File size: 10,942 Bytes
3b609b9 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182 183 184 185 186 187 188 189 190 191 192 193 194 195 196 197 198 199 200 201 202 203 204 205 206 207 208 209 210 211 212 213 214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237 238 239 240 241 242 243 244 245 246 247 248 249 250 |
# Copyright 2024-present the HuggingFace Inc. team.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
# http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.
import warnings
from typing import List, Optional, Tuple
import torch
import torch.nn as nn
import torch.nn.functional as F
from transformers.pytorch_utils import Conv1D
from peft.tuners.tuners_utils import BaseTunerLayer, check_adapters_to_merge
from peft.utils.other import transpose
class VBLoRALayer(BaseTunerLayer):
# List all names of layers that may contain adapter weights
adapter_layer_names = ("vblora_logits_A", "vblora_logits_B", "vblora_vector_bank")
def __init__(self, base_layer: nn.Module, **kwargs):
self.base_layer = base_layer
self.r = {}
self.topk = {}
self.vblora_dropout = nn.ModuleDict({})
# For storing vector scale
self.vblora_logits_A = nn.ParameterDict({})
self.vblora_logits_B = nn.ParameterDict({})
# Mark the weight as unmerged
self._disable_adapters = False
self.merged_adapters = []
base_layer = self.get_base_layer()
if isinstance(base_layer, nn.Linear):
in_features, out_features = base_layer.in_features, base_layer.out_features
elif isinstance(base_layer, Conv1D):
in_features, out_features = (
base_layer.weight.ds_shape if hasattr(base_layer.weight, "ds_shape") else base_layer.weight.shape
)
self.in_features = in_features
self.out_features = out_features
self.kwargs = kwargs
@property
def merged(self) -> bool:
return bool(self.merged_adapters)
def update_layer(
self,
adapter_name: str,
vblora_vector_bank,
r: int,
topk: int,
num_vectors: int,
vector_length: float,
vblora_dropout: float = 0.0,
init_logits_std: float = 0.01,
):
if r <= 0:
raise ValueError(f"`r` {r} should be a positive integer value")
if topk <= 0:
raise ValueError(f"`topk` {topk} should be a positive integer value")
if self.in_features % vector_length != 0:
raise ValueError(f"`in_features` {self.in_features} must be divisible by `vector_length` {vector_length}")
if self.out_features % vector_length != 0:
raise ValueError(
f"`out_features` {self.out_features} must be divisible by `vector_length` {vector_length}"
)
self.r[adapter_name] = r
self.topk[adapter_name] = topk
if vblora_dropout > 0.0:
vblora_dropout_layer = nn.Dropout(p=vblora_dropout)
else:
vblora_dropout_layer = nn.Identity()
self.vblora_dropout.update(nn.ModuleDict({adapter_name: vblora_dropout_layer}))
self.vblora_logits_A[adapter_name] = nn.Parameter(
torch.zeros(r, self.in_features // vector_length, num_vectors), requires_grad=True
)
self.vblora_logits_B[adapter_name] = nn.Parameter(
torch.zeros(self.out_features // vector_length, r, num_vectors), requires_grad=True
)
self.vblora_vector_bank = vblora_vector_bank
self.reset_vblora_logits(adapter_name, init_logits_std)
self._move_adapter_to_device_of_base_layer(adapter_name)
self.set_adapter(self.active_adapters)
def reset_vblora_logits(self, adapter_name, init_logits_std):
if adapter_name in self.vblora_logits_A.keys():
with torch.no_grad():
nn.init.normal_(self.vblora_logits_A[adapter_name], 0, init_logits_std)
nn.init.normal_(self.vblora_logits_B[adapter_name], 0, init_logits_std)
class Linear(nn.Linear, VBLoRALayer):
# VBLoRA implemented in a dense layer
def __init__(
self,
base_layer,
vblora_vector_bank,
adapter_name: str,
r: int,
num_vectors: int,
vector_length: int,
topk: int = 2,
vblora_dropout: float = 0.0,
init_logits_std: float = 0.01,
fan_in_fan_out: bool = False, # Set this to True if the layer to replace stores weight like (fan_in, fan_out)
is_target_conv_1d_layer: bool = False,
**kwargs,
) -> None:
# this gets the init from nn.Linear's super perspective, i.e. nn.Module.__init__, which should always be called
super(nn.Linear, self).__init__()
VBLoRALayer.__init__(self, base_layer, **kwargs)
self.fan_in_fan_out = fan_in_fan_out
self._active_adapter = adapter_name
self.update_layer(
adapter_name, vblora_vector_bank, r, topk, num_vectors, vector_length, vblora_dropout, init_logits_std
)
self.is_target_conv_1d_layer = is_target_conv_1d_layer
def merge(self, safe_merge: bool = False, adapter_names: Optional[List[str]] = None) -> None:
"""
Merge the active adapter weights into the base weights
Args:
safe_merge (`bool`, *optional*):
If True, the merge operation will be performed in a copy of the original weights and check for NaNs
before merging the weights. This is useful if you want to check if the merge operation will produce
NaNs. Defaults to `False`.
adapter_names (`List[str]`, *optional*):
The list of adapter names that should be merged. If None, all active adapters will be merged. Defaults
to `None`.
