Merge branch 'nv_main' of v2.12

build_tools/VERSION.txt

-2.11.0.dev0
+2.13.0.dev0

build_tools/__init__.py

-# Copyright (c) 2022-2025, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+# Copyright (c) 2022-2026, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
 #
 # See LICENSE for license information.
 

build_tools/build_ext.py

-# Copyright (c) 2022-2025, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+# Copyright (c) 2022-2026, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
 #
 # See LICENSE for license information.
 
@@ -63,6 +63,12 @@ class CMakeExtension(setuptools.Extension):
             f"-DCMAKE_BUILD_TYPE={build_type}",
             f"-DCMAKE_INSTALL_PREFIX={install_dir}",
         ]
+        if bool(int(os.getenv("NVTE_USE_CCACHE", "0"))):
+            ccache_bin = os.getenv("NVTE_CCACHE_BIN", "ccache")
+            configure_command += [
+                f"-DCMAKE_CXX_COMPILER_LAUNCHER={ccache_bin}",
+                f"-DCMAKE_CUDA_COMPILER_LAUNCHER={ccache_bin}",
+            ]
         configure_command += self.cmake_flags
 
         import pybind11

build_tools/jax.py

-# Copyright (c) 2022-2025, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+# Copyright (c) 2022-2026, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
 #
 # See LICENSE for license information.
 
@@ -19,8 +19,29 @@ def install_requirements() -> List[str]:
 
 
 def test_requirements() -> List[str]:
-    """Test dependencies for TE/JAX extensions."""
-    return ["numpy"]
+    """Test dependencies for TE/JAX extensions.
+
+    Triton Package Selection:
+        The triton package is selected based on NVTE_USE_PYTORCH_TRITON environment variable:
+
+        Default (NVTE_USE_PYTORCH_TRITON unset or "0"):
+            Returns 'triton' - OpenAI's standard package from PyPI.
+            Install with: pip install triton
+
+        NVTE_USE_PYTORCH_TRITON=1:
+            Returns 'pytorch-triton' - for mixed JAX+PyTorch environments.
+            Install with: pip install pytorch-triton --index-url https://download.pytorch.org/whl/cu121
+
+            Note: Do NOT install pytorch-triton from PyPI directly - that's a placeholder.
+    """
+    use_pytorch_triton = bool(int(os.environ.get("NVTE_USE_PYTORCH_TRITON", "0")))
+
+    triton_package = "pytorch-triton" if use_pytorch_triton else "triton"
+
+    return [
+        "numpy",
+        triton_package,
+    ]
 
 
 def xla_path() -> str:

build_tools/pytorch.py

-# Copyright (c) 2022-2025, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+# Copyright (c) 2022-2026, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
 #
 # See LICENSE for license information.
 
@@ -14,12 +14,17 @@ from typing import List
 
 def install_requirements() -> List[str]:
     """Install dependencies for TE/PyTorch extensions."""
-    return ["torch>=2.1", "einops"] # "onnxscript==0.3.1", "onnx"]
+    return ["torch>=2.1", "einops", "packaging", "pydantic", "nvdlfw-inspect"]
 
 
 def test_requirements() -> List[str]:
-    """Test dependencies for TE/JAX extensions."""
-    return ["numpy", "torchvision", "transformers", "torchao==0.13"]
+    """Test dependencies for TE/PyTorch extensions."""
+    return [
+        "numpy",
+        "torchvision",
+        "transformers",
+        "torchao==0.13",
+    ]
 
 
 def setup_pytorch_extension(

build_tools/te_version.py

-# Copyright (c) 2022-2025, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+# Copyright (c) 2022-2026, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
 #
 # See LICENSE for license information.
 

build_tools/utils.py

-# Copyright (c) 2022-2025, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+# Copyright (c) 2022-2026, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
 #
 # See LICENSE for license information.
 
@@ -279,13 +279,9 @@ def get_cuda_include_dirs() -> Tuple[str, str]:
 
     cuda_root = Path(nvidia.__file__).parent
     return [
-        cuda_root / "cuda_nvcc" / "include",
-        cuda_root / "cublas" / "include",
-        cuda_root / "cuda_runtime" / "include",
-        cuda_root / "cudnn" / "include",
-        cuda_root / "cuda_cccl" / "include",
-        cuda_root / "nvtx" / "include",
-        cuda_root / "cuda_nvrtc" / "include",
+        subdir / "include"
+        for subdir in cuda_root.iterdir()
+        if subdir.is_dir() and (subdir / "include").is_dir()
     ]
 
 

build_tools/wheel_utils/Dockerfile.aarch

-# Copyright (c) 2022-2025, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+# Copyright (c) 2022-2026, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
 #
 # See LICENSE for license information.
 

