If merged this commit doesn the following:

* setup git dir in /workspace/llama31
* Remove subpath from pretrain_llama.py
* Install toml package
* Adjust --gres=gpu:8 to number of user specified devices

Signed-off-by: Sean Smith <seasmith@nvidia.com>

Author: test

large_language_model_pretraining/nemo/Dockerfile

@@ -18,6 +18,7 @@ FROM ${NEMO_BASE_IMAGE} AS nemo-base-image
 RUN pip uninstall transformers -y
 RUN pip install transformers==4.47.1 blobfile==3.0.0
 RUN pip install prettytable==3.12.0
+RUN pip install toml==0.10.2
 RUN pip install git+https://github.com/mlcommons/logging.git@4.1.0-rc3
 
 # setup workspace

large_language_model_pretraining/nemo/pretrain_llama31.py

@@ -75,8 +75,8 @@ def slurm_executor(
         gpus_per_node=devices,
         mem="0",
         exclusive=True,
-        gres="gpu:8",
-        packager=run.GitArchivePackager(subpath="large_language_model_pretraining/nemo", ref="HEAD"),
+        gres=f"gpu:{devices}",
+        packager=run.GitArchivePackager(),
         dependencies=dependencies,
     )
 

large_language_model_pretraining/nemo/run_llama31.sh

@@ -16,6 +16,9 @@
 
 set -e
 
+git init
+git add .
+git commit -sm "First commit"
 git config --global --add safe.directory /workspace/llama31
 
 # Vars without defaults