Update app.py

app.py

@@ -1,42 +1,32 @@
 import gradio as gr
 from datasets import load_dataset
-import numpy as np
+
 import os
 import spaces
 from transformers import AutoModelForCausalLM, AutoTokenizer, TextIteratorStreamer, BitsAndBytesConfig
 import torch
 from threading import Thread
 from sentence_transformers import SentenceTransformer
-from datasets import load_dataset
-import time
+import numpy as np
 
 token = os.environ["HF_TOKEN"]
 ST = SentenceTransformer("mixedbread-ai/mxbai-embed-large-v1")
 
 dataset = load_dataset("Yoxas/statistical_literacyv2")
-data = dataset["train"]
-
-# Convert the list to a numpy array
-embeddings_array = np.array(data["Abstract_Embeddings"])
 
-# Check the shape of the embeddings
-print(embeddings_array.shape)
+data = dataset["train"]
 
-new_data = data.remove_columns("Abstract_Embeddings")
-embeddings_array = [np.fromstring(embedding.replace('[', '').replace(']', '').replace('\n', ''), sep=' ') for embedding in data["Abstract_Embeddings"]]
-embeddings_array = np.array(embeddings_array, dtype=np.float32)
-if len(embeddings_array.shape) == 1:
-    embeddings_array = embeddings_array.reshape(-1, 1)
-embeddings_array = np.ascontiguousarray(embeddings_array)
-new_data = new_data.add_column("Abstract_Embeddings", embeddings_array.tolist())
+# Check the structure of embeddings
+example_embedding = data[0]['Abstract_Embeddings']
+print(f"Example embedding shape: {np.array(example_embedding).shape}")
 
-# Convert the column to a numpy array
-new_data.set_format("numpy", columns=["Abstract_Embeddings"])
+# Ensure embeddings are 2-dimensional
+def ensure_2d_embeddings(embeddings):
+    return [np.atleast_2d(embedding) for embedding in embeddings]
 
-# Initialize the Faiss index
-new_data.add_faiss_index("Abstract_Embeddings")
+# Apply the function to ensure embeddings are 2-dimensional
+data = data.map(lambda example: {'Abstract_Embeddings': ensure_2d_embeddings(example['Abstract_Embeddings'])})
 
-# Now you can use the Dataset with the Faiss index
 data = data.add_faiss_index("Abstract_Embeddings")
 
 model_id = "meta-llama/Meta-Llama-3-8B-Instruct"
@@ -46,7 +36,7 @@ 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)
+tokenizer = AutoTokenizer.from_pretrained(model_id, token=token)
 model = AutoModelForCausalLM.from_pretrained(
     model_id,
     torch_dtype=torch.bfloat16,
@@ -56,58 +46,55 @@ model = AutoModelForCausalLM.from_pretrained(
 )
 terminators = [
     tokenizer.eos_token_id,
-    tokenizer.convert_tokens_to_ids("<|eot_id|>")
+    tokenizer.convert_tokens_to_ids("")
 ]
 
 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."""
 
-
-
-def search(query: str, k: int = 3 ):
+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
+    embedded_query = ST.encode(query)  # embed new query
+    scores, retrieved_examples = data.get_nearest_examples(  # retrieve results
+        "Abstract_Embeddings", embedded_query,  # compare our new embedded query with the dataset embeddings
+        k=k  # get only top k results
     )
     return scores, retrieved_examples
 
-def format_prompt(prompt,retrieved_documents,k):
+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"
+    for idx in range(k):
+        PROMPT += f"{retrieved_documents['text'][idx]}\n"
     return PROMPT
 
-
 @spaces.GPU(duration=150)
-def talk(prompt,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}]
+def talk(prompt, 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"
+        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,
+        input_ids,
+        max_new_tokens=1024,
+        eos_token_id=terminators,
+        do_sample=True,
+        temperature=0.6,
+        top_p=0.9,
     )
     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(
-        input_ids= input_ids,
+        input_ids=input_ids,
         streamer=streamer,
         max_new_tokens=1024,
         do_sample=True,
@@ -124,7 +111,6 @@ def talk(prompt,history):
         print(outputs)
         yield "".join(outputs)
 
-
 TITLE = "# RAG"
 
 DESCRIPTION = """
@@ -136,7 +122,6 @@ Resources used to build this project :
 * chatbot : https://huggingface.co/meta-llama/Meta-Llama-3-8B-Instruct
 """
 
-
 demo = gr.ChatInterface(
     fn=talk,
     chatbot=gr.Chatbot(
@@ -151,6 +136,5 @@ demo = gr.ChatInterface(
     examples=[["what's anarchy ? "]],
     title=TITLE,
     description=DESCRIPTION,
-    
 )
 demo.launch(debug=True)