lingvo

lingvo copied to clipboard

Hi, I'm interested in seeing the timing relationship between layers and operations of the lm denselm128b8x8 model. I was using a v3-8 node and changed the hparams (num_device_per_split, mesh_shape) in...

chadwickleung

Conformer/ ContextNet RNN-Transducer loss computation

4

Does lingvo include the RNN-Transducer loss computation ?

pyyush

Hi @jonathanasdf , @bignamehyp What is the default buffer size for cross device communication in GPipe? Is there a way to modify this?

adis98

pip install downgrades to TF 2.4.0

1

does it really not work for 2.4.1?

bjarthur

Hello, I am trying to run the jupyter notebook example "introduction.ipynb" on "codelabs" folder using Google Colab. However, it gives me the following error: ``` 2021-05-12 03:12:03.347414: I tensorflow/stream_executor/platform/default/dso_loader.cc:48] Successfully...

agemagician

Could not find a version that satisfies the requirement lingvo

22

When i'm trying to install lingvo from pip, something wrong happend: ERROR: Could not find a version that satisfies the requirement lingvo (from versions: none) ERROR: No matching distribution found...

ruby11dog

GPU utilization down to 0% without any error infos

13

Hi, I've been training models for almost two days. Today, the GPU utilization dropped suddenly to 0%, but all GPU memory were still occupied by the experiment. Besides, the experimental...

iamxiaoyubei

For ASR, if I set num_cnn_layers = 0, Invalid argument: Matrix size-incompatible: In[0]: [96,1056], In[1]: [1120,4096]

6

My setting is: ep = p.encoder ep.input_shape = [None, None, 96, 1] ep.lstm_cell_size = 1024 ep.num_lstm_layers = 4 ep.conv_filter_shapes = [] ep.conv_filter_strides = [] ep.num_conv_lstm_layers = 0 ep.num_cnn_layers = 0...

yujunlhz

How to run GPipe in distributed manner?

8

I was able to run GPipe with lm.one_billion_wds.OneBWdsGPipeTransformerWPM in a single node with multiple GPUs. However, I am a little confused about how to run GPipe with multiple nodes (or...

Allen-Czyysx

Why Lingvo is so slow when training Librispeech960Wpm on a host with 6 GPUs

5

my host is one host, 6 GPUs(V100), the speed is about steps/sec: 0.12, so the 800,000 steps would take several months. My running command is as follows: bazel-bin/lingvo/trainer --saver_max_to_keep=3 --worker_gpus=6...

yujunlhz