Remove deprecated examples/llm_autodeploy

.agents/skills/deployment/references/trtllm.md

@@ -42,46 +42,24 @@ llm = LLM(model="<checkpoint_path>", tensor_parallel_size=4)
 
 ## AutoDeploy (for AutoQuant / mixed-precision)
 
-AutoDeploy automates graph transformations for optimized inference. Required for AutoQuant checkpoints.
+AutoDeploy automates graph transformations for optimized inference and is useful for
+AutoQuant / mixed-precision checkpoints. The standalone `examples/llm_autodeploy` example
+was removed in 0.46; use TensorRT-LLM's
+[AutoDeploy](https://github.com/NVIDIA/TensorRT-LLM/tree/main/examples/auto_deploy)
+directly together with a ModelOpt-quantized checkpoint.
 
-### End-to-end script
+### Workflow
 
-```bash
-# Quantize and deploy in one step
-./examples/llm_autodeploy/scripts/run_auto_quant_and_deploy.sh \
-    --hf_ckpt <model_path> \
-    --save_quantized_ckpt <output_path> \
-    --quant fp8,nvfp4 \
-    --effective_bits 4.5
-```
-
-Parameters:
-
-- `--hf_ckpt`: Path to unquantized HuggingFace checkpoint
-- `--save_quantized_ckpt`: Output path for quantized checkpoint
-- `--quant`: Quantization formats (e.g., `fp8,nvfp4`)
-- `--effective_bits`: Target precision (higher = more accuracy for sensitive layers)
-- `--world_size`: Number of GPUs for tensor parallelism
-- `--calib_batch_size`: Calibration batch size (reduce if OOM, default 8)
-
-### AutoDeploy API server
-
-```python
-# examples/llm_autodeploy/api_server.py provides a FastAPI server
-# with OpenAI-compatible endpoints using AutoDeploy
-```
-
-### Test AutoDeploy
-
-```bash
-python examples/llm_autodeploy/api_client.py --prompt "What is AI?" "What is golf?"
-```
+1. Quantize the checkpoint with ModelOpt PTQ (including AutoQuant / mixed precision) via
+   `examples/llm_ptq` (`hf_ptq.py` / `scripts/huggingface_example.sh`), which produces a
+   unified HuggingFace checkpoint with `hf_quant_config.json`.
+2. Deploy that checkpoint with TensorRT-LLM's AutoDeploy backend (see the upstream
+   `examples/auto_deploy` docs for the current API and `trtllm-serve` flags).
 
 ### Notes
 
-- NVFP4 in AutoDeploy requires Blackwell GPUs
-- For Hopper: remove `nvfp4` from `--quant` and set `--effective_bits` above 8.0
-- AutoDeploy supports CUDA graphs, torch compile backends, and KV cache optimization
+- NVFP4 in AutoDeploy requires Blackwell GPUs; on Hopper use FP8 instead.
+- AutoDeploy supports CUDA graphs, torch compile backends, and KV cache optimization.
 
 ## Legacy TRT-LLM Checkpoint (deprecated)
 

.agents/skills/deployment/scripts/deploy.sh

@@ -337,12 +337,10 @@ start_trtllm() {
 
     cat <<TRTEOF
 
-# Option 1: AutoDeploy (recommended)
-./examples/llm_autodeploy/scripts/run_auto_quant_and_deploy.sh \\
-    --hf_ckpt "$MODEL" \\
-    --save_quantized_ckpt <output_path> \\
-    --quant fp8,nvfp4 \\
-    --effective_bits 4.5
+# Option 1: AutoDeploy (recommended for AutoQuant / mixed-precision)
+# Quantize with ModelOpt PTQ (examples/llm_ptq) to produce a unified HF checkpoint,
+# then deploy it with TensorRT-LLM's AutoDeploy backend:
+#   https://github.com/NVIDIA/TensorRT-LLM/tree/main/examples/auto_deploy
 
 # Option 2: Python API
 python3 -c "

.github/CODEOWNERS

@@ -46,7 +46,6 @@ modelopt_recipes @NVIDIA/modelopt-recipes-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

.github/workflows/example_tests.yml

@@ -48,7 +48,7 @@ jobs:
       pip_install_extras: "[hf,dev-test]"
       runner: ${{ startsWith(github.ref, 'refs/heads/pull-request/') && 'linux-amd64-gpu-rtxpro6000-latest-1' || 'linux-amd64-gpu-rtxpro6000-latest-2' }}
 
-  ##### TensorRT-LLM Example Tests (pr/non-pr split: non-pr runs extra autodeploy+eval examples) #####
+  ##### TensorRT-LLM Example Tests (pr/non-pr split: non-pr runs extra eval examples) #####
   trtllm-pr:
     needs: [pr-gate]
     if: startsWith(github.ref, 'refs/heads/pull-request/') && needs.pr-gate.outputs.any_changed == 'true'
@@ -69,7 +69,7 @@ jobs:
     strategy:
       fail-fast: false
       matrix:
-        example: [llm_autodeploy, llm_eval, llm_ptq]
+        example: [llm_eval, llm_ptq]
     uses: ./.github/workflows/_example_tests_runner.yml
     secrets: inherit
     with:

CHANGELOG.rst

@@ -7,6 +7,7 @@ Changelog
 **Backward Breaking Changes**
 
 - Remove the ``examples/diffusers/eval`` image-quality evaluation example (ImageReward / CLIP-IQA / CLIP metrics) and its references in ``examples/diffusers/README.md``. The example was deprecated in 0.45 and is no longer maintained.
+- Remove the deprecated ``examples/llm_autodeploy`` example (deprecated in 0.45). Use TensorRT-LLM's `AutoDeploy <https://github.com/NVIDIA/TensorRT-LLM/tree/main/examples/auto_deploy>`_ directly together with ModelOpt PTQ in ``examples/llm_ptq``.
 
 **Deprecations**