Comments (10)
looks like you are somehow mixing master branch with 0.9 version.
The latest stable version is 0.9 (and docs and pip correspond to it). Could you please switch to 0.9 version and try again. If you want it from GitHub, the release is here: https://github.com/NVIDIA/NeMo/tree/v0.9.0
from nemo.
looks like you are somehow mixing master branch with 0.9 version.
The latest stable version is 0.9 (and docs and pip correspond to it). Could you please switch to 0.9 version and try again. If you want it from GitHub, the release is here: https://github.com/NVIDIA/NeMo/tree/v0.9.0
Thanks that fixed it :D
Now I'm trying to install the language model but I keep getting segmentation fault at random times when setup.sh is run the decoders. Sorry for reporting it on the same issue but do you have any idea what it might be wrong or how to clean install it again? (I tried deleting the decoders generated file and rerun install_decoders.sh but that didnt work)
from nemo.
can you give details about your system (e.g. OS type and RAM size).
"install_decoders.sh" should work on Ubuntu 18.04
from nemo.
can you give details about your system (e.g. OS type and RAM size).
"install_decoders.sh" should work on Ubuntu 18.04
It's Ubuntu 16.04.6 LTS with 32GB ram
Is it supposed to work only on Ubuntu 18.04?
from nemo.
can you please paste the full error?
from nemo.
can you please paste the full error?
I'm gonna paste the last portion before and after the error (the log is a little bigger)
gcc -pthread -B /home/checker/anaconda3/compiler_compat -Wl,--sysroot=/ -Wsign-compare -DNDEBUG -g -fwrapv -O3 -Wall -Wstrict-prototypes -fPIC -I. -Ikenlm -Iopenfst-1.6.3/src/include -IThreadPool -I/home/checker/anaconda3/include/python3.7m -c kenlm/lm/search_trie.cc -o build/temp.linux-x86_64-3.7/kenlm/lm/search_trie.o -O3 -DNDEBUG -DKENLM_MAX_ORDER=6 -std=c++11 -DHAVE_ZLIB
c1plus: warning: command line option ‘-Wstrict-prototypes’ is valid for C/ObjC but not for C++
cc1plus: warning: command line option ‘-Wstrict-prototypes’ is valid for C/ObjC but not for C++
In file included from /usr/include/c++/5/string:52:0,
from /usr/include/c++/5/bits/locale_classes.h:40,
from /usr/include/c++/5/bits/ios_base.h:41,
from /usr/include/c++/5/ios:42,
from /usr/include/c++/5/ostream:38,
from kenlm/util/string_piece.hh:55,
from kenlm/util/fake_ostream.hh:6,
from kenlm/util/string_stream.hh:4,
from kenlm/util/exception.hh:4,
from kenlm/lm/lm_exception.hh:6,
from kenlm/lm/config.hh:4,
from kenlm/lm/quantize.hh:5,
from kenlm/lm/quantize.cc:9:
/usr/include/c++/5/bits/basic_string.h: In substitution of ‘template<class _CharT, class _Traits, class _Alloc> std::__cxx11::basic_string<_CharT, _Traits, _Alloc> std::operator+(const std::__cxx11::basic_string<_CharT, _Traits, _Alloc>&, const std::__cxx11::basic_string<_CharT, _Traits, _Alloc>&) [with _CharT = char; _Traits = std::char_traits; _Alloc = std::allocator]’:
/usr/include/c++/5/system_error:343:50: required from here
/usr/include/c++/5/bits/basic_string.h:4783:5: internal compiler error: Segmentation fault
operator+(const basic_string<_CharT, _Traits, _Alloc>& __lhs,
^
Please submit a full bug report,
with preprocessed source if appropriate.
See file:///usr/share/doc/gcc-5/README.Bugs for instructions.
gcc -pthread -B /home/checker/anaconda3/compiler_compat -Wl,--sysroot=/ -Wsign-compare -DNDEBUG -g -fwrapv -O3 -Wall -Wstrict-prototypes -fPIC -I. -Ikenlm -Iopenfst-1.6.3/src/include -IThreadPool -I/home/checker/anaconda3/include/python3.7m -c kenlm/lm/value_build.cc -o build/temp.linux-x86_64-3.7/kenlm/lm/value_build.o -O3 -DNDEBUG -DKENLM_MAX_ORDER=6 -std=c++11 -DHAVE_ZLIB
cc1plus: warning: command line option ‘-Wstrict-prototypes’ is valid for C/ObjC but not for C++
gcc -pthread -B /home/checker/anaconda3/compiler_compat -Wl,--sysroot=/ -Wsign-compare -DNDEBUG -g -fwrapv -O3 -Wall -Wstrict-prototypes -fPIC -I. -Ikenlm -Iopenfst-1.6.3/src/include -IThreadPool -I/home/checker/anaconda3/include/python3.7m -c kenlm/lm/read_arpa.cc -o build/temp.linux-x86_64-3.7/kenlm/lm/read_arpa.o -O3 -DNDEBUG -DKENLM_MAX_ORDER=6 -std=c++11 -DHAVE_ZLIB
error: command 'gcc' failed with exit status 1
cc1plus: warning: command line option ‘-Wstrict-prototypes’ is valid for C/ObjC but not for C++
Running it again the script compiles for longer but get this segmentation fault at some point and the stops after a while
openfst-1.6.3/src/include/fst/vector-fst.h: In member function ‘void fst::VectorState<Arc, Allocator>::SetArc(const Arc&, size_t) [with A = fst::ArcTpl<fst::LogWeightTpl >; M = std::allocator<fst::ArcTpl<fst::LogWeightTpl > >; fst::VectorState<Arc, Allocator>::Arc = fst::ArcTpl<fst::LogWeightTpl >; size_t = long unsigned int]’:
openfst-1.6.3/src/include/fst/vector-fst.h:86:14: internal compiler error: Segmentation fault
if (arcs_[n].ilabel == 0) --niepsilons_;
^
Please submit a full bug report,
with preprocessed source if appropriate.
