add deepspeed fairscale (#10116)
This commit is contained in:
11
.github/workflows/self-scheduled.yml
vendored
11
.github/workflows/self-scheduled.yml
vendored
@@ -67,7 +67,7 @@ jobs:
|
|||||||
- name: Failure short reports
|
- name: Failure short reports
|
||||||
if: ${{ always() }}
|
if: ${{ always() }}
|
||||||
run: cat reports/tests_torch_gpu_failures_short.txt
|
run: cat reports/tests_torch_gpu_failures_short.txt
|
||||||
|
|
||||||
- name: Run examples tests on GPU
|
- name: Run examples tests on GPU
|
||||||
if: ${{ always() }}
|
if: ${{ always() }}
|
||||||
env:
|
env:
|
||||||
@@ -157,7 +157,7 @@ jobs:
|
|||||||
run: |
|
run: |
|
||||||
source .env/bin/activate
|
source .env/bin/activate
|
||||||
python -m pytest -n 1 --dist=loadfile -s --make-reports=tests_tf_gpu tests
|
python -m pytest -n 1 --dist=loadfile -s --make-reports=tests_tf_gpu tests
|
||||||
|
|
||||||
- name: Failure short reports
|
- name: Failure short reports
|
||||||
if: ${{ always() }}
|
if: ${{ always() }}
|
||||||
run: cat reports/tests_tf_gpu_failures_short.txt
|
run: cat reports/tests_tf_gpu_failures_short.txt
|
||||||
@@ -183,7 +183,7 @@ jobs:
|
|||||||
with:
|
with:
|
||||||
name: run_all_tests_tf_gpu_test_reports
|
name: run_all_tests_tf_gpu_test_reports
|
||||||
path: reports
|
path: reports
|
||||||
|
|
||||||
run_all_tests_torch_multi_gpu:
|
run_all_tests_torch_multi_gpu:
|
||||||
runs-on: [self-hosted, gpu, multi-gpu]
|
runs-on: [self-hosted, gpu, multi-gpu]
|
||||||
steps:
|
steps:
|
||||||
@@ -221,6 +221,8 @@ jobs:
|
|||||||
pip install --upgrade pip
|
pip install --upgrade pip
|
||||||
pip install .[torch,sklearn,testing,onnxruntime,sentencepiece]
|
pip install .[torch,sklearn,testing,onnxruntime,sentencepiece]
|
||||||
pip install git+https://github.com/huggingface/datasets
|
pip install git+https://github.com/huggingface/datasets
|
||||||
|
pip install fairscale
|
||||||
|
pip install deepspeed
|
||||||
pip list
|
pip list
|
||||||
|
|
||||||
- name: Are GPUs recognized by our DL frameworks
|
- name: Are GPUs recognized by our DL frameworks
|
||||||
@@ -342,7 +344,7 @@ jobs:
|
|||||||
run: |
|
run: |
|
||||||
source .env/bin/activate
|
source .env/bin/activate
|
||||||
python -m pytest -n 1 --dist=loadfile -s -m is_pipeline_test --make-reports=tests_tf_pipeline_multi_gpu tests
|
python -m pytest -n 1 --dist=loadfile -s -m is_pipeline_test --make-reports=tests_tf_pipeline_multi_gpu tests
|
||||||
|
|
||||||
- name: Failure short reports
|
- name: Failure short reports
|
||||||
if: ${{ always() }}
|
if: ${{ always() }}
|
||||||
run: cat reports/tests_tf_pipeline_multi_gpu_failures_short.txt
|
run: cat reports/tests_tf_pipeline_multi_gpu_failures_short.txt
|
||||||
@@ -353,4 +355,3 @@ jobs:
|
|||||||
with:
|
with:
|
||||||
name: run_all_tests_tf_multi_gpu_test_reports
|
name: run_all_tests_tf_multi_gpu_test_reports
|
||||||
path: reports
|
path: reports
|
||||||
|
|
||||||
|
|||||||
Reference in New Issue
Block a user