losca commited on
Commit
e0bcfb5
1 Parent(s): 2393e6a

Update README.md

Browse files

Added example usage

Files changed (1) hide show
  1. README.md +23 -0
README.md CHANGED
@@ -15,6 +15,29 @@ This is a finetuning of a MarianMT pretrained on English-Chinese. The target lan
15
  The first phase of training (mixed) is performed on a dataset containing both English-Chinese and English-Vietnamese sentences.
16
  The second phase of training (pure) is performed on a dataset containing only English-Vietnamese sentences.
17
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
18
 
19
  ### Training results
20
 
 
15
  The first phase of training (mixed) is performed on a dataset containing both English-Chinese and English-Vietnamese sentences.
16
  The second phase of training (pure) is performed on a dataset containing only English-Vietnamese sentences.
17
 
18
+ ### Example
19
+ ```
20
+ %%capture
21
+ !pip install transformers transformers[sentencepiece]
22
+
23
+ from transformers import AutoModelForSeq2SeqLM, AutoTokenizer
24
+ # Download the pretrained model for English-Vietnamese available on the hub
25
+ model = AutoModelForSeq2SeqLM.from_pretrained("CLAck/en-vi")
26
+
27
+ tokenizer = AutoTokenizer.from_pretrained("CLAck/en-vi")
28
+ # Download a tokenizer that can tokenize English since the model Tokenizer doesn't know anymore how to do it
29
+ # We used the one coming from the initial model
30
+ # This tokenizer is used to tokenize the input sentence
31
+ tokenizer_en = AutoTokenizer.from_pretrained('Helsinki-NLP/opus-mt-en-zh')
32
+ # These special tokens are needed to reproduce the original tokenizer
33
+ tokenizer_en.add_tokens(["<2zh>", "<2vi>"], special_tokens=True)
34
+
35
+ sentence = "The cat is on the table"
36
+ # This token is needed to identify the target language
37
+ input_sentence = "<2vi> " + sentence
38
+ translated = model.generate(**tokenizer_en(input_sentence, return_tensors="pt", padding=True))
39
+ output_sentence = [tokenizer.decode(t, skip_special_tokens=True) for t in translated]
40
+ ```
41
 
42
  ### Training results
43