Leonard Püttmann commited on
Commit
ffb5497
·
verified ·
1 Parent(s): 0260be1

Upload 2 files

Browse files
Files changed (2) hide show
  1. app.py +6 -2
  2. requirements.txt +1 -2
app.py CHANGED
@@ -1,3 +1,4 @@
 
1
  import gradio as gr
2
  from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
3
  import spacy
@@ -17,20 +18,22 @@ class ModelSingleton:
17
 
18
  # Load translation models and tokenizers
19
  self.tokenizer_en_it = AutoTokenizer.from_pretrained("LeonardPuettmann/Quadrifoglio-mt-en-it")
20
- self.model_en_it = AutoModelForSeq2SeqLM.from_pretrained("LeonardPuettmann/Quadrifoglio-mt-en-it", load_in_8bit=True)
21
 
22
  self.tokenizer_it_en = AutoTokenizer.from_pretrained("LeonardPuettmann/Quadrifoglio-mt-it-en")
23
- self.model_it_en = AutoModelForSeq2SeqLM.from_pretrained("LeonardPuettmann/Quadrifoglio-mt-it-en", load_in_8bit=True)
24
 
25
  self.initialized = True
26
 
27
  model_singleton = ModelSingleton()
28
 
 
29
  def generate_response_en_it(input_text):
30
  input_ids = model_singleton.tokenizer_en_it("translate English to Italian: " + input_text, return_tensors="pt").input_ids
31
  output = model_singleton.model_en_it.generate(input_ids, max_new_tokens=256)
32
  return model_singleton.tokenizer_en_it.decode(output[0], skip_special_tokens=True)
33
 
 
34
  def generate_response_it_en(input_text):
35
  input_ids = model_singleton.tokenizer_it_en("translate Italian to English: " + input_text, return_tensors="pt").input_ids
36
  output = model_singleton.model_it_en.generate(input_ids, max_new_tokens=256)
@@ -67,3 +70,4 @@ iface = gr.Interface(
67
 
68
  # Launch the interface
69
  iface.launch()
 
 
1
+ import spaces
2
  import gradio as gr
3
  from transformers import AutoTokenizer, AutoModelForSeq2SeqLM
4
  import spacy
 
18
 
19
  # Load translation models and tokenizers
20
  self.tokenizer_en_it = AutoTokenizer.from_pretrained("LeonardPuettmann/Quadrifoglio-mt-en-it")
21
+ self.model_en_it = AutoModelForSeq2SeqLM.from_pretrained("LeonardPuettmann/Quadrifoglio-mt-en-it")
22
 
23
  self.tokenizer_it_en = AutoTokenizer.from_pretrained("LeonardPuettmann/Quadrifoglio-mt-it-en")
24
+ self.model_it_en = AutoModelForSeq2SeqLM.from_pretrained("LeonardPuettmann/Quadrifoglio-mt-it-en")
25
 
26
  self.initialized = True
27
 
28
  model_singleton = ModelSingleton()
29
 
30
+ @spaces.GPU(duration=30)
31
  def generate_response_en_it(input_text):
32
  input_ids = model_singleton.tokenizer_en_it("translate English to Italian: " + input_text, return_tensors="pt").input_ids
33
  output = model_singleton.model_en_it.generate(input_ids, max_new_tokens=256)
34
  return model_singleton.tokenizer_en_it.decode(output[0], skip_special_tokens=True)
35
 
36
+ @spaces.GPU(duration=30)
37
  def generate_response_it_en(input_text):
38
  input_ids = model_singleton.tokenizer_it_en("translate Italian to English: " + input_text, return_tensors="pt").input_ids
39
  output = model_singleton.model_it_en.generate(input_ids, max_new_tokens=256)
 
70
 
71
  # Launch the interface
72
  iface.launch()
73
+
requirements.txt CHANGED
@@ -1,8 +1,7 @@
1
- accelerate
2
- bitsandbytes
3
  gradio
4
  transformers
5
  spacy
 
6
  torch
7
  en-core-web-sm @ https://github.com/explosion/spacy-models/releases/download/en_core_web_sm-3.7.1/en_core_web_sm-3.7.1-py3-none-any.whl
8
  it-core-news-sm @ https://github.com/explosion/spacy-models/releases/download/it_core_news_sm-3.8.0/it_core_news_sm-3.8.0-py3-none-any.whl
 
 
 
1
  gradio
2
  transformers
3
  spacy
4
+ spaces
5
  torch
6
  en-core-web-sm @ https://github.com/explosion/spacy-models/releases/download/en_core_web_sm-3.7.1/en_core_web_sm-3.7.1-py3-none-any.whl
7
  it-core-news-sm @ https://github.com/explosion/spacy-models/releases/download/it_core_news_sm-3.8.0/it_core_news_sm-3.8.0-py3-none-any.whl