DeepSpeech fails on CPUs without AVX support #1157

DJ-Hayden · 2018-01-07T23:35:20Z

Hello,

I am running through the standard set-up and was having some trouble on the final step. It appears to be a recurring issue, but the solution I believe is provided hasn't fixed the problem, which is why I'm assuming there is a bug. Very well may be me though so I've tried to describe every step.

I'm using a MacBook Pro 10.11.6 with a 2.4 GHz Intel Core 2 Duo Processor (incredibly old).

Tensorflow has not been installed (instruction don't seem to indicate it as a prerequisite).

Python is 2.7.

I first cloned the repository, but that doesn't seem to be important for what's next. Then I downloaded and unzipped the pre-trained model.

I set up a virtual environment for deepspeech using conda. Then I activated it and ran pip install deepspeech. After installation I ran:

deepspeech location_of_downloaded_pretrained_model.pb location_of_sound.wav

This immediately got the "Illegal instruction: 4" and the code stopped.

After a bit of searching it seems like the problem is AVX2/FMA as documented here (#1057) and here (#1023).

The consensus seemed to be to go to https://tools.taskcluster.net/index/artifacts/project.deepspeech.deepspeech.native_client.master/osx and select deepspeech-0.1.0-cp27-cp27m-macosx_10_10_x86_64.whl (since I'm running python 2.7 with an old Mac).

Then, I uninstalled the current deepspeech with:

pip uninstall deepspeech

and ran:

pip install deepspeech-0.1.0-cp27-cp27m-macosx_10_10_x86_64.whl

while in a folder that contained that file.

Seemed to work, but then I ran:

deepspeech location_of_downloaded_pretrained_model.pb location_of_sound.wav

once more and it still gives me the error "Illegal instruction: 4" and the code stopped.

Have I done something wrong here or is this still a bug?

The text was updated successfully, but these errors were encountered:

reuben · 2018-01-07T23:51:38Z

The steps you followed seem to be correct. Maybe your CPU doesn't support some other instruction set that we compile with. Can you share more information on the CPU model? Specifically the output of this command:

sysctl -n machdep.cpu.brand_string && sysctl -n machdep.cpu.features

DJ-Hayden · 2018-01-07T23:55:51Z

Oh good, I did something right!

I used command:

sysctl -n machdep.cpu.brand_string && sysctl -n machdep.cpu.features

and got

Intel(R) Core(TM)2 Duo CPU     P8600  @ 2.40GHz
FPU VME DE PSE TSC MSR PAE MCE CX8 APIC SEP MTRR PGE MCA CMOV PAT PSE36 CLFSH DS ACPI MMX FXSR SSE SSE2 SS HTT TM PBE SSE3 DTES64 MON DSCPL VMX SMX EST TM2 SSSE3 CX16 TPR PDCM SSE4.1 XSAVE

Is my computer just too bad to even run this? I honestly wouldn't doubt that.

reuben · 2018-01-07T23:58:20Z

I suspect SSE4.2 is the culprit. Let me check.

reuben · 2018-01-08T01:14:34Z

I was wrong, it's AVX, which your CPU does not support. Your best bet is building the bindings from source. The process is documented here: https://github.com/mozilla/DeepSpeech/blob/master/native_client/README.md

DJ-Hayden · 2018-01-08T01:23:59Z

To be clear, I need to follow the "Build Requirements" portion and skip the first sections "Installation, Required Dependencies, Installing the language bindings, Node.JS bindings"?

reuben · 2018-01-08T01:41:39Z

Exactly!

DJ-Hayden · 2018-01-08T04:05:54Z

Okay, a few hours of troubleshooting and I'm stuck again.

First, I followed these instructions: https://github.com/mozilla/DeepSpeech/blob/master/native_client/README.md and https://www.tensorflow.org/install/install_sources

Namely, downloaded the DeepSpeech tensorflow repository, bazel, TensorFlow Python dependencies, etc up to (and including) "Configure the installation" on the https://www.tensorflow.org/install/install_sources website.

Then I configured it by giving it the various paths to python and my site packages (as well as saying no to literally everything it asks about cloud computing, AWS, CUDA, etc).

