menghuanlater / Tianchi2020ChineseMedicineQuestionGeneration

2020 阿里云天池大数据竞赛-中医药文献问题生成挑战赛
26 stars 9 forks source link

WARNING:root:NaN or Inf found in input tensor. #1

Closed zzuyu closed 3 years ago

zzuyu commented 3 years ago

能分享下环境吗? 训练过程中loss消失了

menghuanlater commented 3 years ago

我当时用的conda环境有部分包升级过了,当时用的应该是3.2版本的transformers,1.6版本的pytorch 下面的是我现在的conda环境,你可以把那两个包版本降下去试试

packages in environment at /home/ldmc/anaconda3/envs/anlin_env:

#

Name Version Build Channel

_libgcc_mutex 0.1 main defaults _tflow_select 2.3.0 mkl defaults absl-py 0.9.0 py38_0 defaults antlr4-python3-runtime 4.8 pypi_0 pypi astunparse 1.6.3 py_0 defaults beautifulsoup4 4.9.1 pypi_0 pypi blas 1.0 mkl defaults blinker 1.4 py38_0 defaults brotlipy 0.7.0 py38h7b6447c_1000 defaults c-ares 1.15.0 h7b6447c_1001 defaults ca-certificates 2020.7.22 0 defaults cachetools 4.1.1 py_0 defaults certifi 2020.6.20 py38_0 defaults cffi 1.14.2 py38he30daa8_0 defaults chardet 3.0.4 py38_1003 defaults click 7.1.2 py_0 defaults cryptography 3.1 py38h1ba5d50_0 defaults cudatoolkit 10.1.243 h6bb024c_0 defaults cython 0.29.21 pypi_0 pypi dataclasses 0.6 pypi_0 pypi datasets 1.6.2 pypi_0 pypi dill 0.3.3 pypi_0 pypi docopt 0.6.2 pypi_0 pypi editdistance 0.5.3 pypi_0 pypi faiss-cpu 1.6.3 pypi_0 pypi filelock 3.0.12 pypi_0 pypi flask 1.1.2 pypi_0 pypi fsspec 2021.4.0 pypi_0 pypi future 0.18.2 pypi_0 pypi gast 0.3.3 py_0 defaults google-auth 1.21.0 py_0 defaults google-auth-oauthlib 0.4.1 py_2 defaults google-pasta 0.2.0 py_0 defaults grpcio 1.31.0 py38hf8bcb03_0 defaults h5py 2.10.0 py38hd6299e0_1 defaults hdf5 1.10.6 hb1b8bf9_0 defaults huggingface-hub 0.0.8 pypi_0 pypi hydra-core 1.0.3 pypi_0 pypi idna 2.10 py_0 defaults importlib-metadata 1.7.0 py38_0 defaults importlib-resources 3.3.0 pypi_0 pypi intel-openmp 2020.1 217 defaults itsdangerous 1.1.0 pypi_0 pypi jieba 0.42.1 pypi_0 pypi jinja2 2.11.2 pypi_0 pypi joblib 0.16.0 py_0 defaults keras-preprocessing 1.1.0 py_1 defaults ld_impl_linux-64 2.33.1 h53a641e_7 defaults libedit 3.1.20191231 h14c3975_1 defaults libffi 3.3 he6710b0_2 defaults libgcc-ng 9.1.0 hdf63c60_0 defaults libgfortran-ng 7.3.0 hdf63c60_0 defaults libprotobuf 3.12.4 hd408876_0 defaults libstdcxx-ng 9.1.0 hdf63c60_0 defaults markdown 3.2.2 py38_0 defaults markupsafe 1.1.1 pypi_0 pypi mkl 2020.1 217 defaults mkl-service 2.3.0 py38he904b0f_0 defaults mkl_fft 1.1.0 py38h23d657b_0 defaults mkl_random 1.1.1 py38h0573a6f_0 