[WIP][RFC] Required changes for integration with TorchTitan #82
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
facebook-github-bot
added
the
CLA Signed
fegin
force-pushed
the
chienchin/torchtitan
branch
from
768d014 to
4e5c337
Compare
Summary: We are not going to land this PR, this PR may be further divided into several PRs. Test Plan: Reviewers: Subscribers: Tasks: Tags:
fegin
force-pushed
the
chienchin/torchtitan
branch
from
4e5c337 to
aab9239
Compare
fegin
added a commit
to pytorch/torchtitan
that referenced
this pull request
Jan 31, 2025
**Summary** This is a WIP TorchFT integration PR. **Current Issues** This doesn't work at this moment as there are hanged groups when a new group joins. **Issue 1:** ~Group 0 and group 1 will hang during the first `should_commit` after group 1 applying the pending state_dict from group 0.~ Fixed with: pytorch/torchft#83 **Issue 2:** ~Group 0 and group 1 will pass the `should_commit` but group 0 needs healing which is wrong and the healing process will cause another hang.~ Fixed with: pytorch/torchft#83 **Issue 3:** The byproduct of issue 1 and issue 2: group 1 will continue to print out ``` [rank0]:devgpu051:76838:80357 [0] misc/socket.cc:50 NCCL WARN socketProgress: Connection closed by remote peer devgpu051.cln3.svc.fbinfra.net<33618> ``` ***How to reproduce?*** Using the following the steps in `Reproduce steps` to run 2 groups. Then kill any of the group after both start training. Remember to apply pytorch/torchft#83. **Issue 4:** When there are 3 groups, everyone requests the state dict every step. ***How to reproduce?*** Using the `Reproduce steps` to run 2 groups, then add another group by modifying the command. **Reproduce steps:** 1. Patch TorchFT with pytorch/torchft#82 2. Execute lighthouse 3. Execute the following command in one terminal: ``` TORCHFT_MANAGER_PORT=29520 REPLICA_GROUP_ID=0 CUDA_VISIBLE_DEVICES=0,1 NGPU=2 ./run_llama_train.sh --training.data_parallel_shard_degree=2 --experimental.enable_torchft --experimental.ft_replica_group_id=0 ``` 4. Wait 10 seconds, execute following command in another terminal: ``` TORCHFT_MANAGER_PORT=29522 REPLICA_GROUP_ID=1 CUDA_VISIBLE_DEVICES=2,3 NGPU=2 ./run_llama_train.sh --training.data_parallel_shard_degree=2 --experimental.enable_torchft --experimental.ft_replica_group_id=1 ``` [ghstack-poisoned]
fegin
added a commit
to pytorch/torchtitan
that referenced
this pull request
Jan 31, 2025
**Summary** This is a WIP TorchFT integration PR. **Current Issues** This doesn't work at this moment as there are hanged groups when a new group joins. **Issue 1:** ~Group 0 and group 1 will hang during the first `should_commit` after group 1 applying the pending state_dict from group 0.~ Fixed with: pytorch/torchft#83 **Issue 2:** ~Group 0 and group 1 will pass the `should_commit` but group 0 needs healing which is wrong and the healing process will cause another hang.~ Fixed with: pytorch/torchft#83 **Issue 3:** The byproduct of issue 1 and issue 2: group 1 will continue to print out ``` [rank0]:devgpu051:76838:80357 [0] misc/socket.cc:50 NCCL WARN socketProgress: Connection closed by remote peer devgpu051.cln3.svc.fbinfra.net<33618> ``` ***How to reproduce?*** Using the following the steps in `Reproduce steps` to run 2 groups. Then kill any of the group after both start training. Remember to apply pytorch/torchft#83. **Issue 4:** When there are 3 groups, everyone requests the state dict every step. ***How to reproduce?*** Using the `Reproduce steps` to run 2 groups, then add another group by modifying the command. **Reproduce steps:** 1. Patch TorchFT with pytorch/torchft#82 2. Execute lighthouse 3. Execute the following command in one terminal: ``` TORCHFT_MANAGER_PORT=29520 REPLICA_GROUP_ID=0 CUDA_VISIBLE_DEVICES=0,1 NGPU=2 ./run_llama_train.sh --training.data_parallel_shard_degree=2 --experimental.enable_torchft --experimental.ft_replica_group_id=0 ``` 4. Wait 10 seconds, execute following command in another terminal: ``` TORCHFT_MANAGER_PORT=29522 REPLICA_GROUP_ID=1 CUDA_VISIBLE_DEVICES=2,3 NGPU=2 ./run_llama_train.sh --training.data_parallel_shard_degree=2 --experimental.enable_torchft --experimental.ft_replica_group_id=1 ``` [ghstack-poisoned]
fegin
added a commit
to pytorch/torchtitan
that referenced
this pull request
Feb 3, 2025
**Summary** This is a WIP TorchFT integration PR. **Current Issues** This doesn't work at this moment as there are hanged groups when a new group joins. **Issue 1:** ~Group 0 and group 1 will hang during the first `should_commit` after group 1 applying the pending state_dict from group 0.~ Fixed with: pytorch/torchft#83 **Issue 2:** ~Group 0 and group 1 will pass the `should_commit` but group 0 needs healing which is wrong and the healing process will cause another hang.~ Fixed with: pytorch/torchft#83 **Issue 3:** ~The byproduct of issue 1 and issue 2: group 1 will continue to print out~ ``` [rank0]:devgpu051:76838:80357 [0] misc/socket.cc:50 NCCL WARN socketProgress: Connection closed by remote peer devgpu051.cln3.svc.fbinfra.net<33618> ``` Fixed with pytorch/torchft#91 and several other fixes. **Issue 4:** When there are 3 groups, everyone requests the state dict every step. ***How to reproduce?*** Using the `Reproduce steps` to run 2 groups, then add another group by modifying the command. Seems to be fixed, will need more tests. **Issue 5:** Hang will happen if using functional collective. ***How to reproduce?*** Pull the latest version of this PR and comment out line 41 and uncomment line 42 in `torchtitan/utils.py` **Reproduce steps:** 1. Patch TorchFT with pytorch/torchft#82 2. Execute lighthouse 3. Execute the following command in one terminal: ``` TORCHFT_MANAGER_PORT=29520 REPLICA_GROUP_ID=0 CUDA_VISIBLE_DEVICES=0,1 NGPU=2 ./run_llama_train.sh --training.data_parallel_shard_degree=2 --experimental.enable_torchft --experimental.ft_replica_group_id=0 ``` 4. Wait 10 seconds, execute following command in another terminal: ``` TORCHFT_MANAGER_PORT=29522 REPLICA_GROUP_ID=1 CUDA_VISIBLE_DEVICES=2,3 NGPU=2 ./run_llama_train.sh --training.data_parallel_shard_degree=2 --experimental.enable_torchft --experimental.ft_replica_group_id=1 ``` [ghstack-poisoned]
fegin
added a commit
to pytorch/torchtitan
that referenced
this pull request
Feb 3, 2025
**Summary** This is a WIP TorchFT integration PR. **Current Issues** This doesn't work at this moment as there are hanged groups when a new group joins. **Issue 1:** ~Group 0 and group 1 will hang during the first `should_commit` after group 1 applying the pending state_dict from group 0.~ Fixed with: pytorch/torchft#83 **Issue 2:** ~Group 0 and group 1 will pass the `should_commit` but group 0 needs healing which is wrong and the healing process will cause another hang.~ Fixed with: pytorch/torchft#83 **Issue 3:** ~The byproduct of issue 1 and issue 2: group 1 will continue to print out~ ``` [rank0]:devgpu051:76838:80357 [0] misc/socket.cc:50 NCCL WARN socketProgress: Connection closed by remote peer devgpu051.cln3.svc.fbinfra.net<33618> ``` Fixed with pytorch/torchft#91 and several other fixes. **Issue 4:** When there are 3 groups, everyone requests the state dict every step. ***How to reproduce?*** Using the `Reproduce steps` to run 2 groups, then add another group by modifying the command. Seems to be fixed, will need more tests. **Issue 5:** Hang will happen if using functional collective. ***How to reproduce?*** Pull the latest version of this PR and comment out line 41 and uncomment line 42 in `torchtitan/utils.py` **Reproduce steps:** 1. Patch TorchFT with pytorch/torchft#82 2. Execute lighthouse 3. Execute the following command in one terminal: ``` TORCHFT_MANAGER_PORT=29520 REPLICA_GROUP_ID=0 CUDA_VISIBLE_DEVICES=0,1 NGPU=2 ./run_llama_train.sh --training.data_parallel_shard_degree=2 --experimental.enable_torchft --experimental.ft_replica_group_id=0 ``` 4. Wait 10 seconds, execute following command in another terminal: ``` TORCHFT_MANAGER_PORT=29522 REPLICA_GROUP_ID=1 CUDA_VISIBLE_DEVICES=2,3 NGPU=2 ./run_llama_train.sh --training.data_parallel_shard_degree=2 --experimental.enable_torchft --experimental.ft_replica_group_id=1 ``` [ghstack-poisoned]
fegin
