Skip to content

(Gaudi2) Non-regression tests #151

(Gaudi2) Non-regression tests

(Gaudi2) Non-regression tests #151

name: (Gaudi2) Non-regression tests
on:
workflow_dispatch:
schedule:
- cron: '0 23 * * *' # every Wednesday and Saturday at 1am CET (midnight winter time)
concurrency:
group: ${{ github.workflow }}
jobs:
stable-diffusion:
name: Test Stable Diffusion
runs-on: [self-hosted, linux, x64, gaudi2]
steps:
- name: Checkout
uses: actions/checkout@v2
- name: Pull image
run: |
docker pull vault.habana.ai/gaudi-docker/1.18.0/ubuntu22.04/habanalabs/pytorch-installer-2.4.0:latest
- name: Run tests
run: |
docker run \
--rm \
-v $PWD:/root/workspace \
-v /scratch-1:/data \
--workdir=/root/workspace \
--runtime=habana \
-e HABANA_VISIBLE_DEVICES=all \
-e OMPI_MCA_btl_vader_single_copy_mechanism=none \
-e GAUDI2_CI=1 \
-e HF_HOME=/data \
--cap-add=sys_nice \
--net=host \
--ipc=host \
vault.habana.ai/gaudi-docker/1.18.0/ubuntu22.04/habanalabs/pytorch-installer-2.4.0:latest \
/bin/bash tests/ci/slow_tests_diffusers.sh
deepspeed:
name: Test DeepSpeed models
if: ${{ !cancelled() && (success() || failure()) }}
needs:
- stable-diffusion # run the job when the previous test job is done
runs-on: [self-hosted, linux, x64, gaudi2]
steps:
- name: Checkout
uses: actions/checkout@v2
- name: Pull image
run: |
docker pull vault.habana.ai/gaudi-docker/1.18.0/ubuntu22.04/habanalabs/pytorch-installer-2.4.0:latest
- name: Run tests
run: |
docker run \
--rm \
-v $PWD:/root/workspace \
-v /scratch-1:/data \
--workdir=/root/workspace \
--runtime=habana \
-e HABANA_VISIBLE_DEVICES=all \
-e OMPI_MCA_btl_vader_single_copy_mechanism=none \
-e GAUDI2_CI=1 \
-e HF_HOME=/data \
--cap-add=sys_nice \
--net=host \
--ipc=host \
vault.habana.ai/gaudi-docker/1.18.0/ubuntu22.04/habanalabs/pytorch-installer-2.4.0:latest \
/bin/bash tests/ci/slow_tests_deepspeed.sh ${{ secrets.TEXT_GENERATION_CI_HUB_TOKEN }}
fsdp:
name: Test FSDP models
if: ${{ !cancelled() && (success() || failure()) }}
needs:
- deepspeed # run the job when the previous test job is done
runs-on: [self-hosted, linux, x64, gaudi2]
steps:
- name: Checkout
uses: actions/checkout@v2
- name: Pull image
run: |
docker pull vault.habana.ai/gaudi-docker/1.18.0/ubuntu22.04/habanalabs/pytorch-installer-2.4.0:latest
- name: Run tests
run: |
docker run \
--rm \
-v $PWD:/root/workspace \
-v /scratch-1:/data \
--workdir=/root/workspace \
--runtime=habana \
-e HABANA_VISIBLE_DEVICES=all \
-e OMPI_MCA_btl_vader_single_copy_mechanism=none \
-e GAUDI2_CI=1 \
-e HF_HOME=/data \
--cap-add=sys_nice \
--net=host \
--ipc=host \
vault.habana.ai/gaudi-docker/1.18.0/ubuntu22.04/habanalabs/pytorch-installer-2.4.0:latest \
make slow_tests_fsdp TOKEN=${{ secrets.TEXT_GENERATION_CI_HUB_TOKEN }}
multi-card:
name: Test multi-card models
if: ${{ !cancelled() && (success() || failure()) }}
needs:
- fsdp # run the job when the previous test job is done
runs-on: [self-hosted, linux, x64, gaudi2]
steps:
- name: Checkout
uses: actions/checkout@v2
- name: Pull image
run: |
docker pull vault.habana.ai/gaudi-docker/1.18.0/ubuntu22.04/habanalabs/pytorch-installer-2.4.0:latest
- name: Run tests
run: |
docker run \
--rm \
-v $PWD:/root/workspace \
-v /scratch-1:/data \
--workdir=/root/workspace \
--runtime=habana \
-e HABANA_VISIBLE_DEVICES=all \
-e OMPI_MCA_btl_vader_single_copy_mechanism=none \
-e GAUDI2_CI=1 \
-e HF_HOME=/data \
--cap-add=sys_nice \
--net=host \
--ipc=host \
vault.habana.ai/gaudi-docker/1.18.0/ubuntu22.04/habanalabs/pytorch-installer-2.4.0:latest \
/bin/bash tests/ci/slow_tests_8x.sh
single-card:
