[ci]: pin VSA training regression to H100 (#1591)

Author: macthecadillac

fastvideo/tests/contract/test_modal_fa4_policy.py

@@ -56,6 +56,11 @@ def test_performance_lane_classifies_pull_requests_before_main():
         "BUILDKITE_PULL_REQUEST") < function_strings.index("BUILDKITE_BRANCH")
 
 
+def test_vsa_training_lane_uses_strict_h100_pair():
+    function_strings = _function_strings(PR_TEST, "run_training_tests_VSA")
+    assert "H100!:2" in function_strings.splitlines()
+
+
 def test_pr_model_load_and_training_lanes_disable_fa4():
     lanes = {
         "run_transformer_tests": "pytest ./fastvideo/tests/transformers -vs",

fastvideo/tests/modal/pr_test.py

@@ -229,7 +229,7 @@ def run_training_lora_tests():
     )
 
 
-@app.function(gpu="H100:2",
+@app.function(gpu="H100!:2",
               image=image,
               timeout=900,
               secrets=[wandb_secret, ci_env_secret])

fastvideo/tests/training/VSA/h200_reference_wandb_summary_VSA.json

@@ -18,8 +18,7 @@
 from fastvideo.utils import FlexibleArgumentParser
 
 wandb_name = "test_training_loss_VSA"
-h100_reference_wandb_summary_file = "fastvideo/tests/training/VSA/h100_reference_wandb_summary_VSA.json"
-h200_reference_wandb_summary_file = "fastvideo/tests/training/VSA/h200_reference_wandb_summary_VSA.json"
+reference_wandb_summary_file = "fastvideo/tests/training/VSA/h100_reference_wandb_summary_VSA.json"
 
 NUM_NODES = "1"
 NUM_GPUS_PER_NODE = "2"
@@ -110,12 +109,9 @@ def test_distributed_training():
     summary_file = 'data/wan_finetune_test_VSA/tracker/wandb/latest-run/files/wandb-summary.json'
 
     device_name = torch.cuda.get_device_name()
-    if "H100" in device_name:
-        reference_wandb_summary_file = h100_reference_wandb_summary_file
-    elif "H200" in device_name:
-        reference_wandb_summary_file = h200_reference_wandb_summary_file
-    else:
-        raise ValueError(f"Unknown device: {device_name}")
+    print(f"INFO: device: {device_name}")
+    if "H100" not in device_name:
+        raise ValueError(f"VSA training regression requires H100 GPUs, got: {device_name}")
 
     reference_wandb_summary = json.load(open(reference_wandb_summary_file))
     wandb_summary = json.load(open(summary_file))