added a commit
to pytorch/torchtitan
that referenced
this pull request
Feb 11, 2025
**Summary** This is a WIP TorchFT integration PR. **Current Issues** This doesn't work at this moment as there are hanged groups when a new group joins. **Issue 1:** ~Group 0 and group 1 will hang during the first `should_commit` after group 1 applying the pending state_dict from group 0.~ Fixed with: pytorch/torchft#83 **Issue 2:** ~Group 0 and group 1 will pass the `should_commit` but group 0 needs healing which is wrong and the healing process will cause another hang.~ Fixed with: pytorch/torchft#83 **Issue 3:** ~The byproduct of issue 1 and issue 2: group 1 will continue to print out~ ``` [rank0]:devgpu051:76838:80357 [0] misc/socket.cc:50 NCCL WARN socketProgress: Connection closed by remote peer devgpu051.cln3.svc.fbinfra.net<33618> ``` Fixed with pytorch/torchft#91 and several other fixes. **Issue 4:** When there are 3 groups, everyone requests the state dict every step. ***How to reproduce?*** Using the `Reproduce steps` to run 2 groups, then add another group by modifying the command. Seems to be fixed, will need more tests. **Issue 5:** Hang will happen if using functional collective. ***How to reproduce?*** Pull the latest version of this PR and comment out line 41 and uncomment line 42 in `torchtitan/utils.py` **Reproduce steps:** 1. Patch TorchFT with pytorch/torchft#82 2. Execute lighthouse 3. Execute the following command in one terminal: ``` TORCHFT_MANAGER_PORT=29520 REPLICA_GROUP_ID=0 CUDA_VISIBLE_DEVICES=0,1 NGPU=2 ./run_llama_train.sh --training.data_parallel_shard_degree=2 --experimental.enable_torchft --experimental.ft_replica_group_id=0 ``` 4. Wait 10 seconds, execute following command in another terminal: ``` TORCHFT_MANAGER_PORT=29522 REPLICA_GROUP_ID=1 CUDA_VISIBLE_DEVICES=2,3 NGPU=2 ./run_llama_train.sh --training.data_parallel_shard_degree=2 --experimental.enable_torchft --experimental.ft_replica_group_id=1 ``` ghstack-source-id: bf6f0c51100cf2a1c2fb25a405b6c7592694b323 Pull Request resolved: #834
fegin
added a commit
to pytorch/torchtitan
that referenced
this pull request
Feb 12, 2025
**Summary** This is a WIP TorchFT integration PR. **Current Issues** This doesn't work at this moment as there are hanged groups when a new group joins. **Issue 1:** ~Group 0 and group 1 will hang during the first `should_commit` after group 1 applying the pending state_dict from group 0.~ Fixed with: pytorch/torchft#83 **Issue 2:** ~Group 0 and group 1 will pass the `should_commit` but group 0 needs healing which is wrong and the healing process will cause another hang.~ Fixed with: pytorch/torchft#83 **Issue 3:** ~The byproduct of issue 1 and issue 2: group 1 will continue to print out~ ``` [rank0]:devgpu051:76838:80357 [0] misc/socket.cc:50 NCCL WARN socketProgress: Connection closed by remote peer devgpu051.cln3.svc.fbinfra.net<33618> ``` Fixed with pytorch/torchft#91 and several other fixes. **Issue 4:** When there are 3 groups, everyone requests the state dict every step. ***How to reproduce?*** Using the `Reproduce steps` to run 2 groups, then add another group by modifying the command. Seems to be fixed, will need more tests. **Issue 5:** Hang will happen if using functional collective. ***How to reproduce?*** Pull the latest version of this PR and comment out line 41 and uncomment line 42 in `torchtitan/utils.py` **Reproduce steps:** 1. Patch TorchFT with pytorch/torchft#82 2. Execute lighthouse 3. Execute the following command in one terminal: ``` TORCHFT_MANAGER_PORT=29520 REPLICA_GROUP_ID=0 CUDA_VISIBLE_DEVICES=0,1 NGPU=2 ./run_llama_train.sh --training.data_parallel_shard_degree=2 --experimental.enable_torchft --experimental.ft_replica_group_id=0 ``` 4. Wait 10 seconds, execute following command in another terminal: ``` TORCHFT_MANAGER_PORT=29522 REPLICA_GROUP_ID=1 CUDA_VISIBLE_DEVICES=2,3 NGPU=2 ./run_llama_train.sh --training.data_parallel_shard_degree=2 --experimental.enable_torchft --experimental.ft_replica_group_id=1 ``` ghstack-source-id: 40d49640772abac268fa97147d889aa652559116 Pull Request resolved: #834
fegin
added a commit
to pytorch/torchtitan
that referenced
this pull request
Feb 12, 2025
**Summary** This is a WIP TorchFT integration PR. **Current Issues** This doesn't work at this moment as there are hanged groups when a new group joins. **Issue 1:** ~Group 0 and group 1 will hang during the first `should_commit` after group 1 applying the pending state_dict from group 0.~ Fixed with: pytorch/torchft#83 **Issue 2:** ~Group 0 and group 1 will pass the `should_commit` but group 0 needs healing which is wrong and the healing process will cause another hang.~ Fixed with: pytorch/torchft#83 **Issue 3:** ~The byproduct of issue 1 and issue 2: group 1 will continue to print out~ ``` [rank0]:devgpu051:76838:80357 [0] misc/socket.cc:50 NCCL WARN socketProgress: Connection closed by remote peer devgpu051.cln3.svc.fbinfra.net<33618> ``` Fixed with pytorch/torchft#91 and several other fixes. **Issue 4:** When there are 3 groups, everyone requests the state dict every step. ***How to reproduce?*** Using the `Reproduce steps` to run 2 groups, then add another group by modifying the command. Seems to be fixed, will need more tests. **Issue 5:** Hang will happen if using functional collective. ***How to reproduce?*** Pull the latest version of this PR and comment out line 41 and uncomment line 42 in `torchtitan/utils.py` **Reproduce steps:** 1. Patch TorchFT with pytorch/torchft#82 2. Execute lighthouse 3. Execute the following command in one terminal: ``` TORCHFT_MANAGER_PORT=29520 REPLICA_GROUP_ID=0 CUDA_VISIBLE_DEVICES=0,1 NGPU=2 ./run_llama_train.sh --training.data_parallel_shard_degree=2 --experimental.enable_torchft --experimental.ft_replica_group_id=0 ``` 4. Wait 10 seconds, execute following command in another terminal: ``` TORCHFT_MANAGER_PORT=29522 REPLICA_GROUP_ID=1 CUDA_VISIBLE_DEVICES=2,3 NGPU=2 ./run_llama_train.sh --training.data_parallel_shard_degree=2 --experimental.enable_torchft --experimental.ft_replica_group_id=1 ``` ghstack-source-id: 4e04a58aefdda31329fe06343f1e020d6f4a90c0 Pull Request resolved: #834
fegin
added a commit
to pytorch/torchtitan
that referenced
this pull request
Feb 12, 2025
