Multimodal-CoT / timm /utils /clip_grad.py
cooelf's picture
update
a6dac9a
raw
history blame
796 Bytes
import torch
from timm.utils.agc import adaptive_clip_grad
def dispatch_clip_grad(parameters, value: float, mode: str = 'norm', norm_type: float = 2.0):
""" Dispatch to gradient clipping method
Args:
parameters (Iterable): model parameters to clip
value (float): clipping value/factor/norm, mode dependant
mode (str): clipping mode, one of 'norm', 'value', 'agc'
norm_type (float): p-norm, default 2.0
"""
if mode == 'norm':
torch.nn.utils.clip_grad_norm_(parameters, value, norm_type=norm_type)
elif mode == 'value':
torch.nn.utils.clip_grad_value_(parameters, value)
elif mode == 'agc':
adaptive_clip_grad(parameters, value, norm_type=norm_type)
else:
assert False, f"Unknown clip mode ({mode})."