diff --git a/uv.lock b/uv.lock index 24534b42..20aac2f7 100644 --- a/uv.lock +++ b/uv.lock @@ -1459,15 +1459,15 @@ wheels = [ [[package]] name = "flashinfer-cubin" -version = "0.6.8.post1" +version = "0.6.11.post2" source = { registry = "https://pypi.org/simple" } wheels = [ - { url = "https://files.pythonhosted.org/packages/11/b7/5e3b1a8c67031b421a8bd29c2bc29b900a550bb3392e8bda18bb15b5e476/flashinfer_cubin-0.6.8.post1-py3-none-any.whl", hash = "sha256:43636d4cd39e694a83d76a89f87fefcdf4cecb4c4f7dd22dac25ec368c1e901f", size = 295154113, upload-time = "2026-04-18T18:28:21.738Z" }, + { url = "https://files.pythonhosted.org/packages/29/96/da75a9f61c64c87b16baa339fc8216a6c3743c5d263c555fded30fcbe6f7/flashinfer_cubin-0.6.11.post2-py3-none-any.whl", hash = "sha256:eb01c2801ee31d145bbf7afb2c223150333e602c8208216017b0190b1087b990", size = 360908523, upload-time = "2026-05-14T04:57:41.355Z" }, ] [[package]] name = "flashinfer-python" -version = "0.6.8.post1" +version = "0.6.11.post2" source = { registry = "https://pypi.org/simple" } dependencies = [ { name = "apache-tvm-ffi" }, @@ -1486,9 +1486,9 @@ dependencies = [ { name = "torch", version = "2.11.0+cu129", source = { registry = "https://download.pytorch.org/whl/cu129" }, marker = "sys_platform == 'linux' or sys_platform == 'win32'" }, { name = "tqdm" }, ] -sdist = { url = "https://files.pythonhosted.org/packages/53/1e/2760fef9e74abc4480961048e5790b4c9e955872fb4d7d97900cfddced5a/flashinfer_python-0.6.8.post1.tar.gz", hash = "sha256:b18e4121baf9b93fa9a9f368ba9b981a0342895f50ab9dddc224aeb964ed346f", size = 6675885, upload-time = "2026-04-18T18:28:13.299Z" } +sdist = { url = "https://files.pythonhosted.org/packages/db/53/dbf2157f2bbb96d6f7a6891cf6abfb2e6e18963760a0c53e96c2de5c59db/flashinfer_python-0.6.11.post2.tar.gz", hash = "sha256:e9fdac56aea9f0f58a4e69b0645c54993760d3cc6c7bf5c2df4ce5a0aecc7953", size = 9248515, upload-time = "2026-05-14T04:57:32.83Z" } wheels = [ - { url = "https://files.pythonhosted.org/packages/73/6d/1e8a8533913e33a50a486332ce0673f4fdb860f6eb9ed450327c5c1762cb/flashinfer_python-0.6.8.post1-py3-none-any.whl", hash = "sha256:818f9b8cc2fe66c42a1f6264be4841ac8821ada703685a02cfccb2b5124a710b", size = 9385316, upload-time = "2026-04-18T18:28:10.285Z" }, + { url = "https://files.pythonhosted.org/packages/c7/bc/518b092473f37d904ae07766ad37c772b93da13ea788777b22a80c3f1a7c/flashinfer_python-0.6.11.post2-py3-none-any.whl", hash = "sha256:550cbdb760f9f7ec0e42055e06636b9489d05f1a38989cafd77e6eb820de0138", size = 13746417, upload-time = "2026-05-14T04:57:30.25Z" }, ] [[package]] @@ -1880,6 +1880,37 @@ wheels = [ { url = "https://files.pythonhosted.org/packages/f0/0f/310fb31e39e2d734ccaa2c0fb981ee41f7bd5056ce9bc29b2248bd569169/humanfriendly-10.0-py2.py3-none-any.whl", hash = "sha256:1697e1a8a8f550fd43c2865cd84542fc175a61dcb779b6fee18cf6b6ccba1477", size = 86794, upload-time = "2021-09-17T21:40:39.897Z" }, ] +[[package]] +name = "humming-kernels" +version = "0.1.2" +source = { registry = "https://pypi.org/simple" } +dependencies = [ + { name = "cuda-bindings", version = "12.9.4", source = { registry = "https://pypi.org/simple" }, marker = "sys_platform != 'win32'" }, + { name = "cuda-bindings", version = "13.1.1", source = { registry = "https://pypi.org/simple" }, marker = "sys_platform == 'win32'" }, + { name = "jinja2" }, + { name = "numpy" }, + { name = "nvidia-ml-py" }, + { name = "pyelftools" }, + { name = "safetensors" }, + { name = "tabulate" }, + { name = "torch", version = "2.11.0", source = { registry = "https://pypi.org/simple" }, marker = "sys_platform != 