bump flashinfer to 0.1.5 (ModelCloud#376)

* flashinfer 0.1.5

* Update unit_tests.yml

* Update unit_tests.yml

.github/workflows/unit_tests.yml

@@ -302,7 +302,7 @@ jobs:
       - name: Download wheel
         continue-on-error: true
         run: |
-          file_name=$(curl -s  -F "runid=${{ github.run_id }}" -F "repo=${{ github.repository }}" -F "ref=${{ github.ref }}" -F "fuzz=1" "http://${{ needs.check-vm.outputs.ip }}/gpu/whl/download")
+          file_name=$(curl -s  -F "runid=${{ needs.check-vm.outputs.run_id }}" -F "repo=${{ github.repository }}" -F "ref=${{ github.ref }}" -F "fuzz=1" "http://${{ needs.check-vm.outputs.ip }}/gpu/whl/download")
 
           if echo "$file_name" | grep -q "gptqmodel"; then
               mkdir dist || true
@@ -374,7 +374,7 @@ jobs:
       - list-test-files
       - check-vm
     runs-on: self-hosted
-    if: always() && !cancelled() && (needs.build-intel.result == 'success' || needs.build-amd.result == 'success' || github.event.inputs.artifact_id != '')
+    if: always() && !cancelled() && (needs.build-intel.result == 'success' || needs.build-amd.result == 'success' || github.event.inputs.artifact_id != '') && needs.list-test-files.outputs.cpu-files != '[]'
     timeout-minutes: 40
     container:
       image: ${{ needs.check-vm.outputs.ip }}:5000/modelcloud/gptqmodel:github-ci-v2

tests/test_sglang.py

@@ -16,7 +16,7 @@ class TestLoadSglang(unittest.TestCase):
     @classmethod
     def setUpClass(self):
         # sglang set disable_flashinfer=True still import flashinfer
-        subprocess.check_call([sys.executable, "-m", "pip", "install", "flashinfer==0.1.2", "-i", f"https://flashinfer.ai/whl/cu{torch.version.cuda.replace('.', '')}/torch{'.'.join(torch.__version__.split('.')[:2])}"])
+        subprocess.check_call([sys.executable, "-m", "pip", "install", "flashinfer==0.1.5", "-i", f"https://flashinfer.ai/whl/cu{torch.version.cuda.replace('.', '')}/torch{'.'.join(torch.__version__.split('.')[:2])}"])
         subprocess.check_call([sys.executable, "-m", "pip", "install", "sglang[srt]>=0.2.7"])
 
         self.MODEL_ID = "LnL-AI/TinyLlama-1.1B-Chat-v1.0-GPTQ-4bit"