254 lines
10 KiB
Python
254 lines
10 KiB
Python
import warnings
|
|
import re
|
|
from contextlib import contextmanager
|
|
from dataclasses import asdict
|
|
from enum import Enum
|
|
from typing import List, Optional
|
|
|
|
import torch
|
|
from peft import get_peft_model, PeftConfig, PeftModel, PeftType
|
|
from peft.peft_model import PEFT_TYPE_TO_MODEL_MAPPING
|
|
from peft.tuners.lora import LoraConfig, LoraLayer, LoraModel, Embedding, mark_only_lora_as_trainable, _freeze_adapter
|
|
from peft.mapping import PEFT_TYPE_TO_CONFIG_MAPPING
|
|
from peft.utils.other import (
|
|
_get_submodules,
|
|
TRANSFORMERS_MODELS_TO_LORA_TARGET_MODULES_MAPPING,
|
|
TRANSFORMERS_MODELS_TO_ADALORA_TARGET_MODULES_MAPPING
|
|
)
|
|
|
|
from ..modeling._base import BaseGPTQForCausalLM
|
|
|
|
|
|
class GPTQLoraConfig(LoraConfig):
|
|
injected_fused_attention: bool = False
|
|
|
|
|
|
class GPTQLoraLinear(torch.nn.Linear, LoraLayer):
|
|
def __init__(
|
|
self,
|
|
adapter_name: str,
|
|
linear_module: torch.nn.Linear,
|
|
r: int = 0,
|
|
lora_alpha: int = 1,
|
|
lora_dropout: float = 0.0,
|
|
fan_in_fan_out: bool = False, # Set this to True if the layer to replace stores weight like (fan_in, fan_out)
|
|
**kwargs,
|
|
):
|
|
init_lora_weights = kwargs.pop("init_lora_weights", True)
|
|
|
|
torch.nn.Linear.__init__(self, linear_module.in_features, linear_module.out_features)
|
|
LoraLayer.__init__(self, linear_module.in_features, linear_module.out_features)
|
|
|
|
self.linear_module = linear_module
|
|
|
|
self.weight.requires_grad = False
|
|
self.weight = self.linear_module.weight
|
|
self.bias = self.linear_module.bias
|
|
self.fan_in_fan_out = fan_in_fan_out
|
|
if fan_in_fan_out:
|
|
self.weight.data = self.weight.data.T
|
|
|
|
self.update_layer(adapter_name, r, lora_alpha, lora_dropout, init_lora_weights)
|
|
self.active_adapter = adapter_name
|
|
|
|
def reset_lora_parameters(self, adapter_name):
|
|
if adapter_name in self.lora_A.keys():
|
|
torch.nn.init.xavier_uniform_(self.lora_A[adapter_name].weight)
|
|
torch.nn.init.zeros_(self.lora_B[adapter_name].weight)
|
|
|
|
def merge(self):
|
|
raise NotImplementedError("gptq model not support merge lora adapter")
|
|
|
|
def unmerge(self):
|
|
raise NotImplementedError("gptq model not support unmerge lora adapter")
|
|
|
|
def forward(self, x: torch.Tensor):
|
|
previous_dtype = x.dtype
|
|
if self.active_adapter not in self.lora_A.keys():
|
|
return self.linear_module(x)
|
|
if self.disable_adapters:
|
|
if self.r[self.active_adapter] > 0 and self.merged:
|
|
self.unmerge()
|
|
result = self.linear_module(x)
|
|
elif self.r[self.active_adapter] > 0 and not self.merged:
|
|
result = self.linear_module(x)
|
|
|
|
lora_B = self.lora_B[self.active_adapter]
