zchen/b2txt25
1
0
Fork 0
Code Issues Pull Requests 1 Actions Packages Projects Releases Wiki Activity

Increase batch size from 256 to 512 for training in rnn_args.yaml

Browse Source
This commit is contained in:
Zchen
2025-10-20 00:21:33 +08:00
parent 6e02894a8a
commit e1669b5a4c
1 changed files with 1 additions and 1 deletions
Download Patch File Download Diff File Expand all files Collapse all files

2
model_training_nnn_tpu/rnn_args.yaml
View File

@@ -74,7 +74,7 @@ dataset:
smooth_kernel_std: 2 # standard deviation of the smoothing kernel applied to the data
neural_dim: 512 # dimensionality of the neural data
batch_size: 256 # batch size for training (reduced for TPU memory constraints)
batch_size: 512 # batch size for training (reduced for TPU memory constraints)
n_classes: 41 # number of classes (phonemes) in the dataset
max_seq_elements: 500 # maximum number of sequence elements (phonemes) for any trial
days_per_batch: 4 # number of randomly-selected days to include in each batch