fix

app.py

@@ -22,13 +22,15 @@ device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
 # -----------------------------------------------------------------------------
 tokenizer = AutoTokenizer.from_pretrained(
     MODEL_ID,
-    use_auth_token=HF_TOKEN,
-    trust_remote_code=True
+    token=HF_TOKEN,
+    trust_remote_code=True,
+    torch_dtype=torch.float32,
+    low_cpu_mem_usage=True
 )
 
 model = AutoModelForCausalLM.from_pretrained(
     MODEL_ID,
-    use_auth_token=HF_TOKEN,
+    token=HF_TOKEN,
     torch_dtype=torch.float32,    # CPU 環境では float32
     device_map="auto" if torch.cuda.is_available() else None
 )

requirements.txt

@@ -2,3 +2,5 @@ fastapi
 uvicorn[standard]
 transformers>=4.50.0.dev0
 torch
+accelerate>=0.9.0
+safetensors