🌘w🌖

app.py

@@ -3,7 +3,7 @@ from datasets import load_dataset
 
 import os
 import spaces
-from transformers import AutoModelForCausalLM, AutoTokenizer, TextIteratorStreamer
+from transformers import AutoModelForCausalLM, AutoTokenizer, TextIteratorStreamer, BitsAndBytesConfig
 import torch
 from threading import Thread
 from sentence_transformers import SentenceTransformer
@@ -11,107 +11,156 @@ from datasets import load_dataset
 import time
 
 token = os.environ["HF_TOKEN"]
+ST = SentenceTransformer("mixedbread-ai/mxbai-embed-large-v1")
+
+dataset = load_dataset("not-lain/wikipedia",revision = "embedded")
+
+data = dataset["train"]
+data = data.add_faiss_index("embeddings") # column name that has the embeddings of the dataset
+
+
+model_id = "meta-llama/Meta-Llama-3-8B-Instruct"
+
+# use quantization to lower GPU usage
+bnb_config = BitsAndBytesConfig(
+    load_in_4bit=True, bnb_4bit_use_double_quant=True, bnb_4bit_quant_type="nf4", bnb_4bit_compute_dtype=torch.bfloat16
+)
+
+tokenizer = AutoTokenizer.from_pretrained(model_id,token=token)
 model = AutoModelForCausalLM.from_pretrained(
-    "google/gemma-7b-it",
-    # torch_dtype=torch.float16 if torch.cuda.is_available() else torch.float32,
-    torch_dtype=torch.float16,
-    token=token,
+    model_id,
+    torch_dtype=torch.bfloat16,
+    device_map="auto",
+    quantization_config=bnb_config,
+    token=token
 )
-tokenizer = AutoTokenizer.from_pretrained("google/gemma-7b-it", token=token)
-device = torch.device("cuda")
-model = model.to(device)
-RAG = SentenceTransformer("mixedbread-ai/mxbai-embed-large-v1")
-TOP_K = 1
-HEADER = "\n# RESOURCES:\n"
-# prepare data
-# since data is too big we will only select the first 3K lines
+terminators = [
+    tokenizer.eos_token_id,
+    tokenizer.convert_tokens_to_ids("<|eot_id|>")
+]
 
-data = load_dataset("not-lain/wikipedia-small-3000-embedded", split="train")
+SYS_PROMPT = """You are an assistant for answering questions.
+You are given the extracted parts of a long document and a question. Provide a conversational answer.
+If you don't know the answer, just say "I do not know." Don't make up an answer."""
 
-# index dataset
-data.add_faiss_index("embedding")
 
 
-def search(query: str, k: int = TOP_K):
-    embedded_query = RAG.encode(query)
-    scores, retrieved_examples = data.get_nearest_examples(
-        "embedding", embedded_query, k=k
+def search(query: str, k: int = 3 ):
+    """a function that embeds a new query and returns the most probable results"""
+    embedded_query = ST.encode(query) # embed new query
+    scores, retrieved_examples = data.get_nearest_examples( # retrieve results
+        "embeddings", embedded_query, # compare our new embedded query with the dataset embeddings
+        k=k # get only top k results
     )
-    return retrieved_examples
+    return scores, retrieved_examples
 
-
-def prepare_prompt(query, retrieved_examples):
-    prompt = (
-        f"Query: {query}\nContinue to answer the query in short sentences by using the Search Results:\n"
-    )
-    urls = []
-    titles = retrieved_examples["title"][::-1]
-    texts = retrieved_examples["text"][::-1]
-    urls = retrieved_examples["url"][::-1]
-    titles = titles[::-1]
-    for i in range(TOP_K):
-        prompt += f"* {texts[i]}\n"
-    return prompt, zip(titles, urls)
+def format_prompt(prompt,retrieved_documents,k):
+    """using the retrieved documents we will prompt the model to generate our responses"""
+    PROMPT = f"Question:{prompt}\nContext:"
+    for idx in range(k) :
+        PROMPT+= f"{retrieved_documents['text'][idx]}\n"
+    return PROMPT
 
