NVIDIA · juney-nvidia · Feb 2, 2025 · Feb 2, 2025
diff --git a/cpp/CMakeLists.txt b/cpp/CMakeLists.txt
@@ -567,11 +567,35 @@ print(os.path.dirname(torch.__file__),end='');"
     endif()
   endif()
 
-elseif(NOT WIN32)
-  if(NOT USE_CXX11_ABI)
-    add_compile_options("-D_GLIBCXX_USE_CXX11_ABI=0")
+else()
+  if(NOT "all" IN_LIST CMAKE_CUDA_ARCHITECTURES
+     AND NOT "all-major" IN_LIST CMAKE_CUDA_ARCHITECTURES
+     AND NOT "native" IN_LIST CMAKE_CUDA_ARCHITECTURES)
+    foreach(CUDA_ARCH IN LISTS CMAKE_CUDA_ARCHITECTURES)
+      if(CUDA_ARCH MATCHES "^([1-9][0-9][0-9]?a?)(-real)?$")
+        string(
+          APPEND
+          CMAKE_CUDA_FLAGS
+          " --generate-code=arch=compute_${CMAKE_MATCH_1},code=sm_${CMAKE_MATCH_1}"
+        )
+      endif()
+      if(CUDA_ARCH MATCHES "^([1-9][0-9][0-9]?a?)(-virtual)?$")
+        string(
+          APPEND
+          CMAKE_CUDA_FLAGS
+          " --generate-code=arch=compute_${CMAKE_MATCH_1},code=compute_${CMAKE_MATCH_1}"
+        )
+      endif()
+    endforeach()
+    set(CMAKE_CUDA_ARCHITECTURES OFF)
+  endif()
+
+  if(NOT WIN32)
+    if(NOT USE_CXX11_ABI)
+      add_compile_options("-D_GLIBCXX_USE_CXX11_ABI=0")
+    endif()
+    message(STATUS "Build without PyTorch, USE_CXX11_ABI=${USE_CXX11_ABI}")
   endif()
-  message(STATUS "Build without PyTorch, USE_CXX11_ABI=${USE_CXX11_ABI}")
 endif()
 
