[Cherry-pick] PRs #1648 #1650 #1594 #1269 #1326 #1652 #1651 #1601 #1653 #1558 #1670 #1662 #1677 #1327 #1673 #1676 #1687 #1678 #1691 #1697 #1702 #1704 #1726 #1729

.github/actions/cache-extensions/action.yml

@@ -0,0 +1,24 @@
+name: Cache extensions
+description: Cache and reuse JIT-compiled extensions (e.g. CUDA extensions) across runs.
+
+inputs:
+  cache-key:
+    description: Environment discriminator for the cache key (e.g. GPU arch + container image).
+    required: true
+
+runs:
+  using: composite
+  steps:
+    - shell: bash
+      run: echo "TORCH_EXTENSIONS_DIR=/root/.cache/torch_extensions" >> "$GITHUB_ENV"
+    - id: cache
+      uses: actions/cache@v4
+      with:
+        path: /root/.cache/torch_extensions
+        key: torch-ext-${{ inputs.cache-key }}-${{ hashFiles('modelopt/torch/kernels/quantization/**', 'modelopt/torch/quantization/extensions.py', 'modelopt/torch/utils/cpp_extension.py')
+          }}
+    # On a cache hit, backdate sources below the cached objects so ninja reuses them (touching
+    # the objects instead would desync ninja's .ninja_deps and force a rebuild).
+    - if: steps.cache.outputs.cache-hit == 'true'
+      shell: bash
+      run: find modelopt/torch/kernels/quantization -exec touch -d '2000-01-01' {} +

.github/workflows/_example_tests_runner.yml

@@ -26,7 +26,7 @@ on:
         description: "GitHub runner to use"
         required: false
         type: string
-        default: "linux-amd64-gpu-h100-latest-1"
+        default: "linux-amd64-gpu-rtxpro6000-latest-1"
 
 jobs:
   run-test:
@@ -41,6 +41,9 @@ jobs:
     steps:
       - uses: actions/checkout@v6
       - uses: nv-gha-runners/setup-proxy-cache@main
+      - uses: ./.github/actions/cache-extensions
+        with:
+          cache-key: rtxpro6000-${{ inputs.docker_image }}
       - name: Setup environment variables
         run: |
           echo "LD_LIBRARY_PATH=${LD_LIBRARY_PATH}:/usr/include:/usr/lib/x86_64-linux-gnu:/usr/local/tensorrt/targets/x86_64-linux-gnu/lib" >> $GITHUB_ENV

.github/workflows/example_tests.yml

@@ -35,7 +35,7 @@ jobs:
     strategy:
       fail-fast: false
       matrix:
-        example: [diffusers_sparsity, gpt-oss, llm_distill, llm_qat, llm_sparsity, specdec_bench]
+        example: [gpt-oss, llm_distill, llm_qat, llm_sparsity, specdec_bench]
         include:
           - example: speculative_decoding
             docker_image: "26.01"

.github/workflows/gpu_tests.yml

@@ -45,7 +45,7 @@ jobs:
             timeout: 60
             container_image: nvcr.io/nvidia/nemo:26.04
           - example: gpu_trtllm
-            timeout: 30
+            timeout: 15
             container_image: nvcr.io/nvidia/tensorrt-llm/release:1.3.0rc17
           - example: gpu_vllm
             timeout: 15
@@ -66,6 +66,9 @@ jobs:
         run: apt-get update && apt-get install -y git
       - uses: actions/checkout@v6
       - uses: nv-gha-runners/setup-proxy-cache@main
+      - uses: ./.github/actions/cache-extensions
+        with:
+          cache-key: rtxpro6000-${{ matrix.container_image }}
       - name: Setup environment variables
         run: |
           echo "LD_LIBRARY_PATH=${LD_LIBRARY_PATH}:/usr/include:/usr/lib/x86_64-linux-gnu" >> $GITHUB_ENV

.github/workflows/unit_tests.yml

@@ -1,4 +1,3 @@
-# NOTE: Make sure this file is consistent with .gitlab/tests.yml
 name: Unit tests
 
 on:
@@ -73,6 +72,10 @@ jobs:
           token: ${{ secrets.CODECOV_TOKEN }}
           flags: unit
           fail_ci_if_error: true
+          # Skip GPG/SHASUM integrity check of the Codecov CLI: its key import
+          # intermittently fails (codecov/codecov-action#1876), which would
+          # otherwise hard-fail this required job on a transient infra blip.
+          skip_validation: true
           verbose: true
   windows:
     if: needs.check-file-changes.outputs.any_changed == 'true'

