Skip to content

fix: update Megatron-DeepSpeed tutorial to match current repo structure #7761

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
tohtana merged 5 commits into from
Jan 11, 2026
Merged

fix: update Megatron-DeepSpeed tutorial to match current repo structure #7761

Changes from all commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
a6c0777
fix: update Megatron-DeepSpeed tutorial to match current repo structure
nathon-lee Jan 5, 2026
c8196be
fix: avoid IndexError in BF16_Optimizer.destroy() when using DummyOpt…
nathon-lee Jan 7, 2026
b11fc56
Update megatron.md
nathon-lee Jan 7, 2026
df67bb7
Merge branch 'master' into fix_issue_7757
loadams Jan 7, 2026
8b060da
Merge branch 'master' into fix_issue_7757
tohtana Jan 10, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
20 changes: 10 additions & 10 deletions docs/_tutorials/megatron.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -31,31 +31,31 @@ git submodule update --init --recursive
### Running Unmodified Megatron-LM GPT2 model

* For a single GPU run:
- change `scripts/pretrain_gpt2.sh`, set its `--train-data` argument as `"webtext"`.
- run `bash scripts/pretrain_gpt2.sh`
- change `examples/pretrain_gpt.sh`, set its `--train-data` argument as `"webtext"`.
- run `bash examples/pretrain_gpt.sh`

* For multiple GPUs and/or nodes run:
- change `scripts/pretrain_gpt2_model_parallel.sh`
- change `examples/pretrain_gpt_distributed_with_mp.sh`
- set its `--train-data` argument as `"webtext"`
- `GPUS_PER_NODE` indicates how many GPUs per node involved in the testing
- `NNODES` indicates how many nodes involved in the testing

- run `bash scripts/pretrain_gpt2_model_parallel.sh`
- run `bash examples/pretrain_gpt_distributed_with_mp.sh`


## Enabling DeepSpeed

To use DeepSpeed we will modify three files :

* `arguments.py` : Arguments configurations
* `pretrain_gpt2.py` : Main entry point for training
* `utils.py` : Checkpoint saving and loading utilities
* `megatron/arguments.py` : Arguments configurations
* `pretrain_gpt.py` : Main entry point for training
* `megatron/utils.py` : Checkpoint saving and loading utilities


### Argument Parsing
The first step is adding DeepSpeed arguments to
Megatron-LM GPT2 model, using `deepspeed.add_config_arguments()` in
`arguments.py`.
`megatron/arguments.py`.

```python
def get_args():
Expand Down Expand Up @@ -360,9 +360,9 @@ Megatron-LM GPT2 model with DeepSpeed applied, execute the following command to
start training.

- Single GPU run
- run `bash scripts/ds_pretrain_gpt2.sh`
- run `bash scripts/ds_pretrain_gpt.sh`
- Multiple GPUs/Nodes run
- run `bash scripts/ds_zero2_pretrain_gpt2_model_parallel.sh`
- run `bash scripts/ds_zero2_pretrain_gpt_model_parallel.sh`

## DeepSpeed Evaluation using GPT-2

Expand Down