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

Enable Pydantic mypy checks and convert configs to Pydantic dataclasses (#17599)

Browse Source 
Signed-off-by: Harry Mellor <19981378+hmellor@users.noreply.github.com>
This commit is contained in:
Harry Mellor
2025-05-28 13:46:04 +01:00
committed by GitHub
parent d781930f90
commit 4c2b38ce9e
11 changed files with 115 additions and 102 deletions
Download Patch File Download Diff File Expand all files Collapse all files

14
tests/lora/test_quant_model.py
View File

@@ -24,16 +24,16 @@ if current_platform.is_rocm():
MODELS = [
ModelWithQuantization(
model_path="TheBloke/TinyLlama-1.1B-Chat-v0.3-GPTQ",
quantization="GPTQ"),
quantization="gptq"),
]
else:
MODELS = [
ModelWithQuantization(
model_path="TheBloke/TinyLlama-1.1B-Chat-v0.3-AWQ",
quantization="AWQ"),
quantization="awq"),
ModelWithQuantization(
model_path="TheBloke/TinyLlama-1.1B-Chat-v0.3-GPTQ",
quantization="GPTQ"),
quantization="gptq"),
]
@@ -100,7 +100,7 @@ def test_quant_model_lora(tinyllama_lora_files, model):
"#ff8050",
"#ff8080",
]
elif model.quantization == "AWQ":
elif model.quantization == "awq":
expected_no_lora_output = [
"I'm sorry, I don't understand",
"I'm sorry, I don't understand",
@@ -109,7 +109,7 @@ def test_quant_model_lora(tinyllama_lora_files, model):
"#f07700: A v",
"#f00000: A v",
]
elif model.quantization == "GPTQ":
elif model.quantization == "gptq":
expected_no_lora_output = [
"I'm sorry, I don't have",
"I'm sorry, I don't have",
@@ -122,7 +122,7 @@ def test_quant_model_lora(tinyllama_lora_files, model):
def expect_match(output, expected_output):
# HACK: GPTQ lora outputs are just incredibly unstable.
# Assert that the outputs changed.
if (model.quantization == "GPTQ"
if (model.quantization == "gptq"
and expected_output is expected_lora_output):
assert output != expected_no_lora_output
for i, o in enumerate(output):
@@ -172,7 +172,7 @@ def test_quant_model_tp_equality(tinyllama_lora_files, num_gpus_available,
model):
if num_gpus_available < 2:
pytest.skip(f"Not enough GPUs for tensor parallelism {2}")
if model.quantization == "GPTQ":
if model.quantization == "gptq":
pytest.skip("GPTQ lora outputs are just incredibly unstable")
llm_tp1 = vllm.LLM(
model=model.model_path,