Install torchao

Author: jainapurva

.github/workflows/vllm-benchmark.yml

@@ -292,6 +292,9 @@ jobs:
             -w /tmp/workspace \
             "${DOCKER_IMAGE}"
           )
+          if [[ "${DEVICE_NAME}" == "cuda" ]]; then
+            docker exec -t "${container_name}" bash -c "uv pip install --system --pre torchao --index-url https://download.pytorch.org/whl/nightly/cu128"
+          fi
           docker exec -t "${container_name}" bash -c "cd vllm-benchmarks/vllm && bash .buildkite/nightly-benchmarks/scripts/run-performance-benchmarks.sh"
 
       - name: Authenticate with AWS

vllm-benchmarks/benchmarks/cuda/latency-tests.json

@@ -140,7 +140,7 @@
         }
     },
     {
-        "test_name": "latency_gemma3_12b_it_fp8",
+        "test_name": "latency_gemma3_12b_it_fp8_torchao",
         "parameters": {
             "model": "pytorch/gemma-3-12b-it-FP8",
             "load_format": "dummy",
@@ -149,7 +149,7 @@
         }
     },
     {
-        "test_name": "latency_gemma3_12b_it_int4",
+        "test_name": "latency_gemma3_12b_it_int4_torchao",
         "parameters": {
             "model": "pytorch/gemma-3-12b-it-INT4",
             "load_format": "dummy",
@@ -158,7 +158,7 @@
         }
     },
     {
-        "test_name": "latency_gemma3_12b_it_awq_int4",
+        "test_name": "latency_gemma3_12b_it_awq_int4_torchao",
         "parameters": {
             "model": "pytorch/gemma-3-12b-it-AWQ-INT4",
             "load_format": "dummy",
@@ -167,7 +167,7 @@
         }
     },
     {
-        "test_name": "latency_gemma3_27b_it_fp8",
+        "test_name": "latency_gemma3_27b_it_fp8_torchao",
         "parameters": {
             "model": "pytorch/gemma-3-27b-it-FP8",
             "load_format": "dummy",
@@ -176,7 +176,7 @@
         }
     },
     {
-        "test_name": "latency_gemma3_27b_it_int4",
+        "test_name": "latency_gemma3_27b_it_int4_torchao",
         "parameters": {
             "model": "pytorch/gemma-3-27b-it-INT4",
             "load_format": "dummy",
@@ -185,7 +185,7 @@
         }
     },
     {
-        "test_name": "latency_gemma3_27b_it_awq_int4",
+        "test_name": "latency_gemma3_27b_it_awq_int4_torchao",
         "parameters": {
             "model": "pytorch/gemma-3-27b-it-AWQ-INT4",
             "load_format": "dummy",