CHANGELOG.rst

@@ -24,16 +24,19 @@ Changelog
 **New Features**
 
 - Extend Claude Code agent skills for PTQ, deployment, evaluation, monitoring, and baseline-vs-quantized result comparison. Adds evaluation task references for additional benchmarks, stronger PTQ checkpoint validation gates, and session-scoped workspace/job tracking.
+- Add ``examples/alpamayo`` showing FP8, NVFP4, and AutoQuantize (mixed-precision) quantization of the Alpamayo (formerly Alpamayo-R1) ~10B vision-language-action model, with a joint VLM + diffusion calibration loop and both fake-quant and ``--real-quant`` packed-checkpoint export. See `examples/alpamayo/README.md <https://github.com/NVIDIA/Model-Optimizer/tree/main/examples/alpamayo>`_ for details.
 - Add SLURM Quality of Service (QoS) support to the ModelOpt launcher. Users can set QoS via ``slurm_config.qos`` or ``SLURM_QOS`` and the value is forwarded to ``nemo_run.SlurmExecutor``.
 - Add composable ``$import`` system for recipe YAML configs, enabling reusable config snippets referenced via ``{$import: name}`` markers. All built-in PTQ recipes converted to use imports with shared snippets under ``modelopt_recipes/configs/`` (numeric formats, quant_cfg building blocks, presets). See :ref:`composable-imports`.
 - Add offline DFlash speculative decoding training. Train the draft module from pre-computed base-model hidden states dumped by ``examples/speculative_decoding/collect_hidden_states/compute_hidden_states_hf.py``; base-model transformer layers are deleted after conversion to save memory. Controlled by the auto-derived ``dflash_offline`` flag on ``DFlashConfig`` (derived from ``data_args.offline_data_path``). The dump scripts now share ``collect_hidden_states/common.py`` for aux-layer selection (``--aux-layers eagle|dflash|<list>``) and optional assistant-token ``loss_mask`` for answer-only-loss training.
 - Add support for ``active_params`` (for MoE models) and ``memory_mb`` constraints in Minitron pruning on top of existing ``params`` constraint. You can also provide multiple constraints. See `examples/pruning/README.md <https://github.com/NVIDIA/Model-Optimizer/tree/main/examples/pruning>`_ for more details. The underlying utility functions ``mcore_param_count``, ``mcore_memory_footprint_mb``, and ``print_mcore_model_stats`` in ``modelopt.torch.nas.plugins.megatron_model_stats`` are also available for standalone use to compute parameter counts and memory footprints (weights + KV-cache + Mamba state) for any Megatron-Core model.
 - Add Minitron pruning support for Megatron-Bridge Gemma3 models.
 - Add quantization examples for the Megatron-Bridge framework: post-training quantization (`quantize.py <https://github.com/NVIDIA/Model-Optimizer/blob/main/examples/megatron_bridge/quantize.py>`_), export to a deployable HuggingFace checkpoint (`export.py <https://github.com/NVIDIA/Model-Optimizer/blob/main/examples/megatron_bridge/export.py>`_), and Quantization Aware Distillation (extend existing `distill.py <https://github.com/NVIDIA/Model-Optimizer/blob/main/examples/megatron_bridge/distill.py>`_).
-- Add end-to-end tutorial for Minitron pruning + two-phase distillation (80B @ 8K + 20B @ 32K long-context = 100B tokens) + FP8 PTQ + vLLM deployment for Nemotron-3-Nano-30B-A3B-BF16 (MoE + Mamba-Transformer hybrid) → Pruned 22B/A3.0B active params, along with data blend preparation steps (with tool-calling data) and detailed pruning / data-blend / long-context ablations. See `examples/pruning/minitron/NVIDIA-Nemotron-3-Nano-30B-A3B-BF16/README.md <https://github.com/NVIDIA/Model-Optimizer/tree/main/examples/pruning/minitron/NVIDIA-Nemotron-3-Nano-30B-A3B-BF16/>`_ for details.
+- Add end-to-end optimization tutorial for Minitron pruning + two-phase distillation (80B @ 8K + 20B @ 32K long-context = 100B tokens) + FP8 PTQ + vLLM deployment for Nemotron-3-Nano-30B-A3B-BF16 (MoE + Mamba-Transformer hybrid) → Pruned 22B/A3.0B active params, along with data blend preparation steps (with tool-calling data) and detailed pruning / data-blend / long-context ablations. See `examples/megatron_bridge/tutorials/NVIDIA-Nemotron-3-Nano-30B-A3B-BF16/README.md <https://github.com/NVIDIA/Model-Optimizer/tree/main/examples/megatron_bridge/tutorials/NVIDIA-Nemotron-3-Nano-30B-A3B-BF16/>`_ for details.
 - Add ``--cast_mxfp4_to_nvfp4`` flag to ``examples/llm_ptq/hf_ptq.py`` for closed-form, bit-exact MXFP4 → NVFP4 weight conversion. Supports the GPT-OSS family (``openai/gpt-oss-20b``, ``openai/gpt-oss-120b``). See `examples/llm_ptq/README.md <https://github.com/NVIDIA/Model-Optimizer/tree/main/examples/llm_ptq#mxfp4--nvfp4-cast-for-gpt-oss>`__ for usage.
+- Add ``--cast_mxfp4_to_nvfp4`` flag to ``examples/deepseek/deepseek_v4/quantize_to_nvfp4.py`` for closed-form, bit-exact MXFP4 → NVFP4 conversion of DeepSeek V4 routed-expert weights (mirrors the GPT-OSS cast; w1/w3 share one per-tensor ``scale_2`` for the fused GEMM1). Activation ``input_scale`` still comes from ``--amax_path`` calibration.
