matt/ollama37
1
0
Fork 0
mirror of https://github.com/dogkeeper886/ollama37.git synced 2025-12-20 12:47:00 +00:00
Code Issues Packages Projects Releases Wiki Activity

Add multi-model inference tests for gemma3 12b and 27b

Browse Source 
- TC-INFERENCE-004: gemma3:12b single GPU test
- TC-INFERENCE-005: gemma3:27b dual-GPU test (K80 layer split)
- Each test unloads previous model before loading next
- Workflows unload all 3 model sizes after inference suite
- 27b test verifies both GPUs have memory allocated
This commit is contained in:
Shang Chieh Tseng
2025-12-17 17:01:25 +08:00
parent 22e77e0dde
commit 806232d95f
4 changed files with 290 additions and 6 deletions
Download Patch File Download Diff File Expand all files Collapse all files

8
.github/workflows/full-pipeline.yml vendored
View File

@@ -165,12 +165,14 @@ jobs:
exit 1
fi
- name: Unload test model from VRAM
- name: Unload test models from VRAM
if: always()
run: |
echo "Unloading gemma3:4b from VRAM..."
echo "Unloading all test models from VRAM..."
curl -s http://localhost:11434/api/generate -d '{"model":"gemma3:4b","keep_alive":0}' || true
echo "Model unloaded"
curl -s http://localhost:11434/api/generate -d '{"model":"gemma3:12b","keep_alive":0}' || true
curl -s http://localhost:11434/api/generate -d '{"model":"gemma3:27b","keep_alive":0}' || true
echo "All models unloaded"
- name: Upload inference results
uses: actions/upload-artifact@v4

8
.github/workflows/inference.yml vendored
View File

@@ -111,12 +111,14 @@ jobs:
exit 1
fi
- name: Unload test model from VRAM
- name: Unload test models from VRAM
if: always()
run: |
echo "Unloading gemma3:4b from VRAM..."
echo "Unloading all test models from VRAM..."
curl -s http://localhost:11434/api/generate -d '{"model":"gemma3:4b","keep_alive":0}' || true
echo "Model unloaded"
curl -s http://localhost:11434/api/generate -d '{"model":"gemma3:12b","keep_alive":0}' || true
curl -s http://localhost:11434/api/generate -d '{"model":"gemma3:27b","keep_alive":0}' || true
echo "All models unloaded"
- name: Upload inference results
uses: actions/upload-artifact@v4