|
|
lora_A = self.lora_A[self.active_adapter]
|
|
lora_dropout = self.lora_dropout[self.active_adapter]
|
|
scale = self.scaling[self.active_adapter]
|
|
|
|
x = x.type_as(lora_A.weight.data)
|
|
adapter_result = (lora_B(lora_A(lora_dropout(x))) * scale).type_as(result)
|
|
result += adapter_result
|
|
else:
|
|
result = self.linear_module(x)
|
|
|
|
result = result.to(previous_dtype)
|
|
|
|
return result
|
|
|
|
|
|
class GPTQLoraModel(LoraModel):
|
|
def _find_and_replace(self, adapter_name):
|
|
lora_config = self.peft_config[adapter_name]
|
|
is_target_modules_in_base_model = False
|
|
kwargs = {
|
|
"r": lora_config.r,
|
|
"lora_alpha": lora_config.lora_alpha,
|
|
"lora_dropout": lora_config.lora_dropout,
|
|
"fan_in_fan_out": lora_config.fan_in_fan_out,
|
|
"init_lora_weights": lora_config.init_lora_weights,
|
|
}
|
|
key_list = [key for key, _ in self.model.named_modules()]
|
|
for key in key_list:
|
|
if isinstance(lora_config.target_modules, str):
|
|
target_module_found = re.fullmatch(lora_config.target_modules, key)
|
|
else:
|
|
target_module_found = any(key.endswith(target_key) for target_key in lora_config.target_modules)
|
|
if target_module_found:
|
|
if not is_target_modules_in_base_model:
|
|
is_target_modules_in_base_model = True
|
|
parent, target, target_name = _get_submodules(self.model, key)
|
|
bias = False
|
|
if hasattr(target, "bias"):
|
|
bias = target.bias is not None
|
|
|
|
if isinstance(target, LoraLayer):
|
|
target.update_layer(
|
|
adapter_name,
|
|
lora_config.r,
|
|
lora_config.lora_alpha,
|
|
lora_config.lora_dropout,
|
|
lora_config.init_lora_weights,
|
|
)
|
|
else:
|
|
if isinstance(target, torch.nn.Embedding):
|
|
embedding_kwargs = kwargs.copy()
|
|
embedding_kwargs.pop("fan_in_fan_out", None)
|
|
in_features, out_features = target.num_embeddings, target.embedding_dim
|
|
new_module = Embedding(adapter_name, in_features, out_features, **embedding_kwargs)
|
|
else:
|
|
if isinstance(target, torch.nn.Linear):
|
|
if kwargs["fan_in_fan_out"]:
|
|
warnings.warn(
|
|
"fan_in_fan_out is set to True but the target module is `torch.nn.Linear`. "
|
|
"Setting fan_in_fan_out to False."
|
|
)
|
|
kwargs["fan_in_fan_out"] = lora_config.fan_in_fan_out = False
|
|
else:
|
|
raise ValueError(
|
|
f"Target module {target} is not supported. "
|
|
f"Currently, only `torch.nn.Linear` and its subclasses are supported."
|
|
)
|
|
new_module = GPTQLoraLinear(adapter_name, target, **kwargs)
|
|
|
|
self._replace_module(parent, target_name, new_module, target)
|
|
if not is_target_modules_in_base_model:
|
|
raise ValueError(
|
|
f"Target modules {lora_config.target_modules} not found in the base model. "
|
|
f"Please check the target modules and try again."
