Mel Seto commited on
Commit
108259c
·
1 Parent(s): 6ad01ae

revert to correct model; improve prompt to make more likely to return a 成語

Browse files
Files changed (1) hide show
  1. app.py +2 -2
app.py CHANGED
@@ -16,7 +16,7 @@ USE_MOCK = False # Set False to use the real Cerebras API
16
  if not USE_MOCK:
17
  client = InferenceClient(
18
  provider="cerebras",
19
- model="cerebras/btlm-3b-8k-base",
20
  api_key=os.environ["HF_TOKEN"]
21
  )
22
 
@@ -34,7 +34,7 @@ def generate_idiom_mock(situation: str):
34
  def generate_idiom(situation: str, client):
35
  prompt = f"""
36
  You are a wise assistant. Given a situation, respond with exactly:
37
- 1. A Chinese idiom (成语)
38
  2. Its pinyin
39
  3. A short English explanation
40
 
 
16
  if not USE_MOCK:
17
  client = InferenceClient(
18
  provider="cerebras",
19
+ model="meta-llama/Llama-3.3-70B-Instruct",
20
  api_key=os.environ["HF_TOKEN"]
21
  )
22
 
 
34
  def generate_idiom(situation: str, client):
35
  prompt = f"""
36
  You are a wise assistant. Given a situation, respond with exactly:
37
+ 1. A traditional Chinese idiom (成语), which is different from a 四字格, that conveys the idea of the given situation
38
  2. Its pinyin
39
  3. A short English explanation
40