Mirrors
/
hf_text-generation-inference
mirror of https://github.com/huggingface/text-generation-inference.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

Some small fixes for the Torch 2.4.0 update (#2304) 

* Fix GPTQ autotune data type to be compatible with Torch 2.4.0

* Update poetry lock file

* Fix small PaliGemma logprob differences after the torch update
This commit is contained in:
Daniël de Kok 2024-07-25 13:34:44 +02:00 committed by GitHub
parent 26614057a7
commit 9256d7c38c
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
3 changed files with 50 additions and 91 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

16
integration-tests/models/__snapshots__/test_flash_pali_gemma/test_flash_pali_gemma_two_images.json
View File

@ -8,49 +8,49 @@
"tokens": [
{
"id": 2502,
"logprob": -1.734375,
"logprob": -1.7890625,
"special": false,
"text": "image"
},
{
"id": 2196,
"logprob": -0.5756836,
"logprob": -0.53125,
"special": false,
"text": " result"
},
{
"id": 604,
"logprob": -0.007843018,
"logprob": -0.0077209473,
"special": false,
"text": " for"
},
{
"id": 12254,
"logprob": -1.7167969,
"logprob": -1.703125,
"special": false,
"text": " chicken"
},
{
"id": 611,
"logprob": -0.17053223,
"logprob": -0.21582031,
"special": false,
"text": " on"
},
{
"id": 573,
"logprob": -0.7626953,
"logprob": -0.734375,
"special": false,
"text": " the"
},
{
"id": 8318,
"logprob": -0.02709961,
"logprob": -0.026000977,
"special": false,
"text": " beach"
},
{
"id": 1,
"logprob": -0.20739746,
"logprob": -0.2109375,
"special": true,
"text": "<eos>"
}

123
server/poetry.lock generated
View File

@ -931,20 +931,6 @@ files = [
{file = "iniconfig-2.0.0.tar.gz", hash = "sha256:2d91e135bf72d31a410b17c16da610a82cb55f6b0477d1a902134b24a455b8b3"},
]
[[package]]
name = "intel-openmp"
version = "2021.4.0"
description = "Intel OpenMP* Runtime Library"
optional = true
python-versions = "*"
files = [
{file = "intel_openmp-2021.4.0-py2.py3-none-macosx_10_15_x86_64.macosx_11_0_x86_64.whl", hash = "sha256:41c01e266a7fdb631a7609191709322da2bbf24b252ba763f125dd651bcc7675"},
{file = "intel_openmp-2021.4.0-py2.py3-none-manylinux1_i686.whl", hash = "sha256:3b921236a38384e2016f0f3d65af6732cf2c12918087128a9163225451e776f2"},
{file = "intel_openmp-2021.4.0-py2.py3-none-manylinux1_x86_64.whl", hash = "sha256:e2240ab8d01472fed04f3544a878cda5da16c26232b7ea1b59132dbfb48b186e"},
{file = "intel_openmp-2021.4.0-py2.py3-none-win32.whl", hash = "sha256:6e863d8fd3d7e8ef389d52cf97a50fe2afe1a19247e8c0d168ce021546f96fc9"},
{file = "intel_openmp-2021.4.0-py2.py3-none-win_amd64.whl", hash = "sha256:eef4c8bcc8acefd7f5cd3b9384dbf73d59e2c99fc56545712ded913f43c4a94f"},
]
[[package]]
name = "interegular"
version = "0.3.3"
@ -1153,24 +1139,6 @@ files = [
{file = "MarkupSafe-2.1.5.tar.gz", hash = "sha256:d283d37a890ba4c1ae73ffadf8046435c76e7bc2247bbb63c00bd1a709c6544b"},