 
 @spaces.GPU(duration=150)
-def talk(message, history):
-    print("history, ", history)
-    print("message ", message)
-    print("searching dataset ...")
-    retrieved_examples = search(message)
-    print("preparing prompt ...")
-    message, metadata = prepare_prompt(message, retrieved_examples)
-    resources = HEADER
-    print("preparing metadata ...")
-    for title, url in metadata:
-        resources += f"[{title}]({url}),  "
-    print("preparing chat template ...")
-    chat = []
-    for item in history:
-        chat.append({"role": "user", "content": item[0]})
-        cleaned_past = item[1].split(HEADER)[0]
-        chat.append({"role": "assistant", "content": cleaned_past})
-    chat.append({"role": "user", "content": message})
-    messages = tokenizer.apply_chat_template(
-        chat, tokenize=False, add_generation_prompt=True
+def talk(message,history):
+    k = 1 # number of retrieved documents
+    scores , retrieved_documents = search(prompt, k)
+    formatted_prompt = format_prompt(prompt,retrieved_documents,k)
+    formatted_prompt = formatted_prompt[:2000] # to avoid GPU OOM
+    messages = [{"role":"system","content":SYS_PROMPT},{"role":"user","content":formatted_prompt}]
+    # tell the model to generate
+    input_ids = tokenizer.apply_chat_template(
+      messages,
+      add_generation_prompt=True,
+      return_tensors="pt"
+    ).to(model.device)
+    outputs = model.generate(
+      input_ids,
+      max_new_tokens=1024,
+      eos_token_id=terminators,
+      do_sample=True,
+      temperature=0.6,
+      top_p=0.9,
     )
-    print("chat template prepared, ", messages)
-    print("tokenizing input ...")
-    # Tokenize the messages string
-    model_inputs = tokenizer([messages], return_tensors="pt").to(device)
     streamer = TextIteratorStreamer(
-        tokenizer, timeout=10.0, skip_prompt=True, skip_special_tokens=True
-    )
+            tokenizer, timeout=10.0, skip_prompt=True, skip_special_tokens=True
+        )
     generate_kwargs = dict(
-        model_inputs,
+        input_ids= input_ids,
         streamer=streamer,
         max_new_tokens=1024,
         do_sample=True,
         top_p=0.95,
-        top_k=1000,
         temperature=0.75,
-        num_beams=1,
+        eos_token_id=terminators,
     )
-    print("initializing thread ...")
     t = Thread(target=model.generate, kwargs=generate_kwargs)
     t.start()
-    time.sleep(1)
-    # Initialize an empty string to store the generated text
-    partial_text = ""
-    i = 0
-    while t.is_alive():
-        try:
-            for new_text in streamer:
-                if new_text is not None:
-                    partial_text += new_text
-                    yield partial_text
-        except Exception as e:
-            print(f"retry number {i}\n LOGS:\n")
-            i+=1
-            print(e, e.args)
-    partial_text += resources
-    yield partial_text
+
+    outputs = []
+    for text in streamer:
+        outputs.append(text)
+        print(outputs)
+        yield "".join(outputs)
+    
+# def talk(message, history):
+#     print("history, ", history)
+#     print("message ", message)
+#     print("searching dataset ...")
+#     retrieved_examples = search(message)
+#     print("preparing prompt ...")
+#     message, metadata = prepare_prompt(message, retrieved_examples)
+#     resources = HEADER
+#     print("preparing metadata ...")
+#     for title, url in metadata:
+#         resources += f"[{title}]({url}),  "
+#     print("preparing chat template ...")
+#     chat = []
+#     for item in history:
+#         chat.append({"role": "user", "content": item[0]})
+#         cleaned_past = item[1].split(HEADER)[0]
+#         chat.append({"role": "assistant", "content": cleaned_past})
+#     chat.append({"role": "user", "content": message})
+#     messages = tokenizer.apply_chat_template(
+#         chat, tokenize=False, add_generation_prompt=True
+#     )
+#     print("chat template prepared, ", messages)
+#     print("tokenizing input ...")
+#     # Tokenize the messages string
+#     model_inputs = tokenizer([messages], return_tensors="pt").to(device)
+#     streamer = TextIteratorStreamer(
+#         tokenizer, timeout=10.0, skip_prompt=True, skip_special_tokens=True
+#     )
+    # generate_kwargs = dict(
+    #     model_inputs,
+    #     streamer=streamer,
+    #     max_new_tokens=1024,
+    #     do_sample=True,
+    #     top_p=0.95,
+    #     top_k=1000,
+    #     temperature=0.75,
+    #     num_beams=1,
+    # )
+#     print("initializing thread ...")
+#     t = Thread(target=model.generate, kwargs=generate_kwargs)
+#     t.start()
+#     time.sleep(1)
+#     # Initialize an empty string to store the generated text
+#     partial_text = ""
+#     i = 0
+#     while t.is_alive():
+#         try:
+#             for new_text in streamer:
+#                 if new_text is not None:
+#                     partial_text += new_text
+#                     yield partial_text
+#         except Exception as e:
+#             print(f"retry number {i}\n LOGS:\n")
+#             i+=1
+#             print(e, e.args)
+#     partial_text += resources
+#     yield partial_text
 
 
 TITLE = "# RAG"
@@ -122,11 +171,9 @@ A rag pipeline with a chatbot feature
 Resources used to build this project :
 
 * embedding model : https://huggingface.co/mixedbread-ai/mxbai-embed-large-v1
-* dataset : https://huggingface.co/datasets/not-lain/wikipedia-small-3000-embedded (used mxbai-colbert-large-v1 to create the embedding column )
+* dataset : https://huggingface.co/datasets/not-lain/wikipedia
 * faiss docs : https://huggingface.co/docs/datasets/v2.18.0/en/package_reference/main_classes#datasets.Dataset.add_faiss_index 
-* chatbot : https://huggingface.co/google/gemma-7b-it
-
-If you want to support my work consider clicking on the heart react button ❤️🤗
+* chatbot : https://huggingface.co/meta-llama/Meta-Llama-3-8B-Instruct
 """