Update app.py

app.py

@@ -1,25 +1,23 @@
-from langchain.embeddings.openai import OpenAIEmbeddings
-from langchain.vectorstores import Chroma
-from langchain.text_splitter import CharacterTextSplitter
-from langchain.chains.question_answering import load_qa_chain
-from langchain.llms import OpenAI
-import os
+import torch
+from peft import PeftModel, PeftConfig
+from transformers import AutoModelForCausalLM, AutoTokenizer
+from IPython.display import display, Markdown
 
-with open("guide1.txt") as f:
-    hitchhikersguide = f.read()
+peft_model_id = f"adamtappis/marketing_emails_model"
+config = PeftConfig.from_pretrained(peft_model_id)
+model = AutoModelForCausalLM.from_pretrained(config.base_model_name_or_path, return_dict=True, load_in_8bit=False)
+tokenizer = AutoTokenizer.from_pretrained(config.base_model_name_or_path)
 
-text_splitter = CharacterTextSplitter(chunk_size=1000, chunk_overlap=0, separator = "\n")
-texts = text_splitter.split_text(hitchhikersguide)
+# Load the Lora model
+model = PeftModel.from_pretrained(model, peft_model_id)
 
-embeddings = OpenAIEmbeddings()
+def make_inference(product, description):
+  batch = tokenizer(f"### INSTRUCTION\nBelow is a product and description, please write a marketing email for this product.\n\n### Product:\n{product}\n### Description:\n{description}\n\n### Marketing Email:\n", return_tensors='pt')
 
-docsearch = Chroma.from_texts(texts, embeddings, metadatas=[{"source": str(i)} for i in range(len(texts))]).as_retriever()
+  with torch.cuda.amp.autocast():
+    output_tokens = model.generate(**batch, max_new_tokens=200)
 
-chain = load_qa_chain(OpenAI(temperature=0), chain_type="stuff")
-
-def make_inference(query):
-    docs = docsearch.get_relevant_documents(query)
-    return(chain.run(input_documents=docs, question=query))
+  display(Markdown((tokenizer.decode(output_tokens[0], skip_special_tokens=True))))
 
 if __name__ == "__main__":
     # make a gradio interface
@@ -28,9 +26,10 @@ if __name__ == "__main__":
     gr.Interface(
         make_inference,
         [
-            gr.inputs.Textbox(lines=2, label="Query"),
+            gr.inputs.Textbox(lines=1, label="Product Name"),
+            gr.inputs.Textbox(lines=1, label="Product Description"),
         ],
-        gr.outputs.Textbox(label="Response"),
-        title="🗣️TalkToMyDoc📄",
-        description="🗣️TalkToMyDoc📄 is a tool that allows you to ask questions about a document. In this case - Hitch Hitchhiker's Guide to the Galaxy.",
+        gr.outputs.Textbox(label="Email"),
+        title="🗣️Marketing Email Generator📄",
+        description="🗣️Marketing Email Generator📄 is a tool that allows you to generate marketing emails for different products",
     ).launch()