Switch to fsdp training #81
Merged
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
Signed-off-by: Michael Clifford <[email protected]>
leseb
reviewed
Oct 10, 2024
| @@ -191,7 +191,7 @@ def list_phase1_final_model(): | |||
| export XDG_CACHE_HOME=/tmp | |||
| export HF_HOME=/tmp | |||
| export TRANSFORMERS_CACHE=/tmp | |||
| torchrun --nnodes {nnodes} --nproc_per_node {nproc_per_node} --node_rank \$(RANK) --rdzv_endpoint \$(MASTER_ADDR):\$(MASTER_PORT) -m instructlab.training.main_ds --model_name_or_path={path_to_model} --data_path=/input_data/processed_data/data.jsonl --output_dir=/tmp/model --num_epochs=2 --effective_batch_size=3840 --learning_rate=2e-6 --num_warmup_steps=800 --save_samples=0 --log_level=INFO --max_batch_len=20000 --seed=42 --cpu_offload_optimizer --sharding_strategy=FULL_SHARD --is_granite --checkpoint_at_epoch | |||
| torchrun --nnodes {nnodes} --nproc_per_node {nproc_per_node} --node_rank \$(RANK) --rdzv_endpoint \$(MASTER_ADDR):\$(MASTER_PORT) -m instructlab.training.main_ds --model_name_or_path={path_to_model} --data_path=/input_data/processed_data/data.jsonl --output_dir=/tmp/model --num_epochs=2 --effective_batch_size=3840 --learning_rate=1e-4 --num_warmup_steps=800 --save_samples=0 --log_level=INFO --max_batch_len=20000 --seed=42 --cpu_offload_optimizer --distributed_training_framework fsdp --is_granite --checkpoint_at_epoch | |||
There was a problem hiding this comment.
What is the --learning_rate fix about?
There was a problem hiding this comment.
This was just fixing a typo where I forgot to update the learning_rate as it should be the same value on the master and worker nodes calls to torchrun
Closed
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.
Related to #51
This is PR simply adds the
distributed_training_frameworkparameter to the torchrun call of our PyTorchJob, and sets it tofsdp. This requires that the latest RHEL AI image (1.2) is used since it has the recent updated needed for FSDP.This has been tested on the MOC and seems to work as expected.