This module contains the bits required to use the fastai DataBlock API and/or mid-level data processing pipelines to organize your data for translation tasks
 
What we're running with at the time this documentation was generated:
torch: 1.9.0+cu102
fastai: 2.5.2
transformers: 4.10.0
ds = load_dataset('wmt16', 'de-en', split='train[:1%]')
Reusing dataset wmt16 (/home/wgilliam/.cache/huggingface/datasets/wmt16/de-en/1.0.0/0d9fb3e814712c785176ad8cdb9f465fbe6479000ee6546725db30ad8a8b5f8a)

Translation tokenization, batch transform, and DataBlock methods

Translation tasks attempt to convert text in one language into another

path = Path('./')
wmt_df = pd.DataFrame(ds['translation'], columns=['de', 'en']); len(wmt_df)
45489
wmt_df.head(2)
de en
0 Wiederaufnahme der Sitzungsperiode Resumption of the session
1 Ich erkläre die am Freitag, dem 17. Dezember unterbrochene Sitzungsperiode des Europäischen Parlaments für wiederaufgenommen, wünsche Ihnen nochmals alles Gute zum Jahreswechsel und hoffe, daß Sie schöne Ferien hatten. I declare resumed the session of the European Parliament adjourned on Friday 17 December 1999, and I would like once again to wish you a happy new year in the hope that you enjoyed a pleasant festive period.
pretrained_model_name = "facebook/bart-large-cnn"
model_cls = AutoModelForSeq2SeqLM

hf_arch, hf_config, hf_tokenizer, hf_model = BLURR.get_hf_objects(pretrained_model_name, model_cls=model_cls)
hf_arch, type(hf_tokenizer), type(hf_config), type(hf_model)
('bart',
 transformers.models.bart.tokenization_bart_fast.BartTokenizerFast,
 transformers.models.bart.configuration_bart.BartConfig,
 transformers.models.bart.modeling_bart.BartForConditionalGeneration)
blocks = (HF_Seq2SeqBlock(hf_arch, hf_config, hf_tokenizer, hf_model), noop)
dblock = DataBlock(blocks=blocks, get_x=ColReader('de'), get_y=ColReader('en'), splitter=RandomSplitter())

Two lines! Notice we pass in noop for our targets (e.g. our summaries) because the batch transform will take care of both out inputs and targets.

 
dls = dblock.dataloaders(wmt_df, bs=4)
b = dls.one_batch()
len(b), b[0]['input_ids'].shape, b[1].shape
(2, torch.Size([4, 483]), torch.Size([4, 86]))
dls.show_batch(dataloaders=dls, max_n=2, input_trunc_at=250, target_trunc_at=250)
text target
0 <s> Was nun die Ergebnisse der Verhandlungen über die Anwendung der Artikel 3, 4, 5, 6 und 12 des Interimsabkommens bezüglich Warenhandel, öffentlicher Aufträge, Wettbewerb, Konsultationsmechanismen bei Fragen des geistigen Eigentums und Beilegung vo Although for certain sectors there may be flaws - I am thinking specifically of the textiles sector, where the rules of origin issue causes great concern - the effects will be beneficial for both the European Union and Mexico. For the European Union
1 <s> Somit wirkt sich die wissensbasierte Wirtschaft, zu der wir übergehen, umgekehrt bereits auf unseren Denkansatz bezüglich der europäischen Institutionen aus. Denn wie wir bereits wissen, bedeutet diese neue Methode der Koordinierung zwangsläufig, The knowledge economy which we are moving into is thus already having a retroactive effect on our concept of the European institutions, for this new method of coordination must mean, as we already know, that top-down authoritarian integration and un

Tests

The purpose of the following tests is to ensure as much as possible, that the core DataBlock code above works for the pretrained translation models below. These tests are excluded from the CI workflow because of how long they would take to run and the amount of data that would be required to download.

