Merge branch 'main' into fix_onnx_fp8_scaling

Author: Darth-Kronos

.github/workflows/code_quality.yml

@@ -2,7 +2,7 @@ name: Code Quality
 
 on:
   pull_request:
-    branches: [main, release/*]
+    branches: [main, release/*, feature/*]
   schedule:
     - cron: "0 0 * * *" # Nightly
   workflow_dispatch: # On-demand

.github/workflows/gpu_tests.yml

@@ -61,7 +61,7 @@ jobs:
     if: needs.check-file-changes.outputs.any_changed == 'true'
     # Runner list at https://github.com/nv-gha-runners/enterprise-runner-configuration/blob/main/docs/runner-groups.md
     runs-on: linux-amd64-gpu-l4-latest-1
-    timeout-minutes: 90
+    timeout-minutes: 120
     container: &gpu_container
       image: nvcr.io/nvidia/pytorch:25.06-py3
       env:
@@ -73,15 +73,14 @@ jobs:
       - uses: nv-gha-runners/setup-proxy-cache@main
       - name: Setup environment variables
         run: |
-          echo "LD_LIBRARY_PATH=${LD_LIBRARY_PATH}:/usr/include:/usr/lib/x86_64-linux-gnu:/usr/local/tensorrt/targets/x86_64-linux-gnu/lib" >> $GITHUB_ENV
-          echo "PATH=${PATH}:/usr/local/tensorrt/targets/x86_64-linux-gnu/bin" >> $GITHUB_ENV
+          echo "LD_LIBRARY_PATH=${LD_LIBRARY_PATH}:/usr/include:/usr/lib/x86_64-linux-gnu" >> $GITHUB_ENV
       - name: Run gpu tests
         run: pip install tox-current-env && tox -e py312-cuda12-gpu --current-env
   gpu-tests-non-pr:
     if: ${{ !startsWith(github.ref, 'refs/heads/pull-request/') }}
     # Runner list at https://github.com/nv-gha-runners/enterprise-runner-configuration/blob/main/docs/runner-groups.md
     runs-on: linux-amd64-gpu-h100-latest-1
-    timeout-minutes: 90
+    timeout-minutes: 120
     container: *gpu_container
     steps: *gpu_steps
   gpu-pr-required-check:

.github/workflows/pages.yml

@@ -2,7 +2,7 @@ name: Docs
 
 on:
   pull_request:
-    branches: [main, release/*]
+    branches: [main, release/*, feature/*]
   push:
     branches: [main]
   schedule:

.github/workflows/unit_tests.yml

@@ -3,9 +3,9 @@ name: Unit tests
 
 on:
   pull_request:
-    branches: [main, release/*]
+    branches: [main, release/*, feature/*]
   push:
-    branches: [main, release/*]
+    branches: [main, release/*, feature/*]
     paths:
       - ".github/workflows/unit_tests.yml"
       - "modelopt/**"

.gitlab/tests.yml

@@ -35,9 +35,7 @@ unit:
   tags: [docker, linux, 2-gpu]
   before_script:
     # Add libcudnn*.so and libnv*.so to path
-    - export LD_LIBRARY_PATH="${LD_LIBRARY_PATH}:/usr/include:/usr/lib/x86_64-linux-gnu:/usr/local/tensorrt/targets/x86_64-linux-gnu/lib"
-    # Add trtexec to path
-    - export PATH="${PATH}:/usr/local/tensorrt/targets/x86_64-linux-gnu/bin"
+    - export LD_LIBRARY_PATH="${LD_LIBRARY_PATH}:/usr/include:/usr/lib/x86_64-linux-gnu"
     # Install git-lfs for Daring-Anteater dataset
     - apt-get update && apt-get install -y git-lfs
     - git lfs install --system

.pre-commit-config.yaml

@@ -66,7 +66,7 @@ repos:
           - --comment-style
           - "#"
           - --allow-past-years
-        types: [python, shell]
+        types_or: [python, shell]
         # NOTE: Exclude files that have copyright or license headers from another company or individual
         # since we want to keep those above the license header added by this hook.
         # Instead, we should manually add the license header to those files *after* the original header.

CHANGELOG-Windows.rst

@@ -29,7 +29,7 @@ Model Optimizer Changelog (Windows)
 - **LLM Quantization with Olive:** Enabled LLM quantization through Olive, streamlining model optimization workflows. Refer `example <https://github.com/microsoft/Olive/tree/main/examples/phi3#quantize-models-with-nvidia-tensorrt-model-optimizer>`_
 - **DirectML Deployment Guide:** Added DML deployment guide. Refer :ref:`DirectML_Deployment`.
 - **MMLU Benchmark for Accuracy Evaluations:** Introduced `MMLU benchmarking <https://github.com/NVIDIA/TensorRT-Model-Optimizer/tree/main/examples/windows/accuracy_benchmark/README.md>`_ for accuracy evaluation of ONNX models on DirectML (DML).
-- **Published quantized ONNX models collection:** Published quantized ONNX models at HuggingFace `NVIDIA collections <https://huggingface.co/collections/nvidia/optimized-onnx-models-for-nvidia-rtx-gpus-67373fe7c006ebc1df310613>`_.
+- **Published quantized ONNX models collection:** Published quantized ONNX models at HuggingFace `NVIDIA collections <https://huggingface.co/collections/nvidia/optimized-onnx-models-for-nvidia-rtx-gpus>`_.
 
