Upload inference.py

inference.py

@@ -0,0 +1,12 @@
+from transformers import pipeline
+
+import config
+
+def predict(prompt, model, tokenizer, max_length):
+    pipe = pipeline(task = config.TASK,
+                    model = model,
+                    tokenizer = tokenizer,
+                    max_length = max_length)
+    result = pipe(f"<s>[INST] {prompt} [/INST]")
+    return result[0]['generated_text']
+