Search

huggingface-pytorch-tpuv3x8

train steps
450
batch-size
32
samples/sec
Empty
train-time(sec)
128.7528
x faster
Empty
python xla_spawn.py --num_cores 8 \ language-modeling/run_mlm.py \ --model_name_or_path roberta-base\ --dataset_name wikitext\ --dataset_config_name wikitext-2-raw-v1\ --do_train\ --do_eval\ --output_dir ../test-result\ --per_device_train_batch_size=4
Shell
[INFO|trainer.py:837] 2021-02-20 16:36:23,316 >> ***** Running training ***** [INFO|trainer.py:838] 2021-02-20 16:36:23,316 >> Num examples = 4798 [INFO|trainer.py:839] 2021-02-20 16:36:23,316 >> Num Epochs = 3 [INFO|trainer.py:840] 2021-02-20 16:36:23,316 >> Instantaneous batch size per device = 4 [INFO|trainer.py:841] 2021-02-20 16:36:23,316 >> Total train batch size (w. parallel, distributed & accumulation) = 32 [INFO|trainer.py:842] 2021-02-20 16:36:23,316 >> Gradient Accumulation steps = 1 [INFO|trainer.py:843] 2021-02-20 16:36:23,316 >> Total optimization steps = 450 02/20/2021 16:38:37 - INFO - run_mlm - ***** Train results ***** 02/20/2021 16:38:37 - INFO - run_mlm - epoch = 3.0 02/20/2021 16:38:37 - INFO - run_mlm - train_runtime = 128.7528 02/20/2021 16:38:37 - INFO - run_mlm - train_samples_per_second = 3.495 02/20/2021 16:38:37 - INFO - run_mlm - *** Evaluate *** [INFO|trainer.py:1600] 2021-02-20 16:38:37,436 >> ***** Running Evaluation ***** [INFO|trainer.py:1601] 2021-02-20 16:38:37,436 >> Num examples = 496 [INFO|trainer.py:1602] 2021-02-20 16:38:37,437 >> Batch size = 8 100%|###########################################################################################################################| 8/8 [00:01<00:00, 6.34it/s] 02/20/2021 16:38:39 - INFO - run_mlm - ***** Eval results ***** 02/20/2021 16:38:39 - INFO - run_mlm - perplexity = 3.478115408190154
Shell