tensorflow_fasttext
tensorflow_fasttext copied to clipboard
apcode
Simple embedding based text classifier inspired by fastText, implemented in tensorflow
Hi, I'm trying to train text classifier using `train_langdetect.sh`without horovod installed. And I'm getting this error: ```Traceback (most recent call last): File "/Users/kodlan/tensorflow_venv/lib/python3.7/site-packages/tensorflow_core/python/framework/ops.py", line 1607, in _create_c_op c_op = c_api.TF_FinishOperation(op_desc)...
Hello, during training the following error occurs: tensorflow.python.framework.errors_impl.DataLossError: corrupted record at 0 [[{{node read_batch_features/read/ReaderReadUpToV2}} = ReaderReadUpToV2[_device="/job:localhost/replica:0/task:0/device:CPU:0"](read_batch_features/read/TFRecordReaderV2, read_batch_features/file_name_queue, read_batch_features/read/ReaderReadUpToV2/num_records)]] My environment is: Python 3.6, tensorflow 1.12.0, macOS 10.13 Here is detailed...
``` def GenerateNgrams(words, ngrams): nglist = [] for ng in ngrams: for word in words: nglist.extend([word[n:n+ng] for n in range(len(word)-ng+1)]) return ngli ``` maybe it should like following ``` def...
Should probably get each process to save checkpoint to its own directory. E.g. model_dir/`hvd.local_rank()`/ I think multiple processes are in a race to write checkpoints to the same file.