**Summary** This is a WIP TorchFT integration PR. **Current Issues** This doesn't work at this moment as there are hanged groups when a new group joins. **Issue 1:** ~Group 0 and group 1 will hang during the first `should_commit` after group 1 applying the pending state_dict from group 0.~ Fixed with: pytorch/torchft#83 **Issue 2:** ~Group 0 and group 1 will pass the `should_commit` but group 0 needs healing which is wrong and the healing process will cause another hang.~ Fixed with: pytorch/torchft#83 **Issue 3:** ~The byproduct of issue 1 and issue 2: group 1 will continue to print out~ ``` [rank0]:devgpu051:76838:80357 [0] misc/socket.cc:50 NCCL WARN socketProgress: Connection closed by remote peer devgpu051.cln3.svc.fbinfra.net<33618> ``` Fixed with pytorch/torchft#91 and several other fixes. **Issue 4:** When there are 3 groups, everyone requests the state dict every step. ***How to reproduce?*** Using the `Reproduce steps` to run 2 groups, then add another group by modifying the command. Seems to be fixed, will need more tests. **Issue 5:** Hang will happen if using functional collective. ***How to reproduce?*** Pull the latest version of this PR and comment out line 41 and uncomment line 42 in `torchtitan/utils.py` **Reproduce steps:** 1. Patch TorchFT with pytorch/torchft#82 2. Execute lighthouse 3. Execute the following command in one terminal: ``` TORCHFT_MANAGER_PORT=29520 REPLICA_GROUP_ID=0 CUDA_VISIBLE_DEVICES=0,1 NGPU=2 ./run_llama_train.sh --training.data_parallel_shard_degree=2 --experimental.enable_torchft --experimental.ft_replica_group_id=0 ``` 4. Wait 10 seconds, execute following command in another terminal: ``` TORCHFT_MANAGER_PORT=29522 REPLICA_GROUP_ID=1 CUDA_VISIBLE_DEVICES=2,3 NGPU=2 ./run_llama_train.sh --training.data_parallel_shard_degree=2 --experimental.enable_torchft --experimental.ft_replica_group_id=1 ``` ghstack-source-id: bf6f0c51100cf2a1c2fb25a405b6c7592694b323 Pull Request resolved: #834
fegin
added a commit
to pytorch/torchtitan
that referenced
this pull request
Feb 12, 2025
**Summary** This is a WIP TorchFT integration PR. **Current Issues** This doesn't work at this moment as there are hanged groups when a new group joins. **Issue 1:** ~Group 0 and group 1 will hang during the first `should_commit` after group 1 applying the pending state_dict from group 0.~ Fixed with: pytorch/torchft#83 **Issue 2:** ~Group 0 and group 1 will pass the `should_commit` but group 0 needs healing which is wrong and the healing process will cause another hang.~ Fixed with: pytorch/torchft#83 **Issue 3:** ~The byproduct of issue 1 and issue 2: group 1 will continue to print out~ ``` [rank0]:devgpu051:76838:80357 [0] misc/socket.cc:50 NCCL WARN socketProgress: Connection closed by remote peer devgpu051.cln3.svc.fbinfra.net<33618> ``` Fixed with pytorch/torchft#91 and several other fixes. **Issue 4:** When there are 3 groups, everyone requests the state dict every step. ***How to reproduce?*** Using the `Reproduce steps` to run 2 groups, then add another group by modifying the command. Seems to be fixed, will need more tests. **Issue 5:** Hang will happen if using functional collective. ***How to reproduce?*** Pull the latest version of this PR and comment out line 41 and uncomment line 42 in `torchtitan/utils.py` **Reproduce steps:** 1. Patch TorchFT with pytorch/torchft#82 2. Execute lighthouse 3. Execute the following command in one terminal: ``` TORCHFT_MANAGER_PORT=29520 REPLICA_GROUP_ID=0 CUDA_VISIBLE_DEVICES=0,1 NGPU=2 ./run_llama_train.sh --training.data_parallel_shard_degree=2 --experimental.enable_torchft --experimental.ft_replica_group_id=0 ``` 4. Wait 10 seconds, execute following command in another terminal: ``` TORCHFT_MANAGER_PORT=29522 REPLICA_GROUP_ID=1 CUDA_VISIBLE_DEVICES=2,3 NGPU=2 ./run_llama_train.sh --training.data_parallel_shard_degree=2 --experimental.enable_torchft --experimental.ft_replica_group_id=1 ``` ghstack-source-id: 3690680f53f5a210f96a9f1ed70f5e5d03cb14a1 Pull Request resolved: #834
fegin
added a commit
to pytorch/torchtitan
that referenced
this pull request
Feb 12, 2025
**Summary** This is a WIP TorchFT integration PR. **Current Issues** This doesn't work at this moment as there are hanged groups when a new group joins. **Issue 1:** ~Group 0 and group 1 will hang during the first `should_commit` after group 1 applying the pending state_dict from group 0.~ Fixed with: pytorch/torchft#83 **Issue 2:** ~Group 0 and group 1 will pass the `should_commit` but group 0 needs healing which is wrong and the healing process will cause another hang.~ Fixed with: pytorch/torchft#83 **Issue 3:** ~The byproduct of issue 1 and issue 2: group 1 will continue to print out~ ``` [rank0]:devgpu051:76838:80357 [0] misc/socket.cc:50 NCCL WARN socketProgress: Connection closed by remote peer devgpu051.cln3.svc.fbinfra.net<33618> ``` Fixed with pytorch/torchft#91 and several other fixes. **Issue 4:** When there are 3 groups, everyone requests the state dict every step. ***How to reproduce?*** Using the `Reproduce steps` to run 2 groups, then add another group by modifying the command. Seems to be fixed, will need more tests. **Issue 5:** Hang will happen if using functional collective. ***How to reproduce?*** Pull the latest version of this PR and comment out line 41 and uncomment line 42 in `torchtitan/utils.py` **Reproduce steps:** 1. Patch TorchFT with pytorch/torchft#82 2. Execute lighthouse 3. Execute the following command in one terminal: ``` TORCHFT_MANAGER_PORT=29520 REPLICA_GROUP_ID=0 CUDA_VISIBLE_DEVICES=0,1 NGPU=2 ./run_llama_train.sh --training.data_parallel_shard_degree=2 --experimental.enable_torchft --experimental.ft_replica_group_id=0 ``` 4. Wait 10 seconds, execute following command in another terminal: ``` TORCHFT_MANAGER_PORT=29522 REPLICA_GROUP_ID=1 CUDA_VISIBLE_DEVICES=2,3 NGPU=2 ./run_llama_train.sh --training.data_parallel_shard_degree=2 --experimental.enable_torchft --experimental.ft_replica_group_id=1 ``` ghstack-source-id: 088581cceee2c523f2a4ea358f334a0b1cce3927 Pull Request resolved: #834
fegin
added a commit
to pytorch/torchtitan
that referenced
this pull request
Feb 12, 2025
**Summary** This is a WIP TorchFT integration PR. **Current Issues** This doesn't work at this moment as there are hanged groups when a new group joins. **Issue 1:** ~Group 0 and group 1 will hang during the first `should_commit` after group 1 applying the pending state_dict from group 0.~ Fixed with: pytorch/torchft#83 **Issue 2:** ~Group 0 and group 1 will pass the `should_commit` but group 0 needs healing which is wrong and the healing process will cause another hang.~ Fixed with: pytorch/torchft#83 **Issue 3:** ~The byproduct of issue 1 and issue 2: group 1 will continue to print out~ ``` [rank0]:devgpu051:76838:80357 [0] misc/socket.cc:50 NCCL WARN socketProgress: Connection closed by remote peer devgpu051.cln3.svc.fbinfra.net<33618> ``` Fixed with pytorch/torchft#91 and several other fixes. **Issue 4:** When there are 3 groups, everyone requests the state dict every step. ***How to reproduce?*** Using the `Reproduce steps` to run 2 groups, then add another group by modifying the command. Seems to be fixed, will need more tests. **Issue 5:** Hang will happen if using functional collective. ***How to reproduce?*** Pull the latest version of this PR and comment out line 41 and uncomment line 42 in `torchtitan/utils.py` **Reproduce steps:** 1. Patch TorchFT with pytorch/torchft#82 2. Execute lighthouse 3. Execute the following command in one terminal: ``` TORCHFT_MANAGER_PORT=29520 REPLICA_GROUP_ID=0 CUDA_VISIBLE_DEVICES=0,1 NGPU=2 ./run_llama_train.sh --training.data_parallel_shard_degree=2 --experimental.enable_torchft --experimental.ft_replica_group_id=0 ``` 4. Wait 10 seconds, execute following command in another terminal: ``` TORCHFT_MANAGER_PORT=29522 REPLICA_GROUP_ID=1 CUDA_VISIBLE_DEVICES=2,3 NGPU=2 ./run_llama_train.sh --training.data_parallel_shard_degree=2 --experimental.enable_torchft --experimental.ft_replica_group_id=1 ``` ghstack-source-id: 7f44395ee179185ee7fcd132d2191526b66a6dfb Pull Request resolved: #834
fegin
