maxmustermann/DeepLearningExamples
1
0
Fork 0
Code Issues Pull requests Projects Releases Wiki Activity

updated readme

Browse source
This commit is contained in:
Grzegorz Karch 2019-06-07 08:27:52 -07:00 committed by gkarch
parent 79b9a48bf1
commit dc7cd1ae5b
1 changed files with 5 additions and 5 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

10
PyTorch/SpeechSynthesis/Tacotron2/README.md
View file

@ -153,7 +153,7 @@ and `--waveglow` arguments.
To run inference issue:
```bash
python inference.py --tacotron2 <Tacotron2_checkpoint> --waveglow <WaveGlow_checkpoint> -o output/ -i text.txt --fp16-run
python inference.py --tacotron2 <Tacotron2_checkpoint> --waveglow <WaveGlow_checkpoint> -o output/ -i phrase.txt --fp16-run
```
The speech is generated from text file passed with `-i` argument. To run
inference in mixed precision, use `--fp16-run` flag. The output audio will
@ -310,11 +310,11 @@ To benchmark the inference performance on a batch size=1, run:
* For FP32
```bash
python inference.py --tacotron2 <Tacotron2_checkpoint> --waveglow <WaveGlow_checkpoint> -o output/ --include-warmup --log-file=output/nvlog_fp32.json
python inference.py --tacotron2 <Tacotron2_checkpoint> --waveglow <WaveGlow_checkpoint> -o output/ --include-warmup -i phrase.txt --log-file=output/nvlog_fp32.json
```
* For FP16
```bash
python inference.py --tacotron2 <Tacotron2_checkpoint> --waveglow <WaveGlow_checkpoint> -o output/ --include-warmup --fp16-run --log-file=output/nvlog_fp16.json
python inference.py --tacotron2 <Tacotron2_checkpoint> --waveglow <WaveGlow_checkpoint> -o output/ --include-warmup -i phrase.txt --fp16-run --log-file=output/nvlog_fp16.json
```
The output log files will contain performance numbers for Tacotron 2 model
@ -455,7 +455,7 @@ Results are measured in the number of input tokens per second.
|Number of GPUs|Mixed precision tokens/sec|FP32 tokens/sec|Speed-up with mixed precision|
|---:|---:|---:|---:|
|**1**|130|150|0.87|
|**1**|132|153|0.86|
This table shows the inference performance results for WaveGlow.
@ -463,7 +463,7 @@ Results are measured in the number of output audio samples per second.<sup>1</su
|Number of GPUs|Mixed precision samples/sec|FP32 samples/sec|Speed-up with mixed precision|
|---:|---:|---:|---:|
|**1**|435110|400097|1.09|
|**1**|425379|376037|1.13|
<sup>1</sup>With sampling rate equal to 22050, one second of audio is generated from 22050 samples.