dar-tau commited on
Commit
b80761a
·
verified ·
1 Parent(s): 4333b51

Update app.py

Browse files
Files changed (1) hide show
  1. app.py +4 -2
app.py CHANGED
@@ -4,12 +4,14 @@ import spaces
4
  from transformers import AutoTokenizer, AutoModelForCausalLM, pipeline
5
 
6
 
7
- model_name = "teknium/OpenHermes-2.5-Mistral-7B"
8
  token = os.environ['hf_token']
9
 
10
  pipe = pipeline("text-generation", model=model_name, device="cuda")
11
 
12
 
 
 
13
  system_prompt = '''You are given an input text for a chat interface. Propose auto-completion to the text. You have several roles:
14
  - Fight under-specification: if the user does not provide sufficient context, propose them a set of relevant suggestions.
15
  - Complete text: The text provided to you is in the making. If you have a good idea for how to complete - make suggestions.
@@ -24,7 +26,7 @@ def generate(text):
24
  {'role': 'system', 'content': system_prompt},
25
  {'role': 'user', 'content': text}
26
  ]
27
- return pipe(messages)
28
 
29
 
30
  if __name__ == "__main__":
 
4
  from transformers import AutoTokenizer, AutoModelForCausalLM, pipeline
5
 
6
 
7
+ model_name = "TheBloke/OpenHermes-2.5-Mistral-7B-GGUF"
8
  token = os.environ['hf_token']
9
 
10
  pipe = pipeline("text-generation", model=model_name, device="cuda")
11
 
12
 
13
+ generate_kwargs = {'max_new_tokens': 20}
14
+
15
  system_prompt = '''You are given an input text for a chat interface. Propose auto-completion to the text. You have several roles:
16
  - Fight under-specification: if the user does not provide sufficient context, propose them a set of relevant suggestions.
17
  - Complete text: The text provided to you is in the making. If you have a good idea for how to complete - make suggestions.
 
26
  {'role': 'system', 'content': system_prompt},
27
  {'role': 'user', 'content': text}
28
  ]
29
+ return pipe(messages, generate_kwargs=generate_kwargs)
30
 
31
 
32
  if __name__ == "__main__":