added a commit
to pytorch/torchtitan
that referenced
this pull request
Feb 12, 2025
**Summary** This is a WIP TorchFT integration PR. **Current Issues** This doesn't work at this moment as there are hanged groups when a new group joins. **Issue 1:** ~Group 0 and group 1 will hang during the first `should_commit` after group 1 applying the pending state_dict from group 0.~ Fixed with: pytorch/torchft#83 **Issue 2:** ~Group 0 and group 1 will pass the `should_commit` but group 0 needs healing which is wrong and the healing process will cause another hang.~ Fixed with: pytorch/torchft#83 **Issue 3:** ~The byproduct of issue 1 and issue 2: group 1 will continue to print out~ ``` [rank0]:devgpu051:76838:80357 [0] misc/socket.cc:50 NCCL WARN socketProgress: Connection closed by remote peer devgpu051.cln3.svc.fbinfra.net<33618> ``` Fixed with pytorch/torchft#91 and several other fixes. **Issue 4:** When there are 3 groups, everyone requests the state dict every step. ***How to reproduce?*** Using the `Reproduce steps` to run 2 groups, then add another group by modifying the command. Seems to be fixed, will need more tests. **Issue 5:** Hang will happen if using functional collective. ***How to reproduce?*** Pull the latest version of this PR and comment out line 41 and uncomment line 42 in `torchtitan/utils.py` **Reproduce steps:** 1. Patch TorchFT with pytorch/torchft#82 2. Execute lighthouse 3. Execute the following command in one terminal: ``` TORCHFT_MANAGER_PORT=29520 REPLICA_GROUP_ID=0 CUDA_VISIBLE_DEVICES=0,1 NGPU=2 ./run_llama_train.sh --training.data_parallel_shard_degree=2 --experimental.enable_torchft --experimental.ft_replica_group_id=0 ``` 4. Wait 10 seconds, execute following command in another terminal: ``` TORCHFT_MANAGER_PORT=29522 REPLICA_GROUP_ID=1 CUDA_VISIBLE_DEVICES=2,3 NGPU=2 ./run_llama_train.sh --training.data_parallel_shard_degree=2 --experimental.enable_torchft --experimental.ft_replica_group_id=1 ``` ghstack-source-id: a5168a64719b6dc29df8f138dc8ac3cb7da7c69d Pull Request resolved: #834
fegin
added a commit
to pytorch/torchtitan
that referenced
this pull request
Feb 12, 2025
**Summary** This is a WIP TorchFT integration PR. **Current Issues** This doesn't work at this moment as there are hanged groups when a new group joins. **Issue 1:** ~Group 0 and group 1 will hang during the first `should_commit` after group 1 applying the pending state_dict from group 0.~ Fixed with: pytorch/torchft#83 **Issue 2:** ~Group 0 and group 1 will pass the `should_commit` but group 0 needs healing which is wrong and the healing process will cause another hang.~ Fixed with: pytorch/torchft#83 **Issue 3:** ~The byproduct of issue 1 and issue 2: group 1 will continue to print out~ ``` [rank0]:devgpu051:76838:80357 [0] misc/socket.cc:50 NCCL WARN socketProgress: Connection closed by remote peer devgpu051.cln3.svc.fbinfra.net<33618> ``` Fixed with pytorch/torchft#91 and several other fixes. **Issue 4:** When there are 3 groups, everyone requests the state dict every step. ***How to reproduce?*** Using the `Reproduce steps` to run 2 groups, then add another group by modifying the command. Seems to be fixed, will need more tests. **Issue 5:** Hang will happen if using functional collective. ***How to reproduce?*** Pull the latest version of this PR and comment out line 41 and uncomment line 42 in `torchtitan/utils.py` **Reproduce steps:** 1. Patch TorchFT with pytorch/torchft#82 2. Execute lighthouse 3. Execute the following command in one terminal: ``` TORCHFT_MANAGER_PORT=29520 REPLICA_GROUP_ID=0 CUDA_VISIBLE_DEVICES=0,1 NGPU=2 ./run_llama_train.sh --training.data_parallel_shard_degree=2 --experimental.enable_torchft --experimental.ft_replica_group_id=0 ``` 4. Wait 10 seconds, execute following command in another terminal: ``` TORCHFT_MANAGER_PORT=29522 REPLICA_GROUP_ID=1 CUDA_VISIBLE_DEVICES=2,3 NGPU=2 ./run_llama_train.sh --training.data_parallel_shard_degree=2 --experimental.enable_torchft --experimental.ft_replica_group_id=1 ``` ghstack-source-id: 82766ceaa3874edc3de14528db302b45eb705884 Pull Request resolved: #834
fegin
added a commit
to pytorch/torchtitan
that referenced
this pull request
Feb 12, 2025
**Summary** This is a WIP TorchFT integration PR. **Current Issues** This doesn't work at this moment as there are hanged groups when a new group joins. **Issue 1:** ~Group 0 and group 1 will hang during the first `should_commit` after group 1 applying the pending state_dict from group 0.~ Fixed with: pytorch/torchft#83 **Issue 2:** ~Group 0 and group 1 will pass the `should_commit` but group 0 needs healing which is wrong and the healing process will cause another hang.~ Fixed with: pytorch/torchft#83 **Issue 3:** ~The byproduct of issue 1 and issue 2: group 1 will continue to print out~ ``` [rank0]:devgpu051:76838:80357 [0] misc/socket.cc:50 NCCL WARN socketProgress: Connection closed by remote peer devgpu051.cln3.svc.fbinfra.net<33618> ``` Fixed with pytorch/torchft#91 and several other fixes. **Issue 4:** When there are 3 groups, everyone requests the state dict every step. ***How to reproduce?*** Using the `Reproduce steps` to run 2 groups, then add another group by modifying the command. Seems to be fixed, will need more tests. **Issue 5:** Hang will happen if using functional collective. ***How to reproduce?*** Pull the latest version of this PR and comment out line 41 and uncomment line 42 in `torchtitan/utils.py` **Reproduce steps:** 1. Patch TorchFT with pytorch/torchft#82 2. Execute lighthouse 3. Execute the following command in one terminal: ``` TORCHFT_MANAGER_PORT=29520 REPLICA_GROUP_ID=0 CUDA_VISIBLE_DEVICES=0,1 NGPU=2 ./run_llama_train.sh --training.data_parallel_shard_degree=2 --experimental.enable_torchft --experimental.ft_replica_group_id=0 ``` 4. Wait 10 seconds, execute following command in another terminal: ``` TORCHFT_MANAGER_PORT=29522 REPLICA_GROUP_ID=1 CUDA_VISIBLE_DEVICES=2,3 NGPU=2 ./run_llama_train.sh --training.data_parallel_shard_degree=2 --experimental.enable_torchft --experimental.ft_replica_group_id=1 ``` ghstack-source-id: c90068ba4f5d937f31596de91c5b08416a48b7d3 Pull Request resolved: #834
fegin
added a commit
to pytorch/torchtitan
that referenced
this pull request
Feb 12, 2025
**Summary** This is a WIP TorchFT integration PR. **Current Issues** This doesn't work at this moment as there are hanged groups when a new group joins. **Issue 1:** ~Group 0 and group 1 will hang during the first `should_commit` after group 1 applying the pending state_dict from group 0.~ Fixed with: pytorch/torchft#83 **Issue 2:** ~Group 0 and group 1 will pass the `should_commit` but group 0 needs healing which is wrong and the healing process will cause another hang.~ Fixed with: pytorch/torchft#83 **Issue 3:** ~The byproduct of issue 1 and issue 2: group 1 will continue to print out~ ``` [rank0]:devgpu051:76838:80357 [0] misc/socket.cc:50 NCCL WARN socketProgress: Connection closed by remote peer devgpu051.cln3.svc.fbinfra.net<33618> ``` Fixed with pytorch/torchft#91 and several other fixes. **Issue 4:** When there are 3 groups, everyone requests the state dict every step. ***How to reproduce?*** Using the `Reproduce steps` to run 2 groups, then add another group by modifying the command. Seems to be fixed, will need more tests. **Issue 5:** Hang will happen if using functional collective. ***How to reproduce?*** Pull the latest version of this PR and comment out line 41 and uncomment line 42 in `torchtitan/utils.py` **Reproduce steps:** 1. Patch TorchFT with pytorch/torchft#82 2. Execute lighthouse 3. Execute the following command in one terminal: ``` TORCHFT_MANAGER_PORT=29520 REPLICA_GROUP_ID=0 CUDA_VISIBLE_DEVICES=0,1 NGPU=2 ./run_llama_train.sh --training.data_parallel_shard_degree=2 --experimental.enable_torchft --experimental.ft_replica_group_id=0 ``` 4. Wait 10 seconds, execute following command in another terminal: ``` TORCHFT_MANAGER_PORT=29522 REPLICA_GROUP_ID=1 CUDA_VISIBLE_DEVICES=2,3 NGPU=2 ./run_llama_train.sh --training.data_parallel_shard_degree=2 --experimental.enable_torchft --experimental.ft_replica_group_id=1 ``` ghstack-source-id: 19016cc820b1c2620632130f8e5897d0a7b3bdd6 Pull Request resolved: #834
fegin
