|
|
| """
|
| NeuroReasoner Chain-of-Thought GUI (Final Version)
|
| -----------------------------------------------
|
| A highly refined and premium Streamlit app for exploring
|
| step-by-step reasoning across compatible Hugging Face models,
|
| with enhanced model loading logic, functional multimodal input support,
|
| fixed Streamlit state issues, and preserving all previous features.
|
|
|
| Featuring an advanced, futuristic, layered dark theme, dynamic model-type detection,
|
| enhanced self-consistency sampling & voting display, robust handling with improved feedback,
|
| detailed Chain-of-Thought output presentation including code rendering,
|
| integrated GPU telemetry, and accessibility improvements.
|
|
|
| This version prioritizes loading specific model types (CausalLM, Seq2SeqLM)
|
| for better generation compatibility, falls back to AutoModel, adds a check
|
| for generation capability after loading, includes a minor fix for telemetry,
|
| resolves the Streamlit expander nesting issue by restructuring output display,
|
| ensures full output display, and incorporates extensive futuristic styling.
|
| """
|
| import os
|
| import time
|
| import re
|
| import streamlit as st
|
| import torch
|
| import pynvml
|
| import numpy as np
|
| import logging
|
|
|
| from transformers import (
|
| AutoConfig,
|
| AutoTokenizer,
|
|
|
| AutoModelForCausalLM,
|
| AutoModelForSeq2SeqLM,
|
|
|
| AutoModel,
|
| AutoProcessor,
|
| GenerationConfig,
|
| PretrainedConfig,
|
|
|
|
|
|
|
| GenerationMixin,
|
| )
|
| from collections import Counter
|
| import gc
|
| from typing import Any, Dict, List, Optional, Tuple, Union
|
|
|
| from PIL import Image
|
| import io
|
|
|
|
|
|
|
| logger = logging.getLogger(__name__)
|
|
|
| if not logger.root.handlers:
|
| handler = logging.StreamHandler()
|
| formatter = logging.Formatter('%(asctime)s - %(name)s - %(levelname)s - %(message)s')
|
| handler.setFormatter(formatter)
|
| logger.addHandler(handler)
|
| logger.propagate = False
|
| logger.setLevel(logging.INFO)
|
|
|
|
|
|
|
|
|
|
|
| try:
|
|
|
| from chain_of_thought_wrapper import ChainOfThoughtWrapper
|
| logger.info("Successfully imported ChainOfThoughtWrapper.")
|
| except ImportError:
|
| logger.error("🚨 Fatal Error: `chain_of_thought_wrapper.py` not found. Please ensure the enhanced wrapper script is in the same directory.")
|
| st.error("🚨 Fatal Error: The `chain_of_thought_wrapper.py` script was not found. Please ensure it's in the same directory as the GUI script.")
|
| st.stop()
|
|
|
|
|
|
|
| st.set_page_config(
|
| page_title="🧠 NeuroReasoner CoT (Multimodal)",
|
| page_icon="🧠",
|
| layout="wide",
|
| initial_sidebar_state="expanded",
|
| menu_items={
|
| 'Get Help': 'https://github.com/ayjays132/NeuroReasoner',
|
| 'Report a bug': "https://github.com/ayjays132/NeuroReasoner/issues",
|
| 'About': """
|
| **NeuroReasoner Chain-of-Thought GUI (Final Version)**
|
| Version 2.4 (Futuristic Styling & Output Fix)
|
|
|
| An open-source interface for exploring Chain-of-Thought reasoning with Hugging Face models,
|
| now with enhanced model loading logic, functional multimodal (image + text) input support,
|
| fixed Streamlit state issues, and preserving all previous features.
|
|
|
| \n\n**Key Features:**
|
| - **Futuristic, Layered, Magical Dark Theme**
|
| - Real-time GPU Telemetry
|
| - Efficient Model Caching (including processor)
|
| - Enhanced Model Loading Strategy (Prioritizes CausalLM/Seq2SeqLM)
|
| - Self-Consistency Voting
|
| - Detailed CoT Output Visualization (Markdown/Code support, **Full Output Display Fixed**)
|
| - Robust Generation Controls
|
| - Accessibility improvements
|
| - Functional Multimodal (Image + Text) Input (Requires compatible model)
|
| - Integrated AGI Helper Module Interactions (Conceptual)
|
| - Corrected Streamlit Session State Handling
|
| - **Fixed Expander Nesting Issue**
|
|
|
| **Note on Multimodal Models:** While this version supports multimodal input, you must load
|
| a model from Hugging Face that is designed for multimodal tasks (e.g., LLaVA, InstructBLIP, SmolVLM)
|
| and is compatible with Hugging Face's `AutoProcessor` and model classes designed for generation.
