2024-07-08 14:35:27 +00:00
|
|
|
#!/usr/bin/env python3
|
|
|
|
# -*- coding: utf-8 -*-
|
|
|
|
|
|
|
|
from __future__ import annotations
|
|
|
|
|
2024-07-09 22:26:38 +00:00
|
|
|
from dataclasses import dataclass
|
2024-07-08 14:35:27 +00:00
|
|
|
import logging
|
|
|
|
import argparse
|
|
|
|
import os
|
|
|
|
import sys
|
|
|
|
from pathlib import Path
|
2024-07-09 22:26:38 +00:00
|
|
|
from types import EllipsisType
|
|
|
|
from typing import TYPE_CHECKING, Callable, Iterable, Iterator, Sequence, SupportsIndex, cast
|
2024-07-08 14:35:27 +00:00
|
|
|
|
|
|
|
import torch
|
|
|
|
|
|
|
|
if TYPE_CHECKING:
|
|
|
|
from torch import Tensor
|
|
|
|
|
|
|
|
if 'NO_LOCAL_GGUF' not in os.environ:
|
|
|
|
sys.path.insert(1, str(Path(__file__).parent / 'gguf-py'))
|
|
|
|
import gguf
|
|
|
|
|
|
|
|
# reuse model definitions from convert_hf_to_gguf.py
|
|
|
|
from convert_hf_to_gguf import Model
|
|
|
|
|
|
|
|
logger = logging.getLogger("lora-to-gguf")
|
|
|
|
|
2024-07-08 19:55:41 +00:00
|
|
|
|
2024-07-09 22:26:38 +00:00
|
|
|
@dataclass
|
|
|
|
class PartialLoraTensor:
|
|
|
|
A: Tensor | None = None
|
|
|
|
B: Tensor | None = None
|
|
|
|
|
|
|
|
|
|
|
|
# magic to support tensor shape modifications and splitting
|
|
|
|
class LoraTorchTensor:
|
|
|
|
_lora_A: Tensor
|
|
|
|
_lora_B: Tensor
|
|
|
|
_rank: int
|
|
|
|
|
|
|
|
def __init__(self, A: Tensor, B: Tensor):
|
|
|
|
assert len(A.shape) == len(B.shape)
|
|
|
|
if A.dtype != B.dtype:
|
|
|
|
A = A.to(torch.float32)
|
|
|
|
B = B.to(torch.float32)
|
|
|
|
self._lora_A = A
|
|
|
|
self._lora_B = B
|
|
|
|
assert self._lora_A.shape[-2] == self._lora_B.shape[-1]
|
|
|
|
self._rank = self._lora_B.shape[-1]
|
|
|
|
|
|
|
|
def __getitem__(
|
|
|
|
self,
|
|
|
|
indices: (
|
|
|
|
SupportsIndex
|
|
|
|
| slice
|
|
|
|
| tuple[SupportsIndex | slice | EllipsisType | Tensor, ...]
|
|
|
|
),
|
|
|
|
) -> LoraTorchTensor:
|
|
|
|
shape = self.shape
|
|
|
|
if isinstance(indices, (SupportsIndex, slice)):
|
|
|
|
if len(shape) > 2:
|
|
|
|
return LoraTorchTensor(self._lora_A[indices], self._lora_B[indices])
|
|
|
|
else:
|
|
|
|
raise NotImplementedError
|
|
|
|
elif isinstance(indices, tuple):
|
|
|
|
assert len(indices) > 0
|
|
|
|
if isinstance(indices[-1], EllipsisType):
|
|
|
|
return self[indices[:-1]]
|
|
|
|
# expand ellipsis
|
|
|
|
indices = tuple(
|
|
|
|
u
|
|
|
|
for v in (
|
|
|
|
(
|
|
|
|
(slice(None, None) for _ in range(len(indices) - 1))
|
|
|
|
if isinstance(i, EllipsisType)
|
|
|
|
else (i,)
|
|
|
|
)
|
|
|
|
for i in indices
|
|
|
|
)
|
|
|
|
for u in v
|
|
|
|
)
|
|
|
|
|
|
|
|
if len(indices) < len(shape):
|
|
|
|
indices = (*indices, *(slice(None, None) for _ in range(len(indices), len(shape))))
|
|
|
|
|
|
|
|
# TODO: make sure this is correct
|
|
|
|
# lora_A has a shape which looks like (..., 1, 1, rank, self.shape[-1])
|
|
|
|
indices_A = (
|
|
|
|
*(
|
|
|
|
0 if isinstance(i, SupportsIndex) else slice(None, None)
|
|
|
|
for i in indices[:-2]
|
|
|
|
),
|
|
|
|
slice(None, None),
|
|
|
|
indices[-1],
|
|
|
|
)
|
|
|
|
indices_B = indices[:-1]
|
|
|
|
