Skip to content

Minor fixes for MBridge nemotron deployment #518

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

Open
athitten wants to merge 2 commits into
base: main
Choose a base branch
from
Open

Minor fixes for MBridge nemotron deployment #518

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
6b8878f
Minor fixes for nemotron deployment
athitten Nov 20, 2025
b3566b4
Lint fixes
athitten Nov 21, 2025
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
14 changes: 14 additions & 0 deletions nemo_deploy/llm/inference/inference_base.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -246,6 +246,20 @@ def setup_megatron_model_and_tokenizer_for_inference(
dist_config = DistributedInitConfig(distributed_backend="nccl")
torch_distributed_init(dist_config)
model_config, mlm_args = load_model_config(checkpoint_path)

# Convert attention_backend from string to enum if needed
if hasattr(model_config, "attention_backend") and isinstance(model_config.attention_backend, str):
if model_config.attention_backend == "AttnBackend.fused":
model_config.attention_backend = AttnBackend.fused
elif model_config.attention_backend == "AttnBackend.flash":
model_config.attention_backend = AttnBackend.flash
elif model_config.attention_backend == "AttnBackend.unfused":
model_config.attention_backend = AttnBackend.unfused
elif model_config.attention_backend == "AttnBackend.local":
model_config.attention_backend = AttnBackend.local
elif model_config.attention_backend == "AttnBackend.auto":
model_config.attention_backend = AttnBackend.auto

if tensor_model_parallel_size is not None:
model_config.tensor_model_parallel_size = tensor_model_parallel_size
if pipeline_model_parallel_size is not None:
Expand Down
2 changes: 1 addition & 1 deletion nemo_deploy/llm/inference/tron_utils.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -320,7 +320,7 @@ def _initialize_tp_communicators(model_config: Union[GPTConfig, T5Config], micro
"Tensor Parallel Communication/GEMM Overlap optimization needs 'yaml' and 'transformer_engine' packages"
)

if model_config.tp_comm_overlap_cfg is not None:
if hasattr(model_config, "tp_comm_overlap_cfg") and model_config.tp_comm_overlap_cfg is not None:
with open(model_config.tp_comm_overlap_cfg, "r") as stream:
ub_cfgs = yaml.safe_load(stream)
else:
Expand Down
Loading