need at least 500 tokens for interpretable code-switching performance…

… since it seems to gear towards monolingual English
ctarnold · Mar 22, 2024 · ccb4975 · ccb4975
1 parent b505367
commit ccb4975
Showing 1 changed file with 1 addition and 1 deletion.
diff --git a/mistralPromptingZeroShot.py b/mistralPromptingZeroShot.py
@@ -19,7 +19,7 @@
 model_inputs = tokenizer([prefix + prompt], return_tensors="pt").to(device)
 
 generated_ids = model.generate(**model_inputs, 
-                               max_new_tokens=200, 
+                               max_new_tokens=512, 
                                temperature = 0.7, 
                                do_sample=True)