remove irrelevant code
This commit is contained in:
parent
d11ef0d117
commit
2c4e610743
1 changed files with 1 additions and 71 deletions
70
.github/workflows/llm-harness-evaluation.yml
vendored
70
.github/workflows/llm-harness-evaluation.yml
vendored
|
|
@ -99,11 +99,6 @@ jobs:
|
|||
strategy:
|
||||
fail-fast: false
|
||||
matrix:
|
||||
# include:
|
||||
# python-version: "3.9"
|
||||
# model_name: "stablelm-3b-4e1t"
|
||||
# task: "arc"
|
||||
# precision: "sym_int4" #options: sym_int4, fp4, mixed_fp4, sym_int8, fp8, mixed_fp8
|
||||
python-version: ["3.9"]
|
||||
model_name: ${{ fromJson(needs.set-matrix.outputs.model_name) }}
|
||||
task: ${{ fromJson(needs.set-matrix.outputs.task) }}
|
||||
|
|
@ -138,20 +133,6 @@ jobs:
|
|||
uses: ./.github/actions/llm/setup-llm-env
|
||||
with:
|
||||
extra-dependency: "xpu_2.1"
|
||||
# run: |
|
||||
# retry_count=0
|
||||
# max_retries=1
|
||||
# command="bash ./.github/actions/llm/setup-llm-env --extra-dependency xpu_2.1"
|
||||
# until $command; do
|
||||
# exit_code=$?
|
||||
# echo "Attempt $((retry_count+1)) failed with exit code $exit_code. Retrying..."
|
||||
# retry_count=$((retry_count+1))
|
||||
# if [ "$retry_count" -gt "$max_retries" ]; then
|
||||
# echo "Reached maximum retry attempts. Exiting."
|
||||
# exit $exit_code
|
||||
# fi
|
||||
# sleep 5
|
||||
# done
|
||||
|
||||
- name: Install harness
|
||||
working-directory: ${{ github.workspace }}/python/llm/dev/benchmark/harness/
|
||||
|
|
@ -224,8 +205,6 @@ jobs:
|
|||
if: ${{ always() }}
|
||||
needs: llm-harness-evalution
|
||||
runs-on: ubuntu-latest
|
||||
# env:
|
||||
# OUTPUT_PATH: ${{ needs.llm-harness-evalution.outputs.output_path }}
|
||||
steps:
|
||||
- uses: actions/checkout@v3
|
||||
- name: Set up Python 3.9
|
||||
|
|
@ -252,52 +231,3 @@ jobs:
|
|||
echo ${{ env.OUTPUT_PATH }}
|
||||
ls ${{ env.OUTPUT_PATH }}
|
||||
python ${{ github.workspace }}/python/llm/dev/benchmark/harness/make_table_results.py ${{ env.OUTPUT_PATH }}
|
||||
|
||||
# TODO: add a nightly summary job
|
||||
# llm-harness-summary-nightly:
|
||||
# if: ${{github.event_name == 'schedule' || github.event_name == 'pull_request'}}
|
||||
# needs: llm-harness-evalution
|
||||
# runs-on: '["self-hosted", "llm", "temp-arc01"]'
|
||||
# env:
|
||||
# OUTPUT_PATH: ${{ needs.llm-harness-evalution.outputs.output_path }}
|
||||
# steps:
|
||||
# - uses: actions/checkout@v3
|
||||
# - name: Set up Python 3.9
|
||||
# uses: actions/setup-python@v4
|
||||
# with:
|
||||
# python-version: 3.9
|
||||
|
||||
# - name: Install dependencies
|
||||
# shell: bash
|
||||
# run: |
|
||||
# pip install --upgrade pip
|
||||
# pip install jsonlines pytablewriter regex
|
||||
|
||||
# - name: Download all results for nightly run
|
||||
# if: github.event_name == 'schedule'
|
||||
# uses: actions/download-artifact@v3
|
||||
# with:
|
||||
# name: harness_results
|
||||
# path: /home/arda/harness-action-runners/nightly-accuracy-data/${{ env.OUTPUT_PATH }}
|
||||
|
||||
# - name: Download all results for pull request
|
||||
# if: github.event_name == 'pull_request'
|
||||
# uses: actions/download-artifact@v3
|
||||
# with:
|
||||
# name: harness_results
|
||||
# path: /home/arda/harness-action-runners/pr-accuracy-data/${{ env.OUTPUT_PATH }}
|
||||
|
||||
# - name: Summarize the results for nightly run
|
||||
# if: github.event_name == 'schedule'
|
||||
# shell: bash
|
||||
# run: |
|
||||
# ls /home/arda/harness-action-runners/nightly-accuracy-data/${{ env.OUTPUT_PATH }}
|
||||
# python ${{ github.workspace }}/python/llm/dev/benchmark/harness/make_table_results.py /home/arda/harness-action-runners/nightly-accuracy-data/${{ env.OUTPUT_PATH }}
|
||||
|
||||
# - name: Summarize the results for pull request
|
||||
# if: github.event_name == 'pull_request'
|
||||
# shell: bash
|
||||
# run: |
|
||||
# ls /home/arda/harness-action-runners/pr-accuracy-data/${{ env.OUTPUT_PATH }}
|
||||
# python ${{ github.workspace }}/python/llm/dev/benchmark/harness/make_table_results.py /home/arda/harness-action-runners/pr-accuracy-data/${{ env.OUTPUT_PATH }}
|
||||
|
||||
Loading…
Reference in a new issue