katielink commited on
Commit
3f86222
1 Parent(s): 392414b

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": "0.5.3",
4
  "changelog": {
 
5
  "0.5.3": "remove error dollar symbol in readme",
6
  "0.5.2": "remove the CheckpointLoader from the train.json",
7
  "0.5.1": "add RAM warning",
@@ -24,7 +25,7 @@
24
  "0.1.0": "complete the first version model package",
25
  "0.0.1": "initialize the model package structure"
26
  },
27
- "monai_version": "1.2.0rc6",
28
  "pytorch_version": "1.13.1",
29
  "numpy_version": "1.22.2",
30
  "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": "0.5.4",
4
  "changelog": {
5
+ "0.5.4": "fix the wrong GPU index issue of multi-node",
6
  "0.5.3": "remove error dollar symbol in readme",
7
  "0.5.2": "remove the CheckpointLoader from the train.json",
8
  "0.5.1": "add RAM warning",
 
25
  "0.1.0": "complete the first version model package",
26
  "0.0.1": "initialize the model package structure"
27
  },
28
+ "monai_version": "1.2.0",
29
  "pytorch_version": "1.13.1",
30
  "numpy_version": "1.22.2",
31
  "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)",