Skip to content

Add retry to dataset loading #10

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
gobbleturk merged 7 commits into from
Apr 22, 2023
Merged

Add retry to dataset loading #10

Changes from all commits
Commits
Show all changes
7 commits
Select commit Hold shift + click to select a range
e35ee7e
Add retry to data loading
gobbleturk Apr 21, 2023
04019fa
Add retry to data loading
gobbleturk Apr 21, 2023
78a0e70
Add retry to data loading
gobbleturk Apr 21, 2023
fab0e62
Add retry to data loading
gobbleturk Apr 21, 2023
fd44426
Add retry to data loading
gobbleturk Apr 21, 2023
7208aff
Add retry to data loading
gobbleturk Apr 21, 2023
7eb69fe
Add retry to data loading
gobbleturk Apr 21, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
8 changes: 7 additions & 1 deletion MaxText/multihost_dataloading.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -26,6 +26,7 @@
import os
from typing import Callable, Any, Dict, List, Tuple, Optional
import tensorflow as tf # pylint: disable=g-import-not-at-top
import time
import numpy as np

import jax
Expand Down Expand Up @@ -107,7 +108,12 @@ def get_next_batch_sharded(local_dataset: tf.data.Dataset,
global_mesh: Mesh) -> jax.Array:
"""Splits the host loaded data equally over all devices."""

local_data = local_dataset.next()
try:
local_data = local_dataset.next()
except tf.errors.FailedPreconditionError:
max_logging.log("Failed to get next data batch, retrying")
time.sleep(10)
local_data = local_dataset.next()

# local_devices = jax.local_devices()
local_devices = global_mesh.local_devices
Expand Down