Upload app.py

app.py

@@ -26,54 +26,25 @@ from langchain_community.embeddings import HuggingFaceEmbeddings
 from langchain_community.vectorstores import FAISS
 from langchain.schema.runnable import RunnablePassthrough
 from langchain_core.messages import AIMessage, HumanMessage
+from langchain_community.llms import HuggingFaceEndpoint
 from dotenv import load_dotenv
 
+# Load environment variables from .env file
+load_dotenv()
+
 # Get the API token from environment variable
 api_token = os.getenv("API_TOKEN")
 
-os.environ["PYTORCH_CUDA_ALLOC_CONF"] = "max_split_size_mb:15000"
-
-model_id = "google/gemma-2-9b-it"
-quantization_config = BitsAndBytesConfig(load_in_4bit=True)
+# Define the repository ID and task
+repo_id = "mistralai/Mixtral-8x7B-Instruct-v0.1"
+task = "text-generation"
 
-tokenizer = AutoTokenizer.from_pretrained(
-    model_id,
-    return_tensors="pt",
-    padding=True,
-    truncation=True,
-    trust_remote_code=True,
-)
-tokenizer.pad_token = tokenizer.eos_token
-tokenizer.padding_side = "right"
-
-model = AutoModelForCausalLM.from_pretrained(
-    model_id,
-    quantization_config=quantization_config,
-    device_map="auto",
-    low_cpu_mem_usage=True,
-    pad_token_id=0,
+# Initialize the Hugging Face Endpoint
+chat_model = HuggingFaceEndpoint(
+    huggingfacehub_api_token=api_token,
+    repo_id=repo_id,
+    task=task
 )
-model.config.use_cache = False
-
-# Create a text generation pipeline with specific settings
-pipe = transformers.pipeline(
-    task="text-generation",
-    model=model,
-    tokenizer=tokenizer,
-    torch_dtype=torch.float16,
-    device_map="auto",
-#     do_sample=True,
-#     top_k=10,
-    temperature=0.0,
-    top_p=0.9,
-    num_return_sequences=1,
-    eos_token_id=tokenizer.eos_token_id,
-    max_length=4096,
-    truncation=True,
-)
-
-chat_model = HuggingFacePipeline(pipeline=pipe)
-
 
 template = """
 You are a genius trader with extensive knowledge of the financial and stock markets, capable of providing deep and insightful analysis of financial stocks with remarkable accuracy.