return LoraTorchTensor(self._lora_A[indices_A], self._lora_B[indices_B])
|
|
|
|
else:
|
|
|
|
raise NotImplementedError
|
|
|
|
|
|
|
|
@property
|
|
|
|
def dtype(self) -> torch.dtype:
|
|
|
|
assert self._lora_A.dtype == self._lora_B.dtype
|
|
|
|
return self._lora_A.dtype
|
|
|
|
|
|
|
|
@property
|
|
|
|
def shape(self) -> tuple[int, ...]:
|
|
|
|
return (*self._lora_B.shape[:-1], self._lora_A.shape[-1])
|
|
|
|
|
|
|
|
def size(self, dim=None):
|
|
|
|
assert dim is None
|
|
|
|
return self.shape
|
|
|
|
|
|
|
|
def reshape(self, *shape: int | tuple[int]) -> LoraTorchTensor:
|
|
|
|
if isinstance(shape[0], tuple):
|
|
|
|
new_shape: tuple[int] = shape[0]
|
|
|
|
else:
|
|
|
|
new_shape = cast(tuple[int], shape)
|
|
|
|
orig_shape = self.shape
|
|
|
|
if new_shape[-1] != orig_shape[-1]:
|
|
|
|
raise NotImplementedError
|
|
|
|
return LoraTorchTensor(
|
|
|
|
self._lora_A.reshape((*(1 for _ in new_shape[:-2]), *self._lora_A.shape[-2:])),
|
|
|
|
self._lora_B.reshape((*new_shape[:-1], self._rank)),
|
|
|
|
)
|
|
|
|
|
|
|
|
def reshape_as(self, other: Tensor) -> LoraTorchTensor:
|
|
|
|
return self.reshape(*other.shape)
|
|
|
|
|
|
|
|
def view(self, *size: int) -> LoraTorchTensor:
|
|
|
|
return self.reshape(*size)
|
|
|
|
|
|
|
|
def permute(self, *dims: int) -> LoraTorchTensor:
|
|
|
|
shape = self.shape
|
|
|
|
dims = tuple(dim - len(shape) if dim >= 0 else dim for dim in dims)
|
|
|
|
if dims[-1] == -2 and dims[-2] == -1:
|
|
|
|
return LoraTorchTensor(self._lora_B.permute(*dims), self._lora_A.permute(*dims))
|
|
|
|
else:
|
|
|
|
assert dims[-1] == -1
|
|
|
|
assert all(dim == 1 for dim in self._lora_A.shape[:-2])
|
|
|
|
return LoraTorchTensor(self._lora_A, self._lora_B.permute(*dims))
|
|
|
|
|
|
|
|
def transpose(self, dim0: int, dim1: int) -> LoraTorchTensor:
|
|
|
|
shape = self.shape
|
|
|
|
dims = [i for i in range(len(shape))]
|
|
|
|
dims[dim0], dims[dim1] = dims[dim1], dims[dim0]
|
|
|
|
return self.permute(*dims)
|
|
|
|
|
|
|
|
def swapaxes(self, axis0: int, axis1: int) -> LoraTorchTensor:
|
|
|
|
return self.transpose(axis0, axis1)
|
|
|
|
|
|
|
|
def to(self, *args, **kwargs):
|
|
|
|
return LoraTorchTensor(self._lora_A.to(*args, **kwargs), self._lora_B.to(*args, **kwargs))
|
|
|
|
|
|
|
|
@classmethod
|
|
|
|
def __torch_function__(cls, func: Callable, types, args=(), kwargs=None):
|
|
|
|
del types # unused
|
|
|
|
|
|
|
|
if kwargs is None:
|
|
|
|
kwargs = {}
|
|
|
|
|
|
|
|
if func is torch.permute:
|
|
|
|
return type(args[0]).permute(*args, **kwargs)
|
|
|
|
elif func is torch.reshape:
|
|
|
|
return type(args[0]).reshape(*args, **kwargs)
|
|
|
|
elif func is torch.stack:
|
|
|
|
assert isinstance(args[0], Sequence)
|
|
|
|
dim = kwargs.get("dim", 0)
|
|
|
|
assert dim == 0
|
|
|
|
return LoraTorchTensor(
|
|
|
|
torch.stack([a._lora_A for a in args[0]], dim),
|
|
|
|
torch.stack([b._lora_B for b in args[0]], dim),
|
|
|
|
)
|
|
|
|
elif func is torch.cat:
|
|
|
|
assert isinstance(args[0], Sequence)
|
|
|
|
dim = kwargs.get("dim", 0)
|
|
|
|
assert dim == 0
|
|
|
|
if len(args[0][0].shape) > 2:
|
|
|
|
return LoraTorchTensor(
|
|
|
|