 # Defer UCX/UCXX setup until after USE_CXX11_ABI is well defined, as UCXX will

diff --git a/cpp/tensorrt_llm/batch_manager/aarch64-linux-gnu/libtensorrt_llm_batch_manager_static.a b/cpp/tensorrt_llm/batch_manager/aarch64-linux-gnu/libtensorrt_llm_batch_manager_static.a
diff --git a/...orrt_llm/batch_manager/aarch64-linux-gnu/libtensorrt_llm_batch_manager_static.pre_cxx11.a b/...orrt_llm/batch_manager/aarch64-linux-gnu/libtensorrt_llm_batch_manager_static.pre_cxx11.a
diff --git a/cpp/tensorrt_llm/batch_manager/aarch64-linux-gnu/libtensorrt_llm_ucx_wrapper.so b/cpp/tensorrt_llm/batch_manager/aarch64-linux-gnu/libtensorrt_llm_ucx_wrapper.so
diff --git a/cpp/tensorrt_llm/batch_manager/aarch64-linux-gnu/version.txt b/cpp/tensorrt_llm/batch_manager/aarch64-linux-gnu/version.txt
@@ -1,2 +1,2 @@
-259d50e8c0a765c5a594e74b0c3c0b05 libtensorrt_llm_ucx_wrapper.so
-09df54c0cc99354a60bbc0303e3e8ea33a96bef0 commit
+ca50ae76421863dfebf6080b7f4f6b29 libtensorrt_llm_ucx_wrapper.so
+f8c0381a2bc50ee2739c3d8c2be481b31e5f00bd commit
diff --git a/cpp/tensorrt_llm/batch_manager/x86_64-linux-gnu/libtensorrt_llm_batch_manager_static.a b/cpp/tensorrt_llm/batch_manager/x86_64-linux-gnu/libtensorrt_llm_batch_manager_static.a
diff --git a/...sorrt_llm/batch_manager/x86_64-linux-gnu/libtensorrt_llm_batch_manager_static.pre_cxx11.a b/...sorrt_llm/batch_manager/x86_64-linux-gnu/libtensorrt_llm_batch_manager_static.pre_cxx11.a
diff --git a/cpp/tensorrt_llm/batch_manager/x86_64-linux-gnu/libtensorrt_llm_ucx_wrapper.so b/cpp/tensorrt_llm/batch_manager/x86_64-linux-gnu/libtensorrt_llm_ucx_wrapper.so
diff --git a/cpp/tensorrt_llm/batch_manager/x86_64-linux-gnu/version.txt b/cpp/tensorrt_llm/batch_manager/x86_64-linux-gnu/version.txt
@@ -1,2 +1,2 @@
-09c97d36814ca687352818d4c18aa5a7 libtensorrt_llm_ucx_wrapper.so
-09df54c0cc99354a60bbc0303e3e8ea33a96bef0 commit
+0ec83a0451530fcf0e3f325cdc185043 libtensorrt_llm_ucx_wrapper.so
+f8c0381a2bc50ee2739c3d8c2be481b31e5f00bd commit
diff --git a/cpp/tensorrt_llm/batch_manager/x86_64-windows-msvc/version.txt b/cpp/tensorrt_llm/batch_manager/x86_64-windows-msvc/version.txt
@@ -1,2 +1,2 @@
 41725f3b40ca44954bb9de6e7dcbfd2c tensorrt_llm_batch_manager_static.lib
-09df54c0cc99354a60bbc0303e3e8ea33a96bef0 commit
+f8c0381a2bc50ee2739c3d8c2be481b31e5f00bd commit
diff --git a/cpp/tensorrt_llm/executor/aarch64-linux-gnu/libtensorrt_llm_executor_static.a b/cpp/tensorrt_llm/executor/aarch64-linux-gnu/libtensorrt_llm_executor_static.a
diff --git a/cpp/tensorrt_llm/executor/aarch64-linux-gnu/libtensorrt_llm_executor_static.pre_cxx11.a b/cpp/tensorrt_llm/executor/aarch64-linux-gnu/libtensorrt_llm_executor_static.pre_cxx11.a
diff --git a/cpp/tensorrt_llm/executor/aarch64-linux-gnu/version.txt b/cpp/tensorrt_llm/executor/aarch64-linux-gnu/version.txt
@@ -1,3 +1,3 @@
-a79f07fe3282894afea78ecb23511bde libtensorrt_llm_executor_static.a
-d002feda6db1c3ccd8ba4183122a8414 libtensorrt_llm_executor_static.pre_cxx11.a
-09df54c0cc99354a60bbc0303e3e8ea33a96bef0 commit
+15c05b1921f3f8cbb7bc1e53f189c661 libtensorrt_llm_executor_static.a
+b586d90eac1293ea656ff2db8a35cd92 libtensorrt_llm_executor_static.pre_cxx11.a
