Merge pull request #734 from alibaba-damo-academy/dev_lhn
fix dataset text filter bug
| | |
|---|
| | | mode=mode, |
|---|
| | | ) |
|---|
| | | |
|---|
| | | filter_conf = conf.get('filter_conf', {}) |
|---|
| | | filter_fn = partial(filter, **filter_conf) |
|---|
| | | dataset = FilterIterDataPipe(dataset, fn=filter_fn) |
|---|
| | | |
|---|
| | | if "text" in data_names: |
|---|
| | | vocab = {'vocab': dict, 'seg_dict': seg_dict, 'punc_dict': punc_dict, 'bpe_tokenizer': bpe_tokenizer, 'hw_config': hw_config} |
|---|
| | | tokenize_fn = partial(tokenize, **vocab) |
|---|
| | | dataset = MapperIterDataPipe(dataset, fn=tokenize_fn) |
|---|
| | | |
|---|
| | | filter_conf = conf.get('filter_conf', {}) |
|---|
| | | filter_fn = partial(filter, **filter_conf) |
|---|
| | | dataset = FilterIterDataPipe(dataset, fn=filter_fn) |
|---|
| | | |
|---|
| | | if shuffle: |
|---|
| | | buffer_conf = conf.get('shuffle_conf', {}) |
|---|
| | | buffer_size = buffer_conf['shuffle_size'] |
|---|
