[ROCm][CI] Wait for ROCm VRAM to settle between compiled and eager LL… (#49739)
Signed-off-by: aarushjain29 <Aarushi.Jain2@amd.com>
This commit is contained in:
@@ -9,6 +9,7 @@ import pytest
|
|||||||
import torch
|
import torch
|
||||||
|
|
||||||
from tests.models.utils import check_logprobs_close
|
from tests.models.utils import check_logprobs_close
|
||||||
|
from tests.utils import wait_for_rocm_memory_to_settle
|
||||||
from vllm import LLM, SamplingParams
|
from vllm import LLM, SamplingParams
|
||||||
from vllm.compilation.decorators import support_torch_compile
|
from vllm.compilation.decorators import support_torch_compile
|
||||||
from vllm.config import CompilationConfig, VllmConfig, set_current_vllm_config
|
from vllm.config import CompilationConfig, VllmConfig, set_current_vllm_config
|
||||||
@@ -104,6 +105,7 @@ def test_dynamic_shapes_compilation(
|
|||||||
gc.collect()
|
gc.collect()
|
||||||
torch.accelerator.empty_cache()
|
torch.accelerator.empty_cache()
|
||||||
torch.accelerator.synchronize()
|
torch.accelerator.synchronize()
|
||||||
|
wait_for_rocm_memory_to_settle()
|
||||||
|
|
||||||
eager_model = LLM(model=model_name, enforce_eager=True, max_model_len=1024)
|
eager_model = LLM(model=model_name, enforce_eager=True, max_model_len=1024)
|
||||||
eager_outputs = []
|
eager_outputs = []
|
||||||
|
|||||||
Reference in New Issue
Block a user