 - DeepSeek PTQ (``examples/deepseek/ptq.py``) now defaults to native top-k calibration with post-hoc per-layer peer-max sync of expert ``input_quantizer.amax``; the all-experts path is preserved behind ``--calib_all_experts``.
 - Add NVFP4 W4A16 weight-only quantization (``w4a16_nvfp4``): FP4 weights with group_size=16, BF16 activations, no calibration forward pass required. Use ``mtq.W4A16_NVFP4_CFG`` or ``--qformat w4a16_nvfp4`` in ``hf_ptq.py``. vLLM deployment support is in progress.
+- Add FP8 KV-cache cast variants for the partial-NVFP4 and weight-only general PTQ recipes: ``general/ptq/nvfp4_mlp_only-kv_fp8_cast``, ``general/ptq/nvfp4_experts_only-kv_fp8_cast``, ``general/ptq/nvfp4_omlp_only-kv_fp8_cast``, and ``general/ptq/nvfp4_weight_only-kv_fp8_cast``. These compose the same model-quant configs as their ``-kv_fp8`` siblings with the ``kv_fp8_cast`` unit (constant-amax FP8 KV cache, no KV calibration forward pass).
 - Add Megatron Core export/import mapping for Qwen3-VL (``Qwen3VLForConditionalGeneration``) vision-language models. The mapping handles the ``model.language_model.`` weight prefix used by Qwen3-VL.
 - Add active-MoE cost accounting for ``mtq.auto_quantize`` effective-bits search. Set ``constraints={"effective_bits": ..., "cost_model": "active_moe", "cost": {"active_moe_expert_ratio": ...}}`` to weight routed MoE expert costs by active experts per token while keeping shared experts fully counted. The ``hf_ptq.py`` AutoQuant path exposes this via ``--auto_quantize_cost_model active_moe`` and ``--auto_quantize_active_moe_expert_ratio``.
 - Add ``DATASET_COMBOS`` to ``modelopt.torch.utils.dataset_utils`` — single ``--dataset`` tokens that fan out to multiple registered datasets; per-entry ``num_samples`` is split evenly across the members. Initial combos: ``cnn_nemotron_v2_mix`` (``cnn_dailymail`` + ``nemotron-post-training-dataset-v2``, used by ``hf_ptq.py`` when no ``--dataset`` is provided) and ``nemotron-post-training-v3`` (the seven ``nvidia/Nemotron-*`` SFT datasets added in #1498, mirroring the `nemotron-post-training-v3 collection <https://huggingface.co/collections/nvidia/nemotron-post-training-v3>`_). Combo names are listed by ``get_supported_datasets()`` and surfaced in ``--dataset`` help. ``get_dataset_dataloader`` rejects inputs that mix a combo with one of its member datasets (e.g. ``cnn_dailymail,cnn_nemotron_v2_mix``) to avoid double-sampling, and ``get_dataset_samples`` rejects combo names so callers route through the dataloader. ``hf_ptq.py`` default ``--calib_size`` is bumped from ``512`` to ``1024`` so the total calibration sample count under the new default combo matches the previous two-dataset fallback.
@@ -50,12 +53,16 @@ Changelog
   The legacy FSDP1 accelerate config is removed; ``llm_qat`` now documents FSDP2, DeepSpeed, and DDP backends.
 - The PTQ example scripts ``examples/llm_ptq/hf_ptq.py``, ``examples/llm_ptq/multinode_ptq.py`` and ``examples/megatron_bridge/quantize.py`` now derive their ``--qformat`` / ``--kv_cache_qformat`` (``--quant_cfg`` / ``--kv_cache_quant`` for Megatron-Bridge) CLI vocabularies by discovering the YAML presets under ``modelopt_recipes/configs/ptq/presets/{model,kv}/`` rather than carrying hardcoded ``QUANT_CFG_CHOICES`` / ``KV_QUANT_CFG_CHOICES`` tables. The discovery helper, alias table and ready-built ``QUANT_CFG_CHOICES`` / ``KV_QUANT_CFG_CHOICES`` mappings now live in ``modelopt.recipe.presets`` and are shared by all three scripts. Presets are loaded eagerly into a plain dict at import. Adding a new preset YAML makes it available on the CLI of all three with no script change — note this means each script now accepts every preset under those directories, not just a previously curated subset. All previously-supported short names (``int8_sq``, ``nvfp4_awq``, ``fp8_pb_wo``, ``nvfp4_mse``, ``w4a8_awq``, ``nvfp4_local_hessian``, ``fp8_pc_pt``, ``int8_wo``) keep working via a small deprecation alias table; new formats should be exposed as preset YAMLs (or, longer term, as full ``--recipe`` recipes).
 - Add ``configs/ptq/presets/kv/fp8_cast.yaml`` and ``configs/ptq/presets/kv/nvfp4_cast.yaml``, promoting ``fp8_cast`` / ``nvfp4_cast`` to first-class KV presets composed from the existing ``kv_fp8_cast`` / ``kv_nvfp4_cast`` unit fragments. The previous runtime ``use_constant_amax`` post-edit in ``hf_ptq.py`` is removed; ``use_constant_amax: true`` now lives in the YAML and is therefore authoritative. **Custom (out-of-tree) recipes that target a cast KV format must set ``use_constant_amax: true`` themselves on the ``[kv]_bmm_quantizer`` config** — in-tree recipes already do via the ``kv_*_cast`` units.
+- Add DMD2 distillation for few-step diffusion models in ``examples/diffusers/fastgen/``: distill Qwen-Image into a 4/8-step student via Distribution Matching Distillation. See `examples/diffusers/fastgen/README.md <https://github.com/NVIDIA/Model-Optimizer/tree/main/examples/diffusers/fastgen>`_ for details.
 