torch.cat([a._lora_A for a in args[0]], dim),
|
|
|
|
torch.cat([b._lora_B for b in args[0]], dim),
|
|
|
|
)
|
|
|
|
else:
|
|
|
|
return LoraTorchTensor(
|
|
|
|
args[0][0]._lora_A, # TODO: is this correct? (can't cat over the rank)
|
|
|
|
torch.cat([b._lora_B for b in args[0]], dim),
|
|
|
|
)
|
|
|
|
else:
|
|
|
|
raise NotImplementedError
|
|
|
|
|
|
|
|
|
2024-07-09 22:23:07 +00:00
|
|
|
def get_base_tensor_name(lora_tensor_name: str) -> str:
|
|
|
|
base_name = lora_tensor_name.replace("base_model.model.", "")
|
|
|
|
base_name = base_name.replace(".lora_A.weight", ".weight")
|
|
|
|
base_name = base_name.replace(".lora_B.weight", ".weight")
|
|
|
|
return base_name
|
|
|
|
|
|
|
|
|
2024-07-08 14:35:27 +00:00
|
|
|
def parse_args() -> argparse.Namespace:
|
|
|
|
parser = argparse.ArgumentParser(
|
2024-07-08 19:55:41 +00:00
|
|
|
description="Convert a huggingface PEFT LoRA adapter to a GGML compatible file")
|
2024-07-08 14:35:27 +00:00
|
|
|
parser.add_argument(
|
|
|
|
"--outfile", type=Path,
|
2024-07-08 20:05:35 +00:00
|
|
|
help="path to write to; default: based on input. {ftype} will be replaced by the outtype.",
|
2024-07-08 14:35:27 +00:00
|
|
|
)
|
|
|
|
parser.add_argument(
|
2024-07-08 19:55:41 +00:00
|
|
|
"--outtype", type=str, choices=["f32", "f16", "bf16", "q8_0"], default="f16",
|
|
|
|
help="output format - use f32 for float32, f16 for float16, bf16 for bfloat16, q8_0 for Q8_0",
|
2024-07-08 14:35:27 +00:00
|
|
|
)
|
|
|
|
parser.add_argument(
|
|
|
|
"--bigendian", action="store_true",
|
|
|
|
help="model is executed on big endian machine",
|
|
|
|
)
|
|
|
|
parser.add_argument(
|
|
|
|
"--verbose", action="store_true",
|
|
|
|
help="increase output verbosity",
|
|
|
|
)
|
|
|
|
parser.add_argument(
|
|
|
|
"--base", type=Path, required=True,
|
|
|
|
help="directory containing base model file",
|
|
|
|
)
|
|
|
|
parser.add_argument(
|
|
|
|
"lora_path", type=Path,
|
|
|
|
help="directory containing LoRA adapter file",
|
|
|
|
)
|
|
|
|
|
|
|
|
return parser.parse_args()
|
|
|
|
|
|
|
|
|
|
|
|
if __name__ == '__main__':
|
|
|
|
args = parse_args()
|
|
|
|
logging.basicConfig(level=logging.DEBUG if args.verbose else logging.INFO)
|
|
|
|
|
|
|
|
ftype_map: dict[str, gguf.LlamaFileType] = {
|
|
|
|
"f32": gguf.LlamaFileType.ALL_F32,
|
|
|
|
"f16": gguf.LlamaFileType.MOSTLY_F16,
|
|
|
|
"bf16": gguf.LlamaFileType.MOSTLY_BF16,
|
|
|
|
"q8_0": gguf.LlamaFileType.MOSTLY_Q8_0,
|
|
|
|
}
|
2024-07-08 19:55:41 +00:00
|
|
|
ftype = ftype_map[args.outtype]
|
2024-07-08 14:35:27 +00:00
|
|
|
|
|
|
|
dir_base_model = args.base
|
|
|
|
dir_lora = args.lora_path
|
|
|
|
input_json = os.path.join(dir_lora, "adapter_config.json")
|
2024-07-09 22:26:38 +00:00
|
|
|
input_model = os.path.join(dir_lora, "adapter_model.safetensors")
|
2024-07-08 14:35:27 +00:00
|
|
|
if args.outfile is not None:
|
|
|
|
fname_out = args.outfile
|
|
|
|
else:
|
|
|
|
# output in the same directory as the model by default
|
2024-07-08 20:05:35 +00:00
|
|
|
fname_out = dir_lora / 'ggml-lora-{ftype}.gguf'
|
2024-07-08 14:35:27 +00:00
|
|
|
|
|
|
|
if os.path.exists(input_model):
