Update app.py

app.py

@@ -1,57 +1,47 @@
 import gradio as gr
-from fastai.vision.all import *
 import copy
-import os
 import time
+import llama_cpp
 from llama_cpp import Llama
 from huggingface_hub import hf_hub_download
 
-# Load the LLM model
+
 llm = Llama(
     model_path=hf_hub_download(
         repo_id=os.environ.get("REPO_ID", "TheBloke/Llama-2-7B-Chat-GGML"),
         filename=os.environ.get("MODEL_FILE", "llama-2-7b-chat.ggmlv3.q5_0.bin"),
     ),
     n_ctx=2048,
-    n_gpu_layers=50,  # change n_gpu_layers if you have more or less VRAM
+    n_gpu_layers=50, # change n_gpu_layers if you have more or less VRAM 
 )
 
 history = []
 
 system_message = """
-You are a helpful bird researcher
+You are a helpful, respectful and honest assistant. Always answer as helpfully as possible, while being safe.  Your answers should not include any harmful, unethical, racist, sexist, toxic, dangerous, or illegal content. Please ensure that your responses are socially unbiased and positive in nature.
+If a question does not make any sense, or is not factually coherent, explain why instead of answering something not correct. If you don't know the answer to a question, please don't share false information.
 """
-# The rest of the system message
-
-# Load the Vision Model
-learn = load_learner('export.pkl')
 
-labels = learn.dls.vocab
 
-# Function to generate responses using LLM
 def generate_text(message, history):
     temp = ""
     input_prompt = f"[INST] <<SYS>>\n{system_message}\n<</SYS>>\n\n "
     for interaction in history:
         input_prompt = input_prompt + str(interaction[0]) + " [/INST] " + str(interaction[1]) + " </s><s> [INST] "
 
-    # Truncate or limit the length of the input message
-    max_input_length = 500  # You can adjust this value as needed
-    truncated_message = message[:max_input_length]
-
-    input_prompt = input_prompt + str(truncated_message) + " [/INST] "
+    input_prompt = input_prompt + str(message) + " [/INST] "
 
     output = llm(
         input_prompt,
         temperature=0.15,
         top_p=0.1,
-        top_k=40,
+        top_k=40, 
         repeat_penalty=1.1,
-        max_tokens=4096,
+        max_tokens=1024,
         stop=[
-            "",
-            "",
-            " \n",
+            "<|prompter|>",
+            "<|endoftext|>",
+            "<|endoftext|> \n",
             "ASSISTANT:",
             "USER:",
             "SYSTEM:",
@@ -66,38 +56,24 @@ def generate_text(message, history):
     history = ["init", input_prompt]
 
 
-# Function to predict using the Vision Model and interact with LLM
 def predict(img):
     img = PILImage.create(img)
-    pred, pred_idx, probs = learn.predict(img)
-    bird_predictions = {labels[i]: float(probs[i]) for i in range(len(labels))}
-    
-    # Construct a message for LLM using bird predictions
-    message = "I have detected:\n"
-    for bird, prob in bird_predictions.items():
-        message += f"- {bird}: {prob:.2%}\n"
-    
-    # Generate responses using LLM
-    responses = list(generate_text(message, history))
-    
-    return {"bird_predictions": bird_predictions, "llm_responses": responses}
-
-title = "Bird Detector with LLM"
-description = "Detect birds and interact with LLM."
-examples = [['BIRD.png']]
-interpretation = 'default'
-enable_queue = True
-
-gr.Interface(
-    fn=predict,
-    inputs=gr.inputs.Image(),
-    outputs=[
-        gr.outputs.Label(label="Bird Predictions"),
-        gr.outputs.Textbox(label="LLM Responses")
-    ],
-    title=title,
-    description=description,
-    examples=examples,
-    interpretation=interpretation,
-    enable_queue=enable_queue,
-).launch()
+    pred,pred_idx,probs = learn.predict(img)
+    return {labels[i]: float(probs[i]) for i in range(len(labels))}
+
+title = "Bird Detector"
+description = "Bird Detector."
+examples = ['BIRD.png']
+interpretation='default'
+enable_queue=True
+
+def combined(img, message):
+    prediction = predict(img)
+    response = generate_text(message, history)
+    if "I have detected" in response:
+        response = response.replace("I have detected", f"I have detected {prediction['bird']} in the image.")
+
+    return response
+
+
+gr.Interface(fn=combined,inputs=gr.inputs.Image(shape=(512, 512)),outputs=gr.outputs.Text(),title=title,description=description,examples=examples,interpretation=interpretation,enable_queue=enable_queue).launch()