[ROCm] Fix CI runtime and tests for MI355 DPX (#58244)

Signed-off-by: Andreas Karatzas <Andreas.Karatzas@amd.com>
Co-authored-by: Mahesh Kunreddi <mahesh.kunreddi@amd.com>
Co-authored-by: Claude <noreply@anthropic.com>
Co-authored-by: Andreas Karatzas <Andreas.Karatzas@amd.com>
Co-authored-by: Codex <noreply@openai.com>
Co-authored-by: Andreas Karatzas <akaratza@amd.com>
This commit is contained in:
mkunredd
2026-09-24 23:15:11 -05:00
committed by GitHub
co-authored by Mahesh Kunreddi Claude Andreas Karatzas Codex Andreas Karatzas
parent 44257c2e35
commit 4013f3ad6c
12 changed files with 175 additions and 47 deletions
+6 -4
View File
@@ -364,7 +364,7 @@ def test_end_to_end(model: str):
@create_new_process_for_each_test()
def test_deep_sleep():
def test_deep_sleep(gpu_memory_cleared):
model = "hmellor/tiny-random-LlamaForCausalLM"
free, total = torch.accelerator.get_memory_info()
used_bytes_baseline = total - free # in case other process is running
@@ -395,7 +395,7 @@ def test_deep_sleep():
@create_new_process_for_each_test()
def test_deep_sleep_lora():
def test_deep_sleep_lora(gpu_memory_cleared):
"""Level-2 sleep/wake/reload with enable_lora=True.
LoRA wrapping moves parameters under base_layer and adds LoRA
@@ -488,7 +488,7 @@ def test_deep_sleep_lora_tp2(num_gpus_available, monkeypatch):
@create_new_process_for_each_test()
def test_deep_sleep_async():
def test_deep_sleep_async(gpu_memory_cleared):
async def test():
model = "hmellor/tiny-random-LlamaForCausalLM"
free, total = torch.accelerator.get_memory_info()
@@ -527,7 +527,9 @@ def test_deep_sleep_async():
@requires_fp8
def test_deep_sleep_fp8_kvcache_mrv1(monkeypatch: pytest.MonkeyPatch):
def test_deep_sleep_fp8_kvcache_mrv1(
monkeypatch: pytest.MonkeyPatch, gpu_memory_cleared
):
# Regression test for https://github.com/vllm-project/vllm/pull/28783.
# In particular, verify that MRV1 does not rely on post_kv_cache_wake_up()
# to restore correct output after level-2 sleep.