"""
adapter_names = check_adapters_to_merge(self, adapter_names)
if not adapter_names:
# no adapter to merge
return
for active_adapter in adapter_names:
if active_adapter in self.vblora_logits_A.keys():
base_layer = self.get_base_layer()
if safe_merge:
# Note that safe_merge will be slower than the normal merge
# because of the copy operation.
orig_weights = base_layer.weight.data.clone()
orig_weights += self.get_delta_weight(active_adapter)
if not torch.isfinite(orig_weights).all():
raise ValueError(
f"NaNs detected in the merged weights. The adapter {active_adapter} seems to be broken"
)
base_layer.weight.data = orig_weights
else:
base_layer.weight.data += self.get_delta_weight(active_adapter)
self.merged_adapters.append(active_adapter)
def unmerge(self) -> None:
if not self.merged:
warnings.warn("Already unmerged. Nothing to do.")
return
while len(self.merged_adapters) > 0:
active_adapter = self.merged_adapters.pop()
if active_adapter in self.vblora_logits_A.keys():
self.get_base_layer().weight.data -= self.get_delta_weight(active_adapter)
def _get_low_rank_matrix(self, logits: torch.tensor, vblora_vector_bank, topk) -> torch.Tensor:
top_k_logits, indices = logits.topk(topk, dim=-1)
topk_weights = F.softmax(top_k_logits, dim=-1)
return (topk_weights.unsqueeze(-1) * vblora_vector_bank[indices]).sum(-2)
def _get_lora_matrices(self, adapter, cast_to_fp32=False) -> Tuple[torch.Tensor, torch.Tensor]:
vblora_logits_A = self.vblora_logits_A[adapter]
vblora_logits_B = self.vblora_logits_B[adapter]
# Check for infinity values when training. If found, training was likely resumed from a `save_only_topk_weights` model.
if self.training and vblora_logits_A[0, 0].isinf().any():
raise RuntimeError(
"Found infinity values in VB-LoRA logits. Ensure training was not resumed from a `save_only_topk_weights` model."
)
vblora_vector_bank = self.vblora_vector_bank[adapter].to(vblora_logits_A.device)
topk = self.topk[adapter]
# In case users wants to merge the adapter weights that are in
# float16 while being on CPU, we need to cast the weights to float32, perform the merge and then cast back to
# float16 because the `@` and matmul operation in general is not supported in torch + cpu + fp16.
if cast_to_fp32:
vblora_logits_A = vblora_logits_A.float()
vblora_logits_B = vblora_logits_B.float()
vblora_vector_bank = vblora_vector_bank.float()
# A: (rank, in_tile, vector_length) -> (rank, in_tile x vector_length)
A = self._get_low_rank_matrix(vblora_logits_A, vblora_vector_bank, topk).reshape(vblora_logits_A.shape[0], -1)
# B: (out_tile, rank, vector_length) -> (out_tile, vector_length, rank) -> (out_tile x vector_length, rank)
B = (
self._get_low_rank_matrix(vblora_logits_B, vblora_vector_bank, topk)
.transpose(1, 2)
.reshape(-1, vblora_logits_B.shape[1])
)
return A, B
def get_delta_weight(self, adapter) -> torch.Tensor:
"""
Compute the delta weight for the given adapter.
Args:
adapter (str):
The name of the adapter for which the delta weight should be computed.
"""
device = self.vblora_logits_A[adapter].device
dtype = self.vblora_logits_A[adapter].dtype
cast_to_fp32 = device.type == "cpu" and dtype == torch.float16
A, B = self._get_lora_matrices(adapter, cast_to_fp32)
output_tensor = transpose(B @ A, self.fan_in_fan_out)
return output_tensor
def forward(self, x: torch.Tensor, *args, **kwargs) -> torch.Tensor:
previous_dtype = x.dtype
if self.disable_adapters:
if self.merged:
self.unmerge()
result = self.base_layer(x, *args, **kwargs)
elif self.merged:
result = self.base_layer(x, *args, **kwargs)
else:
result = self.base_layer(x, *args, **kwargs)
for active_adapter in self.active_adapters:
if active_adapter not in self.vblora_logits_A.keys():
continue
A, B = self._get_lora_matrices(active_adapter)
x = x.to(self.vblora_vector_bank[active_adapter].dtype)
dropout = self.vblora_dropout[active_adapter]
result = result + F.linear(F.linear(dropout(x), A), B)
result = result.to(previous_dtype)
return result
|