biondizzle/vllm
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

[Misc] Make SchedulerConfig.max_model_len init-only (#28733)

Browse Source 
Signed-off-by: DarkLight1337 <tlleungac@connect.ust.hk>
This commit is contained in:
Cyrus Leung
2025-11-15 17:59:31 +08:00
committed by GitHub
parent 1ec978c209
commit 638e4196d1
17 changed files with 22 additions and 45 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
tests/kernels/moe/test_batched_moe.py
View File

@@ -40,8 +40,6 @@ NUM_EXPERTS = [8, 64]
TOP_KS = [1, 2, 6]
vllm_config = VllmConfig()
vllm_config.scheduler_config.max_num_seqs = 128
vllm_config.scheduler_config.max_model_len = 8192
@dataclass

2
tests/kernels/moe/test_block_fp8.py
View File

@@ -33,8 +33,6 @@ if current_platform.get_device_capability() < (9, 0):
pytest.skip("FP8 Triton requires CUDA 9.0 or higher", allow_module_level=True)
vllm_config = VllmConfig()
vllm_config.scheduler_config.max_num_seqs = 128
vllm_config.scheduler_config.max_model_len = 8192
# Test configurations
DTYPES = [torch.bfloat16] # [torch.half, torch.bfloat16, torch.float32]

2
tests/kernels/moe/test_block_int8.py
View File

@@ -18,8 +18,6 @@ if current_platform.get_device_capability() < (7, 0):
pytest.skip("INT8 Triton requires CUDA 7.0 or higher", allow_module_level=True)
vllm_config = VllmConfig()
vllm_config.scheduler_config.max_num_seqs = 128
vllm_config.scheduler_config.max_model_len = 8192
DTYPES = [torch.bfloat16]

2
tests/kernels/moe/test_cutlass_moe.py
View File

@@ -42,8 +42,6 @@ MNK_FACTORS = [
]
vllm_config = VllmConfig(parallel_config=ParallelConfig(pipeline_parallel_size=1))
vllm_config.scheduler_config.max_num_seqs = 128
vllm_config.scheduler_config.max_model_len = 8192
@dataclasses.dataclass

2
tests/kernels/moe/test_flashinfer.py
View File

@@ -45,8 +45,6 @@ MNK_FACTORS = [
]
vllm_config = VllmConfig(parallel_config=ParallelConfig(pipeline_parallel_size=1))
vllm_config.scheduler_config.max_num_seqs = 128
vllm_config.scheduler_config.max_model_len = 8192
def quant_fp8_per_tensor_batches(a):

2
tests/kernels/moe/test_moe.py
View File

@@ -81,8 +81,6 @@ FUSED_MOE_WN16_MNK_FACTORS = [
]
vllm_config = VllmConfig()
vllm_config.scheduler_config.max_num_seqs = 128
vllm_config.scheduler_config.max_model_len = 8192
def run_moe_test(

2
tests/kernels/moe/test_pplx_cutlass_moe.py
View File

@@ -192,8 +192,6 @@ def pplx_cutlass_moe(
vllm_config = VllmConfig()
vllm_config.scheduler_config.max_num_seqs = 128
vllm_config.scheduler_config.max_model_len = 8192
def _pplx_moe(

2
tests/kernels/moe/test_pplx_moe.py
View File

@@ -81,8 +81,6 @@ TOP_KS = [1, 2, 6]
DTYPES = [torch.float8_e4m3fn, torch.bfloat16]
vllm_config = VllmConfig()
vllm_config.scheduler_config.max_num_seqs = 128
vllm_config.scheduler_config.max_model_len = 8192
def torch_prepare(

2
tests/kernels/moe/test_triton_moe_ptpc_fp8.py
View File

@@ -18,8 +18,6 @@ if current_platform.get_device_capability() < (9, 0):
pytest.skip("FP8 Triton requires CUDA 9.0 or higher", allow_module_level=True)
vllm_config = VllmConfig()
vllm_config.scheduler_config.max_num_seqs = 128
vllm_config.scheduler_config.max_model_len = 8192
def native_w8a8_per_token_matmul(A, B, As, Bs, output_dtype=torch.float16):