From 8839cf96bf42095780f1414852fb945ecd295de4 Mon Sep 17 00:00:00 2001 From: Martin Popel Date: Tue, 1 Aug 2017 11:20:55 +0200 Subject: [PATCH] better walkthrough README The default batch_size for transformer_base_single_gpu has been decreased to 2048 in transform.py (it used to be 4096 I think). So if it is too much, the user must use a smaller value. --- README.md | 4 ++-- 1 file changed, 2 insertions(+), 2 deletions(-) diff --git a/README.md b/README.md index edd6460d0..bb0f6f534 100644 --- a/README.md +++ b/README.md @@ -89,7 +89,7 @@ t2t-datagen \ --problem=$PROBLEM # Train -# * If you run out of memory, add --hparams='batch_size=2048' or even 1024. +# * If you run out of memory, add --hparams='batch_size=1024'. t2t-trainer \ --data_dir=$DATA_DIR \ --problems=$PROBLEM \ @@ -166,7 +166,7 @@ python -c "from tensor2tensor.models.transformer import Transformer" with `Modality` objects, which are specified per-feature in the dataset/task specification. * Support for multi-GPU machines and synchronous (1 master, many workers) and - asynchrounous (independent workers synchronizing through a parameter server) + asynchronous (independent workers synchronizing through a parameter server) [distributed training](https://github.com/tensorflow/tensor2tensor/tree/master/docs/distributed_training.md). * Easily swap amongst datasets and models by command-line flag with the data generation script `t2t-datagen` and the training script `t2t-trainer`.