un-index commited on
Commit
45d20de
·
1 Parent(s): 0ddb594
Files changed (1) hide show
  1. app.py +4 -3
app.py CHANGED
@@ -121,6 +121,7 @@ def f(context, temperature, top_p, max_length, model_idx, SPACE_VERIFICATION_KEY
121
  "token_max_length": max_length, # 512,
122
  "temperature": temperature,
123
  "top_p": top_p,
 
124
  }
125
 
126
  response = requests.post(
@@ -143,7 +144,7 @@ def f(context, temperature, top_p, max_length, model_idx, SPACE_VERIFICATION_KEY
143
  # todo fix max_length below, maybe there is a max_new_tokens parameter
144
  # try max_length=len(context)+max_length or =len(context)+max_length or make max_length inf or unspecified
145
  # note: added max_new_tokens parameter to see whether it actually works, if not remove,
146
- # if yes, then make max_length infinite because it seems to be counted as max input length, not output
147
  generated_text = generator(context, max_length=400, max_new_tokens=max_length, top_p=top_p, temperature=temperature, num_return_sequences=1)
148
  except Exception as e:
149
  return "Exception while generating text: " + str(e)
@@ -156,7 +157,7 @@ def f(context, temperature, top_p, max_length, model_idx, SPACE_VERIFICATION_KEY
156
  return generated_text[0][0]['generated_text']
157
  except:
158
  return generated_text # was error due to timeout because of not enabling queue in gradio interface?
159
-
160
  # except:
161
  # generated_text = generator(context, max_length=max_length, top_p=top_p, temperature=temperature, num_return_sequences=1)[0]
162
 
@@ -168,7 +169,7 @@ def f(context, temperature, top_p, max_length, model_idx, SPACE_VERIFICATION_KEY
168
 
169
 
170
  except Exception as e:
171
- return f"error with idx{model_idx} : \n"+str(e)
172
 
173
 
174
  iface = gr.Interface(f, [
 
121
  "token_max_length": max_length, # 512,
122
  "temperature": temperature,
123
  "top_p": top_p,
124
+ "max_time": 120.0
125
  }
126
 
127
  response = requests.post(
 
144
  # todo fix max_length below, maybe there is a max_new_tokens parameter
145
  # try max_length=len(context)+max_length or =len(context)+max_length or make max_length inf or unspecified
146
  # note: added max_new_tokens parameter to see whether it actually works, if not remove,
147
+ # TODO if yes, then make max_length infinite because it seems to be counted as max input length, not output
148
  generated_text = generator(context, max_length=400, max_new_tokens=max_length, top_p=top_p, temperature=temperature, num_return_sequences=1)
149
  except Exception as e:
150
  return "Exception while generating text: " + str(e)
 
157
  return generated_text[0][0]['generated_text']
158
  except:
159
  return generated_text # was error due to timeout because of not enabling queue in gradio interface?
160
+ # if it works right now, then that was the reason for the JSON parsing error
161
  # except:
162
  # generated_text = generator(context, max_length=max_length, top_p=top_p, temperature=temperature, num_return_sequences=1)[0]
163
 
 
169
 
170
 
171
  except Exception as e:
172
+ return f"error with idx{model_idx}: "+str(e)
173
 
174
 
175
  iface = gr.Interface(f, [