'linux' and sys_platform != 'win32'" }, + { name = "torch", version = "2.11.0+cu129", source = { registry = "https://download.pytorch.org/whl/cu129" }, marker = "sys_platform == 'linux' or sys_platform == 'win32'" }, + { name = "tqdm" }, + { name = "triton" }, +] +sdist = { url = "https://files.pythonhosted.org/packages/06/f4/e141f45697b7d0d38bfaf8766a7362d8f0136e3cff2620624f24f68e2700/humming_kernels-0.1.2.tar.gz", hash = "sha256:7894c80061c7866591bef12617da720ac4e925636ffc99464af433a5dcb035eb", size = 117251, upload-time = "2026-05-23T16:18:08.084Z" } +wheels = [ + { url = "https://files.pythonhosted.org/packages/6d/41/288bf756d921dbe98982eeb3ec4c20e7cb5224ea6dcb164f2df3d2f68a7f/humming_kernels-0.1.2-py3-none-any.whl", hash = "sha256:f7434b0424946445ef5ad5682bcabf309d97721818ed5bdc4c6f61de3c6b9d2f", size = 160951, upload-time = "2026-05-23T16:18:06.405Z" }, +] + +[package.optional-dependencies] +cu13 = [ + { name = "nvidia-cuda-cccl" }, + { name = "nvidia-cuda-nvcc" }, + { name = "nvidia-cuda-nvrtc" }, + { name = "nvidia-cuda-runtime" }, +] + [[package]] name = "idna" version = "3.15" @@ -2543,15 +2574,15 @@ wheels = [ [[package]] name = "llguidance" -version = "1.3.0" +version = "1.7.6" source = { registry = "https://pypi.org/simple" } -sdist = { url = "https://files.pythonhosted.org/packages/95/48/3f7a9d3ff1b36bba92b5107a3a21286821227afe9ea464736133994d61fb/llguidance-1.3.0.tar.gz", hash = "sha256:861249afd51dc325646834462ea827e57a5c2b2042e108e6aae7059fdad9104d", size = 1070460, upload-time = "2025-10-20T19:58:44.164Z" } +sdist = { url = "https://files.pythonhosted.org/packages/da/91/6bc8bb503dc259e46d253b5424385a54fe06c38a4c7a12befe69a3c2455a/llguidance-1.7.6.tar.gz", hash = "sha256:db7febbe412ed2015501904646750071d7e00e6df7f85c4b956ad4f206fd2df7", size = 1156574, upload-time = "2026-06-03T20:13:25.316Z" } wheels = [ - { url = "https://files.pythonhosted.org/packages/3b/33/be5acb85cd8cdc4afde33d9c234eece9f318e087920255af3c05864cd3e7/llguidance-1.3.0-cp39-abi3-macosx_10_12_x86_64.whl", hash = "sha256:f7685222660a762e481ac633d49cc559c64980fe2ee59c8f932a5bb5cbc0c2c2", size = 3220647, upload-time = "2025-10-20T19:58:42.542Z" }, - { url = "https://files.pythonhosted.org/packages/82/e6/b48bda5b15efeaeb62bd0dba8fc6a01d4ae5457a85dbb5d18632385fe15c/llguidance-1.3.0-cp39-abi3-macosx_11_0_arm64.whl", hash = "sha256:098030ff0687261a3f1bd54cf21fe951fc861d56d37a0671250dd36677eaf224", size = 3099830, upload-time = "2025-10-20T19:58:40.826Z" }, - { url = "https://files.pythonhosted.org/packages/aa/11/44389d3d1526d7a5c38ffd587a5ebc61d7bee443ac1dea95f2089ad58f5f/llguidance-1.3.0-cp39-abi3-manylinux_2_17_aarch64.manylinux2014_aarch64.whl", hash = "sha256:6f6caca5d78db7f76e1fbb0fff8607b861c32d47fa3d5dee2fc49de27ee269df", size = 2835242, upload-time = "2025-10-20T19:58:34.518Z" }, - { url = "https://files.pythonhosted.org/packages/83/a8/1ff2bedb8f9acb46a2d2d603415d272bb622c142ea86f5b95445cc6e366c/llguidance-1.3.0-cp39-abi3-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:bc17e9dd602c3879bf91664a64bf72f54c74dbfbeb24ccfab6a5fe435b12f7aa", size = 3033133, upload-time = "2025-10-20T19:58:38.721Z" }, - { url = "https://files.pythonhosted.org/packages/5a/7e/809349638231f469b9056c0e1bfd924d5ef5558b3b3ec72d093b6fad33b1/llguidance-1.3.0-cp39-abi3-win_amd64.whl", hash = "sha256:1d1cd1c8618d1a13605d3e057c978651e551c8c469b481ee4041f1d6c436002d", size = 2789946, upload-time = "2025-10-20T19:58:45.958Z" }, + { url = "https://files.pythonhosted.org/packages/fa/1d/5a9a13421b1f3f1c1acf82beb63ed72fa4d302e65099b72f4a4fe5a098ab/llguidance-1.7.6-cp39-abi3-macosx_10_12_x86_64.whl", hash = "sha256:eabf4572c8731734c0444c353b9ea06bc5c156986d2ff0a4ec0499159271381f", size = 3227892, upload-time = "2026-06-03T20:13:09.533Z" }, + { url = "https://files.pythonhosted.org/packages/46/fe/bb185f11bad82f2637e3cd8cbf6b200cbb6ed56ac395de47ea05a60d4649/llguidance-1.7.6-cp39-abi3-macosx_11_0_arm64.whl", hash = "sha256:9c54c899db8cb4b4fba128a7d844730066576c70d806c95ada92b2bd2d6ab498", size = 3138127, upload-time = "2026-06-03T20:13:11.649Z" }, + { url = "https://files.pythonhosted.org/packages/51/b9/dc76d7716e04dc7b3427cae52eaa32bd20771382d4d1dd9f4538a9dd2086/llguidance-1.7.6-cp39-abi3-manylinux_2_31_aarch64.whl", hash = "sha256:e70fa25ed550c2b50c2fd70baa9e2808b4ecb859d01e453bd5459aff62ba38c3", size = 2899993, upload-time = "2026-06-03T20:13:13.563Z" }, + { url = "https://files.pythonhosted.org/packages/1a/64/d74336f22242ef94356a456057d4ff1be7c1bc9c7dbc867171c6982a5512/llguidance-1.7.6-cp39-abi3-manylinux_2_31_x86_64.whl", hash = "sha256:ceec951d29a74309984e3be0fe7f5f56c1362434cd937abd517b259a60908b1e", size = 3074809, upload-time = "2026-06-03T20:13:15.498Z" }, + { url = "https://files.pythonhosted.org/packages/49/37/99d700f0e2c83acf25a8d8946b2bee9f5eac47bc530bfbd53ba3126c667f/llguidance-1.7.6-cp39-abi3-win_amd64.whl", hash = "sha256:ace7e81cd31950a87186356ab24bd7f75fbc10a05ca9d9f7f8748f931963f763", size = 2879207, upload-time = "2026-06-03T20:13:23.341Z" }, ] [[package]] @@ -3268,6 +3299,26 @@ wheels = [ { url = "https://files.pythonhosted.org/packages/77/3c/aa88abe01f3be3d1f8f787d1d33dc83e76fec05945f9a28fbb41cfb99cd5/nvidia_cublas_cu12-12.9.1.4-py3-none-manylinux_2_27_x86_64.whl", hash = "sha256:453611eb21a7c1f2c2156ed9f3a45b691deda0440ec550860290dc901af5b4c2", size = 581242350, upload-time = "2025-06-05T20:04:51.979Z" }, ] +[[package]] +name = "nvidia-cuda-cccl" +version = "13.3.3.3.1" +source = { registry = "https://pypi.org/simple" } +wheels = [ + { url = "https://files.pythonhosted.org/packages/5f/7a/9cb8a7fb87a85b11e8753548ae1422be847c5dddf3ca9ff5b080b309e271/nvidia_cuda_cccl-13.3.3.3.1-py3-none-manylinux2014_aarch64.manylinux_2_17_aarch64.whl", hash = "sha256:4dbc9dd84fbaeae267cbd80a9ed76d35171dba78639695dbdff0bae50e4503fa", size = 3453010, upload-time = "2026-05-26T16:27:45.179Z" }, + { url = "https://files.pythonhosted.org/packages/fe/fb/195d50d25ab68a76b817ffc68c45b1fb828598ce35a8e5c1736060628dab/nvidia_cuda_cccl-13.3.3.3.1-py3-none-manylinux2014_x86_64.manylinux_2_17_x86_64.whl", hash = "sha256:40ba1fa0b2c694ddc06cc791ed5c8bdad4638e2735b784960d68ac3086399c97", size = 3453013, upload-time = "2026-05-26T16:28:08.209Z" }, + { url = "https://files.pythonhosted.org/packages/57/44/37cf1596880e7712f357b3f4991cd34d0f322c26e2bc814d1bdeffb2f420/nvidia_cuda_cccl-13.3.3.3.1-py3-none-win_amd64.whl", hash = "sha256:d1ac746f57ab83403f01e64e2b292101caf5b3445babca9f1c1c34f344766adf", size = 3452993, upload-time = "2026-05-26T16:58:59.166Z" }, +] + +[[package]] +name = "nvidia-cuda-crt" +version = "13.3.33" +source = { registry = "https://pypi.org/simple" } +wheels = [ + { url = "https://files.pythonhosted.org/packages/d1/32/5ea57f8cd6ad5df2173d175ac5db4e06edde40028b1b1f6c539ea4c10290/nvidia_cuda_crt-13.3.33-py3-none-manylinux2014_aarch64.manylinux_2_17_aarch64.whl", hash = "sha256:c8c257393f9c9146a85d3644f352be8154843d760031f756e673222c768a4930", size = 157348, upload-time = "2026-05-26T16:28:40.446Z" }, + { url = "https://files.pythonhosted.org/packages/8d/a7/998af901511d5efdc6e42fc597d32a69f34eecf86f1591a9d230ab3ab951/nvidia_cuda_crt-13.3.33-py3-none-manylinux2014_x86_64.manylinux_2_17_x86_64.whl", hash = "sha256:01ff37600c7b880a14cab4ade763b4c10c0ff92f25cc9dca30f0881ce52693c4", size = 157350, upload-time = "2026-05-26T16:29:22.315Z" }, + { url = "https://files.pythonhosted.org/packages/a4/5f/fc8ce6b7719c825e0e519d2922e3b7630238e860222ad3f972dd9b8b7fa9/nvidia_cuda_crt-13.3.33-py3-none-win_amd64.whl", hash = "sha256:7e89c6dbb807a47ee0628907488b158e57c36fa31af3756a8f826a9ec482715f", size = 158284, upload-time = "2026-05-26T16:59:37.309Z" }, +] + [[package]] name = "nvidia-cuda-cupti-cu12" version = "12.9.79" @@ -3277,6 +3328,31 @@ wheels = [ { url = "https://files.pythonhosted.org/packages/c1/2e/b84e32197e33f39907b455b83395a017e697c07a449a2b15fd07fc1c9981/nvidia_cuda_cupti_cu12-12.9.79-py3-none-manylinux_2_25_x86_64.whl", hash = "sha256:096bcf334f13e1984ba36685ad4c1d6347db214de03dbb6eebb237b41d9d934f", size = 10814997, upload-time = "2025-06-05T20:01:10.168Z" }, ] +[[package]] +name = "nvidia-cuda-nvcc" +version = "13.3.33" +source = { registry = "https://pypi.org/simple" } +dependencies = [ + { name = "nvidia-cuda-crt" }, + { name = "nvidia-cuda-runtime" }, + { name = "nvidia-nvvm" }, +] +wheels = [ + { url = "https://files.pythonhosted.org/packages/be/b6/bb07a3a63b5b7b55516366747892abbf3ee62d616684c40bb51e6cbfe956/nvidia_cuda_nvcc-13.3.33-py3-none-manylinux2014_aarch64.manylinux_2_17_aarch64.whl", hash = "sha256:8c348623b1434aebd234da9ec1f81022587ae4995d65c3dc8a7743245cc441f7", size = 39515074, upload-time = "2026-05-26T16:34:28.489Z" }, + { url = "https://files.pythonhosted.org/packages/3f/af/e1b107f034f7c133255c162b922bbad3da5be20ebf76df17662ae4bd31f6/nvidia_cuda_nvcc-13.3.33-py3-none-manylinux2014_x86_64.manylinux_2_17_x86_64.whl", hash = "sha256:53b5f1be1731574368b8be931b77b6313492266c464aef3dd3f431569ce90deb", size = 44943276, upload-time = "2026-05-26T16:35:05.912Z" }, + { url = "https://files.pythonhosted.org/packages/47/c2/831fa54020621a64d44cff47f1ed5eb0611794495fce01c857f6999d76b1/nvidia_cuda_nvcc-13.3.33-py3-none-win_amd64.whl", hash = "sha256:21c93aeef695a81b688137119f9120fe08a67292bf0ad730d94dc2b18bec23f0", size = 32723421, upload-time = "2026-05-26T17:01:47.511Z" }, +] + +[[package]] +name = "nvidia-cuda-nvrtc" +version = "13.3.33" +source = { registry = "https://pypi.org/simple" } +wheels = [ + { url = "https://files.pythonhosted.org/packages/8b/2c/86916c8a34dcdb0c3ddd1c0e30545041bd781184e437b9cb76fcda70560b/nvidia_cuda_nvrtc-13.3.33-py3-none-manylinux2010_x86_64.manylinux_2_12_x86_64.whl", hash = "sha256:82530788b8c6164a54d3fd9ae8bcca8893d397c4aeb998861982a03bbe41e204", size = 51110910, upload-time = "2026-05-26T16:38:16.116Z" }, + { url = "https://files.pythonhosted.org/packages/e7/b6/60a3641111d39ebfcfcd8b8bfd0290d7623c4b8b5f90952c2d84776f8ca4/nvidia_cuda_nvrtc-13.3.33-py3-none-manylinux2014_aarch64.manylinux_2_17_aarch64.whl", hash = "sha256:7b05ecda494c6dabc44231a608b060a71008a730d9dfda932cc508e6d29159e0", size = 49260054, upload-time = "2026-05-26T16:37:51.177Z" }, + { url = "https://files.pythonhosted.org/packages/a1/42/edce72f2c5a0f587168109c867f25f4a9a6cd7289ecf0d68ed2b1070f273/nvidia_cuda_nvrtc-13.3.33-py3-none-win_amd64.whl", hash = "sha256:7d2af818851c0c224d5f92221e9226e51ee23c236df4b51f9194563979c888be", size = 45319163, upload-time = "2026-05-26T17:02:49.217Z" }, +] + [[package]] name = "nvidia-cuda-nvrtc-cu12" version = "12.9.86" @@ -3286,6 +3362,16 @@ wheels = [ { url = "https://files.pythonhosted.org/packages/64/eb/c2295044b8f3b3b08860e2f6a912b702fc92568a167259df5dddb78f325e/nvidia_cuda_nvrtc_cu12-12.9.86-py3-none-manylinux2014_aarch64.manylinux_2_17_aarch64.whl", hash = "sha256:096d4de6bda726415dfaf3198d4f5c522b8e70139c97feef5cd2ca6d4cd9cead", size = 44528905, upload-time = "2025-06-05T20:02:29.754Z" }, ] +[[package]] +name = "nvidia-cuda-runtime" +version = "13.3.29" +source = { registry = "https://pypi.org/simple" } +wheels = [ + { url = "https://files.pythonhosted.org/packages/5f/e5/c1a221c8e6fecd071b80ea44c20fc253ae24f56e15e3f77cfbc3fb76e724/nvidia_cuda_runtime-13.3.29-py3-none-manylinux2014_aarch64.manylinux_2_17_aarch64.whl", hash = "sha256:73291e19c9dd919c140c91bda2f80b0eca487da5ee30a086ef7bc4918ecb90ea", size = 2356574, upload-time = "2026-05-26T16:29:56.333Z" }, + { url = "https://files.pythonhosted.org/packages/97/be/5699b6e642b372f7d24c59c2f41383e2696825e20bab85f7399c7c6a56f7/nvidia_cuda_runtime-13.3.29-py3-none-manylinux2014_x86_64.manylinux_2_17_x86_64.whl", hash = "sha256:e04420616e72f563167a7733272992d7e6df6dc5cb54b2f94f9f1520ea9e30c1", size = 2339786, upload-time = "2026-05-26T16:30:21.584Z" }, + { url = "https://files.pythonhosted.org/packages/d2/27/b53a5e0397842a5c11f0e1a39d4e5b2f22638a4126e83b3c4e196f62c969/nvidia_cuda_runtime-13.3.29-py3-none-win_amd64.whl", hash = "sha256:0667ec61c3d897388efa305ed4f7609ace88849a753ba9c6311d06dca55fff4f", size = 2630354, upload-time = "2026-05-26T17:00:05.389Z" }, +] + [[package]] name = "nvidia-cuda-runtime-cu12" version = "12.8.90" @@ -3407,30 +3493,53 @@ wheels = [ [[package]] name = "nvidia-cutlass-dsl" -version = "4.4.2" +version = "4.5.2" source = { registry = "https://pypi.org/simple" } dependencies = [ { name = "nvidia-cutlass-dsl-libs-base" }, ] wheels = [ - { url = "https://files.pythonhosted.org/packages/a9/03/678dab0383db1ddfc449da216220f40404189eb36eeed9d87a4fa4bdb0e6/nvidia_cutlass_dsl-4.4.2-py3-none-any.whl", hash = "sha256:7cfb9ef19062b055b9372c7a627004724e2755e4c8b16c3cc88807d64501a4ae", size = 10167, upload-time = "2026-03-16T02:18:59.043Z" }, + { url = "https://files.pythonhosted.org/packages/f0/15/575d7df4fe2f3406f1cfc68be72aeff2834f8a696daf1cd5bee8017e4507/nvidia_cutlass_dsl-4.5.2-py3-none-any.whl", hash = "sha256:68ed1b63ca74aae87955012da9dfd7fdaae471329d0028b229b841c7192ccf52", size = 10179, upload-time = "2026-05-25T03:38:56.364Z" }, +] + +[package.optional-dependencies] +cu13 = [ + { name = "nvidia-cutlass-dsl-libs-cu13" }, ] [[package]] name = "nvidia-cutlass-dsl-libs-base" -version = "4.4.2" +version = "4.5.2" +source = { registry = "https://pypi.org/simple" } +dependencies = [ + { name = "cuda-python", version = "12.9.4", source = { registry = "https://pypi.org/simple" }, marker = "sys_platform != 'win32'" }, + { name = "cuda-python", version = "13.1.1", source = { registry = "https://pypi.org/simple" }, marker = "sys_platform == 'win32'" }, + { name = "numpy" }, + { name = "typing-extensions" }, +] +wheels = [ + { url = "https://files.pythonhosted.org/packages/c0/a8/0cca1d11787128c66c0774374d1bb09313352eee11560dd00f36d6d62f36/nvidia_cutlass_dsl_libs_base-4.5.2-cp311-cp311-manylinux_2_28_aarch64.whl", hash = "sha256:cbb555a95c7011e4b3ca328be407299c77d289660adbea22ed515d4406e6949c", size = 75637009, upload-time = "2026-05-25T03:48:37.901Z" }, + { url = "https://files.pythonhosted.org/packages/0c/e0/78eded54b4478ec01a91c75f1b9bc6dc73a2ec205c4fa2fdc25a456f4089/nvidia_cutlass_dsl_libs_base-4.5.2-cp311-cp311-manylinux_2_28_x86_64.whl", hash = "sha256:9117900cba53d3c21a8dacba6bbf3d6e5f269e427a526c320fb44707a0d57363", size = 74511501, upload-time = "2026-05-25T03:52:03.798Z" }, + { url = "https://files.pythonhosted.org/packages/b1/ef/e827e3c67d72adbf4e8f680bdf03b1b67723d9e1ae7c3d0a1751f39f69ce/nvidia_cutlass_dsl_libs_base-4.5.2-cp312-cp312-manylinux_2_28_aarch64.whl", hash = "sha256:d2a3c412287e356fbe48fe9f845d6d33cd35dea5e20d7e4f628c20957967cacd", size = 75643473, upload-time = "2026-05-25T03:49:15.857Z" }, + { url = "https://files.pythonhosted.org/packages/97/68/c1247ab848f26c4ab56e562eea0e3f31fc14c9aaf0d883afaa92d8f05592/nvidia_cutlass_dsl_libs_base-4.5.2-cp312-cp312-manylinux_2_28_x86_64.whl", hash = "sha256:15ef6a59193667e663934ef4873f8ccad37455e9b7c3c419c3072113b8aedf61", size = 74513226, upload-time = "2026-05-25T03:51:32.496Z" }, +] + +[[package]] +name = "nvidia-cutlass-dsl-libs-cu13" +version = "4.5.2" source = { registry = "https://pypi.org/simple" } dependencies = [ { name = "cuda-python", version = "12.9.4", source = { registry = "https://pypi.org/simple" }, marker = "sys_platform != 'win32'" }, { name = "cuda-python", version = "13.1.1", source = { registry = "https://pypi.org/simple" }, marker = "sys_platform == 'win32'" }, { name = "numpy" }, + { name = "nvidia-cutlass-dsl-libs-base" }, { name = "typing-extensions" }, ] wheels = [ - { url = "https://files.pythonhosted.org/packages/60/bf/b9d0fd1ba281b111c941d9616dd9f98a509d84bf35076e60fef27ec7abd6/nvidia_cutlass_dsl_libs_base-4.4.2-cp311-cp311-manylinux_2_28_aarch64.whl", hash = "sha256:261832dafe7579dc83cd3816ab9ea845e3de3737d876c215f01fb4edff1f4473", size = 75476977, upload-time = "2026-03-16T02:26:40.932Z" }, - { url = "https://files.pythonhosted.org/packages/a5/23/86dda6d69a3fc29d0cde2a8b54c056ad69b73a6e5e230e18d906d2ec3b7c/nvidia_cutlass_dsl_libs_base-4.4.2-cp311-cp311-manylinux_2_28_x86_64.whl", hash = "sha256:40c2352b2fcc80789a216cbeb9b2ee10c85c15de839cda8f5c1d18166b8249df", size = 74356100, upload-time = "2026-03-16T02:26:12.778Z" }, - { url = "https://files.pythonhosted.org/packages/8e/7d/0df5e38d11e52cc72095a14d6448bc1c5d0d4b00b069a1189ca417fb225b/nvidia_cutlass_dsl_libs_base-4.4.2-cp312-cp312-manylinux_2_28_aarch64.whl", hash = "sha256:2ec8812eeadcbb6fe20bda2e295ed9c00653f8253b78e33cf0ab65a47b829e73", size = 75473821, upload-time = "2026-03-16T02:27:08.371Z" }, - { url = "https://files.pythonhosted.org/packages/56/98/e264964741d9cc9816625d9600d17a5249fd5cbd8c2d166fb0d0c34dfe5a/nvidia_cutlass_dsl_libs_base-4.4.2-cp312-cp312-manylinux_2_28_x86_64.whl", hash = "sha256:22e37b58f7a6f2f43bba533c4df8a088012122e0b4e9a632eca23937adeafb39", size = 74355593, upload-time = "2026-03-16T02:25:11.762Z" }, + { url = "https://files.pythonhosted.org/packages/34/24/4ad875105f8b834ff0a6dce484c8ac124c292368338b087b993b70288385/nvidia_cutlass_dsl_libs_cu13-4.5.2-cp311-cp311-manylinux_2_28_aarch64.whl", hash = "sha256:f4a7b72147c2efdc7963c64475eac4ed67eb1dd5fdf5b0300daf79319fe9a38a", size = 79081923, upload-time = "2026-05-25T03:40:22.457Z" }, + { url = "https://files.pythonhosted.org/packages/3c/3d/2153608b1f8f594ccfc67daa45a1d0ff600b9e552b1e5662644e6e3ebec3/nvidia_cutlass_dsl_libs_cu13-4.5.2-cp311-cp311-manylinux_2_28_x86_64.whl", hash = "sha256:df61430d6110eea872acb39257042814bf02dcbb1f8d55ea0c5681bb7ce5836a", size = 78759970, upload-time = "2026-05-25T03:43:46.762Z" }, + { url = "https://files.pythonhosted.org/packages/21/e5/aeb570713a7bd6c2cb08102c2ebe6de234ef1bbc276d1af4643266cd71a8/nvidia_cutlass_dsl_libs_cu13-4.5.2-cp312-cp312-manylinux_2_28_aarch64.whl", hash = "sha256:3032405dff28892340f96b467e744a822079cae454dce534fc17b77e85190e42", size = 79084280, upload-time = "2026-05-25T03:40:57.547Z" }, + { url = "https://files.pythonhosted.org/packages/03/60/443e559139da15ab544761ac14f4206dffb981af48cc9856cd5b5b7cf0e7/nvidia_cutlass_dsl_libs_cu13-4.5.2-cp312-cp312-manylinux_2_28_x86_64.whl", hash = "sha256:80f0cd402e0f1d1571e5aed33bfa17dbc9cb90cc5b1352f0f806b4788558e80e", size = 78759198, upload-time = "2026-05-25T03:45:59.297Z" }, ] [[package]] @@ -3478,6 +3587,16 @@ wheels = [ { url = "https://files.pythonhosted.org/packages/c4/e4/82155e4aaedb41621087ba219c95e99c5e417f37a7649b4fb6ec32dcb14d/nvidia_nvtx_cu12-12.9.79-py3-none-manylinux2014_aarch64.manylinux_2_17_aarch64.whl", hash = "sha256:d1f258e752294acdb4f61c3d31fee87bd0f60e459f1e2f624376369b524cd15d", size = 86120, upload-time = "2025-06-05T20:02:51.838Z" }, ] +[[package]] +name = "nvidia-nvvm" +version = "13.3.33" +source = { registry = "https://pypi.org/simple" } +wheels = [ + { url = "https://files.pythonhosted.org/packages/01/8a/f767031dcd0d24c2bbab4b696dbcf004da4f3284e5e4649fc47bc0e2bb78/nvidia_nvvm-13.3.33-py3-none-manylinux2010_x86_64.manylinux_2_12_x86_64.whl", hash = "sha256:aafaf73246b6126bc88f521e5dab1d196395ee87739d9f5b7c39c9fee0ead9c7", size = 69250604, upload-time = "2026-05-26T16:57:56.875Z" }, + { url = "https://files.pythonhosted.org/packages/83/36/ce0d42d3a4465c858c379932f0080d29d22f04383ab79119c7c4f4cdd5ef/nvidia_nvvm-13.3.33-py3-none-manylinux2014_aarch64.manylinux_2_17_aarch64.whl", hash = "sha256:fd74a1c5ef284ba04c1ba75f886404dff953c54731a3a9c7b45e9aedaf1a226b", size = 66984524, upload-time = "2026-05-26T16:57:30.778Z" }, + { url = "https://files.pythonhosted.org/packages/8f/96/4de7a37803d168337ab36f81ecbc496c7c21c9b06ec68ce0ecc381af88d4/nvidia_nvvm-13.3.33-py3-none-win_amd64.whl", hash = "sha256:b1c63cf8972d8a1ff153c5ac4cc7038fe6ef705aa38415f12007b0e5e4c31b79", size = 60175824, upload-time = "2026-05-26T17:13:27.588Z" }, +] + [[package]] name = "omegaconf" version = "2.3.0" @@ -4386,6 +4505,15 @@ wheels = [ { url = "https://files.pythonhosted.org/packages/a6/53/d78dc063216e62fc55f6b2eebb447f6a4b0a59f55c8406376f76bf959b08/pydub-0.25.1-py2.py3-none-any.whl", hash = "sha256:65617e33033874b59d87db603aa1ed450633288aefead953b30bded59cb599a6", size = 32327, upload-time = "2021-03-10T02:09:53.503Z" }, ] +[[package]] +name = "pyelftools" +version = "0.33" +source = { registry = "https://pypi.org/simple" } +sdist = { url = "https://files.pythonhosted.org/packages/a3/11/767522582afab1b884d277de0e6e011640cb9d7292a38694b4b1a1df1ae8/pyelftools-0.33.tar.gz", hash = "sha256:660d82dcbeb8e83d1702bd97f223f761625da06111c0cc988eac6b8ab0c1b61f", size = 15068655, upload-time = "2026-05-29T12:56:22.553Z" } +wheels = [ + { url = "https://files.pythonhosted.org/packages/46/2a/f9697576603dae937727827505a6126a066affb227034e77e6f9068910da/pyelftools-0.33-py3-none-any.whl", hash = "sha256:f215ad5f47d3f1373a21496a6c9e0707c622840d0622f23ff7ce08678b020036", size = 201178, upload-time = "2026-05-29T12:56:20.587Z" }, +] + [[package]] name = "pygments" version = "2.20.0" @@ -6100,7 +6228,7 @@ requires-dist = [ { name = "ultrarag", extras = ["generation"], marker = "extra == 'all'" }, { name = "ultrarag", extras = ["retriever"], marker = "extra == 'all'" }, { name = "uvicorn", marker = "extra == 'retriever'" }, - { name = "vllm", marker = "extra == 'generation'", specifier = ">=0.21.0" }, + { name = "vllm", marker = "extra == 'generation'", specifier = ">=0.22.0" }, ] provides-extras = ["retriever", "generation", "evaluation", "corpus", "all"] @@ -6185,7 +6313,7 @@ wheels = [ [[package]] name = "vllm" -version = "0.21.0" +version = "0.22.0" source = { registry = "https://pypi.org/simple" } dependencies = [ { name = "aiohttp" }, @@ -6205,6 +6333,7 @@ dependencies = [ { name = "flashinfer-cubin" }, { name = "flashinfer-python" }, { name = "gguf" }, + { name = "humming-kernels", extra = ["cu13"] }, { name = "ijson" }, { name = "lark" }, { name = "llguidance", marker = "platform_machine == 'aarch64' or platform_machine == 'arm64' or platform_machine == 'ppc64le' or platform_machine == 'x86_64'" }, @@ -6217,7 +6346,7 @@ dependencies = [ { name = "numba" }, { name = "numpy" }, { name = "nvidia-cudnn-frontend" }, - { name = "nvidia-cutlass-dsl" }, + { name = "nvidia-cutlass-dsl", extra = ["cu13"] }, { name = "openai" }, { name = "openai-harmony" }, { name = "opencv-python-headless" }, @@ -6241,6 +6370,7 @@ dependencies = [ { name = "quack-kernels" }, { name = "regex" }, { name = "requests" }, + { name = "safetensors" }, { name = "sentencepiece" }, { name = "setproctitle" }, { name = "setuptools", marker = "python_full_version >= '3.12'" }, @@ -6259,12 +6389,10 @@ dependencies = [ { name = "watchfiles" }, { name = "xgrammar", marker = "platform_machine == 'aarch64' or platform_machine == 'arm64' or platform_machine == 'ppc64le' or platform_machine == 's390x' or platform_machine == 'x86_64'" }, ] -sdist = { url = "https://files.pythonhosted.org/packages/97/bb/8dbba4136f6851470f4324ac665affe55c0b618341ccc42f35a53c5e708e/vllm-0.21.0.tar.gz", hash = "sha256:05ff89c3e926b88b77d7878e317a659ffba678afc21c1d48952037aa5457f058", size = 34452205, upload-time = "2026-05-15T00:09:15.481Z" } +sdist = { url = "https://files.pythonhosted.org/packages/e2/bf/46631fd8e2e9d81c5abe2ab923e5367754bc0cad685c4ddac1d5d86d91b5/vllm-0.22.0.tar.gz", hash = "sha256:6d41581a9e5288cd69278518a550c6d7ce510ae27a506556a3427d01284be7fe", size = 36239170, upload-time = "2026-05-29T10:35:39.448Z" } wheels = [ - { url = "https://files.pythonhosted.org/packages/ac/58/564b64d17dde6dc31faae836f98313538c152edf88e2a4fb43b9d551a635/vllm-0.21.0-1-cp38-abi3-manylinux_2_24_aarch64.whl", hash = "sha256:dc62135a50dc4b412b4f79549208e782f1665e49e8c13c2d29d2c3d94ff8ac97", size = 239758862, upload-time = "2026-05-15T08:47:06.471Z" }, - { url = "https://files.pythonhosted.org/packages/73/6d/9b78990c9fabc70c7731de6af246a420156dc019f66b48da7c86f509c132/vllm-0.21.0-1-cp38-abi3-manylinux_2_24_x86_64.whl", hash = "sha256:f4a75b1391f44c67dc1ca268f5ffed9f6b7fdbc657c93db64e6892c5d1bc320b", size = 248151215, upload-time = "2026-05-15T08:47:36.846Z" }, - { url = "https://files.pythonhosted.org/packages/59/ae/d78ef0ed561974ea61c6e0786771d3a2a575e22592bd58f2ed52417b9aa2/vllm-0.21.0-cp38-abi3-manylinux_2_24_aarch64.whl", hash = "sha256:d6e63955b595bd2aa364e90f85c0a2e99573e701146db58394da569ddc6f4eea", size = 239758816, upload-time = "2026-05-15T00:08:22.496Z" }, - { url = "https://files.pythonhosted.org/packages/a8/62/8cbf7c943b0aca0538d0f5324848a3f256b8284dd4d881cd65ae106c83d7/vllm-0.21.0-cp38-abi3-manylinux_2_24_x86_64.whl", hash = "sha256:b241b085742cf04a68c82c089d12afe4d9ee729e0c7f81b2b2b9961d36105ee5", size = 248151169, upload-time = "2026-05-15T00:08:53.502Z" }, + { url = "https://files.pythonhosted.org/packages/ed/e8/05a69dbd7416c5a5ac91f51e626fede9ceeabe9c6fe243fc11e2b3e1ad3e/vllm-0.22.0-cp38-abi3-manylinux_2_28_aarch64.whl", hash = "sha256:0fbe1ff32e9ad82c56b002de11b061ca6b5b8a256cd11473946d2222115ed267", size = 252942448, upload-time = "2026-05-29T10:32:10.261Z" }, + { url = "https://files.pythonhosted.org/packages/ca/23/3f7f759763fb9b4cf5787bcb4a43f74904f8e644d53d4fdb4e19654a92fd/vllm-0.22.0-cp38-abi3-manylinux_2_28_x86_64.whl", hash = "sha256:c387a977e35795e8f77b009e019e69722963819c26b55e4a679e09d4279ae35d", size = 261034920, upload-time = "2026-05-29T10:33:10.357Z" }, ] [[package]]