added a commit
to pytorch/torchtitan
that referenced
this pull request
Feb 12, 2025
**Summary** This is a WIP TorchFT integration PR. **Current Issues** This doesn't work at this moment as there are hanged groups when a new group joins. **Issue 1:** ~Group 0 and group 1 will hang during the first `should_commit` after group 1 applying the pending state_dict from group 0.~ Fixed with: pytorch/torchft#83 **Issue 2:** ~Group 0 and group 1 will pass the `should_commit` but group 0 needs healing which is wrong and the healing process will cause another hang.~ Fixed with: pytorch/torchft#83 **Issue 3:** ~The byproduct of issue 1 and issue 2: group 1 will continue to print out~ ``` [rank0]:devgpu051:76838:80357 [0] misc/socket.cc:50 NCCL WARN socketProgress: Connection closed by remote peer devgpu051.cln3.svc.fbinfra.net<33618> ``` Fixed with pytorch/torchft#91 and several other fixes. **Issue 4:** When there are 3 groups, everyone requests the state dict every step. ***How to reproduce?*** Using the `Reproduce steps` to run 2 groups, then add another group by modifying the command. Seems to be fixed, will need more tests. **Issue 5:** Hang will happen if using functional collective. ***How to reproduce?*** Pull the latest version of this PR and comment out line 41 and uncomment line 42 in `torchtitan/utils.py` **Reproduce steps:** 1. Patch TorchFT with pytorch/torchft#82 2. Execute lighthouse 3. Execute the following command in one terminal: ``` TORCHFT_MANAGER_PORT=29520 REPLICA_GROUP_ID=0 CUDA_VISIBLE_DEVICES=0,1 NGPU=2 ./run_llama_train.sh --training.data_parallel_shard_degree=2 --experimental.enable_torchft --experimental.ft_replica_group_id=0 ``` 4. Wait 10 seconds, execute following command in another terminal: ``` TORCHFT_MANAGER_PORT=29522 REPLICA_GROUP_ID=1 CUDA_VISIBLE_DEVICES=2,3 NGPU=2 ./run_llama_train.sh --training.data_parallel_shard_degree=2 --experimental.enable_torchft --experimental.ft_replica_group_id=1 ``` ghstack-source-id: 248182a2375e50362ae83add2126048c9ef5d24d Pull Request resolved: #834
fegin
added a commit
to pytorch/torchtitan
that referenced
this pull request
Feb 13, 2025
**Summary** This is a WIP TorchFT integration PR. **Current Issues** This doesn't work at this moment as there are hanged groups when a new group joins. **Issue 1:** ~Group 0 and group 1 will hang during the first `should_commit` after group 1 applying the pending state_dict from group 0.~ Fixed with: pytorch/torchft#83 **Issue 2:** ~Group 0 and group 1 will pass the `should_commit` but group 0 needs healing which is wrong and the healing process will cause another hang.~ Fixed with: pytorch/torchft#83 **Issue 3:** ~The byproduct of issue 1 and issue 2: group 1 will continue to print out~ ``` [rank0]:devgpu051:76838:80357 [0] misc/socket.cc:50 NCCL WARN socketProgress: Connection closed by remote peer devgpu051.cln3.svc.fbinfra.net<33618> ``` Fixed with pytorch/torchft#91 and several other fixes. **Issue 4:** When there are 3 groups, everyone requests the state dict every step. ***How to reproduce?*** Using the `Reproduce steps` to run 2 groups, then add another group by modifying the command. Seems to be fixed, will need more tests. **Issue 5:** Hang will happen if using functional collective. ***How to reproduce?*** Pull the latest version of this PR and comment out line 41 and uncomment line 42 in `torchtitan/utils.py` **Reproduce steps:** 1. Patch TorchFT with pytorch/torchft#82 2. Execute lighthouse 3. Execute the following command in one terminal: ``` TORCHFT_MANAGER_PORT=29520 REPLICA_GROUP_ID=0 CUDA_VISIBLE_DEVICES=0,1 NGPU=2 ./run_llama_train.sh --training.data_parallel_shard_degree=2 --experimental.enable_torchft --experimental.ft_replica_group_id=0 ``` 4. Wait 10 seconds, execute following command in another terminal: ``` TORCHFT_MANAGER_PORT=29522 REPLICA_GROUP_ID=1 CUDA_VISIBLE_DEVICES=2,3 NGPU=2 ./run_llama_train.sh --training.data_parallel_shard_degree=2 --experimental.enable_torchft --experimental.ft_replica_group_id=1 ``` ghstack-source-id: 65c0469403170606b755b415df0439e020fa3187 Pull Request resolved: #834
fegin
added a commit
to pytorch/torchtitan
that referenced
this pull request
Feb 13, 2025
**Summary** This is a WIP TorchFT integration PR. **Current Issues** This doesn't work at this moment as there are hanged groups when a new group joins. **Issue 1:** ~Group 0 and group 1 will hang during the first `should_commit` after group 1 applying the pending state_dict from group 0.~ Fixed with: pytorch/torchft#83 **Issue 2:** ~Group 0 and group 1 will pass the `should_commit` but group 0 needs healing which is wrong and the healing process will cause another hang.~ Fixed with: pytorch/torchft#83 **Issue 3:** ~The byproduct of issue 1 and issue 2: group 1 will continue to print out~ ``` [rank0]:devgpu051:76838:80357 [0] misc/socket.cc:50 NCCL WARN socketProgress: Connection closed by remote peer devgpu051.cln3.svc.fbinfra.net<33618> ``` Fixed with pytorch/torchft#91 and several other fixes. **Issue 4:** When there are 3 groups, everyone requests the state dict every step. ***How to reproduce?*** Using the `Reproduce steps` to run 2 groups, then add another group by modifying the command. Seems to be fixed, will need more tests. **Issue 5:** Hang will happen if using functional collective. ***How to reproduce?*** Pull the latest version of this PR and comment out line 41 and uncomment line 42 in `torchtitan/utils.py` **Reproduce steps:** 1. Patch TorchFT with pytorch/torchft#82 2. Execute lighthouse 3. Execute the following command in one terminal: ``` TORCHFT_MANAGER_PORT=29520 REPLICA_GROUP_ID=0 CUDA_VISIBLE_DEVICES=0,1 NGPU=2 ./run_llama_train.sh --training.data_parallel_shard_degree=2 --experimental.enable_torchft --experimental.ft_replica_group_id=0 ``` 4. Wait 10 seconds, execute following command in another terminal: ``` TORCHFT_MANAGER_PORT=29522 REPLICA_GROUP_ID=1 CUDA_VISIBLE_DEVICES=2,3 NGPU=2 ./run_llama_train.sh --training.data_parallel_shard_degree=2 --experimental.enable_torchft --experimental.ft_replica_group_id=1 ``` ghstack-source-id: f7410e2ef7b91dd46913bb1213e4b939e4e2f1f1 Pull Request resolved: #834
fegin
added a commit
to pytorch/torchtitan
that referenced
this pull request
Feb 13, 2025
