Open
Conversation
… dynamic inference - Add nemo_deploy/llm/inference/nemo_utils.py which vendors standalone NeMo utilities (MCoreTokenizerWrappper, ckpt path helpers, constants) with no dependency on the nemo package, and re-exports the complex NeMo types (GPTConfig, T5Config, io, set_modelopt_spec_if_exists_in_ckpt) under a single HAVE_NEMO guard. - Remove direct from nemo.* imports from inference_base.py and tron_utils.py; both files now import from the local nemo_utils module instead. - Fix AttributeError in create_mcore_engine: GPTInferenceWrapper was called with (model, inference_context) but the deployed Megatron-LM API expects (model, inference_wrapper_config, inference_context). Add InferenceWrapperConfig built from model.config attributes; MCoreEngine then internally creates a DynamicInferenceContext and switches to DynamicInferenceEngine. Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
- Fix import ordering in test_inference_base.py (ruff I001) - Remove direct nemo imports from inference_base.py, nemo_utils.py, tron_utils.py - Add nemo_io.py with standalone load_context implementation - Remove HAVE_NEMO guard checks now that nemo is no longer a static dependency - Update tests to remove HAVE_NEMO patches and use types.SimpleNamespace
- Remove unused StaticInferenceContext import - Use inner model config for hidden_size/params_dtype instead of outer model - Add buffer_size_gb param to create_mcore_engine and MegatronLLMDeployable Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
oyilmaz-nvidia
requested review from
athitten,
meatybobby and
pthombre
as code owners
Signed-off-by: Onur Yilmaz <oyilmaz@nvidia.com>
Signed-off-by: Charlie Truong <chtruong@nvidia.com>
Signed-off-by: Charlie Truong <chtruong@nvidia.com>
Move the InferenceWrapperConfig import from module level into the body of create_mcore_engine, so pytest can collect test_inference_base.py in the nemo:26.02 container where that megatron-core module path does not exist. GPU-only tests that call create_mcore_engine are skipped in CPU CI, so the import never executes there.
Move the InferenceWrapperConfig import from module level into the body of create_mcore_engine, so pytest can collect test_inference_base.py in the nemo:26.02 container where that megatron-core module path does not exist. GPU-only tests that call create_mcore_engine are skipped in CPU CI, so the import never executes there.
Signed-off-by: Onur Yilmaz <oyilmaz@nvidia.com>
Signed-off-by: Onur Yilmaz <oyilmaz@nvidia.com>
Signed-off-by: Onur Yilmaz <oyilmaz@nvidia.com>
Signed-off-by: Onur Yilmaz <oyilmaz@nvidia.com>
Signed-off-by: Onur Yilmaz <oyilmaz@nvidia.com>
Contributor
Author
|
/ok to test d8ac4f5 |
Contributor
Author
|
/ok to test d8ac4f5 |
Signed-off-by: Onur Yilmaz <oyilmaz@nvidia.com>
Contributor
Author
|
/ok to test c398fee |
Signed-off-by: Onur Yilmaz <oyilmaz@nvidia.com>
Contributor
Author
|
/ok to test f63d2c3 |
This file contains hidden or 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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
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.
No description provided.