Wav2letter: Using Pre-trained model!

[sancheza603]

Hi,

I am just trying to use the pre-trained model. I am getting the following error "ead error: read 217450 blocks instead of 43500464". Snapshot below:

$ ~/usr/bin/luajit ~/wav2letter/test.lua ~/librispeech-glu-highdropout-cpu.bin -progress -show -test dev-clean -save -datadir ~/librispeech-proc/ -dictdir ~/librispeech-proc/ -gfsai
| number of classes (network) = 30
| reloading model
/home/sanchit/usr/bin/luajit: /home/sanchit/usr/share/lua/5.1/torch/File.lua:351: read error: read 217450 blocks instead of 43500464 at /tmp/luarocks_torch-scm-1-1789/torch7/lib/TH/THDiskFile.c:356
stack traceback:
[C]: in function 'read'
/home/sanchit/usr/share/lua/5.1/torch/File.lua:351: in function
[C]: in function 'read'
/home/sanchit/usr/share/lua/5.1/torch/File.lua:351: in function 'readObject'
/home/sanchit/usr/share/lua/5.1/torch/File.lua:369: in function 'readObject'
/home/sanchit/usr/share/lua/5.1/nn/Module.lua:192: in function 'read'
/home/sanchit/usr/share/lua/5.1/torch/File.lua:351: in function 'readObject'
/home/sanchit/usr/share/lua/5.1/torch/File.lua:369: in function 'readObject'
/home/sanchit/usr/share/lua/5.1/torch/File.lua:369: in function 'readObject'
/home/sanchit/usr/share/lua/5.1/nn/Module.lua:192: in function 'read'
/home/sanchit/usr/share/lua/5.1/torch/File.lua:351: in function 'readObject'
/home/sanchit/usr/share/lua/5.1/torch/File.lua:369: in function 'readObject'
.../sanchit/usr/share/lua/5.1/wav2letter/runtime/serial.lua:57: in function 'loadmodel'
/home/sanchit/wav2letter/test.lua:102: in main chunk
[C]: at 0x00404ab0

[VitaliyLi]

Hi, it seems the model file is corrupted. Can you try to re-download the model and run it again?

[sancheza603]

Thanks for your help. Model was indeed corrupted. Re-downloading solved the problem. Now I have a new one.

So, I am trying to use CPU model to make inference on my own flac file. I am not sure whether I am going through the right steps especially in the decoding part. I have created "00000" folder in dev-clean and put my flac in there. Still I am getting this error:

~/usr/bin/luajit ~/wav2letter/test.lua ~/librispeech-glu-highdropout-cpu.bin -progress -show -test dev-clean -save -datadir ~/librispeech-proc/ -dictdir ~/librispeech-proc/ -gfsai
| number of classes (network) = 30
| reloading model
nn.Sequential {
[input -> (1) -> (2) -> (3) -> (4) -> (5) -> (6) -> (7) -> (8) -> (9) -> (10) -> (11) -> (12) -> (13) -> (14) -> (15) -> (16) -> (17) -> (18) -> (19) -> (20) -> (21) -> (22) -> (23) -> (24) -> (25) -> (26) -> (27) -> (28) -> (29) -> (30) -> (31) -> (32) -> (33) -> (34) -> (35) -> (36) -> (37) -> (38) -> (39) -> (40) -> (41) -> (42) -> (43) -> (44) -> (45) -> (46) -> (47) -> (48) -> (49) -> (50) -> (51) -> (52) -> (53) -> (54) -> (55) -> output]
(1): nn.TemporalConvolution
(2): nn.GatedLinearUnit
(3): nn.Dropout(0.200000)
(4): nn.TemporalConvolution
(5): nn.GatedLinearUnit
(6): nn.Dropout(0.214000)
(7): nn.TemporalConvolution
(8): nn.GatedLinearUnit
(9): nn.Dropout(0.228980)
(10): nn.TemporalConvolution
(11): nn.GatedLinearUnit
(12): nn.Dropout(0.245009)
(13): nn.TemporalConvolution
(14): nn.GatedLinearUnit
(15): nn.Dropout(0.262159)
(16): nn.TemporalConvolution
(17): nn.GatedLinearUnit
(18): nn.Dropout(0.280510)
(19): nn.TemporalConvolution
(20): nn.GatedLinearUnit
(21): nn.Dropout(0.300146)
(22): nn.TemporalConvolution
(23): nn.GatedLinearUnit
(24): nn.Dropout(0.321156)
(25): nn.TemporalConvolution
(26): nn.GatedLinearUnit
(27): nn.Dropout(0.343637)
(28): nn.TemporalConvolution
(29): nn.GatedLinearUnit
(30): nn.Dropout(0.367692)
(31): nn.TemporalConvolution
(32): nn.GatedLinearUnit
(33): nn.Dropout(0.393430)
(34): nn.TemporalConvolution
(35): nn.GatedLinearUnit
(36): nn.Dropout(0.420970)
(37): nn.TemporalConvolution
(38): nn.GatedLinearUnit
(39): nn.Dropout(0.450438)
(40): nn.TemporalConvolution
(41): nn.GatedLinearUnit
(42): nn.Dropout(0.481969)
(43): nn.TemporalConvolution
(44): nn.GatedLinearUnit
(45): nn.Dropout(0.515707)
(46): nn.TemporalConvolution
(47): nn.GatedLinearUnit
(48): nn.Dropout(0.551806)
(49): nn.TemporalConvolution
(50): nn.GatedLinearUnit
(51): nn.Dropout(0.590433)
(52): nn.TemporalConvolution
(53): nn.GatedLinearUnit
(54): nn.Dropout(0.590433)
(55): nn.TemporalConvolution
}
| dataset : 0 files found
| dataset : 0 files found
| dataset : 0 files found
/home/sanchit/usr/bin/luajit: /home/sanchit/usr/share/lua/5.1/threads/threads.lua:183: [thread 3 callback] ...it/usr/share/lua/5.1/wav2letter/numberedfilesdataset.lua:71: no file found in nor in
stack traceback:
[C]: in function 'assert'
...it/usr/share/lua/5.1/wav2letter/numberedfilesdataset.lua:71: in function '__init'
/home/sanchit/usr/share/lua/5.1/torch/init.lua:91: in function
[C]: in function 'closure'
/home/sanchit/usr/share/lua/5.1/wav2letter/runtime/data.lua:319: in function 'mapconcat'
/home/sanchit/usr/share/lua/5.1/wav2letter/runtime/data.lua:417: in function 'datasetwithfeatures'
/home/sanchit/usr/share/lua/5.1/wav2letter/runtime/data.lua:468: in function 'closure'
...are/lua/5.1/torchnet/dataset/paralleldatasetiterator.lua:79: in function <...are/lua/5.1/torchnet/dataset/paralleldatasetiterator.lua:78>
[C]: in function 'xpcall'
/home/sanchit/usr/share/lua/5.1/threads/threads.lua:234: in function 'callback'
/home/sanchit/usr/share/lua/5.1/threads/queue.lua:65: in function
[C]: in function 'pcall'
/home/sanchit/usr/share/lua/5.1/threads/queue.lua:40: in function 'dojob'
[string " local Queue = require 'threads.queue'..."]:13: in main chunk
stack traceback:
[C]: in function 'error'
/home/sanchit/usr/share/lua/5.1/threads/threads.lua:183: in function 'dojob'
/home/sanchit/usr/share/lua/5.1/threads/threads.lua:264: in function 'synchronize'
/home/sanchit/usr/share/lua/5.1/threads/threads.lua:142: in function 'specific'
/home/sanchit/usr/share/lua/5.1/threads/threads.lua:125: in function 'Threads'
...are/lua/5.1/torchnet/dataset/paralleldatasetiterator.lua:85: in function '__init'
/home/sanchit/usr/share/lua/5.1/torch/init.lua:91: in function
[C]: in function 'newiterator'
/home/sanchit/wav2letter/test.lua:152: in main chunk
[C]: at 0x00404ab0
| dataset : 0 files found

