Revert exllamav1 (#692)

* Revert "remove exllama in gptqmodel_ext folder (#521)"

This reverts commit 7faaa9f.

* Revert "remove deprecated exllama v1 code (#473)"

This reverts commit eecdeb3

* Revert "remove exllama v1 (#453)"

This reverts commit 5adae36

* fix quant linear

* revert default value

* revert adding parameter

* rearrange import

* update model path

* fix import

* remove use fp16

* remove qbits

* import Enum

* remove optimum

* rename to exllamav1

* rename backend to exllamav1

* rename test name

* update model id

* replace test with lm-eval

* use a correct model

Author: CSY-ModelCloud

gptqmodel/__init__.py

@@ -1,4 +1,5 @@
 from .models import GPTQModel, get_best_device
 from .quantization import BaseQuantizeConfig, QuantizeConfig
 from .utils import BACKEND, get_backend
+from .utils.exllama import exllama_set_max_input_length
 from .version import __version__