defaults multiprocess 0.70.11.1 pypi_0 pypi ncurses 6.2 he6710b0_1 defaults ninja 1.9.0 py38hfd86e86_0 defaults nltk 3.5 pypi_0 pypi numpy 1.18.5 py38ha1c710e_0 defaults numpy-base 1.18.5 py38hde5b4d6_0 defaults oauthlib 3.1.0 py_0 defaults omegaconf 2.0.3 pypi_0 pypi openssl 1.1.1g h7b6447c_0 defaults opt_einsum 3.1.0 py_0 defaults packaging 20.4 pypi_0 pypi pandas 1.2.4 pypi_0 pypi pip 20.1.1 py38_1 defaults portalocker 2.0.0 pypi_0 pypi protobuf 3.13.0 pypi_0 pypi psutil 5.7.3 pypi_0 pypi pyarrow 2.0.0 pypi_0 pypi pyasn1 0.4.8 py_0 defaults pyasn1-modules 0.2.7 py_0 defaults pycparser 2.20 py_2 defaults pyjwt 1.7.1 py38_0 defaults pymysql 0.10.0 pypi_0 pypi pyopenssl 19.1.0 py_1 defaults pyparsing 2.4.7 pypi_0 pypi pysocks 1.7.1 py38_0 defaults python 3.8.3 hcff3b4d_2 https://repo.anaconda.com/pkgs/main python-dateutil 2.8.1 pypi_0 pypi python-levenshtein 0.12.0 pypi_0 pypi pytz 2020.1 pypi_0 pypi pyyaml 5.3.1 pypi_0 pypi readline 8.0 h7b6447c_0 defaults redis 3.5.3 pypi_0 pypi regex 2020.7.14 pypi_0 pypi requests 2.24.0 py_0 defaults requests-oauthlib 1.3.0 py_0 defaults rsa 4.6 py_0 defaults sacrebleu 1.4.14 pypi_0 pypi sacremoses 0.0.43 pypi_0 pypi scikit-learn 0.23.1 py38h423224d_0 defaults scipy 1.5.0 py38h0b6359f_0 defaults selenium 3.141.0 pypi_0 pypi sentencepiece 0.1.91 pypi_0 pypi setuptools 49.2.0 py38_0 defaults six 1.15.0 py_0 defaults soupsieve 2.0.1 pypi_0 pypi sqlite 3.32.3 h62c20be_0 defaults stanfordcorenlp 3.9.1.1 pypi_0 pypi tensorboard 2.2.1 pyh532a8cf_0 defaults tensorboard-plugin-wit 1.6.0 py_0 defaults tensorboardx 2.1 pypi_0 pypi tensorflow 2.2.0 mkl_py38h6d3daf0_0 defaults tensorflow-base 2.2.0 mkl_py38h5059a2d_0 defaults tensorflow-estimator 2.2.0 pyh208ff02_0 defaults termcolor 1.1.0 py38_1 defaults threadpoolctl 2.1.0 pyh5ca1d4c_0 defaults tk 8.6.10 hbc83047_0 defaults tokenizers 0.10.2 pypi_0 pypi torch 1.6.0+cu101 pypi_0 pypi tqdm 4.29.1 pypi_0 pypi transformers 4.5.0 pypi_0 pypi twilio 6.45.0 pypi_0 pypi typing-extensions 3.7.4.3 pypi_0 pypi urllib3 1.25.10 py_0 defaults werkzeug 1.0.1 py_0 defaults wget 3.2 pypi_0 pypi wheel 0.34.2 py38_0 defaults wrapt 1.12.1 py38h7b6447c_1 defaults xxhash 2.0.2 pypi_0 pypi xz 5.2.5 h7b6447c_0 defaults zipp 3.1.0 py_0 defaults zlib 1.2.11 h7b6447c_3 defaults

zzuyu commented 3 years ago

好的,感谢,模型已经跑通了,想用做baseline来做一些其他短句生成的任务,为什么输入的句子长度比较短的时候,会出现WARNING:root:NaN or Inf found in input tensor.这个问题呢?把句子重复几次扩长在输入 这个问题就消失?