prajdabre commited on
Commit
8a9003b
·
1 Parent(s): 836d206

Initial space

Browse files
Files changed (1) hide show
  1. app.py +48 -0
app.py ADDED
@@ -0,0 +1,48 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ import gradio as gr
2
+ from transformers import MBartForConditionalGeneration, AutoModelForSeq2SeqLM
3
+ from transformers import AlbertTokenizer, AutoTokenizer
4
+
5
+ tokenizer = AlbertTokenizer.from_pretrained("ai4bharat/MultiIndicWikiBioSS", do_lower_case=False, use_fast=False, keep_accents=True)
6
+
7
+ # Or use tokenizer = AlbertTokenizer.from_pretrained("ai4bharat/IndicBART-XLSum", do_lower_case=False, use_fast=False, keep_accents=True)
8
+
9
+ # xlsummodel = AutoModelForSeq2SeqLM.from_pretrained("ai4bharat/IndicBART-XLSum")
10
+ qgmodel = AutoModelForSeq2SeqLM.from_pretrained("ai4bharat/MultiIndicQuestionGenerationSS")
11
+ hgmodel = AutoModelForSeq2SeqLM.from_pretrained("ai4bharat/MultiIndicHeadlineGenerationSS")
12
+ ssmodel = AutoModelForSeq2SeqLM.from_pretrained("ai4bharat/MultiIndicSentenceSummarizationSS")
13
+ ppmodel = AutoModelForSeq2SeqLM.from_pretrained("ai4bharat/MultiIndicParaphraseGenerationSS")
14
+ wbmodel = AutoModelForSeq2SeqLM.from_pretrained("ai4bharat/MultiIndicWikiBioSS")
15
+
16
+ # Some initial mapping
17
+ bos_id = tokenizer._convert_token_to_id_with_added_voc("<s>")
18
+ eos_id = tokenizer._convert_token_to_id_with_added_voc("</s>")
19
+ pad_id = tokenizer._convert_token_to_id_with_added_voc("<pad>")
20
+ # To get lang_id use any of ['<2bn>', '<2gu>', '<2hi>', '<2mr>', '<2pa>', '<2ta>', '<2te>']
21
+
22
+
23
+
24
+ def greet(choice, lang, input):
25
+ if choice == "IndicWikiBio":
26
+ model = wbmodel
27
+ elif choice == "IndicHeadlineGeneration":
28
+ model = hgmodel
29
+ elif choice == "IndicParaprasing":
30
+ model = ppmodel
31
+ elif choice == "IndicSentenceSummarization":
32
+ model = ssmodel
33
+ elif choice == "IndicQuestionGeneration":
34
+ model = qgmodel
35
+
36
+
37
+ inp = tokenizer(input.strip() + " </s> <2" + lang + ">", add_special_tokens=False, return_tensors="pt", padding=True).input_ids
38
+ model_output=model.generate(inp, use_cache=True, num_beams=1, max_length=100, min_length=1, early_stopping=True, pad_token_id=pad_id, bos_token_id=bos_id, eos_token_id=eos_id, decoder_start_token_id=tokenizer._convert_token_to_id_with_added_voc("<2"+lang+">"))
39
+
40
+
41
+ # Decode to get output strings
42
+
43
+ decoded_output=tokenizer.decode(model_output[0], skip_special_tokens=True, clean_up_tokenization_spaces=False)
44
+
45
+ return decoded_output
46
+
47
+ iface = gr.Interface(fn=greet, inputs=[gr.inputs.Dropdown("IndicWikiBio", "IndicHeadlineGeneration", "IndicParaprasing", "IndicSentenceSummarization", "IndicQuestionGeneration"), gr.inputs.Dropdown("as","bn", "gu", "hi", "kn", "ml", "mr", "or", "pa", "ta", "te"), "text"], outputs="text")
48
+ iface.launch()