,5 +1,5 @@
1
  {
2
- "device": "$torch.device(f'cuda:{dist.get_rank()}')",
3
  "network": {
4
  "_target_": "torch.nn.parallel.DistributedDataParallel",
5
  "module": "$@network_def.to(@device)",
 
1
  {
2
+ "device": "$torch.device('cuda:' + os.environ['LOCAL_RANK'])",
3
  "network": {
4
  "_target_": "torch.nn.parallel.DistributedDataParallel",
5
  "module": "$@network_def.to(@device)",