**Summary** This is a WIP TorchFT integration PR. **Current Issues** This doesn't work at this moment as there are hanged groups when a new group joins. **Issue 1:** ~Group 0 and group 1 will hang during the first `should_commit` after group 1 applying the pending state_dict from group 0.~ Fixed with: pytorch/torchft#83 **Issue 2:** ~Group 0 and group 1 will pass the `should_commit` but group 0 needs healing which is wrong and the healing process will cause another hang.~ Fixed with: pytorch/torchft#83 **Issue 3:** ~The byproduct of issue 1 and issue 2: group 1 will continue to print out~ ``` [rank0]:devgpu051:76838:80357 [0] misc/socket.cc:50 NCCL WARN socketProgress: Connection closed by remote peer devgpu051.cln3.svc.fbinfra.net<33618> ``` Fixed with pytorch/torchft#91 and several other fixes. **Issue 4:** When there are 3 groups, everyone requests the state dict every step. ***How to reproduce?*** Using the `Reproduce steps` to run 2 groups, then add another group by modifying the command. Seems to be fixed, will need more tests. **Issue 5:** Hang will happen if using functional collective. ***How to reproduce?*** Pull the latest version of this PR and comment out line 41 and uncomment line 42 in `torchtitan/utils.py` **Reproduce steps:** 1. Patch TorchFT with pytorch/torchft#82 2. Execute lighthouse 3. Execute the following command in one terminal: ``` TORCHFT_MANAGER_PORT=29520 REPLICA_GROUP_ID=0 CUDA_VISIBLE_DEVICES=0,1 NGPU=2 ./run_llama_train.sh --training.data_parallel_shard_degree=2 --experimental.enable_torchft --experimental.ft_replica_group_id=0 ``` 4. Wait 10 seconds, execute following command in another terminal: ``` TORCHFT_MANAGER_PORT=29522 REPLICA_GROUP_ID=1 CUDA_VISIBLE_DEVICES=2,3 NGPU=2 ./run_llama_train.sh --training.data_parallel_shard_degree=2 --experimental.enable_torchft --experimental.ft_replica_group_id=1 ``` ghstack-source-id: 531d305c0cb5d161012cbcf4d53d0266e09c7b21 Pull Request resolved: #834
fegin
added a commit
to pytorch/torchtitan
that referenced
this pull request
Feb 13, 2025
**Summary** This is a WIP TorchFT integration PR. **Current Issues** This doesn't work at this moment as there are hanged groups when a new group joins. **Issue 1:** ~Group 0 and group 1 will hang during the first `should_commit` after group 1 applying the pending state_dict from group 0.~ Fixed with: pytorch/torchft#83 **Issue 2:** ~Group 0 and group 1 will pass the `should_commit` but group 0 needs healing which is wrong and the healing process will cause another hang.~ Fixed with: pytorch/torchft#83 **Issue 3:** ~The byproduct of issue 1 and issue 2: group 1 will continue to print out~ ``` [rank0]:devgpu051:76838:80357 [0] misc/socket.cc:50 NCCL WARN socketProgress: Connection closed by remote peer devgpu051.cln3.svc.fbinfra.net<33618> ``` Fixed with pytorch/torchft#91 and several other fixes. **Issue 4:** When there are 3 groups, everyone requests the state dict every step. ***How to reproduce?*** Using the `Reproduce steps` to run 2 groups, then add another group by modifying the command. Seems to be fixed, will need more tests. **Issue 5:** Hang will happen if using functional collective. ***How to reproduce?*** Pull the latest version of this PR and comment out line 41 and uncomment line 42 in `torchtitan/utils.py` **Reproduce steps:** 1. Patch TorchFT with pytorch/torchft#82 2. Execute lighthouse 3. Execute the following command in one terminal: ``` TORCHFT_MANAGER_PORT=29520 REPLICA_GROUP_ID=0 CUDA_VISIBLE_DEVICES=0,1 NGPU=2 ./run_llama_train.sh --training.data_parallel_shard_degree=2 --experimental.enable_torchft --experimental.ft_replica_group_id=0 ``` 4. Wait 10 seconds, execute following command in another terminal: ``` TORCHFT_MANAGER_PORT=29522 REPLICA_GROUP_ID=1 CUDA_VISIBLE_DEVICES=2,3 NGPU=2 ./run_llama_train.sh --training.data_parallel_shard_degree=2 --experimental.enable_torchft --experimental.ft_replica_group_id=1 ``` ghstack-source-id: deba15cc9de1f8a332448994ce44665e9d2e6790 Pull Request resolved: #834
fegin
added a commit
to pytorch/torchtitan
that referenced
this pull request
Feb 13, 2025
**Summary** This is a WIP TorchFT integration PR. **Current Issues** This doesn't work at this moment as there are hanged groups when a new group joins. **Issue 1:** ~Group 0 and group 1 will hang during the first `should_commit` after group 1 applying the pending state_dict from group 0.~ Fixed with: pytorch/torchft#83 **Issue 2:** ~Group 0 and group 1 will pass the `should_commit` but group 0 needs healing which is wrong and the healing process will cause another hang.~ Fixed with: pytorch/torchft#83 **Issue 3:** ~The byproduct of issue 1 and issue 2: group 1 will continue to print out~ ``` [rank0]:devgpu051:76838:80357 [0] misc/socket.cc:50 NCCL WARN socketProgress: Connection closed by remote peer devgpu051.cln3.svc.fbinfra.net<33618> ``` Fixed with pytorch/torchft#91 and several other fixes. **Issue 4:** When there are 3 groups, everyone requests the state dict every step. ***How to reproduce?*** Using the `Reproduce steps` to run 2 groups, then add another group by modifying the command. Seems to be fixed, will need more tests. **Issue 5:** Hang will happen if using functional collective. ***How to reproduce?*** Pull the latest version of this PR and comment out line 41 and uncomment line 42 in `torchtitan/utils.py` **Reproduce steps:** 1. Patch TorchFT with pytorch/torchft#82 2. Execute lighthouse 3. Execute the following command in one terminal: ``` TORCHFT_MANAGER_PORT=29520 REPLICA_GROUP_ID=0 CUDA_VISIBLE_DEVICES=0,1 NGPU=2 ./run_llama_train.sh --training.data_parallel_shard_degree=2 --experimental.enable_torchft --experimental.ft_replica_group_id=0 ``` 4. Wait 10 seconds, execute following command in another terminal: ``` TORCHFT_MANAGER_PORT=29522 REPLICA_GROUP_ID=1 CUDA_VISIBLE_DEVICES=2,3 NGPU=2 ./run_llama_train.sh --training.data_parallel_shard_degree=2 --experimental.enable_torchft --experimental.ft_replica_group_id=1 ``` ghstack-source-id: 26184a1937925c5ebf3b8c61c6a9f15418cca6b5 Pull Request resolved: #834
fegin
added a commit
to pytorch/torchtitan
that referenced
this pull request
Feb 13, 2025
**Summary** This is a WIP TorchFT integration PR. **Current Issues** This doesn't work at this moment as there are hanged groups when a new group joins. **Issue 1:** ~Group 0 and group 1 will hang during the first `should_commit` after group 1 applying the pending state_dict from group 0.~ Fixed with: pytorch/torchft#83 **Issue 2:** ~Group 0 and group 1 will pass the `should_commit` but group 0 needs healing which is wrong and the healing process will cause another hang.~ Fixed with: pytorch/torchft#83 **Issue 3:** ~The byproduct of issue 1 and issue 2: group 1 will continue to print out~ ``` [rank0]:devgpu051:76838:80357 [0] misc/socket.cc:50 NCCL WARN socketProgress: Connection closed by remote peer devgpu051.cln3.svc.fbinfra.net<33618> ``` Fixed with pytorch/torchft#91 and several other fixes. **Issue 4:** When there are 3 groups, everyone requests the state dict every step. ***How to reproduce?*** Using the `Reproduce steps` to run 2 groups, then add another group by modifying the command. Seems to be fixed, will need more tests. **Issue 5:** Hang will happen if using functional collective. ***How to reproduce?*** Pull the latest version of this PR and comment out line 41 and uncomment line 42 in `torchtitan/utils.py` **Reproduce steps:** 1. Patch TorchFT with pytorch/torchft#82 2. Execute lighthouse 3. Execute the following command in one terminal: ``` TORCHFT_MANAGER_PORT=29520 REPLICA_GROUP_ID=0 CUDA_VISIBLE_DEVICES=0,1 NGPU=2 ./run_llama_train.sh --training.data_parallel_shard_degree=2 --experimental.enable_torchft --experimental.ft_replica_group_id=0 ``` 4. Wait 10 seconds, execute following command in another terminal: ``` TORCHFT_MANAGER_PORT=29522 REPLICA_GROUP_ID=1 CUDA_VISIBLE_DEVICES=2,3 NGPU=2 ./run_llama_train.sh --training.data_parallel_shard_degree=2 --experimental.enable_torchft --experimental.ft_replica_group_id=1 ``` ghstack-source-id: 11e74f8fafa225d64376e317de39b362a8013471 Pull Request resolved: #834
fegin
