Merge branch 'main' of github.com:NVIDIA/TensorRT-Model-Optimizer into chenjiel/deprecate_trtllm_build_2

Signed-off-by: Chenjie Luo <[email protected]>

Author: cjluo-nv

.github/CODEOWNERS

@@ -32,24 +32,24 @@ modelopt/torch/utils @NVIDIA/modelopt-torch-utils-codeowners
 # Examples
 /docker @NVIDIA/modelopt-docker-codeowners
 /README.md @NVIDIA/modelopt-examples-codeowners
-examples @NVIDIA/modelopt-examples-codeowners
-examples/chained_optimizations @NVIDIA/modelopt-torch-nas-prune-codeowners
-examples/cnn_qat @NVIDIA/modelopt-examples-cnn_qat-codeowners
-examples/deepseek @NVIDIA/modelopt-deploy-codeowners
-examples/diffusers @NVIDIA/modelopt-examples-diffusers-codeowners
-examples/gpt-oss @NVIDIA/modelopt-examples-gpt-oss-codeowners
-examples/llm_autodeploy @NVIDIA/modelopt-deploy-codeowners
-examples/llm_distill @NVIDIA/modelopt-torch-distill-codeowners
-examples/llm_eval @NVIDIA/modelopt-examples-llm_ptq-codeowners
-examples/llm_ptq @NVIDIA/modelopt-examples-llm_ptq-codeowners
-examples/llm_qat @NVIDIA/modelopt-examples-llm_qat-codeowners
-examples/llm_sparsity @NVIDIA/modelopt-torch-sparsity-codeowners
-examples/megatron-lm @NVIDIA/modelopt-examples-megatron-codeowners
-examples/model_hub @NVIDIA/modelopt-examples-model_hub-codeowners
-examples/nemo_run @NVIDIA/modelopt-examples-megatron-codeowners
-examples/onnx_ptq @NVIDIA/modelopt-onnx-codeowners
-examples/pruning @NVIDIA/modelopt-torch-nas-prune-codeowners
-examples/speculative_decoding @NVIDIA/modelopt-torch-speculative-codeowners
-examples/vlm_eval @NVIDIA/modelopt-examples-vlm-codeowners
-examples/vlm_ptq @NVIDIA/modelopt-examples-vlm-codeowners
-examples/windows @NVIDIA/modelopt-windows-codeowners
+/examples @NVIDIA/modelopt-examples-codeowners
+/examples/chained_optimizations @NVIDIA/modelopt-torch-nas-prune-codeowners
+/examples/cnn_qat @NVIDIA/modelopt-examples-cnn_qat-codeowners
+/examples/deepseek @NVIDIA/modelopt-deploy-codeowners
+/examples/diffusers @NVIDIA/modelopt-examples-diffusers-codeowners
+/examples/gpt-oss @NVIDIA/modelopt-examples-gpt-oss-codeowners
+/examples/llm_autodeploy @NVIDIA/modelopt-deploy-codeowners
+/examples/llm_distill @NVIDIA/modelopt-torch-distill-codeowners
+/examples/llm_eval @NVIDIA/modelopt-examples-llm_ptq-codeowners
+/examples/llm_ptq @NVIDIA/modelopt-examples-llm_ptq-codeowners
+/examples/llm_qat @NVIDIA/modelopt-examples-llm_qat-codeowners
+/examples/llm_sparsity @NVIDIA/modelopt-torch-sparsity-codeowners
+/examples/megatron-lm @NVIDIA/modelopt-examples-megatron-codeowners
+/examples/model_hub @NVIDIA/modelopt-examples-model_hub-codeowners
+/examples/nemo_run @NVIDIA/modelopt-examples-megatron-codeowners
+/examples/onnx_ptq @NVIDIA/modelopt-onnx-codeowners
+/examples/pruning @NVIDIA/modelopt-torch-nas-prune-codeowners
+/examples/speculative_decoding @NVIDIA/modelopt-torch-speculative-codeowners
+/examples/vlm_eval @NVIDIA/modelopt-examples-vlm-codeowners
+/examples/vlm_ptq @NVIDIA/modelopt-examples-vlm-codeowners
+/examples/windows @NVIDIA/modelopt-windows-codeowners

.github/workflows/unit_tests.yml

@@ -4,13 +4,6 @@ name: Unit tests
 on:
   pull_request:
     branches: [main, release/*]
-    paths:
-      - ".github/workflows/unit_tests.yml"
-      - "modelopt/**"
-      - "tests/unit/**"
-      - "pyproject.toml"
-      - "setup.py"
-      - "tox.ini"
   push:
     branches: [main, release/*]
     paths:

CHANGELOG.rst

@@ -5,6 +5,7 @@ Model Optimizer Changelog (Linux)
 ^^^^^^^^^^^^^^^^^
 
 **Deprecations**
+- Deprecated ``quantize_mode`` argument in ``examples/onnx_ptq/evaluate.py`` to support strongly typing. Use ``engine_precision`` instead.
 