Note: Feel free to modify the code below to test whatever pretrained translation models you are working with ... and if any of your pretrained summarization models fail, please submit a github issue (or a PR if you'd like to fix it yourself)

[ model_type for model_type in BLURR.get_models(task='ConditionalGeneration') 
 if (not model_type.startswith('TF')) ]
['BartForConditionalGeneration',
 'BigBirdPegasusForConditionalGeneration',
 'BlenderbotForConditionalGeneration',
 'BlenderbotSmallForConditionalGeneration',
 'FSMTForConditionalGeneration',
 'LEDForConditionalGeneration',
 'M2M100ForConditionalGeneration',
 'MBartForConditionalGeneration',
 'MT5ForConditionalGeneration',
 'PegasusForConditionalGeneration',
 'ProphetNetForConditionalGeneration',
 'Speech2TextForConditionalGeneration',
 'T5ForConditionalGeneration',
 'XLMProphetNetForConditionalGeneration']
pretrained_model_names = [
    'facebook/bart-base',
    'facebook/wmt19-de-en',                      # FSMT
    'Helsinki-NLP/opus-mt-de-en',                # MarianMT
    'sshleifer/tiny-mbart',
    'google/mt5-small',
    't5-small'
]
path = Path('./')
wmt_df = pd.DataFrame(ds['translation'], columns=['de', 'en'])
#hide_output
model_cls = AutoModelForSeq2SeqLM
bsz = 2
seq_sz = 128
trg_seq_sz = 128

test_results = []
for model_name in pretrained_model_names:
    error=None
    
    print(f'=== {model_name} ===\n')
    
    hf_tok_kwargs = {}
    if (model_name == 'sshleifer/tiny-mbart'):
        hf_tok_kwargs['src_lang'], hf_tok_kwargs['tgt_lang'] = "de_DE", "en_XX"
            
    hf_arch, hf_config, hf_tokenizer, hf_model = BLURR.get_hf_objects(model_name, 
                                                                      model_cls=model_cls, 
                                                                      tokenizer_kwargs=hf_tok_kwargs)
    
    print(f'architecture:\t{hf_arch}\ntokenizer:\t{type(hf_tokenizer).__name__}\n')
    
    # not all architectures include a native pad_token (e.g., gpt2, ctrl, etc...), so we add one here
    if (hf_tokenizer.pad_token is None): 
        hf_tokenizer.add_special_tokens({'pad_token': '<pad>'})  
        hf_config.pad_token_id = hf_tokenizer.get_vocab()['<pad>']
        hf_model.resize_token_embeddings(len(hf_tokenizer)) 
    
    before_batch_tfm = HF_Seq2SeqBeforeBatchTransform(hf_arch, hf_config, hf_tokenizer, hf_model,
                                                      padding='max_length', 
                                                      max_length=seq_sz,
                                                      max_target_length=trg_seq_sz)
    
    def add_t5_prefix(inp): return f'translate German to English: {inp}' if (hf_arch == 't5') else inp
    
    blocks = (HF_Seq2SeqBlock(before_batch_tfm=before_batch_tfm), noop)
    dblock = DataBlock(blocks=blocks, 
                   get_x=Pipeline([ColReader('de'), add_t5_prefix]), 
                   get_y=ColReader('en'), 
                   splitter=RandomSplitter())

    dls = dblock.dataloaders(wmt_df, bs=bsz) 
    b = dls.one_batch()
    
    try:
        print('*** TESTING DataLoaders ***\n')
        test_eq(len(b), 2)
        test_eq(len(b[0]['input_ids']), bsz)
        test_eq(b[0]['input_ids'].shape, torch.Size([bsz, seq_sz]))
        test_eq(len(b[1]), bsz)
        test_eq(b[1].shape, torch.Size([bsz, trg_seq_sz]))

        if (hasattr(hf_tokenizer, 'add_prefix_space')):
             test_eq(hf_tokenizer.add_prefix_space, True)
            
        test_results.append((hf_arch, type(hf_tokenizer).__name__, model_name, 'PASSED', ''))
        dls.show_batch(dataloaders=dls, max_n=2, input_trunc_at=1000)
        
    except Exception as err:
        test_results.append((hf_arch, type(hf_tokenizer).__name__, model_name, 'FAILED', err))
arch tokenizer model_name result error
0 bart BartTokenizerFast facebook/bart-base PASSED
1 fsmt FSMTTokenizer facebook/wmt19-de-en PASSED
2 marian MarianTokenizer Helsinki-NLP/opus-mt-de-en PASSED
3 mbart MBartTokenizerFast sshleifer/tiny-mbart PASSED
4 mt5 T5TokenizerFast google/mt5-small PASSED
5 t5 T5TokenizerFast t5-small PASSED

Summary

This module includes the fundamental data preprocessing bits to use Blurr for translation.