Then I put both DeepSpeech and tensorflow-master in the same directory. Tensorflow-master is the direct download from the DeepSpeech website, so it contains a folder named tensorflow as well.

Within tensorflow-master, I ran:

ln -s ../DeepSpeech/native_client ./

then ran:

bazel build -c opt --copt=-O3 //tensorflow:libtensorflow_cc.so //tensorflow:libtensorflow_framework.so //native_client:deepspeech //native_client:deepspeech_utils //native_client:libctc_decoder_with_kenlm.so //native_client:generate_trie

Which failed.

I tried it again by cd'ing into the tensorflow folder within the tensorflow-master folder and ran:

bazel build -c opt --copt=-O3 //tensorflow:libtensorflow_cc.so //tensorflow:libtensorflow_framework.so //native_client:deepspeech //native_client:deepspeech_utils //native_client:libctc_decoder_with_kenlm.so //native_client:generate_trie

That failed too. The error message is:

ERROR: (path stuff)/local_config_sycl/sycl/BUILD:4:1: First argument of 'load' must be a label and start with either '//', ':', or '@'. Use --incompatible_load_argument_is_label=false to temporarily disable this check.
ERROR: /(path stuff)/local_config_sycl/sycl/BUILD:6:1: First argument of 'load' must be a label and start with either '//', ':', or '@'. Use --incompatible_load_argument_is_label=false to temporarily disable this check.
ERROR: /(path stuff)/local_config_sycl/sycl/BUILD:30:9: Traceback (most recent call last):
	File "/(path stuff)/local_config_sycl/sycl/BUILD", line 27
		cc_library(name = "syclrt", srcs = [sycl_libr...")], <3 more arguments>)
	File "/(path stuff)/local_config_sycl/sycl/BUILD", line 30, in cc_library
		sycl_library_path
name 'sycl_library_path' is not defined
ERROR: /(path stuff)/local_config_sycl/sycl/BUILD:39:1: Target '@local_config_sycl//sycl:using_sycl' contains an error and its package is in error and referenced by '@local_config_sycl//sycl:sycl'
ERROR: /(other path stuff)/tensorflow-master/third_party/eigen3/BUILD:20:1: Target '@local_config_sycl//sycl:sycl' contains an error and its package is in error and referenced by '//third_party/eigen3:eigen3'
ERROR: Analysis of target '//tensorflow:libtensorflow_cc.so' failed; build aborted: Loading failed
INFO: Elapsed time: 80.050s
FAILED: Build did NOT complete successfully (34 packages loaded)

lissyx · 2018-01-08T10:42:16Z

@DJ-Hayden can you make sure you used our tensorflow: https://github.com/mozilla/tensorflow ? And Bazel 0.5.4 ? Besides, bazel build should be done in the tensorflow directory, as you did :)

cahya-wirawan · 2018-01-08T13:32:03Z

@DJ-Hayden I had the same problem with my old MBP, so I installed the precompiled deepspeech from taskcluster as follow and it works now:
pip install https://index.taskcluster.net/v1/task/project.deepspeech.deepspeech.native_client.master.osx/artifacts/public/deepspeech-0.1.0-cp27-cp27m-macosx_10_10_x86_64.whl
(maybe you need also the option --upgrade if you already installed deepspeech)

DJ-Hayden · 2018-01-08T14:16:34Z

@lissyx Ah shoot, I definitely have the mozilla tensorflow but the latest build of bazel (0.9.0 via homebrew). I'll use 0.5.4 and check it out.

@cahya-wirawan I did that (albeit via the download the whl file to my laptop method) and got Illegal Instruction: 4. I uninstalled deepspeech, then used pip install https://index.taskcluster.net/v1/task/project.deepspeech.deepspeech.native_client.master.osx/artifacts/public/deepspeech-0.1.0-cp27-cp27m-macosx_10_10_x86_64.whl, and after it downloaded it still said "Illegal Instruction: 4" when I tried to run inference on the pre-trained model.