|
|
|
|
# lazy import load_file only if lora is in safetensors format.
|
|
|
|
from safetensors.torch import load_file
|
2024-07-09 22:26:38 +00:00
|
|
|
|
2024-07-08 14:35:27 +00:00
|
|
|
lora_model = load_file(input_model, device="cpu")
|
2024-07-09 22:26:38 +00:00
|
|
|
else:
|
|
|
|
input_model = os.path.join(dir_lora, "adapter_model.bin")
|
|
|
|
lora_model = torch.load(input_model, map_location="cpu", weights_only=True)
|
2024-07-08 14:35:27 +00:00
|
|
|
|
|
|
|
# load base model
|
|
|
|
logger.info(f"Loading base model: {dir_base_model.name}")
|
|
|
|
hparams = Model.load_hparams(dir_base_model)
|
|
|
|
with torch.inference_mode():
|
|
|
|
try:
|
|
|
|
model_class = Model.from_model_architecture(hparams["architectures"][0])
|
|
|
|
except NotImplementedError:
|
|
|
|
logger.error(f"Model {hparams['architectures'][0]} is not supported")
|
|
|
|
sys.exit(1)
|
|
|
|
|
2024-07-09 22:26:38 +00:00
|
|
|
class LoraModel(model_class):
|
|
|
|
model_arch = model_class.model_arch
|
2024-07-08 14:35:27 +00:00
|
|
|
|
2024-07-09 22:26:38 +00:00
|
|
|
def get_tensors(self) -> Iterator[tuple[str, Tensor]]:
|
|
|
|
tensor_map: dict[str, PartialLoraTensor] = {}
|
2024-07-08 14:35:27 +00:00
|
|
|
|
2024-07-09 22:26:38 +00:00
|
|
|
for name, tensor in lora_model.items():
|
|
|
|
base_name = get_base_tensor_name(name)
|
|
|
|
is_lora_a = ".lora_A.weight" in name
|
|
|
|
is_lora_b = ".lora_B.weight" in name
|
|
|
|
if not is_lora_a and not is_lora_b:
|
|
|
|
if ".base_layer.weight" in name:
|
|
|
|
continue
|
|
|
|
logger.error(f"Unexpected name '{name}': Not a lora_A or lora_B tensor")
|
|
|
|
sys.exit(1)
|
2024-07-08 14:35:27 +00:00
|
|
|
|
2024-07-09 22:26:38 +00:00
|
|
|
if base_name in tensor_map:
|
|
|
|
if is_lora_a:
|
|
|
|
tensor_map[base_name].A = tensor
|
|
|
|
else:
|
|
|
|
tensor_map[base_name].B = tensor
|
|
|
|
else:
|
|
|
|
if is_lora_a:
|
|
|
|
tensor_map[base_name] = PartialLoraTensor(A=tensor)
|
|
|
|
else:
|
|
|
|
tensor_map[base_name] = PartialLoraTensor(B=tensor)
|
2024-07-08 14:35:27 +00:00
|
|
|
|
2024-07-09 22:26:38 +00:00
|
|
|
for name, tensor in tensor_map.items():
|
|
|
|
assert tensor.A is not None
|
|
|
|
assert tensor.B is not None
|
|
|
|
yield (name, cast(torch.Tensor, LoraTorchTensor(tensor.A, tensor.B)))
|
2024-07-08 15:05:17 +00:00
|
|
|
|
2024-07-09 22:26:38 +00:00
|
|
|
def modify_tensors(self, data_torch: Tensor, name: str, bid: int | None) -> Iterable[tuple[str, Tensor]]:
|
|
|
|
dest = super().modify_tensors(data_torch, name, bid)
|
|
|
|
for dest_name, dest_data in dest:
|
|
|
|
assert isinstance(dest_data, LoraTorchTensor)
|
|
|
|
# logger.info(f"{orig_name} --> {dest_name}")
|
|
|
|
yield (dest_name + ".lora_a", dest_data._lora_A)
|
|
|
|
yield (dest_name + ".lora_b", dest_data._lora_B)
|
2024-07-09 22:23:07 +00:00
|
|
|
|
2024-07-09 22:26:38 +00:00
|
|
|
model_instance = LoraModel(dir_base_model, ftype, fname_out, args.bigendian, False, False, None)
|
|
|
|
logger.info("Set model parameters")
|
|
|
|
model_instance.set_gguf_parameters()
|
2024-07-09 22:23:07 +00:00
|
|
|
|
2024-07-09 22:26:38 +00:00
|
|
|
# adapter_config = json.load(input_json)
|
|
|
|
model_instance.gguf_writer.add_string("training.type", "finetune_lora")
|
2024-07-08 19:55:41 +00:00
|
|
|
|
2024-07-09 22:26:38 +00:00
|
|
|
model_instance.gguf_writer.add_quantization_version(gguf.GGML_QUANT_VERSION)
|
|
|
|
logger.info("Exporting model...")
|
|
|
|
model_instance.write()
|
|
|
|
logger.info(f"Model successfully exported to {model_instance.fname_out}")
|