moon-hotel / transformertranslation Goto Github PK

if key_padding_mask is not None: attn_output_weights = attn_output_weights.view(bsz, num_heads, tgt_len, src_len) # 变成 [batch_size, num_heads, tgt_len, src_len]的形状 attn_output_weights = attn_output_weights.masked_fill( key_padding_mask.unsqueeze(1).unsqueeze(2), # 扩展维度，从[batch_size,src_len]变成[batch_size,1,1,src_len] float('-inf')) # attn_output_weights = attn_output_weights.view(bsz * num_heads, tgt_len, src_len) # [batch_size * num_heads, tgt_len, src_len]

中的 attn_output_weights = attn_output_weights.masked_fill(padding_mask.unsqueeze(1).unsqueeze(2), float('-inf')) 是否应该是 attn_output_weights = attn_output_weights.masked_fill(padding_mask.unsqueeze(1).unsqueeze(2) == False, float('-inf')) ？？

在使用python train.py的时候 报错
rnatives. (Triggered internally at /Users/runner/work/pytorch/pytorch/pytorch/torch/csrc/tensor/python_tensor.cpp:453.)
_C._set_default_tensor_type(t)
Traceback (most recent call last):
File "/Users/wangwan/Desktop/TransformerTranslation-main/train.py", line 152, in
train_model(config)
File "/Users/wangwan/Desktop/TransformerTranslation-main/train.py", line 51, in train_model
data_loader = LoadEnglishGermanDataset(config.train_corpus_file_paths,
File "/Users/wangwan/Desktop/TransformerTranslation-main/utils/data_helpers.py", line 43, in init
self.tokenizer = tokenizer()
File "/Users/wangwan/Desktop/TransformerTranslation-main/utils/data_helpers.py", line 13, in my_tokenizer
tokenizer['de'] = get_tokenizer('spacy', language='de_core_news_sm') # 德语
File "/Users/wangwan/opt/anaconda3/lib/python3.9/site-packages/torchtext/data/utils.py", line 114, in get_tokenizer
spacy = spacy.load(language)
File "/Users/wangwan/opt/anaconda3/lib/python3.9/site-packages/spacy/init.py", line 50, in load
return util.load_model(
File "/Users/wangwan/opt/anaconda3/lib/python3.9/site-packages/spacy/util.py", line 324, in load_model
return load_model_from_package(name, **kwargs)
File "/Users/wangwan/opt/anaconda3/lib/python3.9/site-packages/spacy/util.py", line 357, in load_model_from_package
return cls.load(vocab=vocab, disable=disable, exclude=exclude, config=config)
File "/Users/wangwan/opt/anaconda3/lib/python3.9/site-packages/de_core_news_sm/init.py", line 10, in load
return load_model_from_init_py(file, **overrides)
File "/Users/wangwan/opt/anaconda3/lib/python3.9/site-packages/spacy/util.py", line 517, in load_model_from_init_py
return load_model_from_path(
File "/Users/wangwan/opt/anaconda3/lib/python3.9/site-packages/spacy/util.py", line 392, in load_model_from_path
nlp = load_model_from_config(config, vocab=vocab, disable=disable, exclude=exclude)
File "/Users/wangwan/opt/anaconda3/lib/python3.9/site-packages/spacy/util.py", line 429, in load_model_from_config
nlp = lang_cls.from_config(
File "/Users/wangwan/opt/anaconda3/lib/python3.9/site-packages/spacy/language.py", line 1672, in from_config
nlp.add_pipe(
File "/Users/wangwan/opt/anaconda3/lib/python3.9/site-packages/spacy/language.py", line 774, in add_pipe
pipe_component = self.create_pipe(
File "/Users/wangwan/opt/anaconda3/lib/python3.9/site-packages/spacy/language.py", line 660, in create_pipe
resolved = registry.resolve(cfg, validate=validate)
File "/Users/wangwan/opt/anaconda3/lib/python3.9/site-packages/thinc/config.py", line 746, in resolve
resolved, _ = cls._make(
File "/Users/wangwan/opt/anaconda3/lib/python3.9/site-packages/thinc/config.py", line 795, in _make
filled, _, resolved = cls._fill(
File "/Users/wangwan/opt/anaconda3/lib/python3.9/site-packages/thinc/config.py", line 850, in _fill
filled[key], validation[v_key], final[key] = cls._fill(
File "/Users/wangwan/opt/anaconda3/lib/python3.9/site-packages/thinc/config.py", line 849, in _fill
promise_schema = cls.make_promise_schema(value, resolve=resolve)
File "/Users/wangwan/opt/anaconda3/lib/python3.9/site-packages/thinc/config.py", line 1057, in make_promise_schema
return create_model("ArgModel", **sig_args)
File "pydantic/main.py", line 990, in pydantic.main.create_model
File "pydantic/main.py", line 299, in pydantic.main.ModelMetaclass.new
File "pydantic/fields.py", line 411, in pydantic.fields.ModelField.infer
File "pydantic/fields.py", line 342, in pydantic.fields.ModelField.init
File "pydantic/fields.py", line 451, in pydantic.fields.ModelField.prepare
File "pydantic/fields.py", line 550, in pydantic.fields.ModelField._type_analysis
File "/Users/wangwan/opt/anaconda3/lib/python3.9/typing.py", line 852, in subclasscheck
return issubclass(cls, self.origin)
TypeError: issubclass() arg 1 must be a class