DJ-Hayden · 2018-01-08T16:06:55Z

My new steps are:

Download DeepSpeech and Mozilla's fork of tensorflow. Place DeepSpeech within tensorflow-master folder (not the tensorflow folder within tensorflow-master)
Install mozilla's version of tensowflow by using https://www.tensorflow.org/install/install_sources#PrepareMac
2a) install bazel 0.5.4 using https://docs.bazel.build/versions/master/install-os-x.html (Installing using binary installer)
2b) use export PATH="$PATH:$HOME/bin" to let PATH know where bazel is
2c) pip install six numpy wheel (which just confirms I have these installed)
2d) cd into tensorflow-master folder from mozilla's version of tensorflow to run ./configure; again say no to literally everything; I noticed that this never really asks where bazel is but I think 2b takes care of that

3a) cd into actual tensorflow folder (within the tensorflow-master folder) and check thatcd ../DeepSpeech/native_client ./correctly gets me to the native_client folder within DeepSpeech
3b) cd back into actual tensorflow folder (within the tensorflow-master folder) and runln -s ../DeepSpeech/native_client ./to create an alias which alerts subsequent programs that’s where to find DeepSpeech

run:
bazel build -c opt --copt=-O3 //tensorflow:libtensorflow_cc.so //tensorflow:libtensorflow_framework.so //native_client:deepspeech //native_client:deepspeech_utils //native_client:libctc_decoder_with_kenlm.so //native_client:generate_trie
and get:
ERROR: no such package 'native_client': BUILD file not found on package path.

I'm doing something obviously wrong here but cd ../DeepSpeech/native_client ./ works fine, so I'm not sure why the alias isn't working at step 4). I thought I may need to run the ./configure with DeepSpeech already in the folder with tensorflow-master (instead of after, like I initially did), but doing that returned the same error.

lissyx · 2018-01-08T16:15:00Z

@DJ-Hayden Can you make sure the symlink is good? It's possible GNU / BSD ln -s have different behaviors.

DJ-Hayden · 2018-01-08T16:17:22Z

@lissyx double-clicking on it within finder correctly takes me to the native_folder within DeepSpeech. It does already contain quite a bit of files so I'm not sure if that's the problem or not.

lissyx · 2018-01-08T16:18:24Z

@DJ-Hayden I'd prefer ls -hal and ls -hal native_client/ inside your tensorflow dir

DJ-Hayden · 2018-01-08T16:22:53Z

@lissyx ls -hal lists many things including

lrwxr-xr-x   1 <user> staff    27B Jan  8 10:54 native_client -> ../DeepSpeech/native_client

andls -hal native_client/lists many things, all of which are in the /DeepSpeech/native_client folder

reuben · 2018-01-08T16:28:22Z

I think your folder structure is incorrect. It should be:

DeepSpeech
  - DeepSpeech.py
  - native_client
  - etc.
mozilla-tensorflow
  - tensorflow
  - tools
  - native_client -> ../DeepSpeech/native_client
  - etc.

You should invoke bazel from the root of the mozilla-tensorflow repository, aka cd mozilla-tensorflow, not cd mozilla-tensorflow/tensorflow.

DJ-Hayden · 2018-01-08T18:24:30Z

@reuben I thought that's what I was originally doing, but perhaps I got one of the steps slightly wrong. It is now running (albeit with a large swathe of warnings related to:

unused typedef 'VerifySizesAreEqual' [-Wunused-local-typedef]
  typedef char VerifySizesAreEqual[sizeof(Dest) == sizeof(Source) ? 1 : -1]

and:

in includes attribute of cc_library rule //tensorflow/core:framework_headers_lib: '../../external/nsync/public' resolves to 'external/nsync/public' not below the relative path of its package 'tensorflow/core'. This will be an error in the future. Since this rule was created by the macro 'cc_header_only_library', the error might have been caused by the macro implementation in <path>/tensorflow-master/tensorflow/tensorflow.bzl:1064:30

There's quite a few other warnings (too many to list) but nothing has broken yet. It has been a few hours and it says " [2,405 / 2,410] Still waiting for 2 jobs to complete: " but that number has been increasing as the code runs. Once finished I'll cd into the DeepSpeech/native_client directory and run make deepspeech.

Once that is done, should I do the source activate deepspeech-venv, cd to where deepspeech is in the native_client directory, then run the prebuilt model command? Or does the command "make deepspeech" effectively do the same as pip install deepspeech such that I can just run deepspeech from any folder within the console?