See file:///usr/share/doc/gcc-5/README.Bugs for instructions.
gcc -pthread -B /home/checker/anaconda3/compiler_compat -Wl,--sysroot=/ -Wsign-compare -DNDEBUG -g -fwrapv -O3 -Wall -Wstrict-prototypes -fPIC -I. -Ikenlm -Iopenfst-1.6.3/src/include -IThreadPool -I/home/checker/anaconda3/include/python3.7m -c ctc_greedy_decoder.cpp -o build/temp.linux-x86_64-3.7/ctc_greedy_decoder.o -O3 -DNDEBUG -DKENLM_MAX_ORDER=6 -std=c++11 -DHAVE_ZLIB
error: command 'gcc' failed with exit status 1
from nemo.
Looks like gcc 5.4.0 bug
from nemo.
I think @borisgin is right, 16.04 comes with an older gcc. Could you please update your gcc and try again?
from nemo.
Looks like gcc 5.4.0 bug
I think @borisgin is right, 16.04 comes with an older gcc. Could you please update your gcc and try again?
Yes the first answer in the link fixed it thank you! Now the final (hopefully) problem I'm facing is building the 6-gram LM for librispeech by calling ./build_6-gram_OpenSLR_lm.sh
After that the lm model starts to build but after some minutes I get this error
/build_6-gram_OpenSLR_lm.sh
Requirement already satisfied: pandas in /home/checker/anaconda3/lib/python3.7/site-packages (0.25.3)
Requirement already satisfied: python-dateutil>=2.6.1 in /home/checker/anaconda3/lib/python3.7/site-packages (from pandas) (2.8.0)
Requirement already satisfied: numpy>=1.13.3 in /home/checker/anaconda3/lib/python3.7/site-packages (from pandas) (1.18.0)
Requirement already satisfied: pytz>=2017.2 in /home/checker/anaconda3/lib/python3.7/site-packages (from pandas) (2019.3)
Requirement already satisfied: six>=1.5 in /home/checker/anaconda3/lib/python3.7/site-packages (from python-dateutil>=2.6.1->pandas) (1.13.0)
decoders/kenlm/build/bin/lmplz --text language_model/6-gram.txt --arpa language_model/6-gram.arpa --o 6
=== 1/5 Counting and sorting n-grams ===
Reading /home/checker/speech_proj/NeMo/scripts/language_model/6-gram.txt
----5---10---15---20---25---30---35---40---45---50---55---60---65---70---75---80---85---90---95--100
******************************************************************************************Last input should have been poison.
Aborted (core dumped)
decoders/kenlm/build/bin/build_binary trie -q 8 -b 7 -a 256 language_model/6-gram.arpa language_model/6-gram-lm.binary
Reading language_model/6-gram.arpa
----5---10---15---20---25---30---35---40---45---50---55---60---65---70---75---80---85---90---95--100
End of file Byte: 0
ERROR
from nemo.
closing as this is related to the old version
from nemo.
Related Issues (20)
- TypeError: EncDecCTCModel.transcribe() got an unexpected keyword argument 'logprobs' HOT 2
- NeMo License Discrepancy?
- Consider refactoring CTC greedy decoding HOT 5
- Memory is fully eaten and training quit with errors for 40k hours ASR training HOT 5
- The error in loading Llama pretrain checkpoint for NeVa(LLAVA) HOT 3
- File not found in the github repo
- Streaming example provided for Hinglish doesnt work. HOT 5
- eval_trainer.predict() gives AttributeError: 'PipelineMixedPrecisionPlugin' object has no attribute '_desired_input_dtype' error HOT 1
- Error when using packed sequence and gradient checkpoint: save_for_backward can only save variables, but argument 5 is of type PakedSeqParams HOT 1
- Warning: nvfuser is no longer supported in torch script HOT 1
- How to specify the rank,delta and dropout values while LORA finetuning
- Inference discrepancies after merging weights into a LoRA model HOT 1
- Feature Normalization in the ASR preprocessor is too slow. HOT 3
- Conflict between precision and plugins arguments in Trainer HOT 1
- Error response from daemon: unauhorized: authentication required HOT 1
- Why use two types of names? spe refers to the Google sentencepiece library tokenizer. bpe for SentencePiece tokenizer HOT 2
- Error in coverting Mixtral-7B hf checkpoint to Nemo HOT 1
- Error while exporting to TensorRTLLM format - AttributeError: 'NoneType' object has no attribute 'get' HOT 3
- `megatron_gpt_finetuning.py` does not work `max_epochs` HOT 1
- Is frame marblenet VAD still supported?
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from nemo.