 
 \* *This version includes experimental features such as TensorRT deployment of ONNX INT4 models, PyTorch quantization and sparsity. These are currently unverified on Windows.*

CHANGELOG.rst

@@ -1,18 +1,20 @@
 Model Optimizer Changelog (Linux)
 =================================
 
-0.39 (2025-11-xx)
+0.39 (2025-11-07)
 ^^^^^^^^^^^^^^^^^
 
-**Deprecations**
-
 **New Features**
 
 - Add flag ``op_types_to_exclude_fp16`` in ONNX quantization to exclude ops from being converted to FP16/BF16. Alternatively, for custom TensorRT ops, this can also be done by indicating ``'fp32'`` precision in ``trt_plugins_precision``.
 - Add LoRA mode support for MCore in a new peft submodule: ``modelopt.torch.peft.update_model(model, LORA_CFG)``.
 - Support PTQ and fakequant in vLLM for fast evaluation of arbitrary quantization formats. See ``examples/vllm_serve`` for more details.
-- Add support for ``nemotron-post-training-dataset-v2`` and ``nemotron-post-training-dataset-v1`` in ``examples/llm_ptq``. Default to a mix of ``cnn_dailymail`` and ``nemotron-post-training-dataset-v2`` if no dataset is specified.
+- Add support for ``nemotron-post-training-dataset-v2`` and ``nemotron-post-training-dataset-v1`` in ``examples/llm_ptq``. Default to a mix of ``cnn_dailymail`` and ``nemotron-post-training-dataset-v2`` (gated dataset accessed using ``HF_TOKEN`` environment variable) if no dataset is specified.
 - Allow specifying ``calib_seq`` in ``examples/llm_ptq`` to set the maximum sequence length for calibration.
+- Add support for MCore MoE PTQ/QAT/QAD.
+- Add support for multi-node PTQ and export with FSDP2 in ``examples/llm_ptq/multinode_ptq.py``. See `examples/llm_ptq/README.md <https://github.com/NVIDIA/TensorRT-Model-Optimizer/tree/main/examples/llm_ptq#multi-node-post-training-quantization-with-fsdp2>`_ for more details.
+- Add support for Nemotron Nano VL v1 & v2 models in FP8/NVFP4 PTQ workflow.
+- Add flags ``nodes_to_include`` and ``op_types_to_include`` in AutoCast to force-include nodes in low precision, even if they would otherwise be excluded by other rules.
 
 **Documentation**
 

README.md

@@ -98,7 +98,7 @@ more fine-grained control on installed dependencies or for alternative docker im
 
 ## Pre-Quantized Checkpoints
 
-- Ready-to-deploy checkpoints \[[🤗 Hugging Face - Nvidia TensorRT Model Optimizer Collection](https://huggingface.co/collections/nvidia/model-optimizer-66aa84f7966b3150262481a4)\]
+- Ready-to-deploy checkpoints \[[🤗 Hugging Face - Nvidia TensorRT Model Optimizer Collection](https://huggingface.co/collections/nvidia/inference-optimized-checkpoints-with-model-optimizer)\]
 - Deployable on [TensorRT-LLM](https://github.com/NVIDIA/TensorRT-LLM), [vLLM](https://github.com/vllm-project/vllm) and [SGLang](https://github.com/sgl-project/sglang)
 - More models coming soon!
 

docs/source/deployment/2_directml.rst

@@ -42,4 +42,4 @@ For further details and examples, please refer to the `ONNX Runtime documentatio
 Collection of optimized ONNX models
 ===================================
 
-The ready-to-deploy optimized ONNX models from ModelOpt-Windows are available at HuggingFace `NVIDIA collections <https://huggingface.co/collections/nvidia/optimized-onnx-models-for-nvidia-rtx-gpus-67373fe7c006ebc1df310613>`_. These models can be deployed using DirectML backend. Follow the instructions provided along with the published models for deployment.
+The ready-to-deploy optimized ONNX models from ModelOpt-Windows are available at HuggingFace `NVIDIA collections <https://huggingface.co/collections/nvidia/optimized-onnx-models-for-nvidia-rtx-gpus>`_. These models can be deployed using DirectML backend. Follow the instructions provided along with the published models for deployment.