name: Test single-card models
if: ${{ !cancelled() && (success() || failure()) }}
needs:
- deepspeed
- multi-card # run the job when the previous test jobs are done
runs-on: [self-hosted, linux, x64, gaudi2]
steps:
- name: Checkout
uses: actions/checkout@v2
- name: Pull image
run: |
docker pull vault.habana.ai/gaudi-docker/1.18.0/ubuntu22.04/habanalabs/pytorch-installer-2.4.0:latest
- name: Run tests
run: |
docker run \
--rm \
-v $PWD:/root/workspace \
-v /scratch-1:/data \
--workdir=/root/workspace \
--runtime=habana \
-e HABANA_VISIBLE_DEVICES=all \
-e OMPI_MCA_btl_vader_single_copy_mechanism=none \
-e GAUDI2_CI=1 \
-e RUN_ALBERT_XXL_1X=1 \
-e HF_HOME=/data \
--cap-add=sys_nice \
--net=host \
--ipc=host \
vault.habana.ai/gaudi-docker/1.18.0/ubuntu22.04/habanalabs/pytorch-installer-2.4.0:latest \
/bin/bash tests/ci/slow_tests_1x.sh
text-generation:
name: Test text-generation example
if: ${{ !cancelled() && (success() || failure()) }}
needs:
- deepspeed
- multi-card
- single-card # run the job when the previous test jobs are done
runs-on: [self-hosted, linux, x64, gaudi2]
steps:
- name: Checkout
uses: actions/checkout@v2
- name: Pull image
run: |
docker pull vault.habana.ai/gaudi-docker/1.18.0/ubuntu22.04/habanalabs/pytorch-installer-2.4.0:latest
- name: Run tests
run: |
docker run \
--rm \
-v $PWD:/root/workspace \
-v /scratch-1:/data \
--workdir=/root/workspace \
--runtime=habana \
-e HABANA_VISIBLE_DEVICES=all \
-e OMPI_MCA_btl_vader_single_copy_mechanism=none \
-e GAUDI2_CI=1 \
-e HF_HOME=/data \
--cap-add=sys_nice \
--net=host \
--ipc=host \
vault.habana.ai/gaudi-docker/1.18.0/ubuntu22.04/habanalabs/pytorch-installer-2.4.0:latest \
make slow_tests_text_generation_example TOKEN=${{ secrets.TEXT_GENERATION_CI_HUB_TOKEN }}
trl:
name: Test TRL integration
if: ${{ !cancelled() && (success() || failure()) }}
needs:
- text-generation
runs-on: [self-hosted, linux, x64, gaudi2]
steps:
- name: Checkout
uses: actions/checkout@v2
- name: Pull image
run: |
docker pull vault.habana.ai/gaudi-docker/1.18.0/ubuntu22.04/habanalabs/pytorch-installer-2.4.0:latest
- name: Run tests
run: |
docker run \
--rm \
-v $PWD:/root/workspace \
-v /scratch-1:/data \
--workdir=/root/workspace \
--runtime=habana \
-e HABANA_VISIBLE_DEVICES=all \
-e OMPI_MCA_btl_vader_single_copy_mechanism=none \
-e GAUDI2_CI=1 \
-e HF_HOME=/data \
--cap-add=sys_nice \
--net=host \
--ipc=host \
vault.habana.ai/gaudi-docker/1.18.0/ubuntu22.04/habanalabs/pytorch-installer-2.4.0:latest \
/bin/bash tests/ci/slow_tests_trl.sh
sentence-transformers:
name: Test Sentence Transformers integration
if: ${{ !cancelled() && (success() || failure()) }}
needs:
- trl
runs-on: [self-hosted, linux, x64, gaudi2]
steps:
- name: Checkout Optimum Habana
uses: actions/checkout@v2
with:
repository: 'huggingface/optimum-habana'
path: optimum-habana
- name: Checkout Sentence Transformers
uses: actions/checkout@v2
with:
repository: 'UKPLab/sentence-transformers'
path: sentence-transformers
- name: Pull image
run: |
docker pull vault.habana.ai/gaudi-docker/1.18.0/ubuntu22.04/habanalabs/pytorch-installer-2.4.0:latest
- name: Run tests
run: |
docker run \
--rm \
-v $PWD:/root/workspace \
-v /scratch-1:/data \
--workdir=/root/workspace \
--runtime=habana \
-e HABANA_VISIBLE_DEVICES=all \
-e OMPI_MCA_btl_vader_single_copy_mechanism=none \
-e GAUDI2_CI=1 \
-e HF_HOME=/data \
--cap-add=sys_nice \
--net=host \
--ipc=host \
vault.habana.ai/gaudi-docker/1.18.0/ubuntu22.04/habanalabs/pytorch-installer-2.4.0:latest \
/bin/bash optimum-habana/tests/ci/sentence_transformers.sh