 - TRT-LLM's TRT backend in ``examples/llm_ptq`` and ``examples/vlm_ptq``. Tasks ``build`` and ``benchmark`` support are removed and replaced with ``quant``. For performance evaluation, please use ``trtllm-bench`` directly.
 - ``--export_fmt`` flag in ``examples/llm_ptq`` is removed. By default we export to the unified Hugging Face checkpoint format.
@@ -13,6 +14,7 @@ Model Optimizer Changelog (Linux)
 **Bug Fixes**
 
 **New Features**
+- ``high_precision_dtype`` default to fp16 in ONNX quantization, i.e. quantized output model weights are now FP16 by default.
 
 0.35 (2025-09-04)
 ^^^^^^^^^^^^^^^^^
@@ -25,6 +27,7 @@ Model Optimizer Changelog (Linux)
 **Bug Fixes**
 
 - Fix attention head ranking logic for pruning Megatron Core GPT models.
+- Upgrade TensorRT-LLM dependency to 1.1.0rc2.
 
 **New Features**
 

docker/Dockerfile

@@ -1,4 +1,4 @@
-FROM nvcr.io/nvidia/tensorrt-llm/release:1.0.0rc6
+FROM nvcr.io/nvidia/tensorrt-llm/release:1.1.0rc2.post2
 
 ARG PIP_EXTRA_INDEX_URL="https://pypi.nvidia.com"
 ENV PIP_EXTRA_INDEX_URL=$PIP_EXTRA_INDEX_URL \

examples/diffusers/quantization/diffusion_trt.py

@@ -105,7 +105,11 @@ def main():
 
     image_name = args.save_image_as if args.save_image_as else f"{args.model}.png"
 
-    pipe = PipelineManager.create_pipeline_from(MODEL_ID[args.model], dtype_map[args.model_dtype])
+    pipe = PipelineManager.create_pipeline_from(
+        MODEL_ID[args.model],
+        dtype_map[args.model_dtype],
+        override_model_path=args.override_model_path,
+    )
 
     # Save the backbone of the pipeline and move it to the GPU
     add_embedding = None

examples/diffusers/quantization/quantize.py

@@ -309,7 +309,9 @@ def __init__(self, config: ModelConfig, logger: logging.Logger):
 
     @staticmethod
     def create_pipeline_from(
-        model_type: ModelType, torch_dtype: torch.dtype = torch.bfloat16
+        model_type: ModelType,
+        torch_dtype: torch.dtype = torch.bfloat16,
+        override_model_path: str | None = None,
     ) -> DiffusionPipeline:
         """
         Create and return an appropriate pipeline based on configuration.
@@ -321,7 +323,9 @@ def create_pipeline_from(
             ValueError: If model type is unsupported
         """
         try:
-            model_id = MODEL_REGISTRY[model_type]
+            model_id = (
+                MODEL_REGISTRY[model_type] if override_model_path is None else override_model_path
+            )
             if model_type == ModelType.SD3_MEDIUM:
                 pipe = StableDiffusion3Pipeline.from_pretrained(model_id, torch_dtype=torch_dtype)
             elif model_type in [ModelType.FLUX_DEV, ModelType.FLUX_SCHNELL]:

examples/diffusers/quantization/requirements.txt

@@ -1,4 +1,5 @@
 cuda-python
+diffusers<=0.34.0
 nvtx
 onnx_graphsurgeon
 opencv-python>=4.8.1.78,<4.12.0.88

examples/llm_ptq/hf_ptq.py

@@ -722,7 +722,7 @@ def output_decode(generated_ids, input_shape):
     )
     parser.add_argument(
         "--verbose",
-        help="Print verbose output (e.g. quantization summary). Disable by --no_verbose.",
+        help="Print verbose output (e.g. quantization summary). Disable by --no-verbose.",
         default=True,
         action=argparse.BooleanOptionalAction,
     )