frozenleaves/pytorch
1
0
Fork 0
mirror of https://github.com/pytorch/pytorch.git synced 2025-10-20 21:14:14 +08:00
Code Issues Packages Projects Releases Wiki Activity

Update HF version to commit hash (6c26faa) (#107400)

Browse Source 
Some [errors](https://ossci-raw-job-status.s3.amazonaws.com/log/15968424899) in the [torchinductor hf benchmarks](https://hud.pytorch.org/benchmark/huggingface/inductor_aot_inductor?startTime=Thu,%2010%20Aug%202023%2018:05:47%20GMT&stopTime=Thu,%2017%20Aug%202023%2018:05:47%20GMT&granularity=hour&mode=inference&dtype=bfloat16&lBranch=main&lCommit=384e0d104fd077d31efafc564129660e9b7a0f25&rBranch=main&rCommit=03414081ff7ee011e17ee10f9ddb2584811bf965) should be fixed in the most recent release (for example, this [line](c036c814f4/src/transformers/models/opt/modeling_opt.py (L688)) no longer exists). Additionally, I landed a [commit (6c26faa)](6c26faa159) to the HF transformers repro to fix one of the graph breaks. This PR results in [76% pass rate for the export + aot inductor HF benchmark!](https://hud.pytorch.org/benchmark/compilers?startTime=Thu%2C%2010%20Aug%202023%2022%3A45%3A09%20GMT&stopTime=Thu%2C%2017%20Aug%202023%2022%3A45%3A09%20GMT&granularity=hour&suite=torchbench&mode=inference&dtype=bfloat16&lBranch=angelayi/hf_version&lCommit=0accaaca2fa70ca2f78c1a587dd4b6750448dd90&rBranch=main&rCommit=03414081ff7ee011e17ee10f9ddb2584811bf965)

Pull Request resolved: https://github.com/pytorch/pytorch/pull/107400
Approved by: https://github.com/ezyang, https://github.com/desertfire, https://github.com/malfet
This commit is contained in:
angelayi
2023-09-12 15:25:26 +00:00
committed by PyTorch MergeBot
parent 58391aeaf1
commit c3945b5f84
3 changed files with 4 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
.ci/docker/ci_commit_pins/huggingface.txt
View File

@ -1 +1 @@
4.27.4
6c26faa159b79a42d7fa46cb66e2d21523351987

4
.ci/docker/common/install_inductor_benchmark_deps.sh
View File

@ -6,9 +6,9 @@ source "$(dirname "${BASH_SOURCE[0]}")/common_utils.sh"
function install_huggingface() {
local version
version=$(get_pinned_commit huggingface)
commit=$(get_pinned_commit huggingface)
pip_install pandas==2.0.3
pip_install "transformers==${version}"
pip_install "git+https://github.com/huggingface/transformers@${commit}"
}
function install_timm() {

1
benchmarks/dynamo/common.py
View File

@ -225,6 +225,7 @@ CI_SKIP[CI("inductor", training=False, device="cpu")] = [
"resnet50_quantized_qat", # Eager model failed to run(Quantize only works on Float Tensor, got Double)
"sage", # does not work with fp32
# Huggingface
"GPT2ForSequenceClassification", # Accuracy https://github.com/pytorch/pytorch/issues/109019
"MBartForConditionalGeneration", # Accuracy https://github.com/pytorch/pytorch/issues/94793
"PLBartForConditionalGeneration", # Accuracy https://github.com/pytorch/pytorch/issues/94794
# TIMM