[V0 Deprecation] Remove unused swap_space parameter (#36216)
Signed-off-by: majiayu000 <1835304752@qq.com> Co-authored-by: mcelrath
This commit is contained in:
@@ -794,7 +794,6 @@ class VllmRunner:
|
||||
tensor_parallel_size: int = 1,
|
||||
block_size: int = 16 if not torch.xpu.is_available() else 64,
|
||||
enable_chunked_prefill: bool | None = False,
|
||||
swap_space: int = 4,
|
||||
enforce_eager: bool | None = False,
|
||||
# Set this to avoid hanging issue
|
||||
default_torch_num_threads: int | None = None,
|
||||
@@ -831,7 +830,6 @@ class VllmRunner:
|
||||
trust_remote_code=trust_remote_code,
|
||||
dtype=dtype,
|
||||
seed=seed,
|
||||
swap_space=swap_space,
|
||||
enforce_eager=enforce_eager,
|
||||
disable_log_stats=disable_log_stats,
|
||||
tensor_parallel_size=tensor_parallel_size,
|
||||
|
||||
@@ -22,7 +22,7 @@ prompts = [
|
||||
|
||||
sampling_params = SamplingParams(temperature=0.8, top_p=0.95)
|
||||
|
||||
# set different `gpu_memory_utilization` and `swap_space` for different ranks,
|
||||
# set different `gpu_memory_utilization` for different ranks,
|
||||
# to test if all ranks agree on the same kv cache configuration.
|
||||
llm = LLM(
|
||||
model="facebook/opt-125m",
|
||||
@@ -30,7 +30,6 @@ llm = LLM(
|
||||
pipeline_parallel_size=int(os.getenv("PP_SIZE", 1)),
|
||||
distributed_executor_backend="external_launcher",
|
||||
gpu_memory_utilization=random.uniform(0.7, 0.9),
|
||||
swap_space=random.randint(1, 4),
|
||||
seed=0,
|
||||
)
|
||||
|
||||
|
||||
@@ -28,7 +28,7 @@ if dp_size > 1:
|
||||
|
||||
sampling_params = SamplingParams(temperature=0.8, top_p=0.95)
|
||||
|
||||
# set different `gpu_memory_utilization` and `swap_space` for different ranks,
|
||||
# set different `gpu_memory_utilization` for different ranks,
|
||||
# to test if all ranks agree on the same kv cache configuration.
|
||||
llm = LLM(
|
||||
model="microsoft/Phi-mini-MoE-instruct",
|
||||
@@ -37,7 +37,6 @@ llm = LLM(
|
||||
enable_expert_parallel=int(os.getenv("ENABLE_EP", "0")) == 1,
|
||||
distributed_executor_backend="external_launcher",
|
||||
gpu_memory_utilization=random.uniform(0.7, 0.9),
|
||||
swap_space=random.randint(1, 4),
|
||||
seed=0,
|
||||
)
|
||||
|
||||
|
||||
@@ -64,7 +64,6 @@ def test_worker_apply_lora(qwen3_lora_files):
|
||||
device_config=DeviceConfig("cuda"),
|
||||
cache_config=CacheConfig(
|
||||
block_size=16,
|
||||
swap_space=0,
|
||||
cache_dtype="auto",
|
||||
),
|
||||
lora_config=LoRAConfig(
|
||||
|
||||
@@ -182,7 +182,6 @@ def create_vllm_config(
|
||||
cache_config = CacheConfig(
|
||||
block_size=block_size,
|
||||
cache_dtype="auto",
|
||||
swap_space=0,
|
||||
)
|
||||
# Set cache blocks for testing
|
||||
# (these may be set during initialization normally)
|
||||
|
||||
@@ -1776,7 +1776,6 @@ def create_scheduler_with_priority(
|
||||
cache_config = CacheConfig(
|
||||
block_size=block_size,
|
||||
gpu_memory_utilization=0.9,
|
||||
swap_space=0,
|
||||
cache_dtype="auto",
|
||||
enable_prefix_caching=enable_prefix_caching,
|
||||
)
|
||||
@@ -3726,7 +3725,6 @@ def _create_encoder_decoder_scheduler(
|
||||
cache_config = CacheConfig(
|
||||
block_size=block_size,
|
||||
gpu_memory_utilization=0.9,
|
||||
swap_space=0,
|
||||
cache_dtype="auto",
|
||||
enable_prefix_caching=False,
|
||||
)
|
||||
|
||||
@@ -94,7 +94,6 @@ def create_scheduler(
|
||||
cache_config = CacheConfig(
|
||||
block_size=block_size,
|
||||
gpu_memory_utilization=0.9,
|
||||
swap_space=0,
|
||||
cache_dtype="auto",
|
||||
enable_prefix_caching=enable_prefix_caching,
|
||||
)
|
||||
|
||||
@@ -506,7 +506,6 @@ def test_encoder_instance_zero_kv_cache(
|
||||
cache_config = CacheConfig(
|
||||
block_size=16,
|
||||
gpu_memory_utilization=gpu_memory_utilization,
|
||||
swap_space=0,
|
||||
cache_dtype="auto",
|
||||
enable_prefix_caching=enable_prefix_caching,
|
||||
)
|
||||
|
||||
@@ -206,7 +206,6 @@ def create_vllm_config(
|
||||
cache_config = CacheConfig(
|
||||
block_size=block_size,
|
||||
gpu_memory_utilization=0.9,
|
||||
swap_space=0,
|
||||
cache_dtype="auto",
|
||||
enable_prefix_caching=True,
|
||||
)
|
||||
|
||||
@@ -118,7 +118,6 @@ def create_vllm_config(
|
||||
cache_config = CacheConfig(
|
||||
block_size=block_size,
|
||||
gpu_memory_utilization=0.9,
|
||||
swap_space=0,
|
||||
cache_dtype=cache_dtype,
|
||||
enable_prefix_caching=True,
|
||||
)
|
||||
|
||||
@@ -96,7 +96,6 @@ def get_vllm_config():
|
||||
cache_config = CacheConfig(
|
||||
block_size=BLOCK_SIZE,
|
||||
gpu_memory_utilization=0.9,
|
||||
swap_space=0,
|
||||
cache_dtype="auto",
|
||||
)
|
||||
parallel_config = ParallelConfig()
|
||||
@@ -809,7 +808,6 @@ def test_hybrid_attention_mamba_tensor_shapes():
|
||||
cache_config = CacheConfig(
|
||||
block_size=BLOCK_SIZE,
|
||||
gpu_memory_utilization=0.9,
|
||||
swap_space=0,
|
||||
cache_dtype="auto",
|
||||
)
|
||||
parallel_config = ParallelConfig()
|
||||
@@ -1242,7 +1240,6 @@ def test_cudagraph_sizes_capped_for_mamba_cache():
|
||||
cache_config = CacheConfig(
|
||||
block_size=BLOCK_SIZE,
|
||||
gpu_memory_utilization=0.9,
|
||||
swap_space=0,
|
||||
cache_dtype="auto",
|
||||
)
|
||||
parallel_config = ParallelConfig()
|
||||
|
||||
Reference in New Issue
Block a user