a

app.py

@@ -4,7 +4,7 @@ from huggingface_hub import InferenceClient
 """
 Copied from inference in colab notebook
 """
-# from unsloth.chat_templates import get_chat_template
+from unsloth.chat_templates import get_chat_template
 from unsloth import FastLanguageModel
 
 # IMPORTING MODEL AND TOKENIZER ————————
@@ -20,10 +20,10 @@ model, tokenizer = FastLanguageModel.from_pretrained(
     load_in_4bit = load_in_4bit,
 )
 
-# tokenizer = get_chat_template(
-#     tokenizer,
-#     chat_template = "llama-3.1",
-# )
+tokenizer = get_chat_template(
+    tokenizer,
+    chat_template = "llama-3.1",
+)
 FastLanguageModel.for_inference(model) # Enable native 2x faster inference
 
 # RUNNING INFERENCE ————————————————————————

requirements.txt

@@ -1,3 +1,6 @@
 huggingface_hub==0.25.2
 
-unsloth
+unsloth
+torch==2.0.1+cpu --force-reinstall
+torchvision==0.15.2+cpu --force-reinstall
+-f https://download.pytorch.org/whl/cpu