 **Bug Fixes**
 
 - In Megatron-Core only do EP amax sync for routed expert weights if ``sync_expert_weight_amax=True``. Previously EP amax sync would sync routed expert weights across EP ranks even when ``sync_expert_weight_amax`` was False.
 - Fix Megatron-Core HF importer to load fused ``TELayerNormColumnParallelLinear.layer_norm_weight`` from HF for GPT-family models (Qwen3 etc.) under ``--export-default-te-spec``. Importer now prefers per-context keys ``fused_input_layernorm`` / ``fused_pre_mlp_layernorm`` (fallback ``fused_norm`` for Nemotron-H backward compatibility); ``mcore_qwen.py`` provides the new rules. Without this fix, post-prune MMLU sat at chance.
 - Fix ONNX AutoCast ``keep_io_types=True`` sanity-check failure (``Unexpected type in I/O tensor ...``) when a network input/output is an empty tensor (a dimension of size 0). Such tensors were "fake-cast" (retyped in place) to the low precision type; because the value-info map aliases the ``graph.input``/``graph.output`` ``ValueInfoProto``, this silently changed the model's I/O type. AutoCast now inserts a real ``Cast`` for protected I/O tensors instead.
+- Fix INT8 entropy calibration of fp16 ONNX models raising ``ValueError: Too many bins for data range`` on numpy >= 2.0. ``_collect_value`` in ``modelopt.onnx.quantization.ort_patching`` now casts the histogram range endpoints to Python float so bin edges are computed in float64, instead of inheriting the fp16 dtype of an activation tensor with a small range (which collapsed the 128-bin linspace under NEP-50 promotion).
+- Fix the GPT-OSS MXFP4 → NVFP4 PTQ path in ``examples/llm_ptq/hf_ptq.py`` (used with ``--cast_mxfp4_to_nvfp4``). ``get_model`` now loads native MXFP4 checkpoints (``openai/gpt-oss-*``) dequantized to BF16 ``GptOssExperts`` via ``Mxfp4Config(dequantize=True)`` on a sequential device map. This fixes a CUDA illegal-memory access during the multi-GPU dequant load and the ``NotImplementedError`` for experts type ``Mxfp4GptOssExperts`` during unified HF export (the packed-kernel experts wrapper, used when the optional ``kernels`` package is installed, is unsupported by export); ``kernels`` is no longer required. The ``--cast_mxfp4_to_nvfp4`` step now also resolves a HF Hub ID ``--pyt_ckpt_path`` to its local snapshot directory instead of failing with ``FileNotFoundError``.
+- Fix ``_QuantGptOssExperts`` / ``_QuantLlama4TextExperts`` static-block NVFP4 weight calibration raising ``ValueError: Input shape has changed`` during the calibration forward. These experts quantize their weights transposed (``_transposed_quantize``); ``iter_weights_for_calibration`` now yields the same transposed view so weight-only calibration and the forward agree on the block-quant shape (and the export ``_amax`` orientation).
 