+f8c0381a2bc50ee2739c3d8c2be481b31e5f00bd commit
diff --git a/cpp/tensorrt_llm/executor/x86_64-linux-gnu/libtensorrt_llm_executor_static.a b/cpp/tensorrt_llm/executor/x86_64-linux-gnu/libtensorrt_llm_executor_static.a
diff --git a/cpp/tensorrt_llm/executor/x86_64-linux-gnu/libtensorrt_llm_executor_static.pre_cxx11.a b/cpp/tensorrt_llm/executor/x86_64-linux-gnu/libtensorrt_llm_executor_static.pre_cxx11.a
diff --git a/cpp/tensorrt_llm/executor/x86_64-linux-gnu/version.txt b/cpp/tensorrt_llm/executor/x86_64-linux-gnu/version.txt
@@ -1,3 +1,3 @@
-1b610c42593967293eabcd0323a2b879 libtensorrt_llm_executor_static.a
-116fc9bd13edd0c1187b12bbdedf9873 libtensorrt_llm_executor_static.pre_cxx11.a
-09df54c0cc99354a60bbc0303e3e8ea33a96bef0 commit
+d573456348a96fd7a97aa832f1113986 libtensorrt_llm_executor_static.a
+b3ba7776e3b5bb6e750e8412fc3b5c60 libtensorrt_llm_executor_static.pre_cxx11.a
+f8c0381a2bc50ee2739c3d8c2be481b31e5f00bd commit
diff --git a/cpp/tensorrt_llm/executor/x86_64-windows-msvc/tensorrt_llm_executor_static.lib b/cpp/tensorrt_llm/executor/x86_64-windows-msvc/tensorrt_llm_executor_static.lib
diff --git a/cpp/tensorrt_llm/executor/x86_64-windows-msvc/version.txt b/cpp/tensorrt_llm/executor/x86_64-windows-msvc/version.txt
@@ -1,2 +1,2 @@
-b4bb2037715954763241d8ee0b9bef8b tensorrt_llm_executor_static.lib
-09df54c0cc99354a60bbc0303e3e8ea33a96bef0 commit
+5beaadd32fc3dd25770746016b293229 tensorrt_llm_executor_static.lib
+f8c0381a2bc50ee2739c3d8c2be481b31e5f00bd commit
diff --git a/...tention/decoderXQAImplJIT/nvrtcWrapper/aarch64-linux-gnu/libtensorrt_llm_nvrtc_wrapper.so b/...tention/decoderXQAImplJIT/nvrtcWrapper/aarch64-linux-gnu/libtensorrt_llm_nvrtc_wrapper.so
diff --git a/...oderMaskedMultiheadAttention/decoderXQAImplJIT/nvrtcWrapper/aarch64-linux-gnu/version.txt b/...oderMaskedMultiheadAttention/decoderXQAImplJIT/nvrtcWrapper/aarch64-linux-gnu/version.txt
@@ -1,2 +1,2 @@
-95bc050a40ef1436a8d01f7c241c3010 libtensorrt_llm_nvrtc_wrapper.so
-09df54c0cc99354a60bbc0303e3e8ea33a96bef0 commit
+200a2d19469277b9906a00f7da83fd04 libtensorrt_llm_nvrtc_wrapper.so
+f8c0381a2bc50ee2739c3d8c2be481b31e5f00bd commit
diff --git a/...ttention/decoderXQAImplJIT/nvrtcWrapper/x86_64-linux-gnu/libtensorrt_llm_nvrtc_wrapper.so b/...ttention/decoderXQAImplJIT/nvrtcWrapper/x86_64-linux-gnu/libtensorrt_llm_nvrtc_wrapper.so
diff --git a/...coderMaskedMultiheadAttention/decoderXQAImplJIT/nvrtcWrapper/x86_64-linux-gnu/version.txt b/...coderMaskedMultiheadAttention/decoderXQAImplJIT/nvrtcWrapper/x86_64-linux-gnu/version.txt
@@ -1,2 +1,2 @@
-46ff65c35b24eba2c0f9e2dd98298471 libtensorrt_llm_nvrtc_wrapper.so
-09df54c0cc99354a60bbc0303e3e8ea33a96bef0 commit
+20761e50ba44b91a7a2d1f8d1c5c780b libtensorrt_llm_nvrtc_wrapper.so
+f8c0381a2bc50ee2739c3d8c2be481b31e5f00bd commit
diff --git a/...erMaskedMultiheadAttention/decoderXQAImplJIT/nvrtcWrapper/x86_64-windows-msvc/version.txt b/...erMaskedMultiheadAttention/decoderXQAImplJIT/nvrtcWrapper/x86_64-windows-msvc/version.txt
@@ -1,3 +1,3 @@
 3082017cee538017c343567d938bb106 tensorrt_llm_nvrtc_wrapper.lib