DJ-Hayden · 2018-01-08T18:53:00Z

The bazel build finished.

I then tried to make deepspeech but it didn't recognize the directory initially. I looked at the makefile and it seems it's hardcoded as tensorflow/tensorflow and not tensorflow-master/tensorflow. I changed my tensorflow-master folder to tensorflow and reran make deepspeech in the appropriate place. It then said sox.h wasn't installed, so I didbrew install sox.

After that, it gave the following error:

/bin/sh: pkg-config: command not found
/bin/sh: pkg-config: command not found
client.cc:117:23: warning: 'tmpnam' is deprecated: This function is provided for compatibility reasons only. Due to security concerns inherent
      in the design of tmpnam(3), it is highly recommended that you use mkstemp(3) instead. [-Wdeprecated-declarations]
  char* output_name = tmpnam(NULL);
                      ^
/usr/include/stdio.h:276:7: note: 'tmpnam' has been explicitly marked deprecated here
char    *tmpnam(char *);
         ^
1 warning generated.
ld: warning: directory not found for option '-L/my_path/tensorflow/bazel-bin/tensorflow'
ld: library not found for -ldeepspeech
clang: error: linker command failed with exit code 1 (use -v to see invocation)
make: *** [deepspeech] Error 1

I feel like we're really close, but there's something wrong. Those pathway warning are correct in that they are not found in the directory.

reuben · 2018-01-08T19:07:20Z

Looks like $TFDIR is pointing to the wrong place. Did you set it?

DJ-Hayden · 2018-01-08T19:19:28Z

I didn't set it, but my_path/tensorflow(previously tensorflow-master)/bazel-bin only contains a reference to native_client, not tensorflow, as the error suggests:

directory not found for option '-L/my_path/tensorflow/bazel-bin/tensorflow'

reuben · 2018-01-08T19:25:53Z

Weird… You should have a bazel-bin/tensorflow as a result of building libtensorflow_cc and libtensorflow_framework.

DJ-Hayden · 2018-01-08T19:30:41Z

Ah, that sounds like the topic of a lot of warnings I was getting. The very first thing that came up (as mentioned above) is:

my_path/tensorflow/tensorflow/core/BUILD:1784:1: in includes attribute of cc_library rule //tensorflow/core:framework_headers_lib: '../../external/nsync/public' resolves to 'external/nsync/public' not below the relative path of its package 'tensorflow/core'. This will be an error in the future. Since this rule was created by the macro 'cc_header_only_library', the error might have been caused by the macro implementation in my_path/tensorflow/tensorflow/tensorflow.bzl:1064:30.

Good news is that I just received a new Asus that will be running Ubuntu, so that's one way to solve the problem.

reuben · 2018-01-08T19:31:50Z

Can you try building again? Things should be in cache this time so should not take as long. Hopefully that'll create the binaries normally. I see that warning on my system as well, but it doesn't break the build.

DJ-Hayden · 2018-01-08T19:56:14Z

Doing that now. Still a lot of warnings along the lines of:

/Applications/Xcode.app/Contents/Developer/Toolchains/XcodeDefault.xctoolchain/usr/bin/ranlib: file: bazel-out/darwin_x86_64-opt/bin/external/protobuf_archive/libprotobuf_lite.a(arenastring.o) has no symbols

I'll monitor other warnings and see if any are specifically about libtensorflow_cc

DJ-Hayden · 2018-01-08T21:38:33Z

Finally finished. The last lines are:

INFO: From Linking tensorflow/libtensorflow_cc.so:
clang: warning: argument unused during compilation: '-pthread

There were a number of warnings related to:

INFO: From Linking tensorflow/cc/libcc_ops_internal.lo:
/Applications/Xcode.app/Contents/Developer/Toolchains/XcodeDefault.xctoolchain/usr/bin/ranlib: file: bazel-out/darwin_x86_64-opt/bin/tensorflow/cc/libcc_ops_internal.lo(audio_ops_internal.o) has no symbols

DJ-Hayden · 2018-01-08T23:24:25Z

Oh, and when trying to make deepspeech, I get the following error:

Undefined symbols for architecture x86_64:
  "_sox_add_effect", referenced from:
      _main in client-ebd38b.o
  "_sox_close", referenced from:
      _main in client-ebd38b.o
  "_sox_create_effect", referenced from:
      _main in client-ebd38b.o
  "_sox_create_effects_chain", referenced from:
      _main in client-ebd38b.o
  "_sox_delete_effects_chain", referenced from:
      _main in client-ebd38b.o
  "_sox_effect_options", referenced from:
      _main in client-ebd38b.o
  "_sox_find_effect", referenced from:
      _main in client-ebd38b.o
  "_sox_flow_effects", referenced from:
      _main in client-ebd38b.o
  "_sox_init", referenced from:
      _main in client-ebd38b.o
  "_sox_open_read", referenced from:
      _main in client-ebd38b.o
  "_sox_open_write", referenced from:
      _main in client-ebd38b.o
  "_sox_quit", referenced from:
      _main in client-ebd38b.o
ld: symbol(s) not found for architecture x86_64

cahya-wirawan · 2018-01-09T10:44:21Z

@DJ-Hayden did you try to use python 3.5 and install deepspeech from taskcluster? because this is what I have in my old MBP and it works fine without recompiling anything which is very cumbersome as you see.

reuben · 2018-01-09T11:00:38Z

Can you try doing brew install pkg-config and then retrying make deepspeech?

DJ-Hayden · 2018-01-09T16:11:56Z

Last night I ran:

brew install libtool
brew install wget
brew install automake
brew install autoconf

and this morning I ran:

brew install pkg-config
make deepspeech

It seemingly worked, so I ran (when in the native client folder):

/my_path/DeepSpeech/native_client/deepspeech cd ../models/output_graph.pb cd ../models/my_audio.wav cd ../models/alphabet.txt cd ../models/lm.binary models/trie

It states:

dyld: Library not loaded:  @rpath/libdeepspeech.so
  Referenced from: /my_path/DeepSpeech/native_client/deepspeech
  Reason: image not found
Trace/BPT trap: 5

lissyx · 2018-01-09T16:21:03Z

@DJ-Hayden Can you post you build commands? It seems you have some empty library name. otool -L deepspeech would help also. The deepspeech make target should take care of proper renaming:

DeepSpeech/native_client/Makefile

Lines 23 to 27 in a03604d

    
           ifeq ($(OS),Darwin) 
        
           	install_name_tool -change $$TASKCLUSTER_TASK_DIR/homebrew/opt/sox/lib/libsox.3.dylib @rpath/libsox.3.dylib deepspeech 
        
           	install_name_tool -change bazel-out/local-opt/bin/native_client/libdeepspeech.so @rpath/libdeepspeech.so deepspeech 
        
           	install_name_tool -change bazel-out/local-opt/bin/native_client/libdeepspeech_utils.so @rpath/libdeepspeech_utils.so deepspeech 
        
           endif

DJ-Hayden · 2018-01-09T16:24:13Z

Sure thing!

otool -L deepspeech
deepspeech:
	@rpath/libdeepspeech.so (compatibility version 0.0.0, current version 0.0.0)
	@rpath/libdeepspeech_utils.so (compatibility version 0.0.0, current version 0.0.0)
	@rpath/libtensorflow_cc.so (compatibility version 0.0.0, current version 0.0.0)
	@rpath/libtensorflow_framework.so (compatibility version 0.0.0, current version 0.0.0)
	/usr/local/opt/sox/lib/libsox.3.dylib (compatibility version 4.0.0, current version 4.0.0)
	/usr/lib/libc++.1.dylib (compatibility version 1.0.0, current version 120.1.0)
	/usr/lib/libSystem.B.dylib (compatibility version 1.0.0, current version 1226.10.1)

and build was the standard:

bazel build -c opt --copt=-O3 //tensorflow:libtensorflow_cc.so //tensorflow:libtensorflow_framework.so //native_client:deepspeech //native_client:deepspeech_utils //native_client:libctc_decoder_with_kenlm.so //native_client:generate_trie

(which ran with a plethora of warnings)

lissyx · 2018-01-09T16:25:49Z

@DJ-Hayden Can you use proper formatting? it's really strange that you have only .so, that should not even happen, but I cannot tell if it's not just Github reformatting your input.

Also, we need your make statement.

