HF token

.gitignore

@@ -0,0 +1 @@
+.env

app.py

@@ -8,6 +8,9 @@ from fastapi.responses import StreamingResponse
 from pydantic import BaseModel
 import json
 from typing import List, Literal
+import os
+
+HF_TOKEN = os.getenv("HF_TOKEN")
 
 MODEL = "meta-llama/Llama-3.2-1B-Instruct"
 
@@ -16,6 +19,7 @@ app = FastAPI()
 # base model and tokenizer
 base_model = AutoModelForCausalLM.from_pretrained(
     MODEL,
+    token=HF_TOKEN,
     torch_dtype=torch.float32, #huggingface free tier only has cpu
     device_map="cpu",
     low_cpu_mem_usage=True
@@ -23,7 +27,7 @@ base_model = AutoModelForCausalLM.from_pretrained(
 
 base_model.config.use_cache = True
 
-tokenizer = AutoTokenizer.from_pretrained(MODEL)
+tokenizer = AutoTokenizer.from_pretrained(MODEL, token=HF)
 
 if tokenizer.pad_token is None:
     tokenizer.pad_token = tokenizer.eos_token