 b9b4bf6a2d38abae1d3e038ad73b3890 tensorrt_llm_nvrtc_wrapper.dll
-09df54c0cc99354a60bbc0303e3e8ea33a96bef0 commit
+f8c0381a2bc50ee2739c3d8c2be481b31e5f00bd commit
diff --git a/...ernal_cutlass_kernels/aarch64-linux-gnu/libtensorrt_llm_internal_cutlass_kernels_static.a b/...ernal_cutlass_kernels/aarch64-linux-gnu/libtensorrt_llm_internal_cutlass_kernels_static.a
diff --git a/...ass_kernels/aarch64-linux-gnu/libtensorrt_llm_internal_cutlass_kernels_static.pre_cxx11.a b/...ass_kernels/aarch64-linux-gnu/libtensorrt_llm_internal_cutlass_kernels_static.pre_cxx11.a
diff --git a/cpp/tensorrt_llm/kernels/internal_cutlass_kernels/aarch64-linux-gnu/version.txt b/cpp/tensorrt_llm/kernels/internal_cutlass_kernels/aarch64-linux-gnu/version.txt
@@ -1,3 +1,3 @@
-99b53fa9c0f84d86373517d0c3805980 libtensorrt_llm_internal_cutlass_kernels_static.a
-ff3a0e032808705c8f47b92e823ceab4 libtensorrt_llm_internal_cutlass_kernels_static.pre_cxx11.a
-09df54c0cc99354a60bbc0303e3e8ea33a96bef0 commit
+32e01331abfcacf5b70854104ca4bf20 libtensorrt_llm_internal_cutlass_kernels_static.a
+662d22cefd410c1851ac701e6e3bcbbf libtensorrt_llm_internal_cutlass_kernels_static.pre_cxx11.a
+f8c0381a2bc50ee2739c3d8c2be481b31e5f00bd commit
diff --git a/...ternal_cutlass_kernels/x86_64-linux-gnu/libtensorrt_llm_internal_cutlass_kernels_static.a b/...ternal_cutlass_kernels/x86_64-linux-gnu/libtensorrt_llm_internal_cutlass_kernels_static.a
diff --git a/...lass_kernels/x86_64-linux-gnu/libtensorrt_llm_internal_cutlass_kernels_static.pre_cxx11.a b/...lass_kernels/x86_64-linux-gnu/libtensorrt_llm_internal_cutlass_kernels_static.pre_cxx11.a
diff --git a/cpp/tensorrt_llm/kernels/internal_cutlass_kernels/x86_64-linux-gnu/version.txt b/cpp/tensorrt_llm/kernels/internal_cutlass_kernels/x86_64-linux-gnu/version.txt
@@ -1,3 +1,3 @@
-6acaf9e2ddb3d473ab679f5b2aea20d1 libtensorrt_llm_internal_cutlass_kernels_static.a
-18f56758fcb9688d91dc9bb240656812 libtensorrt_llm_internal_cutlass_kernels_static.pre_cxx11.a
-09df54c0cc99354a60bbc0303e3e8ea33a96bef0 commit
+d8af682c4274543b06992255e727f52e libtensorrt_llm_internal_cutlass_kernels_static.a
+c797baf2a0a7538eb8f75e0f898ae208 libtensorrt_llm_internal_cutlass_kernels_static.pre_cxx11.a
+f8c0381a2bc50ee2739c3d8c2be481b31e5f00bd commit
diff --git a/...rnal_cutlass_kernels/x86_64-windows-msvc/tensorrt_llm_internal_cutlass_kernels_static.lib b/...rnal_cutlass_kernels/x86_64-windows-msvc/tensorrt_llm_internal_cutlass_kernels_static.lib
diff --git a/cpp/tensorrt_llm/kernels/internal_cutlass_kernels/x86_64-windows-msvc/version.txt b/cpp/tensorrt_llm/kernels/internal_cutlass_kernels/x86_64-windows-msvc/version.txt
@@ -1,2 +1,2 @@
-1494bc30b146375aa38d00956d277a2f tensorrt_llm_internal_cutlass_kernels_static.lib
-09df54c0cc99354a60bbc0303e3e8ea33a96bef0 commit
+ab51496e515622f560f4b989ed1d7e63 tensorrt_llm_internal_cutlass_kernels_static.lib
+f8c0381a2bc50ee2739c3d8c2be481b31e5f00bd commit
diff --git a/docs/source/installation/windows.md b/docs/source/installation/windows.md
@@ -56,7 +56,7 @@ TensorRT-LLM on Windows only supports single-GPU execution.
   before installing TensorRT-LLM with the following command.
 