DJ-Hayden · 2018-01-09T16:30:41Z

@lissyx Wasn't sure how, but now I see it and am editing all of my previous comments.

My make statement was simply

make deepspeech

while in the native_client folder

lissyx · 2018-01-09T16:41:35Z

@DJ-Hayden Ok, that makes more sense now. It's just that you need to help dyld locate your libraries. Please use ARGS="../models/output_graph.pb ../models/vox-machina.wav ../models/alphabet.txt ../models/lm.binary models/trie" make run

DJ-Hayden · 2018-01-09T17:28:46Z

Awesome, thanks so much! I think it's finally working. I ran:

ARGS="../models/output_graph.pb ../models/my_audio.wav ../models/alphabet.txt ../models/lm.binary models/trie" make run

which output:

DYLD_LIBRARY_PATH=/my_path/tensorflow/bazel-bin/tensorflow:/Users/my_path/tensorflow/bazel-bin/native_client: ./deepspeech ../models/output_graph.pb ../models/my_audio.wav ../models/alphabet.txt ../models/lm.binary models/trie
2018-01-09 11:44:41.342291: I tensorflow/core/platform/cpu_feature_guard.cc:137] Your CPU supports instructions that this TensorFlow binary was not compiled to use: SSE4.1

Error: Can't parse trie file, invalid header. Try updating your trie file.
input: multi-channel effect drained asymmetrically!

a obbushdtatsunshsenhamsohimestofarerermsholtfromtolthebdromsouwolddre

I presume "a obbushdtatsunshsenhamsohimestofarerermsholtfromtolthebdromsouwolddre" is the output, which I believe was due to the minute-long file I used and seems to be a known issue with longer files. I'll try a shorter five second version.

In the future, will I need to do:

ARGS="../models/output_graph.pb ../models/my_audio.wav ../models/alphabet.txt ../models/lm.binary models/trie" make run

For every piece of audio I'd like to run? Or now that I've "made" it can I go back to the documented way of:

/my_path/DeepSpeech/native_client/deepspeech cd ../models/output_graph.pb cd ../models/my_audio.wav cd ../models/alphabet.txt cd ../models/lm.binary models/trie

lissyx · 2018-01-09T17:44:41Z

@DJ-Hayden Either you continue to make run, or you copy the libraries altogether with the deepspeech binary :). Everything looks like so far it's working as expected. You might be hitting issue #1156 as a side effect related to your parsing error about the trie file. At the same time, it seems you should rebuild (sorry, but the long bazel build) with -march=native added, since you have the warning about SSE4.1.

I guess you can close the issue now :)

DJ-Hayden · 2018-01-09T17:50:22Z

@lissyx @reuben Thanks for all the help! When configuring I did use -march=native, but maybe that got lost in translation somewhere. I can rebuild that later.

For now though, before I close the issue, how would I go about copying the libraries with the deepspeech binary? I'm at a bit of a loss as to why, in the native folder:

ARGS="../models/output_graph.pb ../models/my_audio.wav ../models/alphabet.txt ../models/lm.binary models/trie" make run

works but:

/my_path/DeepSpeech/native_client/deepspeech cd ../models/output_graph.pb cd ../models/my_audio.wav cd ../models/alphabet.txt cd ../models/lm.binary models/trie

(again, in the native_client folder) doesn't. I used the full path name to deepspeech because deepspeech alone says command not found.

lissyx · 2018-01-09T17:53:03Z

Just pick (cp):

	@rpath/libdeepspeech.so (compatibility version 0.0.0, current version 0.0.0)
	@rpath/libdeepspeech_utils.so (compatibility version 0.0.0, current version 0.0.0)
	@rpath/libtensorflow_cc.so (compatibility version 0.0.0, current version 0.0.0)
	@rpath/libtensorflow_framework.so (compatibility version 0.0.0, current version 0.0.0)

From $TFDIR/bazel-bin/native_client and $TFDIR/bazel-bin/tensorflow

Regarding -march, it seems you have to add it manually again when you issue bazel build:

$ bazel build -c opt --copt=-O3 --copt=-march=native [...]