build_tools/wheel_utils/Dockerfile.x86

-# Copyright (c) 2022-2025, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+# Copyright (c) 2022-2026, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
 #
 # See LICENSE for license information.
 

build_tools/wheel_utils/build_wheels.sh

-# Copyright (c) 2022-2025, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+# Copyright (c) 2022-2026, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
 #
 # See LICENSE for license information.
 

build_tools/wheel_utils/launch_aarch.sh

-# Copyright (c) 2022-2025, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+# Copyright (c) 2022-2026, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
 #
 # See LICENSE for license information.
 

build_tools/wheel_utils/launch_x86.sh

-# Copyright (c) 2022-2025, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+# Copyright (c) 2022-2026, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
 #
 # See LICENSE for license information.
 

docs/_static/css/output-style.css

+/* Custom styling for program output blocks */
+
+.program-output {
+    background-color: #f8f9fa;
+    padding: 0;  /* No padding at all */
+    margin: 0;  /* No margins at all */
+    border-radius: 0;  /* No rounded corners */
+    font-family: 'Courier New', monospace;
+    font-size: 14px;
+    line-height: 1.5;
+    width: 100%;
+    max-width: 100%;
+}
+
+.program-output pre {
+    margin: 0;
+    padding: 0;
+    background: transparent !important;
+    border: none !important;
+    color: #2c3e50;
+    width: 100%;
+}
+
+.program-output .highlight {
+    background: transparent !important;
+    margin: 0;
+    width: 100%;
+}
+
+/* Alternative lighter style */
+.output-block {
+    background-color: #fafbfc;
+    border: 1px solid #e1e4e8;
+    padding: 10px 14px;
+    margin: 10px 0;
+    border-radius: 3px;
+    font-family: 'SF Mono', 'Consolas', monospace;
+    font-size: 13px;
+    color: #24292e;
+}
+
+/* Console-like output style */
+.console-output {
+    background-color: #1e1e1e;
+    border-left: 3px solid #76b900;
+    padding: 14px 18px;
+    margin: 12px 0;
+    border-radius: 5px;
+    font-family: 'Fira Code', 'Consolas', monospace;
+    font-size: 13px;
+    color: #d4d4d4;
+    box-shadow: 0 2px 4px rgba(0,0,0,0.1);
+}
+
+.console-output pre {
+    margin: 0;
+    color: #d4d4d4;
+    background: transparent !important;
+}
+

docs/_static/css/rtabs.css

+/* Custom styling for sphinx-tabs */
+
+.sphinx-tabs {
+    margin-bottom: 1rem;
+}
+
+.sphinx-tabs-tab {
+    background-color: #f4f4f4;
+    border: 1px solid #ccc;
+    border-bottom: none;
+    padding: 0.5rem 1rem;
+    margin-right: 0.5rem;
+    cursor: pointer;
+    font-weight: 500;
+    transition: background-color 0.2s;
+}
+
+.sphinx-tabs-tab:hover {
+    background-color: #e0e0e0;
+}
+
+.sphinx-tabs-tab[aria-selected="true"] {
+    background-color: #76b900; /* NVIDIA green */
+    color: white;
+    border-color: #76b900;
+    margin-right: 0.5rem;
+}
+
+.sphinx-tabs-panel {
+    border: 1px solid #ccc;
+    padding: 1rem;
+    background-color: #f9f9f9;
+}
+
+/* Dark mode support for RTD theme */
+.rst-content .sphinx-tabs-tab {
+    color: #333;
+}
+
+.rst-content .sphinx-tabs-tab[aria-selected="true"] {
+    color: white;
+}
+

docs/api/c/activation.rst

 ..
-    Copyright (c) 2022-2025, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+    Copyright (c) 2022-2026, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
 
     See LICENSE for license information.
 

docs/api/c/cast.rst

 ..
-    Copyright (c) 2022-2025, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+    Copyright (c) 2022-2026, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
 
     See LICENSE for license information.
 

docs/api/c/cast_transpose_noop.rst

 ..
-    Copyright (c) 2022-2025, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+    Copyright (c) 2022-2026, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
 
     See LICENSE for license information.
 

docs/api/c/cudnn.rst

 ..
-    Copyright (c) 2022-2025, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+    Copyright (c) 2022-2026, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
 
     See LICENSE for license information.
 

docs/api/c/fused_attn.rst

 ..
-    Copyright (c) 2022-2025, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+    Copyright (c) 2022-2026, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
 
     See LICENSE for license information.
 

docs/api/c/fused_rope.rst

 ..
-    Copyright (c) 2022-2025, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
+    Copyright (c) 2022-2026, NVIDIA CORPORATION & AFFILIATES. All rights reserved.
 
     See LICENSE for license information.