Update app.py

app.py

@@ -2,9 +2,14 @@ import torch
 import gradio as gr
 from transformers import AutoModelForCausalLM, AutoTokenizer
 import os
-os.environ['HF_HOME'] = '/tmp/huggingface'
-os.environ['TORCH_HOME'] = '/tmp/torch_cache'
-
+export HF_HOME=/data/.cache
+export HF_HUB_CACHE=/data/.cache/huggingface/hub
+export TRANSFORMERS_CACHE=/data/.cache/huggingface/hub
+export HF_DATASETS_CACHE=/data/.cache/huggingface/datasets
+export XDG_CACHE_HOME=/data/.cache
+export TORCH_HOME=/data/.cache/torch
+export TMPDIR=/data/tmp
+mkdir -p /data/.cache /data/tmp
 MODEL_PATH = "zai-org/GLM-4.7-Flash"
 
 # 1. Load the model and tokenizer globally for efficiency