File size: 3,284 Bytes
2ea2835 | 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 | import os
from typing import Optional
import numpy as np
import random
import torch
def _placeholder(value, placeholder: str) -> Optional[str]:
"""Return ``value`` if it is set and not equal to ``placeholder``, else ``None``."""
if value is None:
return None
s = str(value)
if not s or s == placeholder:
return None
return s
def init_wandb_from_config(config) -> None:
"""Initialise ``wandb`` from ``config`` and environment variables.
The shipped YAMLs use placeholder strings (``WANDB_HOST``, ``WANDB_KEY``
etc.) for the wandb fields so the configs can be committed without
leaking secrets. This helper treats those placeholders as unset and
falls back to standard ``WANDB_API_KEY``/``WANDB_ENTITY``/``WANDB_PROJECT``
environment variables (or any pre-existing ``wandb login``).
"""
import wandb
from omegaconf import OmegaConf
wb_key = os.environ.get("WANDB_API_KEY") or _placeholder(
getattr(config, "wandb_key", None), "WANDB_KEY"
)
wb_host = _placeholder(getattr(config, "wandb_host", None), "WANDB_HOST")
wb_entity = os.environ.get("WANDB_ENTITY") or _placeholder(
getattr(config, "wandb_entity", None), "WANDB_ENTITY"
)
wb_project = os.environ.get("WANDB_PROJECT") or _placeholder(
getattr(config, "wandb_project", None), "WANDB_PROJECT"
)
wb_name = os.environ.get("WANDB_NAME") or getattr(config, "config_name", None)
wb_run_group = os.environ.get("WANDB_RUN_GROUP") or _placeholder(
getattr(config, "wandb_run_group", None), "WANDB_RUN_GROUP"
)
if wb_key:
login_kwargs = {"key": wb_key}
if wb_host:
login_kwargs["host"] = wb_host
wandb.login(**login_kwargs)
init_kwargs = dict(
config=OmegaConf.to_container(config, resolve=True),
name=wb_name,
mode="online",
dir=getattr(config, "wandb_save_dir", "") or None,
)
if wb_entity:
init_kwargs["entity"] = wb_entity
if wb_project:
init_kwargs["project"] = wb_project
if wb_run_group:
init_kwargs["group"] = wb_run_group
wandb.init(**init_kwargs)
def set_seed(seed: int, deterministic: bool = False):
"""
Helper function for reproducible behavior to set the seed in `random`, `numpy`, `torch`.
Args:
seed (`int`):
The seed to set.
deterministic (`bool`, *optional*, defaults to `False`):
Whether to use deterministic algorithms where available. Can slow down training.
"""
random.seed(seed)
np.random.seed(seed)
torch.manual_seed(seed)
torch.cuda.manual_seed_all(seed)
if deterministic:
torch.use_deterministic_algorithms(True)
def merge_dict_list(dict_list):
if len(dict_list) == 1:
return dict_list[0]
merged_dict = {}
for k, v in dict_list[0].items():
if isinstance(v, torch.Tensor):
if v.ndim == 0:
merged_dict[k] = torch.stack([d[k] for d in dict_list], dim=0)
else:
merged_dict[k] = torch.cat([d[k] for d in dict_list], dim=0)
else:
# for non-tensor values, we just copy the value from the first item
merged_dict[k] = v
return merged_dict
|