]
[[package]]
name = "mkl"
version = "2021.4.0"
description = "Intel® oneAPI Math Kernel Library"
optional = true
python-versions = "*"
files = [
{file = "mkl-2021.4.0-py2.py3-none-macosx_10_15_x86_64.macosx_11_0_x86_64.whl", hash = "sha256:67460f5cd7e30e405b54d70d1ed3ca78118370b65f7327d495e9c8847705e2fb"},
{file = "mkl-2021.4.0-py2.py3-none-manylinux1_i686.whl", hash = "sha256:636d07d90e68ccc9630c654d47ce9fdeb036bb46e2b193b3a9ac8cfea683cce5"},
{file = "mkl-2021.4.0-py2.py3-none-manylinux1_x86_64.whl", hash = "sha256:398dbf2b0d12acaf54117a5210e8f191827f373d362d796091d161f610c1ebfb"},
{file = "mkl-2021.4.0-py2.py3-none-win32.whl", hash = "sha256:439c640b269a5668134e3dcbcea4350459c4a8bc46469669b2d67e07e3d330e8"},
{file = "mkl-2021.4.0-py2.py3-none-win_amd64.whl", hash = "sha256:ceef3cafce4c009dd25f65d7ad0d833a0fbadc3d8903991ec92351fe5de1e718"},
]
[package.dependencies]
intel-openmp = "==2021.*"
tbb = "==2021.*"
[[package]]
name = "mpmath"
version = "1.3.0"
@ -1465,12 +1433,13 @@ files = [
[[package]]
name = "nvidia-cudnn-cu12"
version = "8.9.2.26"
version = "9.1.0.70"
description = "cuDNN runtime libraries"
optional = true
python-versions = ">=3"
files = [
{file = "nvidia_cudnn_cu12-8.9.2.26-py3-none-manylinux1_x86_64.whl", hash = "sha256:5ccb288774fdfb07a7e7025ffec286971c06d8d7b4fb162525334616d7629ff9"},
{file = "nvidia_cudnn_cu12-9.1.0.70-py3-none-manylinux2014_x86_64.whl", hash = "sha256:165764f44ef8c61fcdfdfdbe769d687e06374059fbb388b6c89ecb0e28793a6f"},
{file = "nvidia_cudnn_cu12-9.1.0.70-py3-none-win_amd64.whl", hash = "sha256:6278562929433d68365a07a4a1546c237ba2849852c0d4b2262a486e805b977a"},
]
[package.dependencies]
@ -2841,19 +2810,6 @@ mpmath = ">=1.1.0,<1.4"
[package.extras]
dev = ["hypothesis (>=6.70.0)", "pytest (>=7.1.0)"]
[[package]]
name = "tbb"
version = "2021.13.0"
description = "Intel® oneAPI Threading Building Blocks (oneTBB)"
optional = true
python-versions = "*"
files = [
{file = "tbb-2021.13.0-py2.py3-none-manylinux1_i686.whl", hash = "sha256:a2567725329639519d46d92a2634cf61e76601dac2f777a05686fea546c4fe4f"},
{file = "tbb-2021.13.0-py2.py3-none-manylinux1_x86_64.whl", hash = "sha256:aaf667e92849adb012b8874d6393282afc318aca4407fc62f912ee30a22da46a"},
{file = "tbb-2021.13.0-py3-none-win32.whl", hash = "sha256:6669d26703e9943f6164c6407bd4a237a45007e79b8d3832fe6999576eaaa9ef"},
{file = "tbb-2021.13.0-py3-none-win_amd64.whl", hash = "sha256:3528a53e4bbe64b07a6112b4c5a00ff3c61924ee46c9c68e004a1ac7ad1f09c3"},
]
[[package]]
name = "texttable"
version = "1.7.0"
@ -2995,44 +2951,43 @@ files = [
[[package]]
name = "torch"
version = "2.3.1"
version = "2.4.0"
description = "Tensors and Dynamic neural networks in Python with strong GPU acceleration"
optional = true
python-versions = ">=3.8.0"
files = [