|
|
)
|
|
|
|
def _replace_module(self, parent_module, child_name, new_module, old_module):
|
|
setattr(parent_module, child_name, new_module)
|
|
if not isinstance(new_module, GPTQLoraLinear):
|
|
new_module.weight = old_module.weight
|
|
if hasattr(old_module, "bias"):
|
|
if old_module.bias is not None:
|
|
new_module.bias = old_module.bias
|
|
|
|
if getattr(old_module, "state", None) is not None:
|
|
new_module.state = old_module.state
|
|
new_module.to(old_module.weight.device)
|
|
|
|
# dispatch to correct device
|
|
for name, module in new_module.named_modules():
|
|
if "lora_" in name:
|
|
module.to(old_module.weight.device)
|
|
|
|
def merge_adapter(self):
|
|
raise NotImplementedError("gptq model not support merge lora adapter")
|
|
|
|
def unmerge_adapter(self):
|
|
raise NotImplementedError("gptq model not support unmerge lora adapter")
|
|
|
|
def merge_and_unload(self):
|
|
raise NotImplementedError("gptq model not support merge and unload")
|
|
|
|
|
|
def find_all_linear_names(model: BaseGPTQForCausalLM, ignore: Optional[List[str]] = None, ignore_lm_head: bool = True):
|
|
if not ignore:
|
|
ignore = []
|
|
lm_head_name = model.lm_head_name
|
|
if ignore_lm_head and lm_head_name not in ignore:
|
|
ignore.append(lm_head_name)
|
|
results = set()
|
|
for n, m in model.named_modules():
|
|
if isinstance(m, torch.nn.Linear):
|
|
res = n.split('.')[-1]
|
|
if res not in ignore:
|
|
results.add(res)
|
|
return list(results)
|
|
|
|
|
|
@contextmanager
|
|
def hijack_peft_mappings():
|
|
PEFT_TYPE_TO_CONFIG_MAPPING[PeftType.LORA] = GPTQLoraConfig
|
|
PEFT_TYPE_TO_MODEL_MAPPING[PeftType.LORA] = GPTQLoraModel
|
|
|
|
try:
|
|
yield
|
|
except:
|
|
PEFT_TYPE_TO_CONFIG_MAPPING[PeftType.LORA] = LoraConfig
|
|
PEFT_TYPE_TO_MODEL_MAPPING[PeftType.LORA] = LoraModel
|
|
raise
|
|
finally:
|
|
PEFT_TYPE_TO_CONFIG_MAPPING[PeftType.LORA] = LoraConfig
|
|
PEFT_TYPE_TO_MODEL_MAPPING[PeftType.LORA] = LoraModel
|
|
|
|
|
|
def get_gptq_peft_model(
|
|
model: BaseGPTQForCausalLM,
|
|
peft_config: PeftConfig = None,
|
|
model_id: str = None,
|
|
adapter_name: str = "default",
|
|
auto_find_all_linears: bool = True,
|
|
train_mode: bool = False
|
|
):
|
|
if (
|
|
model.fused_attn_module_type is not None and not model.injected_fused_attention
|
|
):
|
|
warnings.warn(
|
|
"You can just ignore this warning if the peft type you use isn't lora.\n"
|
|
f"{model.__class__.__name__} supports injecting fused attention but not enables this time. "
|
|
"If you are training lora adapters, you must also disable fused attention injection when loading quantized "
|
|
"base model at inference time, otherwise adapters may not be added to base model properly. "
|
|
"If you are loading lora adapters to do inference, you can reference to adapter's config file to check "
|
|
"whether the adapters are trained using base model that not enable fused attention injection."
|
|
)
|
|
|
|
if train_mode and not peft_config:
|
|
raise ValueError("peft_config not specified when in train mode.")
|
|
if not train_mode and not model_id:
|
|
raise ValueError("model_id(where to load adapters) not specified when in inference mode.")
|
|
if train_mode and peft_config.peft_type == PeftType.LORA and auto_find_all_linears:
|
|
peft_config.target_modules = find_all_linear_names(model, ignore_lm_head=True)
|
|
|
|
if isinstance(peft_config, LoraConfig) and not isinstance(peft_config, GPTQLoraConfig):
|
|
peft_config = GPTQLoraConfig(**peft_config.to_dict())
|
|
peft_config.injected_fused_attention = model.injected_fused_attention
|
|
|
|
with hijack_peft_mappings():
|
|
try:
|
|
if train_mode:
|
|
peft_model = get_peft_model(model.model, peft_config)
|
|
else:
|
|
peft_model = PeftModel.from_pretrained(model.model, model_id, adapter_name)
|
|
except:
|
|
raise NotImplementedError(f"auto_gptq not support {peft_config.peft_type.value} peft type yet.")
|
|
|
|
return peft_model
|
|
|
|
|
|
__all__ = ["get_gptq_peft_model"]
|