   ```bash
-  pip install tensorrt_llm==0.17.0 --extra-index-url https://download.pytorch.org/whl/
+  pip install tensorrt_llm==0.17.0.post1 --extra-index-url https://download.pytorch.org/whl/
   ```
 
   Run the following command to verify that your TensorRT-LLM installation is working properly.

diff --git a/docs/source/release-notes.md b/docs/source/release-notes.md
@@ -10,7 +10,7 @@ All published functionality in the Release Notes has been fully tested and verif
 ### Key Features and Enhancements
   - **Blackwell support**
     - Added support for B200.
-    - Added support for gb20x Windows Subsystem for Linux (WSL) for limited models.
+    - Added support for GeForce RTX 50 series using Windows Subsystem for Linux (WSL) for limited models.
     - Added NVFP4 Gemm support for Llama and Mixtral models.
     - Added NVFP4 support for the `LLM` API and `trtllm-bench` command.
     - GB200 NVL is not fully supported.

diff --git a/examples/baichuan/requirements.txt b/examples/baichuan/requirements.txt
@@ -1,4 +1,4 @@
-tensorrt_llm==0.17.0
+tensorrt_llm==0.17.0.post1
 datasets~=2.15.0
 evaluate~=0.4.1
 rouge_score~=0.1.2

diff --git a/examples/bloom/requirements.txt b/examples/bloom/requirements.txt
@@ -1,4 +1,4 @@
-tensorrt_llm==0.17.0
+tensorrt_llm==0.17.0.post1
 datasets~=2.14.5
 evaluate~=0.4.1
 rouge_score~=0.1.2

diff --git a/examples/chatglm/requirements.txt b/examples/chatglm/requirements.txt
@@ -1,4 +1,4 @@
-tensorrt_llm==0.17.0
+tensorrt_llm==0.17.0.post1
 datasets~=2.14.5
 evaluate~=0.4.1
 protobuf

diff --git a/examples/commandr/requirements.txt b/examples/commandr/requirements.txt
@@ -1,4 +1,4 @@
-tensorrt_llm==0.17.0
+tensorrt_llm==0.17.0.post1
 datasets==2.14.6
 evaluate~=0.4.1
 rouge_score~=0.1.2
diff --git a/examples/dbrx/requirements.txt b/examples/dbrx/requirements.txt
@@ -1,4 +1,4 @@
-tensorrt_llm==0.17.0
+tensorrt_llm==0.17.0.post1
 datasets~=2.14.5
 evaluate~=0.4.1
 rouge_score~=0.1.2

diff --git a/examples/deepseek_v1/requirements.txt b/examples/deepseek_v1/requirements.txt
@@ -1,4 +1,4 @@
-tensorrt_llm==0.17.0
+tensorrt_llm==0.17.0.post1
 datasets~=2.14.6
 evaluate~=0.4.1
 rouge_score~=0.1.2
diff --git a/examples/draft_target_model/requirements.txt b/examples/draft_target_model/requirements.txt
@@ -1,4 +1,4 @@
-tensorrt_llm==0.17.0
+tensorrt_llm==0.17.0.post1
 datasets~=2.14.5
 rouge_score~=0.1.2
 sentencepiece>=0.1.99

diff --git a/examples/eagle/requirements.txt b/examples/eagle/requirements.txt
@@ -1,4 +1,4 @@
-tensorrt_llm==0.17.0
+tensorrt_llm==0.17.0.post1
 datasets~=2.14.5
 rouge_score~=0.1.2
 SentencePiece~=0.1.99

diff --git a/examples/falcon/requirements.txt b/examples/falcon/requirements.txt
@@ -1,4 +1,4 @@
-tensorrt_llm==0.17.0
+tensorrt_llm==0.17.0.post1
 transformers>=4.31.0
 datasets~=2.14.5
 evaluate~=0.4.1

diff --git a/examples/gemma/requirements.txt b/examples/gemma/requirements.txt
@@ -2,7 +2,7 @@
 # WAR the new posting of "nvidia-cudnn-cu12~=9.0".
 # "jax[cuda12_pip]~=0.4.19" specifies "nvidia-cudnn-cu12>=8.9" but actually requires "nvidia-cudnn-cu12~=8.9".
 nvidia-cudnn-cu12~=8.9; platform_machine == "x86_64"
-tensorrt_llm==0.17.0
+tensorrt_llm==0.17.0.post1
 flax~=0.8.0
 # jax[cuda12_pip]~=0.4.19; platform_system != "Windows"
 jax~=0.4.19; platform_system == "Windows"

diff --git a/examples/gpt/requirements.txt b/examples/gpt/requirements.txt
@@ -1,4 +1,4 @@
-tensorrt_llm==0.17.0
+tensorrt_llm==0.17.0.post1
 datasets~=2.14.5
 evaluate~=0.4.1
 rouge_score~=0.1.2

