Improve configs - ModelConfig (#17130)
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
This commit is contained in:
+1
-1
@@ -738,7 +738,7 @@ class VllmRunner:
|
||||
- `block_size`: Set to `16` instead of `None` to reduce memory usage.
|
||||
- `enable_chunked_prefill`: Set to `False` instead of `None` for
|
||||
test reproducibility.
|
||||
- `enforce_eager`: Set to `False` instead of `None` to test CUDA graph.
|
||||
- `enforce_eager`: Set to `False` to test CUDA graph.
|
||||
"""
|
||||
|
||||
def __init__(
|
||||
|
||||
@@ -8,7 +8,7 @@ from typing import Literal, Optional
|
||||
|
||||
import pytest
|
||||
|
||||
from vllm.config import PoolerConfig, config
|
||||
from vllm.config import config
|
||||
from vllm.engine.arg_utils import (EngineArgs, contains_type, get_kwargs,
|
||||
get_type, is_not_builtin, is_type,
|
||||
literal_to_kwargs, nullable_kvs,
|
||||
@@ -222,17 +222,6 @@ def test_prefix_cache_default():
|
||||
assert not engine_args.enable_prefix_caching
|
||||
|
||||
|
||||
def test_valid_pooling_config():
|
||||
parser = EngineArgs.add_cli_args(FlexibleArgumentParser())
|
||||
args = parser.parse_args([
|
||||
'--override-pooler-config',
|
||||
'{"pooling_type": "MEAN"}',
|
||||
])
|
||||
engine_args = EngineArgs.from_cli_args(args=args)
|
||||
assert engine_args.override_pooler_config == PoolerConfig(
|
||||
pooling_type="MEAN", )
|
||||
|
||||
|
||||
@pytest.mark.parametrize(
|
||||
("arg"),
|
||||
[
|
||||
|
||||
@@ -14,7 +14,7 @@ import torch.nn.functional as F
|
||||
from vllm.model_executor.layers.linear import LinearBase # noqa: E501
|
||||
from vllm.model_executor.layers.linear import UnquantizedLinearMethod
|
||||
from vllm.model_executor.layers.quantization import (
|
||||
get_quantization_config, register_quantization_config)
|
||||
QuantizationMethods, get_quantization_config, register_quantization_config)
|
||||
from vllm.model_executor.layers.quantization.base_config import ( # noqa: E501
|
||||
QuantizationConfig)
|
||||
|
||||
@@ -54,7 +54,7 @@ class CustomQuantConfig(QuantizationConfig):
|
||||
"""Initialize the quantization config."""
|
||||
self.num_bits = num_bits
|
||||
|
||||
def get_name(self) -> str:
|
||||
def get_name(self) -> QuantizationMethods:
|
||||
"""Name of the quantization method."""
|
||||
return "custom_quant"
|
||||
|
||||
|
||||
@@ -185,7 +185,7 @@ def test_get_pooling_config():
|
||||
revision=None,
|
||||
)
|
||||
|
||||
pooling_config = model_config._init_pooler_config(None)
|
||||
pooling_config = model_config._init_pooler_config()
|
||||
assert pooling_config is not None
|
||||
|
||||
assert pooling_config.normalize
|
||||
@@ -205,11 +205,12 @@ def test_get_pooling_config_from_args():
|
||||
dtype="float16",
|
||||
revision=None)
|
||||
|
||||
override_config = PoolerConfig(pooling_type='CLS', normalize=True)
|
||||
override_pooler_config = PoolerConfig(pooling_type='CLS', normalize=True)
|
||||
model_config.override_pooler_config = override_pooler_config
|
||||
|
||||
pooling_config = model_config._init_pooler_config(override_config)
|
||||
pooling_config = model_config._init_pooler_config()
|
||||
assert pooling_config is not None
|
||||
assert asdict(pooling_config) == asdict(override_config)
|
||||
assert asdict(pooling_config) == asdict(override_pooler_config)
|
||||
|
||||
|
||||
@pytest.mark.skipif(current_platform.is_rocm(),
|
||||
|
||||
Reference in New Issue
Block a user