[ci]: retire issue 1586 handoff

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

@@ -1,18 +0,0 @@
-{
-  "step_time": 0.8274769484996796,
-  "avg_step_time": 1.4095150619745254,
-  "timing/optimizer_step": 0.02941332757472992,
-  "grad_norm": 0.24477368593215942,
-  "timing/forward_backward": 0.693624597042799,
-  "_timestamp": 1765590605.595408,
-  "timing/prepare_dit_inputs": 0.009604107588529587,
-  "learning_rate": 1e-05,
-  "train_loss": 0.30774877965450287,
-  "_step": 5,
-  "timing/normalize_input": 0.0003825277090072632,
-  "vsa_sparsity": 0.05,
-  "timing/clip_grad_norm": 0.025074530392885208,
-  "_runtime": 90.44591733,
-  "timing/get_next_batch": 0.0055067278444767,
-  "timing/reduce_loss": 0.00021916255354881287
-}

fastvideo/tests/training/VSA/test_training_loss_VSA.py

@@ -18,8 +18,7 @@ from fastvideo.fastvideo_args import FastVideoArgs, TrainingArgs
 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))