下面是翻译结果

德语：Eine Gruppe von Menschen steht vor einem Iglu.
翻译： A group of people standing in front of an outdoor studio . 
英语：A group of people are facing an igloo.


德语：Ein Mann in einem blauen Hemd steht auf einer Leiter und putzt ein Fenster.
翻译： A man in a shirt standing on a ladder and cleaning a ladder . 
英语：A man in a blue shirt is standing on a ladder cleaning a window.

需要更多的train data吗？

你就不能给个训练结果吗，还要我们自己训练，挺麻烦的

MyTransformer.py的 multi_head_attention_forward()里，这里是不是错了？
attn_output_weights = attn_output_weights.masked_fill(
key_padding_mask.unsqueeze(1).unsqueeze(2),
float('-inf')) #

应该填充的位置是key_padding_mask.unsqueeze(1).unsqueeze(2) == 0的地方吧。 少了一个 == 0

I can't find this in config.please

运行报错了：init() got an unexpected keyword argument 'specials'
看了下好像是utils\data_helpers.py中return Vocab(counter, specials=specials, min_freq=min_freq)这个函数报错。求解答

我是python小白，运行代码说 RuntimeError: Token Eine not found and default index is not set ，但我用print查找过有这个Eine字符， 请问是什么问题？

=======================================================================
49 data_loader.load_train_val_test_data(config.train_corpus_file_paths,
50 config.val_corpus_file_paths,
---> 51 config.test_corpus_file_paths)
52 logging.info("############初始化模型############")
53 translation_model = TranslationModel(src_vocab_size=len(data_loader.de_vocab),

/tmp/ipykernel_27/774184140.py in load_train_val_test_data(self, train_file_paths, val_file_paths, test_file_paths)
42 print('load traning data done')
43
---> 44 val_data = self.data_process(val_file_paths)
45 print('load val data done')
46

/tmp/ipykernel_27/774184140.py in data_process(self, filepaths)
25 for (raw_de, raw_en) in tqdm(zip(raw_de_iter, raw_en_iter),ncols=80):
26 de_tensor_ = torch.tensor([self.de_vocab[token] for token in
---> 27 self.tokenizer'de'], dtype=torch.long)
28 en_tensor_ = torch.tensor([self.en_vocab[token] for token in
29 self.tokenizer'en'], dtype=torch.long)

/tmp/ipykernel_27/774184140.py in (.0)
24 logging.info(f"### 正在将数据集 {filepaths} 转换成 Token ID ")
25 for (raw_de, raw_en) in tqdm(zip(raw_de_iter, raw_en_iter),ncols=80):
---> 26 de_tensor_ = torch.tensor([self.de_vocab[token] for token in
27 self.tokenizer'de'], dtype=torch.long)
28 en_tensor_ = torch.tensor([self.en_vocab[token] for token in

/opt/conda/lib/python3.7/site-packages/torchtext/vocab/vocab.py in getitem(self, token)
62 The index corresponding to the associated token.
63 """
---> 64 return self.vocab[token]
65
66 @torch.jit.export

RuntimeError: Token Eine not found and default index is not set

在这个位置：

    loss_fn = torch.nn.CrossEntropyLoss(ignore_index=data_loader.PAD_IDX)

    optimizer = torch.optim.Adam(translation_model.parameters(),
                                 lr=0.,
                                 betas=(config.beta1, config.beta2), eps=config.epsilon)

按照 README 配置环境后，执行 python train.py 过程中会报如下这种错误：

de_tensor_ = torch.tensor([self.de_vocab[token] for token in
File "/usr/local/lib/python3.9/site-packages/torchtext/vocab/vocab.py", line 64, in getitem
return self.vocab[token]
RuntimeError: Token Bude not found and default index is not set

请问是有可能是哪一步没弄对吗？

torchtext == [0.12.0]