Update app.py

app.py

@@ -14,17 +14,16 @@ from collections import defaultdict
 from datetime import date, datetime, timedelta
 from transformers import AutoTokenizer, AutoModelForCausalLM, TextStreamer
 
-os.environ['HF_HOME'] = '/data/.huggingface'
 
-access_token = "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"
-
-finnhub_client = finnhub.Client(api_key="ckrn2mpr01qiu0ijr9hgckrn2mpr01qiu0ijr9i0")
+access_token = os.environ["HF_TOKEN"]
+finnhub_client = finnhub.Client(api_key=os.environ["FINNHUB_API_KEY"])
 
 base_model = AutoModelForCausalLM.from_pretrained(
     'meta-llama/Llama-2-7b-chat-hf',
     token=access_token,
     trust_remote_code=True, 
     device_map="auto",
+    torch_dtype=torch.float16,
     offload_folder="offload/"
 )
 model = PeftModel.from_pretrained(
@@ -32,11 +31,8 @@ model = PeftModel.from_pretrained(
     'FinGPT/fingpt-forecaster_dow30_llama2-7b_lora',
     offload_folder="offload/"
 )
-model = model.half()
 model = model.eval()
 
-print_gpu_utilization()
-
 tokenizer = AutoTokenizer.from_pretrained(
     'meta-llama/Llama-2-7b-chat-hf',
     token=access_token