Update app.py

app.py

@@ -15,10 +15,10 @@ device = "cuda" if torch.cuda.is_available() else "cpu"
 
 # Define variables 
 temperature = 0.0  
-max_new_tokens = 3000
+max_new_tokens = 1500
 top_p = 0.95
 repetition_penalty = 1.0
-min_new_tokens = 1000
+min_new_tokens = 800
 early_stopping = False
 
 model_name = "PleIAs/Pleias-Rag"
@@ -36,7 +36,7 @@ model.to(device)
 # Set tokenizer configuration
 tokenizer.pad_token = tokenizer.eos_token
 tokenizer.pad_token_id = tokenizer.eos_token_id
-tokenizer.eos_token = "<|end_of_text|>"
+tokenizer.eos_token = "<|answer_end|>"
 
 # Connect to the LanceDB database
 db = lancedb.connect("content 5/lancedb_data")