DJ-Hayden · 2018-01-09T18:16:23Z

Sweet, thanks again for all the help!

lock · 2019-01-03T04:52:48Z

This thread has been automatically locked since there has not been any recent activity after it was closed. Please open a new issue for related bugs.

reuben changed the title ~~Using Non-AVX2/FMA Builds~~ DeepSpeech fails on CPUs without AVX support Jan 8, 2018

DJ-Hayden closed this as completed Jan 8, 2018

DJ-Hayden reopened this Jan 8, 2018

DJ-Hayden closed this as completed Jan 9, 2018

mozilla deleted a comment from jehoshua7 Jan 20, 2018

lock bot locked and limited conversation to collaborators Jan 3, 2019

DeepSpeech fails on CPUs without AVX support #1157

DeepSpeech fails on CPUs without AVX support #1157

Comments

DJ-Hayden commented Jan 7, 2018 • edited Loading

reuben commented Jan 7, 2018

DJ-Hayden commented Jan 7, 2018 • edited Loading

reuben commented Jan 7, 2018

reuben commented Jan 8, 2018

DJ-Hayden commented Jan 8, 2018

reuben commented Jan 8, 2018

DJ-Hayden commented Jan 8, 2018 • edited Loading

lissyx commented Jan 8, 2018

cahya-wirawan commented Jan 8, 2018 • edited Loading

DJ-Hayden commented Jan 8, 2018 • edited Loading

DJ-Hayden commented Jan 8, 2018 • edited Loading

lissyx commented Jan 8, 2018

DJ-Hayden commented Jan 8, 2018

lissyx commented Jan 8, 2018

DJ-Hayden commented Jan 8, 2018 • edited Loading

reuben commented Jan 8, 2018

DJ-Hayden commented Jan 8, 2018 • edited Loading

DJ-Hayden commented Jan 8, 2018 • edited Loading

reuben commented Jan 8, 2018

DJ-Hayden commented Jan 8, 2018 • edited Loading

reuben commented Jan 8, 2018

DJ-Hayden commented Jan 8, 2018 • edited Loading

reuben commented Jan 8, 2018

DJ-Hayden commented Jan 8, 2018 • edited Loading

DJ-Hayden commented Jan 8, 2018 • edited Loading

DJ-Hayden commented Jan 8, 2018 • edited Loading

cahya-wirawan commented Jan 9, 2018

reuben commented Jan 9, 2018

DJ-Hayden commented Jan 9, 2018 • edited Loading

lissyx commented Jan 9, 2018

DJ-Hayden commented Jan 9, 2018 • edited by lissyx Loading

lissyx commented Jan 9, 2018

DJ-Hayden commented Jan 9, 2018

lissyx commented Jan 9, 2018

DJ-Hayden commented Jan 9, 2018 • edited Loading

lissyx commented Jan 9, 2018

DJ-Hayden commented Jan 9, 2018 • edited Loading

lissyx commented Jan 9, 2018

DJ-Hayden commented Jan 9, 2018

lock bot commented Jan 3, 2019

DJ-Hayden commented Jan 7, 2018 •

edited

Loading

DJ-Hayden commented Jan 7, 2018 •

edited

Loading

DJ-Hayden commented Jan 8, 2018 •

edited

Loading

cahya-wirawan commented Jan 8, 2018 •

edited

Loading

DJ-Hayden commented Jan 8, 2018 •

edited

Loading

DJ-Hayden commented Jan 8, 2018 •

edited

Loading

DJ-Hayden commented Jan 8, 2018 •

edited

Loading

DJ-Hayden commented Jan 8, 2018 •

edited

Loading

DJ-Hayden commented Jan 8, 2018 •

edited

Loading

DJ-Hayden commented Jan 8, 2018 •

edited

Loading

DJ-Hayden commented Jan 8, 2018 •

edited

Loading

DJ-Hayden commented Jan 8, 2018 •

edited

Loading

DJ-Hayden commented Jan 8, 2018 •

edited

Loading

DJ-Hayden commented Jan 8, 2018 •

edited

Loading

DJ-Hayden commented Jan 9, 2018 •

edited

Loading

DJ-Hayden commented Jan 9, 2018 •

edited by lissyx

Loading

DJ-Hayden commented Jan 9, 2018 •

edited

Loading

DJ-Hayden commented Jan 9, 2018 •

edited

Loading