Upload folder using huggingface_hub

.gitattributes

@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+faissdb/index.faiss filter=lfs diff=lfs merge=lfs -text

faissdb/index.faiss

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a86c1246d6477f34bca9db9b8ab3bb1a10e70db706b34676599e2fac1474101d
+size 151471149

faissdb/index.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5d8750363e543e852f0764682a184645dfe7d51d76394feb685b4ffee57dfcdd
+size 55501392

gradio_app.py

@@ -1,23 +1,22 @@
-# %%
-
 from threading import Thread
 
 import gradio as gr
+from langchain.callbacks.manager import CallbackManager
+from langchain.callbacks.streaming_stdout import StreamingStdOutCallbackHandler
+from langchain.chains import RetrievalQA
+from langchain.embeddings import HuggingFaceBgeEmbeddings
+from langchain.llms import HuggingFaceTextGenInference
+from langchain.prompts import PromptTemplate
+from langchain.vectorstores import FAISS
 # import torch
 from text_generation import Client, InferenceAPIClient
 
 client = Client("http://20.83.177.108:8080")
 
 
-# text = ""
-# for response in client.generate_stream("What is Deep Learning?", max_new_tokens=20):
-#     if not response.token.special:
-#         text += response.token.text
-# print(text)
-
-
-def run_generation(user_text, top_p, temperature, top_k, max_new_tokens):
+def run_generation_stream(user_text, f, max_new_tokens, temperature):
     # Get the model and tokenizer, and tokenize the user text.
+    print('called stream')
 
     if len(user_text.strip()) == 0:
         print('blank')
@@ -29,7 +28,7 @@ def run_generation(user_text, top_p, temperature, top_k, max_new_tokens):
     ### Response: """
 
     text = ""
-    for response in client.generate_stream(user_text, max_new_tokens=max_new_tokens, repetition_penalty=1.05):
+    for response in client.generate_stream(user_text, max_new_tokens=max_new_tokens, repetition_penalty=1.05, temperature=temperature):
         if not response.token.special:
             text += response.token.text
             yield text
@@ -41,6 +40,69 @@ def reset_textbox():
     return gr.update(value='')
 
 
+model_name = "BAAI/bge-base-en"
+# set True to compute cosine similarity
+encode_kwargs = {'normalize_embeddings': True}
+
+model_norm = HuggingFaceBgeEmbeddings(
+    model_name=model_name,
+    encode_kwargs=encode_kwargs
+)
+
+
+vectordb = FAISS.load_local('faissdb', embeddings=model_norm)
+retriever = vectordb.as_retriever(
+    search_type='similarity', search_kwargs={"k": 5})
+
+
+# relating to refer to Indian Penal Code(IPC), CrPC(Code of Criminal Procedure) for most cases and therefore laws
+prompt_template = """You are an expert legal assistant with extensive knowledge about Indian law. Your task is to respond to the given query in a factually correct and consise manner unless asked for a detailed explanation. Assume the query is asked by a common man unless explicitly specified otherwise, therefore no special acts or laws like ones for railway , army , police would apply to them. Use the following pieces of context to answer the question at the end. If you don't know the answer, just say that you don't know, don't try to make up an answer.
+
+{context}
+
+Question: {question}
+Response:"""
+
+
+PROMPT = PromptTemplate(
+    template=prompt_template, input_variables=["context", "question"]
+)
+
+
+def run_generation(query, factual, max_tokens, temperature):
+    print('called non stream')
+
+    llm = HuggingFaceTextGenInference(
+        inference_server_url="http://20.83.177.108:8080/",
+        max_new_tokens=max_tokens,
+        top_k=10,
+        top_p=0.95,
+        typical_p=0.95,
+        temperature=temperature,
+        streaming=True if factual else False,
+        # repetition_penalty=1.1,
+    )
+
+    qa_chain = RetrievalQA.from_chain_type(llm=llm,
+                                           chain_type_kwargs={
+                                               "prompt": PROMPT},
+                                           retriever=retriever,
+                                           return_source_documents=True,
+                                           )
+
+    # text = ""
+    # if factual:
+    #     response = llm(query, callbacks=[StreamingStdOutCallbackHandler()])
+    #     print(response)
+    #     # text += response
+    #     yield response
+
+    # else:
+    llm_response = qa_chain(query)
+    print(llm_response['result'])
+    return llm_response['result']
+
+
 with gr.Blocks() as demo:
     with gr.Row():
         with gr.Column(scale=4):
@@ -54,21 +116,23 @@ with gr.Blocks() as demo:
 
         with gr.Column(scale=1):
             max_new_tokens = gr.Slider(
-                minimum=1, maximum=1000, value=250, step=10, interactive=True, label="Max New Tokens",
-            )
-            top_p = gr.Slider(
-                minimum=0.05, maximum=1.0, value=0.95, step=0.05, interactive=True, label="Top-p (nucleus sampling)",
-            )
-            top_k = gr.Slider(
-                minimum=1, maximum=50, value=50, step=1, interactive=True, label="Top-k",
+                minimum=1, maximum=1000, value=50, step=10, interactive=True, label="Number of words to generate",
             )
             temperature = gr.Slider(
-                minimum=0.1, maximum=1.0, value=0.8, step=0.1, interactive=True, label="Temperature",
+                minimum=0.1, maximum=1.0, value=0.6, step=0.1, interactive=True, label="Randomness(can be between 0-1, 0 being least random)",
             )
-
-    user_text.submit(run_generation, [
-                     user_text, top_p, temperature, top_k, max_new_tokens], model_output)
-    button_submit.click(run_generation, [
-                        user_text, top_p, temperature, top_k, max_new_tokens], model_output)
+            factual = gr.Checkbox(
+                label='Turn on to get factually correct answers')
+
+    # user_text.submit(run_generation, [
+    #                  user_text, top_p, temperature, top_k, max_new_tokens], model_output)
+    # button_submit.click(run_generation, [
+    #                     user_text, top_p, temperature, top_k, max_new_tokens], model_output)
+
+    # user_text.submit(run_generation, [
+    #     user_text, factual, max_new_tokens, temperature], model_output)
+    print('fac', factual.value)
+    button_submit.click(run_generation if factual.value else run_generation_stream, [
+                        user_text, factual, max_new_tokens, temperature], model_output)
 
     demo.queue(max_size=32).launch(enable_queue=True)

requirements.txt

@@ -1,2 +1,4 @@
 text-generation
 gradio
+langchain
+faiss-cpu