{file = "torch-2.3.1-cp310-cp310-manylinux1_x86_64.whl", hash = "sha256:605a25b23944be5ab7c3467e843580e1d888b8066e5aaf17ff7bf9cc30001cc3"},
{file = "torch-2.3.1-cp310-cp310-manylinux2014_aarch64.whl", hash = "sha256:f2357eb0965583a0954d6f9ad005bba0091f956aef879822274b1bcdb11bd308"},
{file = "torch-2.3.1-cp310-cp310-win_amd64.whl", hash = "sha256:32b05fe0d1ada7f69c9f86c14ff69b0ef1957a5a54199bacba63d22d8fab720b"},
{file = "torch-2.3.1-cp310-none-macosx_11_0_arm64.whl", hash = "sha256:7c09a94362778428484bcf995f6004b04952106aee0ef45ff0b4bab484f5498d"},
{file = "torch-2.3.1-cp311-cp311-manylinux1_x86_64.whl", hash = "sha256:b2ec81b61bb094ea4a9dee1cd3f7b76a44555375719ad29f05c0ca8ef596ad39"},
{file = "torch-2.3.1-cp311-cp311-manylinux2014_aarch64.whl", hash = "sha256:490cc3d917d1fe0bd027057dfe9941dc1d6d8e3cae76140f5dd9a7e5bc7130ab"},
{file = "torch-2.3.1-cp311-cp311-win_amd64.whl", hash = "sha256:5802530783bd465fe66c2df99123c9a54be06da118fbd785a25ab0a88123758a"},
{file = "torch-2.3.1-cp311-none-macosx_11_0_arm64.whl", hash = "sha256:a7dd4ed388ad1f3d502bf09453d5fe596c7b121de7e0cfaca1e2017782e9bbac"},
{file = "torch-2.3.1-cp312-cp312-manylinux1_x86_64.whl", hash = "sha256:a486c0b1976a118805fc7c9641d02df7afbb0c21e6b555d3bb985c9f9601b61a"},
{file = "torch-2.3.1-cp312-cp312-manylinux2014_aarch64.whl", hash = "sha256:224259821fe3e4c6f7edf1528e4fe4ac779c77addaa74215eb0b63a5c474d66c"},
{file = "torch-2.3.1-cp312-cp312-win_amd64.whl", hash = "sha256:e5fdccbf6f1334b2203a61a0e03821d5845f1421defe311dabeae2fc8fbeac2d"},
{file = "torch-2.3.1-cp312-none-macosx_11_0_arm64.whl", hash = "sha256:3c333dc2ebc189561514eda06e81df22bf8fb64e2384746b2cb9f04f96d1d4c8"},
{file = "torch-2.3.1-cp38-cp38-manylinux1_x86_64.whl", hash = "sha256:07e9ba746832b8d069cacb45f312cadd8ad02b81ea527ec9766c0e7404bb3feb"},
{file = "torch-2.3.1-cp38-cp38-manylinux2014_aarch64.whl", hash = "sha256:462d1c07dbf6bb5d9d2f3316fee73a24f3d12cd8dacf681ad46ef6418f7f6626"},
{file = "torch-2.3.1-cp38-cp38-win_amd64.whl", hash = "sha256:ff60bf7ce3de1d43ad3f6969983f321a31f0a45df3690921720bcad6a8596cc4"},
{file = "torch-2.3.1-cp38-none-macosx_11_0_arm64.whl", hash = "sha256:bee0bd33dc58aa8fc8a7527876e9b9a0e812ad08122054a5bff2ce5abf005b10"},
{file = "torch-2.3.1-cp39-cp39-manylinux1_x86_64.whl", hash = "sha256:aaa872abde9a3d4f91580f6396d54888620f4a0b92e3976a6034759df4b961ad"},
{file = "torch-2.3.1-cp39-cp39-manylinux2014_aarch64.whl", hash = "sha256:3d7a7f7ef21a7520510553dc3938b0c57c116a7daee20736a9e25cbc0e832bdc"},
{file = "torch-2.3.1-cp39-cp39-win_amd64.whl", hash = "sha256:4777f6cefa0c2b5fa87223c213e7b6f417cf254a45e5829be4ccd1b2a4ee1011"},
{file = "torch-2.3.1-cp39-none-macosx_11_0_arm64.whl", hash = "sha256:2bb5af780c55be68fe100feb0528d2edebace1d55cb2e351de735809ba7391eb"},
{file = "torch-2.4.0-cp310-cp310-manylinux1_x86_64.whl", hash = "sha256:4ed94583e244af51d6a8d28701ca5a9e02d1219e782f5a01dd401f90af17d8ac"},
