ctopd / kdflow /arguments /fsdp_args.py
codemaivanngu's picture
chore(upstream): import Cross-Tokenizer-OPD source snapshot
0ca529a unverified
Raw History Blame Contribute Delete
387 Bytes
from dataclasses import dataclass, field
from typing import Optional
@dataclass
class FSDPArguments:
""" Arguments for fsdp backend."""
fsdp_size: int = field(
default=-1,
metadata={"help": "FSDP shard size (for HSDP)."}
)
cpu_offload: bool = field(
default=False,
metadata={"help": "Offload Adam optimizer to GPU."}
)