Code Update: 11 out of 20 correct.

agents.py

@@ -46,7 +46,7 @@ def create_web_agent():
                      tools = [FinalAnswerTool(),
                               GoogleSearchTool(),
                               DuckDuckGoSearchTool(),
-                              VisitWebpageTool(max_output_length = 75000),
+                              VisitWebpageTool(max_output_length = 100000),
                               WikipediaSearchTool(user_agent = "FinalAssignmentResearchBot ([email protected])",
                                                   language = "en",
                                                   content_type = "text",
@@ -97,6 +97,6 @@ def create_manager_agent():
                      planning_interval = 3,
                      verbosity_level = 2,
                      stream_outputs = True,
-                     max_steps = 20,
+                     max_steps = 25,
                      provide_run_summary = True,
                      managed_agents = [vision_agent, web_agent])

app.py

@@ -117,7 +117,7 @@ def run_and_submit_all( profile: gr.OAuthProfile | None):
             # Run Manager Agent
             submitted_answer = manager_agent.run(get_manager_agent_prompt(question_text, file_prompt))
 
-            # Basic verification...convert both to string...            
+            # Basic verification...convert both to string...shouldn't happen as output...but occassionally does ;-)         
             if type(submitted_answer) is list or type(submitted_answer) is dict:
                 submitted_answer = str(submitted_answer)
             

tooling.py

@@ -1,5 +1,3 @@
-# https://github.com/huggingface/smolagents/blob/v1.17.0/src/smolagents/default_tools.py#L479
-
 # Import Modules
 import os
 import pandas as pd
@@ -10,11 +8,9 @@ import re
 import torch
 from transformers import AutoProcessor, AutoModelForVision2Seq
 from smolagents import tool, Tool
-from smolagents.tools import PipelineTool
 from transformers import WhisperProcessor, WhisperForConditionalGeneration
 import librosa
 import numpy as np
- 
 
 gradio_main_instructions =  """
 **Instructions:**
@@ -126,12 +122,14 @@ def vision_language_tool(question: str, file_name: str) -> str:
     """
     
     prompt = f"""
+# Objective:
 You are provided with an image.
 
-Answer the following question about the image very specifically and in detail:
+Answer the following question about the image very specifically and in detail. Think step by step.
 
-{question}"""
-    print(f"vlt: {os.listdir('./')}")
+# Question:
+{question}
+"""
     conversation = [
         {
             "role": "user",
@@ -145,13 +143,13 @@ Answer the following question about the image very specifically and in detail:
                                                   return_tensors = "pt").to(device)
 
 
-    # autoregressively complete prompt
+    # Generate
     model_output = vision_model.generate(**inputs, 
-                                         max_new_tokens = 1024,
-                                         temperature = 0.2,
+                                         max_new_tokens = 2048,
+                                         temperature = 0.5,
                                          do_sample = True,
-                                         top_p = 0.975,
-                                         top_k = 75,
+                                         top_p = 0.98,
+                                         top_k = 80,
                                          min_p = 0.05,
                                          repetition_penalty = 1.15)
     answer = vision_processor.decode(model_output[0], skip_special_tokens = True)