{file = "torch-2.4.0-cp310-cp310-manylinux2014_aarch64.whl", hash = "sha256:c4ca297b7bd58b506bfd6e78ffd14eb97c0e7797dcd7965df62f50bb575d8954"},
{file = "torch-2.4.0-cp310-cp310-win_amd64.whl", hash = "sha256:2497cbc7b3c951d69b276ca51fe01c2865db67040ac67f5fc20b03e41d16ea4a"},
{file = "torch-2.4.0-cp310-none-macosx_11_0_arm64.whl", hash = "sha256:685418ab93730efbee71528821ff54005596970dd497bf03c89204fb7e3f71de"},
{file = "torch-2.4.0-cp311-cp311-manylinux1_x86_64.whl", hash = "sha256:e743adadd8c8152bb8373543964551a7cb7cc20ba898dc8f9c0cdbe47c283de0"},
{file = "torch-2.4.0-cp311-cp311-manylinux2014_aarch64.whl", hash = "sha256:7334325c0292cbd5c2eac085f449bf57d3690932eac37027e193ba775703c9e6"},
{file = "torch-2.4.0-cp311-cp311-win_amd64.whl", hash = "sha256:97730014da4c57ffacb3c09298c6ce05400606e890bd7a05008d13dd086e46b1"},
{file = "torch-2.4.0-cp311-none-macosx_11_0_arm64.whl", hash = "sha256:f169b4ea6dc93b3a33319611fcc47dc1406e4dd539844dcbd2dec4c1b96e166d"},
{file = "torch-2.4.0-cp312-cp312-manylinux1_x86_64.whl", hash = "sha256:997084a0f9784d2a89095a6dc67c7925e21bf25dea0b3d069b41195016ccfcbb"},
{file = "torch-2.4.0-cp312-cp312-manylinux2014_aarch64.whl", hash = "sha256:bc3988e8b36d1e8b998d143255d9408d8c75da4ab6dd0dcfd23b623dfb0f0f57"},
{file = "torch-2.4.0-cp312-cp312-win_amd64.whl", hash = "sha256:3374128bbf7e62cdaed6c237bfd39809fbcfaa576bee91e904706840c3f2195c"},
{file = "torch-2.4.0-cp312-none-macosx_11_0_arm64.whl", hash = "sha256:91aaf00bfe1ffa44dc5b52809d9a95129fca10212eca3ac26420eb11727c6288"},
{file = "torch-2.4.0-cp38-cp38-manylinux1_x86_64.whl", hash = "sha256:cc30457ea5489c62747d3306438af00c606b509d78822a88f804202ba63111ed"},
{file = "torch-2.4.0-cp38-cp38-manylinux2014_aarch64.whl", hash = "sha256:a046491aaf96d1215e65e1fa85911ef2ded6d49ea34c8df4d0638879f2402eef"},
{file = "torch-2.4.0-cp38-cp38-win_amd64.whl", hash = "sha256:688eec9240f3ce775f22e1e1a5ab9894f3d5fe60f3f586deb7dbd23a46a83916"},
{file = "torch-2.4.0-cp38-none-macosx_11_0_arm64.whl", hash = "sha256:3af4de2a618fb065e78404c4ba27a818a7b7957eaeff28c6c66ce7fb504b68b8"},
{file = "torch-2.4.0-cp39-cp39-manylinux1_x86_64.whl", hash = "sha256:618808d3f610d5f180e47a697d4ec90b810953bb1e020f424b2ac7fb0884b545"},
{file = "torch-2.4.0-cp39-cp39-manylinux2014_aarch64.whl", hash = "sha256:ed765d232d23566052ba83632ec73a4fccde00b4c94ad45d63b471b09d63b7a7"},
{file = "torch-2.4.0-cp39-cp39-win_amd64.whl", hash = "sha256:a2feb98ac470109472fb10dfef38622a7ee08482a16c357863ebc7bc7db7c8f7"},
{file = "torch-2.4.0-cp39-none-macosx_11_0_arm64.whl", hash = "sha256:8940fc8b97a4c61fdb5d46a368f21f4a3a562a17879e932eb51a5ec62310cb31"},
]
[package.dependencies]
filelock = "*"
fsspec = "*"
jinja2 = "*"
mkl = {version = ">=2021.1.1,<=2021.4.0", markers = "platform_system == \"Windows\""}
networkx = "*"
nvidia-cublas-cu12 = {version = "12.1.3.1", markers = "platform_system == \"Linux\" and platform_machine == \"x86_64\""}
