selee
0 Followers · 0 Following
  • Joined on 2024-08-26
Repositories
5
 Projects Packages Public Activity Starred Repositories
selee pushed to main at selee/kubeflow-trainer-v2-tf-custom-... 2026-03-03 15:37:02 +09:00
5633426a5e Update README with training results, TFJob comparison, and full troubleshooting
selee pushed to main at selee/kubeflow-trainer-v2-tf-custom-... 2026-03-03 15:17:35 +09:00
a80f3bd97e Fix: all workers must call model.evaluate() together
selee pushed to main at selee/kubeflow-trainer-v2-tf-custom-... 2026-03-03 15:13:07 +09:00
1c915cab65 Rewrite README with latest findings and troubleshooting
selee pushed to main at selee/kubeflow-trainer-v2-tf-custom-... 2026-03-03 15:07:58 +09:00
9b1cd02a87 Fix Pod naming pattern and TF_CONFIG initialization order
selee pushed to main at selee/kubeflow-trainer-v2-tf-custom-... 2026-03-03 13:35:51 +09:00
a96f0fbc22 Remove hardcoded namespace from TrainJob metadata
selee pushed to main at selee/kubeflow-trainer-v2-tf-custom-... 2026-03-03 13:25:08 +09:00
9e92f2e3d4 Revert "Revert "Remove NUM_WORKERS env var, auto-detect workers via DNS probing""
selee pushed to main at selee/kubeflow-trainer-v2-tf-custom-... 2026-03-03 13:15:26 +09:00
5f6dec6799 Revert "Remove NUM_WORKERS env var, auto-detect workers via DNS probing"
selee pushed to main at selee/kubeflow-trainer-v2-tf-custom-... 2026-03-03 11:34:09 +09:00
0428780d9b Remove NUM_WORKERS env var, auto-detect workers via DNS probing
selee pushed to main at selee/kubeflow-trainer-v2-tf-custom-... 2026-03-03 10:24:23 +09:00
63fbcf8085 Add TensorFlow custom runtime for Kubeflow Trainer v2 with Volcano integration
selee created branch main in selee/kubeflow-trainer-v2-tf-custom-... 2026-03-03 10:24:23 +09:00
selee created repository selee/kubeflow-trainer-v2-tf-custom-... 2026-03-03 10:19:19 +09:00
selee pushed to main at selee/kubeflow_trainer_with_volcano 2026-02-10 16:40:50 +09:00
a5171437a2 Add CIFAR-10 dataset pre-placement requirement to README
selee pushed to main at selee/kubeflow_trainer_with_volcano 2026-02-10 16:38:35 +09:00
965b87ced3 Update README: torchrun usage guide, verified training results
selee pushed to main at selee/kubeflow_trainer_with_volcano 2026-02-10 16:36:45 +09:00
7cd7c1af7e Fix torchrun injection: add ancestor-step label, use torchrun command
selee pushed to main at selee/kubeflow_trainer_with_volcano 2026-02-10 16:26:32 +09:00
2b1ad10c2c Remove command from Runtime, move to TrainJob trainer.command
selee pushed to main at selee/kubeflow_trainer_with_volcano 2026-02-10 16:23:13 +09:00
00529c5ca9 Comment out JAX runtime, update README with install results
selee pushed to main at selee/kubeflow_trainer_with_volcano 2026-02-10 16:10:11 +09:00
146baa41e8 Add ConfigMap training script, update IB/NCCL config in README
selee pushed to main at selee/kubeflow_trainer_with_volcano 2026-02-10 15:57:06 +09:00
f62425843b Update integration config for InfiniBand GPU environment
selee pushed to main at selee/kubeflow_trainer_with_volcano 2026-02-10 15:44:51 +09:00
6b72590351 Add resource descriptions to README, set GPU to 8 per node
selee pushed to main at selee/kubeflow_trainer_with_volcano 2026-02-10 15:27:19 +09:00
6b413fe7d7 Update README: InfiniBand GPU 3-node config, simplify YAML examples