added a commit
to pytorch/torchtitan
that referenced
this pull request
Feb 13, 2025
**Summary** This is a WIP TorchFT integration PR. **Current Issues** This doesn't work at this moment as there are hanged groups when a new group joins. **Issue 1:** ~Group 0 and group 1 will hang during the first `should_commit` after group 1 applying the pending state_dict from group 0.~ Fixed with: pytorch/torchft#83 **Issue 2:** ~Group 0 and group 1 will pass the `should_commit` but group 0 needs healing which is wrong and the healing process will cause another hang.~ Fixed with: pytorch/torchft#83 **Issue 3:** ~The byproduct of issue 1 and issue 2: group 1 will continue to print out~ ``` [rank0]:devgpu051:76838:80357 [0] misc/socket.cc:50 NCCL WARN socketProgress: Connection closed by remote peer devgpu051.cln3.svc.fbinfra.net<33618> ``` Fixed with pytorch/torchft#91 and several other fixes. **Issue 4:** When there are 3 groups, everyone requests the state dict every step. ***How to reproduce?*** Using the `Reproduce steps` to run 2 groups, then add another group by modifying the command. Seems to be fixed, will need more tests. **Issue 5:** Hang will happen if using functional collective. ***How to reproduce?*** Pull the latest version of this PR and comment out line 41 and uncomment line 42 in `torchtitan/utils.py` **Reproduce steps:** 1. Patch TorchFT with pytorch/torchft#82 2. Execute lighthouse 3. Execute the following command in one terminal: ``` TORCHFT_MANAGER_PORT=29520 REPLICA_GROUP_ID=0 CUDA_VISIBLE_DEVICES=0,1 NGPU=2 ./run_llama_train.sh --training.data_parallel_shard_degree=2 --experimental.enable_torchft --experimental.ft_replica_group_id=0 ``` 4. Wait 10 seconds, execute following command in another terminal: ``` TORCHFT_MANAGER_PORT=29522 REPLICA_GROUP_ID=1 CUDA_VISIBLE_DEVICES=2,3 NGPU=2 ./run_llama_train.sh --training.data_parallel_shard_degree=2 --experimental.enable_torchft --experimental.ft_replica_group_id=1 ``` ghstack-source-id: 11e74f8fafa225d64376e317de39b362a8013471 Pull Request resolved: #834
fegin
added a commit
to pytorch/torchtitan
that referenced
this pull request
Feb 13, 2025
**Summary** This is a WIP TorchFT integration PR. **Current Issues** This doesn't work at this moment as there are hanged groups when a new group joins. **Issue 1:** ~Group 0 and group 1 will hang during the first `should_commit` after group 1 applying the pending state_dict from group 0.~ Fixed with: pytorch/torchft#83 **Issue 2:** ~Group 0 and group 1 will pass the `should_commit` but group 0 needs healing which is wrong and the healing process will cause another hang.~ Fixed with: pytorch/torchft#83 **Issue 3:** ~The byproduct of issue 1 and issue 2: group 1 will continue to print out~ ``` [rank0]:devgpu051:76838:80357 [0] misc/socket.cc:50 NCCL WARN socketProgress: Connection closed by remote peer devgpu051.cln3.svc.fbinfra.net<33618> ``` Fixed with pytorch/torchft#91 and several other fixes. **Issue 4:** When there are 3 groups, everyone requests the state dict every step. ***How to reproduce?*** Using the `Reproduce steps` to run 2 groups, then add another group by modifying the command. Seems to be fixed, will need more tests. **Issue 5:** Hang will happen if using functional collective. ***How to reproduce?*** Pull the latest version of this PR and comment out line 41 and uncomment line 42 in `torchtitan/utils.py` **Reproduce steps:** 1. Patch TorchFT with pytorch/torchft#82 2. Execute lighthouse 3. Execute the following command in one terminal: ``` TORCHFT_MANAGER_PORT=29520 REPLICA_GROUP_ID=0 CUDA_VISIBLE_DEVICES=0,1 NGPU=2 ./run_llama_train.sh --training.data_parallel_shard_degree=2 --experimental.enable_torchft --experimental.ft_replica_group_id=0 ``` 4. Wait 10 seconds, execute following command in another terminal: ``` TORCHFT_MANAGER_PORT=29522 REPLICA_GROUP_ID=1 CUDA_VISIBLE_DEVICES=2,3 NGPU=2 ./run_llama_train.sh --training.data_parallel_shard_degree=2 --experimental.enable_torchft --experimental.ft_replica_group_id=1 ``` ghstack-source-id: faaff7d35797af17e91def949776f06f208948bc Pull Request resolved: #834
fegin
added a commit
to pytorch/torchtitan
that referenced
this pull request
Feb 13, 2025
**Summary** This is a WIP TorchFT integration PR. **Current Issues** This doesn't work at this moment as there are hanged groups when a new group joins. **Issue 1:** ~Group 0 and group 1 will hang during the first `should_commit` after group 1 applying the pending state_dict from group 0.~ Fixed with: pytorch/torchft#83 **Issue 2:** ~Group 0 and group 1 will pass the `should_commit` but group 0 needs healing which is wrong and the healing process will cause another hang.~ Fixed with: pytorch/torchft#83 **Issue 3:** ~The byproduct of issue 1 and issue 2: group 1 will continue to print out~ ``` [rank0]:devgpu051:76838:80357 [0] misc/socket.cc:50 NCCL WARN socketProgress: Connection closed by remote peer devgpu051.cln3.svc.fbinfra.net<33618> ``` Fixed with pytorch/torchft#91 and several other fixes. **Issue 4:** When there are 3 groups, everyone requests the state dict every step. ***How to reproduce?*** Using the `Reproduce steps` to run 2 groups, then add another group by modifying the command. Seems to be fixed, will need more tests. **Issue 5:** Hang will happen if using functional collective. ***How to reproduce?*** Pull the latest version of this PR and comment out line 41 and uncomment line 42 in `torchtitan/utils.py` **Reproduce steps:** 1. Patch TorchFT with pytorch/torchft#82 2. Execute lighthouse 3. Execute the following command in one terminal: ``` TORCHFT_MANAGER_PORT=29520 REPLICA_GROUP_ID=0 CUDA_VISIBLE_DEVICES=0,1 NGPU=2 ./run_llama_train.sh --training.data_parallel_shard_degree=2 --experimental.enable_torchft --experimental.ft_replica_group_id=0 ``` 4. Wait 10 seconds, execute following command in another terminal: ``` TORCHFT_MANAGER_PORT=29522 REPLICA_GROUP_ID=1 CUDA_VISIBLE_DEVICES=2,3 NGPU=2 ./run_llama_train.sh --training.data_parallel_shard_degree=2 --experimental.enable_torchft --experimental.ft_replica_group_id=1 ``` ghstack-source-id: 72e155127b8389cf4b0f1398b6ef438f68a15062 Pull Request resolved: #834
fegin
added a commit
to pytorch/torchtitan
that referenced
this pull request
Feb 13, 2025