nvidia-cuda-cupti-cu12 = {version = "12.1.105", markers = "platform_system == \"Linux\" and platform_machine == \"x86_64\""}
nvidia-cuda-nvrtc-cu12 = {version = "12.1.105", markers = "platform_system == \"Linux\" and platform_machine == \"x86_64\""}
nvidia-cuda-runtime-cu12 = {version = "12.1.105", markers = "platform_system == \"Linux\" and platform_machine == \"x86_64\""}
nvidia-cudnn-cu12 = {version = "8.9.2.26", markers = "platform_system == \"Linux\" and platform_machine == \"x86_64\""}
nvidia-cudnn-cu12 = {version = "9.1.0.70", markers = "platform_system == \"Linux\" and platform_machine == \"x86_64\""}
nvidia-cufft-cu12 = {version = "11.0.2.54", markers = "platform_system == \"Linux\" and platform_machine == \"x86_64\""}
nvidia-curand-cu12 = {version = "10.3.2.106", markers = "platform_system == \"Linux\" and platform_machine == \"x86_64\""}
nvidia-cusolver-cu12 = {version = "11.4.5.107", markers = "platform_system == \"Linux\" and platform_machine == \"x86_64\""}
@ -3040,12 +2995,12 @@ nvidia-cusparse-cu12 = {version = "12.1.0.106", markers = "platform_system == \"
nvidia-nccl-cu12 = {version = "2.20.5", markers = "platform_system == \"Linux\" and platform_machine == \"x86_64\""}
nvidia-nvtx-cu12 = {version = "12.1.105", markers = "platform_system == \"Linux\" and platform_machine == \"x86_64\""}
sympy = "*"
triton = {version = "2.3.1", markers = "platform_system == \"Linux\" and platform_machine == \"x86_64\" and python_version < \"3.12\""}
triton = {version = "3.0.0", markers = "platform_system == \"Linux\" and platform_machine == \"x86_64\" and python_version < \"3.13\""}
typing-extensions = ">=4.8.0"
[package.extras]
opt-einsum = ["opt-einsum (>=3.3)"]
optree = ["optree (>=0.9.1)"]
optree = ["optree (>=0.11.0)"]
[[package]]
name = "tqdm"
@ -3137,17 +3092,21 @@ vision = ["Pillow (>=10.0.1,<=15.0)"]
[[package]]
name = "triton"
version = "2.3.1"
version = "3.0.0"
description = "A language and compiler for custom Deep Learning operations"
optional = true
python-versions = "*"
files = [
{file = "triton-2.3.1-cp310-cp310-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:3c84595cbe5e546b1b290d2a58b1494df5a2ef066dd890655e5b8a8a92205c33"},
{file = "triton-2.3.1-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:c9d64ae33bcb3a7a18081e3a746e8cf87ca8623ca13d2c362413ce7a486f893e"},
{file = "triton-2.3.1-cp312-cp312-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:eaf80e8761a9e3498aa92e7bf83a085b31959c61f5e8ac14eedd018df6fccd10"},
{file = "triton-2.3.1-cp37-cp37m-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:b13bf35a2b659af7159bf78e92798dc62d877aa991de723937329e2d382f1991"},
{file = "triton-2.3.1-cp38-cp38-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:63381e35ded3304704ea867ffde3b7cfc42c16a55b3062d41e017ef510433d66"},
