f7897def96
Co-authored-by: gemini-code-assist[bot] <176961590+gemini-code-assist[bot]@users.noreply.github.com>
80 lines
2.9 KiB
Python
80 lines
2.9 KiB
Python
import unittest
|
|
from unittest.mock import MagicMock
|
|
|
|
from sglang.srt.configs.model_config import ModelConfig
|
|
|
|
|
|
class MockHfConfig:
|
|
def __init__(self, quant_config=None):
|
|
self.quantization_config = quant_config
|
|
self.architectures = ["LlamaForCausalLM"]
|
|
self.model_type = "llama"
|
|
|
|
|
|
class TestQuantLogString(unittest.TestCase):
|
|
def test_qwen_fp8_config(self):
|
|
# Example from Qwen/Qwen3-4B-Thinking-2507-FP8
|
|
quant_config = {
|
|
"activation_scheme": "dynamic",
|
|
"modules_to_not_convert": ["lm_head"],
|
|
"fmt": "e4m3",
|
|
"quant_method": "fp8",
|
|
"weight_block_size": [128, 128],
|
|
}
|
|
|
|
# Create a raw instance
|
|
model_config = ModelConfig.__new__(ModelConfig)
|
|
model_config._parse_quant_hf_config = MagicMock(return_value=quant_config)
|
|
|
|
expected = "quant=fp8, fmt=e4m3"
|
|
result = model_config.get_quantization_config_log_str()
|
|
print(f"\n[Test Qwen FP8] Result: {result}")
|
|
self.assertEqual(result, expected)
|
|
|
|
def test_llama_gptq_int4_config(self):
|
|
# Example from hugging-quants/Meta-Llama-3.1-8B-Instruct-GPTQ-INT4
|
|
quant_config = {"bits": 4, "quant_method": "gptq", "group_size": 128}
|
|
model_config = ModelConfig.__new__(ModelConfig)
|
|
model_config._parse_quant_hf_config = MagicMock(return_value=quant_config)
|
|
|
|
expected = "quant=gptq, bits=4"
|
|
result = model_config.get_quantization_config_log_str()
|
|
print(f"\n[Test Llama GPTQ] Result: {result}")
|
|
self.assertEqual(result, expected)
|
|
|
|
def test_awq_config(self):
|
|
quant_config = {
|
|
"quant_method": "awq",
|
|
"bits": 4,
|
|
"group_size": 128,
|
|
}
|
|
model_config = ModelConfig.__new__(ModelConfig)
|
|
model_config._parse_quant_hf_config = MagicMock(return_value=quant_config)
|
|
|
|
expected = "quant=awq, bits=4"
|
|
result = model_config.get_quantization_config_log_str()
|
|
print(f"\n[Test AWQ] Result: {result}")
|
|
self.assertEqual(result, expected)
|
|
|
|
def test_modelopt_nvfp4(self):
|
|
quant_config = {"quant_method": "modelopt_fp4", "quant_algo": "NVFP4"}
|
|
model_config = ModelConfig.__new__(ModelConfig)
|
|
model_config._parse_quant_hf_config = MagicMock(return_value=quant_config)
|
|
|
|
expected = "quant=modelopt_fp4, quant_algo=NVFP4"
|
|
result = model_config.get_quantization_config_log_str()
|
|
print(f"\n[Test ModelOpt] Result: {result}")
|
|
self.assertEqual(result, expected)
|
|
|
|
def test_no_quant_config(self):
|
|
model_config = ModelConfig.__new__(ModelConfig)
|
|
model_config._parse_quant_hf_config = MagicMock(return_value=None)
|
|
|
|
result = model_config.get_quantization_config_log_str()
|
|
print(f"\n[Test No Quant] Result: {result}")
|
|
self.assertIsNone(result)
|
|
|
|
|
|
if __name__ == "__main__":
|
|
unittest.main()
|