Skip to content

[CI] fix sglang test name & add status logs & remove exllama packing test #1312

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 5 commits into from
Feb 21, 2025
Merged

[CI] fix sglang test name & add status logs & remove exllama packing test #1312

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
5 commits
Select commit Hold shift + click to select a range
179fc6b
[CI] fix sglang test name
CSY-ModelCloud Feb 21, 2025
c926734
[CI] don't pass gpu to ipex
CSY-ModelCloud Feb 21, 2025
c19a94d
[CI] print status after getting gpu
CSY-ModelCloud Feb 21, 2025
7994f20
remove packing test for exllama
CSY-ModelCloud Feb 21, 2025
7b7d96e
add triton & torch test back
CSY-ModelCloud Feb 21, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
9 changes: 8 additions & 1 deletion .github/workflows/unit_tests.yml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -417,6 +417,7 @@ jobs:
if [ "$gpu_id" -lt 0 ]; then
echo "http://${{ needs.check-vm.outputs.ip }}/gpu/get?id=${{ github.run_id }}&timestamp=$timestamp&test=${{ matrix.test_script }}&runner=${RUNNER_NAME}&exclusive=${{ github.event.inputs.exclusive-gpu }} returned $gpu_id"
echo "No available GPU, waiting 5 seconds..."
curl http://${{ needs.check-vm.outputs.ip }}/gpu/status2
sleep 5
else
echo "Allocated GPU ID: $gpu_id"
Expand All @@ -428,6 +429,7 @@ jobs:
echo "CUDA_VISIBLE_DEVICES=$gpu_id" >> $GITHUB_ENV
echo "STEP_TIMESTAMP=$timestamp" >> $GITHUB_ENV
echo "CUDA_VISIBLE_DEVICES set to $gpu_id, timestamp=$timestamp"
curl http://${{ needs.check-vm.outputs.ip }}/gpu/status2

- name: Run tests
if: ${{ (!github.event.inputs.test_names || contains(github.event.inputs.test_names, matrix.test_script)) && !cancelled() }}
Expand Down Expand Up @@ -563,7 +565,7 @@ jobs:
uv pip install colorlog
fi

if [[ "${{ matrix.test_script }}" == "test_sglang.py" ]]; then
if [[ "${{ matrix.test_script }}" == "test_sglang" ]]; then
uv pip install transformers==4.48.3
fi

Expand Down Expand Up @@ -616,6 +618,7 @@ jobs:
if [ "$gpu_id" -lt 0 ]; then
echo "http://${{ needs.check-vm.outputs.ip }}/gpu/get?id=${{ github.run_id }}&timestamp=$timestamp&test=${{ matrix.test_script }}&runner=${RUNNER_NAME}&exclusive=${{ github.event.inputs.exclusive-gpu }} returned $gpu_id"
echo "No available GPU, waiting 5 seconds..."
curl http://${{ needs.check-vm.outputs.ip }}/gpu/status2
sleep 5
else
echo "Allocated GPU ID: $gpu_id"
Expand All @@ -627,10 +630,14 @@ jobs:
echo "CUDA_VISIBLE_DEVICES=$gpu_id" >> $GITHUB_ENV
echo "STEP_TIMESTAMP=$timestamp" >> $GITHUB_ENV
echo "CUDA_VISIBLE_DEVICES set to $gpu_id, timestamp=$timestamp"
curl http://${{ needs.check-vm.outputs.ip }}/gpu/status2

- name: Run tests
if: ${{ (!github.event.inputs.test_names || contains(github.event.inputs.test_names, matrix.test_script)) && !cancelled() }}
run: |
if [[ "${{ matrix.test_script }}" == *ipex* ]]; then
export CUDA_VISIBLE_DEVICES=""
fi
if [[ "${{ matrix.test_script }}" == *xpu* ]]; then
export CUDA_VISIBLE_DEVICES=""
source /etc/profile.d/pyenv.sh && pyenv activate xpu
Expand Down
19 changes: 3 additions & 16 deletions tests/test_packing.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -100,15 +100,6 @@ def pack(self, qlinearCls):
return qlinear

def test_compare_exllama_triton_torch(self):
# validate exllama packer
exllama_linear = self.pack(ExllamaQuantLinear)

dequantized_weight, dequantized_qzeros = dequantize_4bits_weight(exllama_linear)
dequantized_weight = dequantized_weight.to(torch.float16)

self.assertTrue(torch.equal(dequantized_weight, self.linear.weight))
self.assertTrue(torch.all(dequantized_qzeros == 8))

triton_linear = self.pack(TritonV2QuantLinear)

dequantized_weight, dequantized_qzeros = dequantize_4bits_weight(triton_linear)
Expand All @@ -117,10 +108,6 @@ def test_compare_exllama_triton_torch(self):
self.assertTrue(torch.equal(dequantized_weight, self.linear.weight))
self.assertTrue(torch.all(dequantized_qzeros == 8))

self.assertTrue(torch.allclose(exllama_linear.qweight, triton_linear.qweight))
self.assertTrue(torch.allclose(exllama_linear.scales, triton_linear.scales))
self.assertTrue(torch.allclose(exllama_linear.qzeros, triton_linear.qzeros))

# validate torch packer
torch_linear = self.pack(TorchQuantLinear)

Expand All @@ -130,6 +117,6 @@ def test_compare_exllama_triton_torch(self):
self.assertTrue(torch.equal(dequantized_weight, self.linear.weight))
self.assertTrue(torch.all(dequantized_qzeros == 8))

self.assertTrue(torch.allclose(exllama_linear.qweight, torch_linear.qweight))
self.assertTrue(torch.allclose(exllama_linear.scales, torch_linear.scales))
self.assertTrue(torch.allclose(exllama_linear.qzeros, torch_linear.qzeros))
self.assertTrue(torch.allclose(triton_linear.qweight, torch_linear.qweight))
self.assertTrue(torch.allclose(triton_linear.scales, torch_linear.scales))
self.assertTrue(torch.allclose(triton_linear.qzeros, torch_linear.qzeros))