{file = "triton-2.3.1-cp39-cp39-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:1d968264523c7a07911c8fb51b4e0d1b920204dae71491b1fe7b01b62a31e124"},
{file = "triton-3.0.0-1-cp310-cp310-manylinux2014_x86_64.manylinux_2_17_x86_64.whl", hash = "sha256:e1efef76935b2febc365bfadf74bcb65a6f959a9872e5bddf44cc9e0adce1e1a"},
{file = "triton-3.0.0-1-cp311-cp311-manylinux2014_x86_64.manylinux_2_17_x86_64.whl", hash = "sha256:5ce8520437c602fb633f1324cc3871c47bee3b67acf9756c1a66309b60e3216c"},
{file = "triton-3.0.0-1-cp312-cp312-manylinux2014_x86_64.manylinux_2_17_x86_64.whl", hash = "sha256:34e509deb77f1c067d8640725ef00c5cbfcb2052a1a3cb6a6d343841f92624eb"},
{file = "triton-3.0.0-1-cp38-cp38-manylinux2014_x86_64.manylinux_2_17_x86_64.whl", hash = "sha256:bcbf3b1c48af6a28011a5c40a5b3b9b5330530c3827716b5fbf6d7adcc1e53e9"},
{file = "triton-3.0.0-1-cp39-cp39-manylinux2014_x86_64.manylinux_2_17_x86_64.whl", hash = "sha256:6e5727202f7078c56f91ff13ad0c1abab14a0e7f2c87e91b12b6f64f3e8ae609"},
{file = "triton-3.0.0-cp310-cp310-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:39b052da883351fdf6be3d93cedae6db3b8e3988d3b09ed221bccecfa9612230"},
{file = "triton-3.0.0-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:cd34f19a8582af96e6291d4afce25dac08cb2a5d218c599163761e8e0827208e"},
{file = "triton-3.0.0-cp312-cp312-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:0d5e10de8c011adeb7c878c6ce0dd6073b14367749e34467f1cff2bde1b78253"},
{file = "triton-3.0.0-cp38-cp38-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:e8903767951bf86ec960b4fe4e21bc970055afc65e9d57e916d79ae3c93665e3"},
{file = "triton-3.0.0-cp39-cp39-manylinux_2_17_x86_64.manylinux2014_x86_64.whl", hash = "sha256:41004fb1ae9a53fcb3e970745feb87f0e3c94c6ce1ba86e95fa3b8537894bef7"},
]
[package.dependencies]
@ -3155,8 +3114,8 @@ filelock = "*"
[package.extras]
build = ["cmake (>=3.20)", "lit"]
tests = ["autopep8", "flake8", "isort", "numpy", "pytest", "scipy (>=1.7.1)", "torch"]
tutorials = ["matplotlib", "pandas", "tabulate", "torch"]
tests = ["autopep8", "flake8", "isort", "llnl-hatchet", "numpy", "pytest", "scipy (>=1.7.1)"]
tutorials = ["matplotlib", "pandas", "tabulate"]
[[package]]
name = "typer"
@ -3556,4 +3515,4 @@ torch = ["torch"]
[metadata]
lock-version = "2.0"
python-versions = ">=3.9,<3.13"
content-hash = "52d2dda4603443ee76ce880f3f1fdf7f044e7ba6df9973d230f52b0fbc0dbf24"
content-hash = "c94bbdf8131750891fb3f7132066718534129d85a4c09126d8d01c2de6c72798"

2
server/text_generation_server/layers/gptq/custom_autotune.py
View File

@ -91,7 +91,7 @@ class Autotuner(triton.KernelInterface):
kernel_call, quantiles=(0.5, 0.2, 0.8), rep=40
)
except triton.OutOfResources:
return (float("inf"), float("inf"), float("inf"))
return [float("inf"), float("inf"), float("inf")]
def run(self, *args, **kwargs):
self.nargs = dict(zip(self.arg_names, args))