メモリ効率のいいfp8化スクリプト。 / Memory efficient fp8 conversion script.

mem_eff_fp8_convert.py

import json
from pathlib import Path
import torch
from tqdm import tqdm
import struct
from typing import Dict, Any
import sys

# input file
if(len(sys.argv) < 3):
    print("Usage: mem_eff_fp8_convert.py {fp16 model path} {output path}")
    sys.exit(1)

path = sys.argv[1]
output =sys.argv[2]
model_file = Path(path)

class MemoryEfficientSafeOpen:
    # does not support metadata loading
    def __init__(self, filename):
        self.filename = filename
        self.header, self.header_size = self._read_header()
        self.file = open(filename, "rb")

    def __enter__(self):
        return self

    def __exit__(self, exc_type, exc_val, exc_tb):
        self.file.close()

    def keys(self):
        return [k for k in self.header.keys() if k != "__metadata__"]

    def get_tensor(self, key):
        if key not in self.header:
            raise KeyError(f"Tensor '{key}' not found in the file")

        metadata = self.header[key]
        offset_start, offset_end = metadata["data_offsets"]

        if offset_start == offset_end:
            tensor_bytes = None
        else:
            # adjust offset by header size
            self.file.seek(self.header_size + 8 + offset_start)
            tensor_bytes = self.file.read(offset_end - offset_start)

        return self._deserialize_tensor(tensor_bytes, metadata)

    def _read_header(self):
        with open(self.filename, "rb") as f:
            header_size = struct.unpack("<Q", f.read(8))[0]
            header_json = f.read(header_size).decode("utf-8")
            return json.loads(header_json), header_size

    def _deserialize_tensor(self, tensor_bytes, metadata):
        dtype = self._get_torch_dtype(metadata["dtype"])
        shape = metadata["shape"]

        if tensor_bytes is None:
            byte_tensor = torch.empty(0, dtype=torch.uint8)
        else:
            tensor_bytes = bytearray(tensor_bytes)  # make it writable
            byte_tensor = torch.frombuffer(tensor_bytes, dtype=torch.uint8)

        # process float8 types
        if metadata["dtype"] in ["F8_E5M2", "F8_E4M3"]:
            return self._convert_float8(byte_tensor, metadata["dtype"], shape)

        # convert to the target dtype and reshape
        return byte_tensor.view(dtype).reshape(shape)

    @staticmethod
    def _get_torch_dtype(dtype_str):
        dtype_map = {
            "F64": torch.float64,
            "F32": torch.float32,
            "F16": torch.float16,
            "BF16": torch.bfloat16,
            "I64": torch.int64,
            "I32": torch.int32,
            "I16": torch.int16,
            "I8": torch.int8,
            "U8": torch.uint8,
            "BOOL": torch.bool,
        }
        # add float8 types if available
        if hasattr(torch, "float8_e5m2"):
            dtype_map["F8_E5M2"] = torch.float8_e5m2
        if hasattr(torch, "float8_e4m3fn"):
            dtype_map["F8_E4M3"] = torch.float8_e4m3fn
        return dtype_map.get(dtype_str)

    @staticmethod
    def _convert_float8(byte_tensor, dtype_str, shape):
        if dtype_str == "F8_E5M2" and hasattr(torch, "float8_e5m2"):
            return byte_tensor.view(torch.float8_e5m2).reshape(shape)
        elif dtype_str == "F8_E4M3" and hasattr(torch, "float8_e4m3fn"):
            return byte_tensor.view(torch.float8_e4m3fn).reshape(shape)
        else:
            # # convert to float16 if float8 is not supported
            # print(f"Warning: {dtype_str} is not supported in this PyTorch version. Converting to float16.")
            # return byte_tensor.view(torch.uint8).to(torch.float16).reshape(shape)
            raise ValueError(f"Unsupported float8 type: {dtype_str} (upgrade PyTorch to support float8 types)")


def mem_eff_save_file(tensors: Dict[str, torch.Tensor], filename: str, metadata: Dict[str, Any] = None):
    _TYPES = {
        torch.float64: "F64",
        torch.float32: "F32",
        torch.float16: "F16",
        torch.bfloat16: "BF16",
        torch.int64: "I64",
        torch.int32: "I32",
        torch.int16: "I16",
        torch.int8: "I8",
        torch.uint8: "U8",
        torch.bool: "BOOL",
        getattr(torch, "float8_e5m2", None): "F8_E5M2",
        getattr(torch, "float8_e4m3fn", None): "F8_E4M3",
    }
    _ALIGN = 256