**Summary** This is a WIP TorchFT integration PR. **Current Issues** This doesn't work at this moment as there are hanged groups when a new group joins. **Issue 1:** ~Group 0 and group 1 will hang during the first `should_commit` after group 1 applying the pending state_dict from group 0.~ Fixed with: pytorch/torchft#83 **Issue 2:** ~Group 0 and group 1 will pass the `should_commit` but group 0 needs healing which is wrong and the healing process will cause another hang.~ Fixed with: pytorch/torchft#83 **Issue 3:** ~The byproduct of issue 1 and issue 2: group 1 will continue to print out~ ``` [rank0]:devgpu051:76838:80357 [0] misc/socket.cc:50 NCCL WARN socketProgress: Connection closed by remote peer devgpu051.cln3.svc.fbinfra.net<33618> ``` Fixed with pytorch/torchft#91 and several other fixes. **Issue 4:** When there are 3 groups, everyone requests the state dict every step. ***How to reproduce?*** Using the `Reproduce steps` to run 2 groups, then add another group by modifying the command. Seems to be fixed, will need more tests. **Issue 5:** Hang will happen if using functional collective. ***How to reproduce?*** Pull the latest version of this PR and comment out line 41 and uncomment line 42 in `torchtitan/utils.py` **Reproduce steps:** 1. Patch TorchFT with pytorch/torchft#82 2. Execute lighthouse 3. Execute the following command in one terminal: ``` TORCHFT_MANAGER_PORT=29520 REPLICA_GROUP_ID=0 CUDA_VISIBLE_DEVICES=0,1 NGPU=2 ./run_llama_train.sh --training.data_parallel_shard_degree=2 --experimental.enable_torchft --experimental.ft_replica_group_id=0 ``` 4. Wait 10 seconds, execute following command in another terminal: ``` TORCHFT_MANAGER_PORT=29522 REPLICA_GROUP_ID=1 CUDA_VISIBLE_DEVICES=2,3 NGPU=2 ./run_llama_train.sh --training.data_parallel_shard_degree=2 --experimental.enable_torchft --experimental.ft_replica_group_id=1 ``` ghstack-source-id: f7caa9f8608ae4a7fc60dcc58dbc35287879911a Pull Request resolved: #834
fegin
added a commit
to pytorch/torchtitan
that referenced
this pull request
Feb 13, 2025
**Summary** This is a WIP TorchFT integration PR. **Current Issues** This doesn't work at this moment as there are hanged groups when a new group joins. **Issue 1:** ~Group 0 and group 1 will hang during the first `should_commit` after group 1 applying the pending state_dict from group 0.~ Fixed with: pytorch/torchft#83 **Issue 2:** ~Group 0 and group 1 will pass the `should_commit` but group 0 needs healing which is wrong and the healing process will cause another hang.~ Fixed with: pytorch/torchft#83 **Issue 3:** ~The byproduct of issue 1 and issue 2: group 1 will continue to print out~ ``` [rank0]:devgpu051:76838:80357 [0] misc/socket.cc:50 NCCL WARN socketProgress: Connection closed by remote peer devgpu051.cln3.svc.fbinfra.net<33618> ``` Fixed with pytorch/torchft#91 and several other fixes. **Issue 4:** When there are 3 groups, everyone requests the state dict every step. ***How to reproduce?*** Using the `Reproduce steps` to run 2 groups, then add another group by modifying the command. Seems to be fixed, will need more tests. **Issue 5:** Hang will happen if using functional collective. ***How to reproduce?*** Pull the latest version of this PR and comment out line 41 and uncomment line 42 in `torchtitan/utils.py` **Reproduce steps:** 1. Patch TorchFT with pytorch/torchft#82 2. Execute lighthouse 3. Execute the following command in one terminal: ``` TORCHFT_MANAGER_PORT=29520 REPLICA_GROUP_ID=0 CUDA_VISIBLE_DEVICES=0,1 NGPU=2 ./run_llama_train.sh --training.data_parallel_shard_degree=2 --experimental.enable_torchft --experimental.ft_replica_group_id=0 ``` 4. Wait 10 seconds, execute following command in another terminal: ``` TORCHFT_MANAGER_PORT=29522 REPLICA_GROUP_ID=1 CUDA_VISIBLE_DEVICES=2,3 NGPU=2 ./run_llama_train.sh --training.data_parallel_shard_degree=2 --experimental.enable_torchft --experimental.ft_replica_group_id=1 ``` ghstack-source-id: 9e2c0c609c04182976e2c4d591a29df5c4253125 Pull Request resolved: #834
fegin
added a commit
to pytorch/torchtitan
that referenced
this pull request
Feb 13, 2025
**Summary** This is a WIP TorchFT integration PR. **Current Issues** This doesn't work at this moment as there are hanged groups when a new group joins. **Issue 1:** ~Group 0 and group 1 will hang during the first `should_commit` after group 1 applying the pending state_dict from group 0.~ Fixed with: pytorch/torchft#83 **Issue 2:** ~Group 0 and group 1 will pass the `should_commit` but group 0 needs healing which is wrong and the healing process will cause another hang.~ Fixed with: pytorch/torchft#83 **Issue 3:** ~The byproduct of issue 1 and issue 2: group 1 will continue to print out~ ``` [rank0]:devgpu051:76838:80357 [0] misc/socket.cc:50 NCCL WARN socketProgress: Connection closed by remote peer devgpu051.cln3.svc.fbinfra.net<33618> ``` Fixed with pytorch/torchft#91 and several other fixes. **Issue 4:** When there are 3 groups, everyone requests the state dict every step. ***How to reproduce?*** Using the `Reproduce steps` to run 2 groups, then add another group by modifying the command. Seems to be fixed, will need more tests. **Issue 5:** Hang will happen if using functional collective. ***How to reproduce?*** Pull the latest version of this PR and comment out line 41 and uncomment line 42 in `torchtitan/utils.py` **Reproduce steps:** 1. Patch TorchFT with pytorch/torchft#82 2. Execute lighthouse 3. Execute the following command in one terminal: ``` TORCHFT_MANAGER_PORT=29520 REPLICA_GROUP_ID=0 CUDA_VISIBLE_DEVICES=0,1 NGPU=2 ./run_llama_train.sh --training.data_parallel_shard_degree=2 --experimental.enable_torchft --experimental.ft_replica_group_id=0 ``` 4. Wait 10 seconds, execute following command in another terminal: ``` TORCHFT_MANAGER_PORT=29522 REPLICA_GROUP_ID=1 CUDA_VISIBLE_DEVICES=2,3 NGPU=2 ./run_llama_train.sh --training.data_parallel_shard_degree=2 --experimental.enable_torchft --experimental.ft_replica_group_id=1 ``` ghstack-source-id: 1c5c3953695085b0fe9d02458c074bd63666b2e2 Pull Request resolved: #834
fegin
added a commit
to pytorch/torchtitan
that referenced
this pull request
Feb 14, 2025
**Summary** This is a WIP TorchFT integration PR. **Current Issues** This doesn't work at this moment as there are hanged groups when a new group joins. **Issue 1:** ~Group 0 and group 1 will hang during the first `should_commit` after group 1 applying the pending state_dict from group 0.~ Fixed with: pytorch/torchft#83 **Issue 2:** ~Group 0 and group 1 will pass the `should_commit` but group 0 needs healing which is wrong and the healing process will cause another hang.~ Fixed with: pytorch/torchft#83 **Issue 3:** ~The byproduct of issue 1 and issue 2: group 1 will continue to print out~ ``` [rank0]:devgpu051:76838:80357 [0] misc/socket.cc:50 NCCL WARN socketProgress: Connection closed by remote peer devgpu051.cln3.svc.fbinfra.net<33618> ``` Fixed with pytorch/torchft#91 and several other fixes. **Issue 4:** When there are 3 groups, everyone requests the state dict every step. ***How to reproduce?*** Using the `Reproduce steps` to run 2 groups, then add another group by modifying the command. Seems to be fixed, will need more tests. **Issue 5:** Hang will happen if using functional collective. ***How to reproduce?*** Pull the latest version of this PR and comment out line 41 and uncomment line 42 in `torchtitan/utils.py` **Reproduce steps:** 1. Patch TorchFT with pytorch/torchft#82 2. Execute lighthouse 3. Execute the following command in one terminal: ``` TORCHFT_MANAGER_PORT=29520 REPLICA_GROUP_ID=0 CUDA_VISIBLE_DEVICES=0,1 NGPU=2 ./run_llama_train.sh --training.data_parallel_shard_degree=2 --experimental.enable_torchft --experimental.ft_replica_group_id=0 ``` 4. Wait 10 seconds, execute following command in another terminal: ``` TORCHFT_MANAGER_PORT=29522 REPLICA_GROUP_ID=1 CUDA_VISIBLE_DEVICES=2,3 NGPU=2 ./run_llama_train.sh --training.data_parallel_shard_degree=2 --experimental.enable_torchft --experimental.ft_replica_group_id=1 ``` ghstack-source-id: 5cb94c697e0053266f6ca08c1a441dfe364c215f Pull Request resolved: #834
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Summary:
We are not going to land this PR, this PR may be further divided into several PRs.