fine, i'll set my own cache dir

app.py

@@ -3,7 +3,6 @@ from __future__ import print_function, division, unicode_literals
 import gradio as gr
 
 import sys
-from os import environ
 from os.path import abspath, dirname
 
 import json
@@ -13,9 +12,9 @@ from torchmoji.sentence_tokenizer import SentenceTokenizer
 from torchmoji.model_def import torchmoji_emojis
 from transformers import AutoModel, AutoTokenizer
 model_name = "Pendrokar/TorchMoji"
-model = AutoModel.from_pretrained(model_name)
+model = AutoModel.from_pretrained(model_name, cache_dir="~/.cache/huggingface/hub/")
 tokenizer = AutoTokenizer.from_pretrained(model_name)
-model_path = environ['HF_HUB_CACHE'] + "{model_name}/pytorch_model.bin"
+model_path = "~/.cache/huggingface/hub/{model_name}/pytorch_model.bin"
 vocab_path = './' + model_name + "/vocabulary.json"
 
 def top_elements(array, k):