    def validate_metadata(metadata: Dict[str, Any]) -> Dict[str, str]:
        validated = {}
        for key, value in metadata.items():
            if not isinstance(key, str):
                raise ValueError(f"Metadata key must be a string, got {type(key)}")
            if not isinstance(value, str):
                print(f"Warning: Metadata value for key '{key}' is not a string. Converting to string.")
                validated[key] = str(value)
            else:
                validated[key] = value
        return validated

    header = {}
    offset = 0
    if metadata:
        header["__metadata__"] = validate_metadata(metadata)
    for k, v in tensors.items():
        if v.numel() == 0:  # empty tensor
            header[k] = {"dtype": _TYPES[v.dtype], "shape": list(v.shape), "data_offsets": [offset, offset]}
        else:
            size = v.numel() * v.element_size()
            header[k] = {"dtype": _TYPES[v.dtype], "shape": list(v.shape), "data_offsets": [offset, offset + size]}
            offset += size

    hjson = json.dumps(header).encode("utf-8")
    hjson += b" " * (-(len(hjson) + 8) % _ALIGN)

    with open(filename, "wb") as f:
        f.write(struct.pack("<Q", len(hjson)))
        f.write(hjson)

        for k, v in tensors.items():
            if v.numel() == 0:
                continue
            if v.is_cuda:
                # Direct GPU to disk save
                with torch.cuda.device(v.device):
                    if v.dim() == 0:  # if scalar, need to add a dimension to work with view
                        v = v.unsqueeze(0)
                    tensor_bytes = v.contiguous().view(torch.uint8)
                    tensor_bytes.cpu().numpy().tofile(f)
            else:
                # CPU tensor save
                if v.dim() == 0:  # if scalar, need to add a dimension to work with view
                    v = v.unsqueeze(0)
                v.contiguous().view(torch.uint8).numpy().tofile(f)


# read safetensors metadata
def read_safetensors_metadata(path: str):
    with open(path, 'rb') as f:
        header_size = int.from_bytes(f.read(8), 'little')
        header_json = f.read(header_size).decode('utf-8')
        header = json.loads(header_json)
        metadata = header.get('__metadata__', {})
        return metadata

metadata = read_safetensors_metadata(path)
print(json.dumps(metadata, indent=4)) #show metadata

sd_pruned = dict() #initialize empty dict

with MemoryEfficientSafeOpen(path) as reader:
    keys = reader.keys()
    for key in tqdm(keys): #for each key in the safetensors file
        sd_pruned[key] = reader.get_tensor(key).to(torch.float8_e4m3fn) #convert to fp8

# save the pruned safetensors file
mem_eff_save_file(sd_pruned, output, metadata={"format": "pt", **metadata})

スペシャルサンクス:

• Kijai: 画像生成モデルのfp8化スクリプトの作者様。
• kohya-ss: メインメモリを消費しないsafetensorsファイル保存スクリプトの作者様。
  お二方のコードがなければ、このスクリプトを作ることは出来ませんでした。多大な感謝を申し上げます。
  また、開発に協力してくださったharuharu-1105氏にも感謝を申し上げさせていただきます。

Special Thanks:

• Kijai: Author of a script to convert image generation model to fp8.
• kohya-ss: Author of a script to save safetensors files without consuming main memory.
  Without their code, I would not have been able to create this script. A huge thanks goes out to them.
  I would also like to thank haruharu-1105 for his help in the development.

アップデート： kohya-ss氏のmem_eff_safeopen.pyを組み込み、更にメモリ効率の良い変換を実現しました。Animagine XL 3.1の場合、メモリ使用量3.4GBで変換できます。
ライセンス： kohya-ss氏のmem_eff_safeopen.pyとmem_eff_save_file.pyはApache 2.0ライセンスで提供されています。

Update: I incorporated kohya-ss's mem_eff_safeopen.py for even more memory-efficient conversion.
For example, Animagine XL 3.1 can convert to fp8 with 3.4 GB of memory usage.
License: kohya-ss's mem_eff_safeopen.py and mem_eff_save_file.py are provided under the Apache 2.0 license.

Good work!

Thank you!