Update app.py

app.py

@@ -34,8 +34,20 @@ device = "cuda" if torch.cuda.is_available() else "cpu"
 
 #tokenizer = LlamaTokenizerFast.from_pretrained(REPO_NAME, token=HF_TOKEN)
 #tokenizer = LlamaTokenizer.from_pretrained(REPO_NAME, use_auth_token=HF_TOKEN)
-tokenizer = AutoTokenizer.from_pretrained("microsoft/Phi-3-mini-4k-instruct")
 
+#tokenizer = AutoTokenizer.from_pretrained("microsoft/Phi-3-mini-4k-instruct")
+tokenizer = LlamaTokenizerFast.from_pretrained(
+    "microsoft/Phi-3-mini-4k-instruct",
+    token=HF_TOKEN
+)
+
+
+# model = AutoModelForCausalLM.from_pretrained(
+#     REPO_NAME,
+#     use_auth_token=HF_TOKEN,
+#     torch_dtype=torch.float16 if torch.cuda.is_available() else torch.float32,
+#     device_map="auto"
+# )
 model = AutoModelForCausalLM.from_pretrained(
     REPO_NAME,
     use_auth_token=HF_TOKEN,