diff --git a/examples/gptj/requirements.txt b/examples/gptj/requirements.txt
@@ -1,4 +1,4 @@
-tensorrt_llm==0.17.0
+tensorrt_llm==0.17.0.post1
 datasets~=2.14.5
 evaluate~=0.4.1
 rouge_score~=0.1.2
diff --git a/examples/gptneox/requirements.txt b/examples/gptneox/requirements.txt
@@ -1,4 +1,4 @@
-tensorrt_llm==0.17.0
+tensorrt_llm==0.17.0.post1
 datasets~=2.14.5
 rouge_score~=0.1.2
 evaluate~=0.4.1
diff --git a/examples/grok/requirements.txt b/examples/grok/requirements.txt
@@ -1,5 +1,5 @@
 -f https://storage.googleapis.com/jax-releases/jax_cuda_releases.html
-tensorrt_llm==0.17.0
+tensorrt_llm==0.17.0.post1
 datasets==2.14.6
 evaluate~=0.4.1
 rouge_score~=0.1.2

diff --git a/examples/internlm/requirements.txt b/examples/internlm/requirements.txt
@@ -1,4 +1,4 @@
-tensorrt_llm==0.17.0
+tensorrt_llm==0.17.0.post1
 datasets==2.14.5
 rouge_score~=0.1.2
 sentencepiece>=0.1.99

diff --git a/examples/jais/requirements.txt b/examples/jais/requirements.txt
@@ -1,4 +1,4 @@
-tensorrt_llm==0.17.0
+tensorrt_llm==0.17.0.post1
 datasets~=2.14.5
 evaluate~=0.4.1
 rouge_score~=0.1.2

diff --git a/examples/llama/requirements.txt b/examples/llama/requirements.txt
@@ -1,4 +1,4 @@
-tensorrt_llm==0.17.0
+tensorrt_llm==0.17.0.post1
 transformers>=4.43.0
 datasets==2.14.6
 evaluate~=0.4.1

diff --git a/examples/lookahead/requirements.txt b/examples/lookahead/requirements.txt
@@ -1,4 +1,4 @@
-tensorrt_llm==0.17.0
+tensorrt_llm==0.17.0.post1
 datasets~=2.14.5
 rouge_score~=0.1.2
 sentencepiece>=0.1.99

diff --git a/examples/mamba/requirements.txt b/examples/mamba/requirements.txt
@@ -1,4 +1,4 @@
-tensorrt_llm==0.17.0
+tensorrt_llm==0.17.0.post1
 transformers>=4.39.0
 datasets~=2.14.5
 evaluate

diff --git a/examples/medusa/requirements.txt b/examples/medusa/requirements.txt
@@ -1,4 +1,4 @@
-tensorrt_llm==0.17.0
+tensorrt_llm==0.17.0.post1
 datasets~=2.14.5
 rouge_score~=0.1.2
 sentencepiece>=0.1.99

diff --git a/examples/mixtral/requirements.txt b/examples/mixtral/requirements.txt
@@ -1,3 +1,3 @@
-tensorrt_llm==0.17.0
+tensorrt_llm==0.17.0.post1
 transformers==4.38.2
 accelerate==0.25.0
diff --git a/examples/mpt/requirements.txt b/examples/mpt/requirements.txt
@@ -1,4 +1,4 @@
-tensorrt_llm==0.17.0
+tensorrt_llm==0.17.0.post1
 datasets~=2.14.5
 evaluate~=0.4.1
 rouge_score~=0.1.2
diff --git a/examples/nemotron/requirements.txt b/examples/nemotron/requirements.txt
@@ -1,4 +1,4 @@
-tensorrt_llm==0.17.0
+tensorrt_llm==0.17.0.post1
 nemo-toolkit[all]==2.0.0rc1
 megatron-core==0.8.0
 datasets~=2.14.5

diff --git a/examples/opt/requirements.txt b/examples/opt/requirements.txt
@@ -1,4 +1,4 @@
-tensorrt_llm==0.17.0
+tensorrt_llm==0.17.0.post1
 datasets~=2.14.5
 evaluate~=0.4.1
 rouge_score~=0.1.2