28 lines
1.2 KiB
Python
28 lines
1.2 KiB
Python
from typing import Any, Dict
|
|
import torch
|
|
import bitsandbytes as bnb
|
|
from transformers.quantizers.quantizer_bnb_4bit import Bnb4BitHfQuantizer, get_module_from_name
|
|
from transformers.modeling_utils import PreTrainedModel
|
|
|
|
# CogVLM stores inv_freq in the state dictionary but it is not in models._parameters so it cannot be quantized
|
|
# was patched in transformers for other models here: https://github.com/huggingface/transformers/pull/28837/files but cog is not part of transformers
|
|
def _patched_check_quantized_param(
|
|
self, model: "PreTrainedModel", param_value: "torch.Tensor", param_name: str, state_dict: Dict[str, Any], **kwargs
|
|
) -> bool:
|
|
|
|
# if "inv_freq" in param_name: # detect failure case
|
|
# print("check_quantized_param", param_name)
|
|
|
|
module, tensor_name = get_module_from_name(model, param_name)
|
|
if ("inv_freq" == tensor_name): # the fix
|
|
return False
|
|
if isinstance(module._parameters[tensor_name], bnb.nn.Params4bit): # will throw key error for inv_freq
|
|
return True
|
|
elif isinstance(module, bnb.nn.Linear4bit) and tensor_name == "bias":
|
|
return True
|
|
else:
|
|
return False
|
|
|
|
def patch_cog():
|
|
Bnb4BitHfQuantizer.check_quantized_param = _patched_check_quantized_param
|