# SPDX-License-Identifier: Apache-2.0 import unittest from transformers import PretrainedConfig from sglang.srt.layers.modelopt_utils import canonicalize_modelopt_quant_algo from sglang.srt.layers.quantization.base_config import QuantizationConfig from sglang.srt.model_loader.checkpoint_quantization import ( CheckpointQuantSpec, resolve_checkpoint_quant_spec, ) from sglang.test.ci.ci_register import register_cpu_ci from sglang.test.test_utils import CustomTestCase register_cpu_ci(est_time=11, suite="base-a-test-cpu") class TestResolveCheckpointQuantSpec(CustomTestCase): def test_modelopt_quant_algo_canonicalization(self): cases = { "FP8": "modelopt_fp8", "mxfp8": "mxfp8", "NVFP4": "modelopt_fp4", "NVFP4_AWQ": "modelopt_fp4", "W4A16_NVFP4": "modelopt_fp4", "FP8_FAKE": None, "MIXED_PRECISION": None, None: None, } for quant_algo, expected in cases.items(): with self.subTest(quant_algo=quant_algo): self.assertEqual(canonicalize_modelopt_quant_algo(quant_algo), expected) def test_srt_modelopt_override_uses_the_exact_algorithm_allowlist(self): self.assertEqual( QuantizationConfig._modelopt_override_quantization_method( {"quant_algo": "NVFP4"}, "modelopt" ), "modelopt_fp4", ) self.assertIsNone( QuantizationConfig._modelopt_override_quantization_method( {"quant_algo": "NVFP4_FAKE"}, "modelopt" ) ) def test_top_level_quantization_config(self): config = { "quantization_config": { "quant_method": "fp8", "activation_scheme": "dynamic", } } spec = resolve_checkpoint_quant_spec(config) self.assertEqual( spec, CheckpointQuantSpec( declared_method="fp8", config={"quant_method": "fp8", "activation_scheme": "dynamic"}, source="quantization_config", ), ) def test_text_config_fallback_supports_pretrained_configs(self): config = PretrainedConfig( text_config=PretrainedConfig( quantization_config={"quant_method": "gptq", "bits": 4} ), compression_config={"quant_method": "compressed-tensors"}, ) spec = resolve_checkpoint_quant_spec(config) self.assertIsNotNone(spec) self.assertEqual(spec.declared_method, "gptq") self.assertEqual(spec.source, "text_config.quantization_config") def test_compression_config_fallback(self): config = PretrainedConfig( compression_config={"quant_method": "compressed-tensors"} ) spec = resolve_checkpoint_quant_spec(config) self.assertIsNotNone(spec) self.assertEqual(spec.declared_method, "compressed-tensors") self.assertEqual(spec.source, "compression_config") def test_modelopt_quant_algo_does_not_infer_declared_method(self): config = { "quantization_config": { "quant_algo": "FP8", "exclude_modules": ["lm_head"], } } spec = resolve_checkpoint_quant_spec(config) self.assertIsNotNone(spec) self.assertIsNone(spec.declared_method) self.assertEqual(spec.config["quant_algo"], "FP8") def test_lookup_priority_matches_srt_loader(self): config = { "quantization_config": {}, "text_config": {"quantization_config": {"quant_method": "gptq"}}, "compression_config": {"quant_method": "compressed-tensors"}, } spec = resolve_checkpoint_quant_spec(config) self.assertIsNotNone(spec) self.assertEqual(spec.config, {}) self.assertEqual(spec.source, "quantization_config") def test_metadata_is_deep_copied(self): metadata = {"quant_method": "fp8", "modules_to_not_convert": ["lm_head"]} spec = resolve_checkpoint_quant_spec({"quantization_config": metadata}) self.assertIsNotNone(spec) spec.config["modules_to_not_convert"].append("embed_tokens") self.assertEqual(metadata["modules_to_not_convert"], ["lm_head"]) def test_missing_metadata_returns_none(self): self.assertIsNone(resolve_checkpoint_quant_spec({"model_type": "qwen3_vl"})) def test_invalid_metadata_type_has_clear_error(self): with self.assertRaisesRegex(TypeError, "quantization_config must be a mapping"): resolve_checkpoint_quant_spec({"quantization_config": "fp8"}) if __name__ == "__main__": unittest.main()