fix unit test quantlinear name (ModelCloud#138)

Author: PZS-ModelCloud

tests/test_lm_head.py

@@ -35,7 +35,7 @@ def test_load(self):
         model = GPTQModel.from_quantized(self.MODEL_ID, use_safetensors=True, device=self.DEVICE)
 
        # validate lm_head is loaded as quantized layer
-        assert model.model.lm_head.__class__.__name__ == "QuantLinear"
+        assert model.model.lm_head.__class__.__name__ == "ExllamaV2QuantLinear"
 
         res = model.model.generate(
             **inputs, num_beams=1, min_new_tokens=1, max_new_tokens=128, repetition_penalty=1.25

tests/test_q4_cuda.py

@@ -9,7 +9,7 @@
 import unittest  # noqa: E402
 
 import torch  # noqa: E402
-from gptqmodel.nn_modules.qlinear.qlinear_cuda_old import QuantLinear as QuantLinearCudaOld  # noqa: E402
+from gptqmodel.nn_modules.qlinear.qlinear_cuda_old import CudaOldQuantLinear # noqa: E402
 from parameterized import parameterized  # noqa: E402
 
 try:
@@ -561,7 +561,7 @@ def test_cuda_old(self, use_half2: bool):
         device = "cuda"
 
         weight_dtype = torch.float16 if use_half2 else torch.float32
-        linear = QuantLinearCudaOld(
+        linear = CudaOldQuantLinear(
             bits=4,
             group_size=group_size,
             desc_act=False,

tests/test_q4_exallama.py

@@ -11,7 +11,7 @@
 import torch  # noqa: E402
 from gptqmodel import GPTQModel, exllama_set_max_input_length  # noqa: E402
 from gptqmodel.models._const import EXLLAMA_DEFAULT_MAX_INPUT_LENGTH  # noqa: E402
-from gptqmodel.nn_modules.qlinear.qlinear_exllama import QuantLinear  # noqa: E402
+from gptqmodel.nn_modules.qlinear.qlinear_exllama import ExllamaQuantLinear  # noqa: E402
 from gptqmodel.quantization import FORMAT
 from gptqmodel.utils.importer import select_quant_linear  # noqa: E402
 from gptqmodel.utils.model import gptqmodel_post_init  # noqa: E402
@@ -1078,7 +1078,7 @@ def test_exllama(self):
             outfeatures=n,
             bias=False,
         )
-        self.assertTrue(isinstance(linear, QuantLinear))
+        self.assertTrue(isinstance(linear, ExllamaQuantLinear))
 
         torch.manual_seed(42)
 

tests/test_q4_exallama_v2.py

@@ -8,7 +8,7 @@
 
 import torch  # noqa: E402
 from gptqmodel import Backend, GPTQModel  # noqa: E402
-from gptqmodel.nn_modules.qlinear.qlinear_exllamav2 import QuantLinear  # noqa: E402
+from gptqmodel.nn_modules.qlinear.qlinear_exllamav2 import ExllamaV2QuantLinear  # noqa: E402
 from gptqmodel.quantization import FORMAT
 from gptqmodel.utils.importer import select_quant_linear  # noqa: E402
 from gptqmodel.utils.model import gptqmodel_post_init  # noqa: E402
@@ -46,7 +46,7 @@ def test_exllamav2(self):
             bias=False,
         )
 
-        self.assertTrue(isinstance(linear, QuantLinear))
+        self.assertTrue(isinstance(linear, ExllamaV2QuantLinear))
 
         torch.manual_seed(42)
 

tests/test_q4_marlin.py

@@ -8,7 +8,7 @@
 
 import torch  # noqa: E402
 from gptqmodel import Backend, GPTQModel  # noqa: E402
-from gptqmodel.nn_modules.qlinear.qlinear_marlin import QuantLinear as MarlinQuantLinear  # noqa: E402
+from gptqmodel.nn_modules.qlinear.qlinear_marlin import MarlinQuantLinear  # noqa: E402
 from transformers import AutoTokenizer  # noqa: E402
 
 

tests/test_q4_triton.py

@@ -8,7 +8,7 @@
 
 import torch  # noqa: E402
 from gptqmodel import Backend, GPTQModel  # noqa: E402
-from gptqmodel.nn_modules.qlinear.qlinear_tritonv2 import QuantLinear as TritonV2QuantLinear  # noqa: E402
+from gptqmodel.nn_modules.qlinear.qlinear_tritonv2 import TritonV2QuantLinear  # noqa: E402
 from transformers import AutoTokenizer  # noqa: E402
 
 GENERATE_EVAL_SIZE = 100

tests/test_repacking.py

@@ -12,8 +12,8 @@
 import torch.nn as nn # noqa: E402
 import gptqmodel_marlin_cuda # noqa: E402
 # isort: on
-from gptqmodel.nn_modules.qlinear.qlinear_cuda_old import QuantLinear as CudaOldQuantLinear  # noqa: E402
-from gptqmodel.nn_modules.qlinear.qlinear_marlin import QuantLinear as MarlinQuantLinear  # noqa: E402
+from gptqmodel.nn_modules.qlinear.qlinear_cuda_old import CudaOldQuantLinear  # noqa: E402
+from gptqmodel.nn_modules.qlinear.qlinear_marlin import MarlinQuantLinear  # noqa: E402
 from gptqmodel.nn_modules.qlinear.qlinear_marlin import _get_perms, dequantize_weight  # noqa: E402