[sancheza603]

So after looking at some common issues, I realized that I might have missed a couple of data pre-processing steps.

So I commented out all other folders other than dev-clean in create.lua and ran:
$ ~/usr/bin/luajit ~/wav2letter/data/librispeech/create.lua ~/LibriSpeech ~/librispeech-proc

And I got the following error:
| parsing SPEAKERS.TXT for gender...
| analyzing /home/sanchit/LibriSpeech/dev-clean...
| writing /home/sanchit/librispeech-proc/dev-clean...
/home/sanchit/usr/bin/luajit: could not open file (File contains data in an unimplemented format.)
stack traceback:
[C]: at 0x7f70348ab7a0
[C]: in function 'SndFile'
/home/sanchit/wav2letter/data/librispeech/create.lua:16: in function 'copytoflac'
/home/sanchit/wav2letter/data/librispeech/create.lua:95: in function 'createidx'
/home/sanchit/wav2letter/data/librispeech/create.lua:150: in main chunk
[C]: at 0x00404ab0

All of them are .flac files.

PS: As mentioned above, all I want is to transcribe a .flac file of my own.

[saisrinivas047]

Hi @sanchit-beginnings
are you able to transcribe your own .flac files ?
did you also include .trans.txt in dev-clean folder ?

[sancheza603]

Hi @saisrinivas047 , still no clue about the data-processing! Was hoping @VitaliyLi to help out!

[vineelpratap]

Can you try the fixes mentioned in #3

[bolt163]

"
So after looking at some common issues, I realized that I might have missed a couple of data pre-processing steps.

So I commented out all other folders other than dev-clean in create.lua and ran:
$ ~/usr/bin/luajit ~/wav2letter/data/librispeech/create.lua ~/LibriSpeech ~/librispeech-proc

And I got the following error:
| parsing SPEAKERS.TXT for gender...
| analyzing /home/sanchit/LibriSpeech/dev-clean...
| writing /home/sanchit/librispeech-proc/dev-clean...
/home/sanchit/usr/bin/luajit: could not open file (File contains data in an unimplemented format.)
stack traceback:
/home/sanchit/wav2letter/data/librispeech/create.lua:16: in function 'copytoflac'
/home/sanchit/wav2letter/data/librispeech/create.lua:95: in function 'createidx'
/home/sanchit/wav2letter/data/librispeech/create.lua:150: in main chunk
All of them are .flac files.

PS: As mentioned above, all I want is to transcribe a .flac file of my own.

"

——under macosx，i did't meet the problem, maybe because the libsndfile I installed by the way *brew install libsndfile*,
but I also meet the same problem under the linux platform, which was installed from the libsndfile source code, i can't figure out why。。。。is there any pit？？

[mooncaptain]

http://www.mega-nerd.com/libsndfile/#Download
First, download the latest source.
Unpack it somewhere.
Open Terminal and cd to the directory you unpacked.
Configure it: ./configure
Build it: make -j8
Install it (requires administrative rights): sudo make -j8 install

And: U need to check your flac before u do above steps