 **Deprecations**
 
@@ -74,7 +81,7 @@ Changelog
 - Add Puzzletron - a new algorithm for heterogeneous pruning of LLM and VLM models. See `examples/puzzletron/README.md <https://github.com/NVIDIA/Model-Optimizer/tree/main/examples/puzzletron>`_ for more details.
 - Added iterator interface using CalibrationDataReader in ONNX quantization workflow.
 - Add N:M sparse softmax support to the Triton flash attention kernel (``modelopt.torch.kernels.common.attention.triton_fa``). See `examples/llm_sparsity/attention_sparsity/README.md <https://github.com/NVIDIA/Model-Optimizer/tree/main/examples/llm_sparsity/attention_sparsity>`_ for usage.
-- Add skip-softmax skipping to the Triton flash attention kernel (``modelopt.torch.kernels.common.attention.triton_fa``). See `examples/llm_sparsity/attention_sparsity/README.md <https://github.com/NVIDIA/Model-Optimizer/tree/main/examples/llm_sparsity/attention_sparsity>`_ for usage.
+- Add skip-softmax skipping to the Triton flash attention kernel for both language models and video diffusion models (``modelopt.torch.kernels.common.attention.triton_fa``). See `examples/llm_sparsity/attention_sparsity/README.md <https://github.com/NVIDIA/Model-Optimizer/tree/main/examples/llm_sparsity/attention_sparsity>`_ and `examples/diffusers/sparsity/ <https://github.com/NVIDIA/Model-Optimizer/tree/main/examples/diffusers/sparsity>`_ for usage.
 - Add Video Sparse Attention (VSA) method for video diffusion models (``modelopt.torch.sparsity.attention_sparsity``). VSA uses 3D block tiling with a two-branch architecture for attention speedup.
 - Enable PTQ workflow for the Step3.5-Flash MoE model with NVFP4 W4A4 + FP8 KV cache quantization. See `modelopt_recipes/models/Step3.5-Flash/nvfp4-mlp-only.yaml <https://github.com/NVIDIA/Model-Optimizer/blob/main/modelopt_recipes/models/Step3.5-Flash/nvfp4-mlp-only.yaml>`_ for more details.
 - Add support for vLLM fakequant reload using ModelOpt state for HF models. See `examples/vllm_serve/README.md <https://github.com/NVIDIA/Model-Optimizer/tree/main/examples/vllm_serve#load-qatptq-model-and-serve-in-vllm-wip>`_ for more details.

README.md

@@ -26,7 +26,7 @@ Model Optimizer is also integrated with [NVIDIA Megatron-Bridge](https://github.
 
 ## Latest News
 
-- [2026/05/27] [**End-to-end Minitron workflow for Nemotron-3-Nano-30B-A3B**](./examples/pruning/minitron/NVIDIA-Nemotron-3-Nano-30B-A3B-BF16): Pruning + two-phase distillation + FP8 quantization achieving 1.64× vLLM throughput and 2.6× memory reduction.
+- [2026/05/27] [**End-to-end optimization tutorial for Nemotron-3-Nano-30B-A3B**](./examples/megatron_bridge/tutorials/NVIDIA-Nemotron-3-Nano-30B-A3B-BF16): Pruning + distillation (with long context extension) + FP8 quantization achieving 2.6× vLLM throughput and 2.6× memory reduction.
 - [2026/05/13] [**Puzzletron**](./examples/puzzletron): A new algorithm for heterogeneous pruning & NAS of LLM and VLM models.
 - [2026/04/15] Customer story: [Domyn compresses Colosseum-355B → 260B using ModelOpt's Minitron pruning + distillation](https://www.domyn.com/blog/domyn-large-the-journey-of-a-european-sovereign-ai-model-for-regulated-industries)
 - [2026/03/17] Customer story: [Bielik.AI builds Bielik Minitron 7B (33% smaller, 50% faster, 90% quality retained) using ModelOpt's Minitron pruning + distillation](https://bielik.ai/en/nvidia-gtc-bielik-minitron-premiere/)