|
| Loading a text-only model will disable the image processing logic for that session. Loading
|
| a model that does not have a language modeling head (like the base `GPT2Model` instead of
|
| `GPT2LMHeadModel`) will result in a generation error. This version attempts to load
|
| generation-compatible classes more robustly.
|
|
|
| Developed by [Your Name/Org Here - Optional]
|
| Built with Streamlit, Hugging Face Transformers, and pynvml.
|
| """
|
| }
|
| )
|
|
|
|
|
| st.markdown("""
|
| <style>
|
| /* Basic Reset and Body/App Styling */
|
| html, body, [data-testid="stAppViewContainer"] {
|
| margin: 0 !important;
|
| padding: 0 !important;
|
| box-sizing: border-box !important;
|
| font-family: 'Segoe UI', Roboto, Arial, sans-serif !important;
|
| color: #EDF0F5 !important; /* Refined off-white text */
|
| line-height: 1.62 !important;
|
| letter-spacing: 0.018rem !important;
|
| }
|
|
|
| .stApp {
|
| background: linear-gradient(160deg, #050508 0%, #0A0A10 50%, #0D0D15 100%); /* More complex gradient */
|
| color: #EDF0F5; /* Refined off-white text with slight blue tint */
|
| font-family: 'Segoe UI', Roboto, Arial, sans-serif;
|
| padding: 0;
|
| margin: 0;
|
| letter-spacing: 0.018rem; /* Enhanced letter spacing */
|
| line-height: 1.62;
|
| }
|
| .stApp::before {
|
| content: "";
|
| position: absolute;
|
| top: 0;
|
| left: 0;
|
| width: 100%;
|
| height: 100%;
|
| background: url("data:image/svg+xml,%3Csvg width='100' height='100' viewBox='0 0 100 100' xmlns='http://www.w3.org/2000/svg'%3E%3Cpath d='M11 18c3.866 0 7-3.134 7-7s-3.134-7-7-7-7 3.134-7 7 3.134 7 7 7zm48 25c3.866 0 7-3.134 7-7s-3.134-7-7-7-7 3.134-7 7 3.134 7 7 7zm-43-7c1.657 0 3-1.343 3-3s-1.343-3-3-3-3 1.343-3 3 1.343 3 3 3zm63 31c1.657 0 3-1.343 3-3s-1.343-3-3-3-3 1.343-3 3 1.343 3 3 3zM34 90c1.657 0 3-1.343 3-3s-1.343-3-3-3-3 1.343-3 3 1.343 3 3 3zm56-76c1.657 0 3-1.343 3-3s-1.343-3-3-3-3 1.343-3 3 1.343 3 3 3zM12 86c2.21 0 4-1.79 4-4s-1.79-4-4-4-4 1.79-4 4 1.79 4 4 4zm28-65c2.21 0 4-1.79 4-4s-1.79-4-4-4-4 1.79-4 4 1.79 4 4 4zm23-11c2.76 0 5-2.24 5-5s-2.24-5-5-5-5 2.24-5 5 2.24 5 5 5zm-6 60c2.21 0 4-1.79 4-4s-1.79-4-4-4-4 1.79-4 4 1.79 4 4 4zm29 22c2.76 0 5-2.24 5-5s-2.24-5-5-5-5 2.24-5 5 2.24 5 5 5zM32 63c2.76 0 5-2.24 5-5s-2.24-5-5-5-5 2.24-5 5 2.24 5 5 5zm57-13c2.76 0 5-2.24 5-5s-2.24-5-5-5-5 2.24-5 5 2.24 5 5 5zm-9-21c1.105 0 2-.895 2-2s-.895-2-2-2-2 .895-2 2 .895 2 2 2zM60 91c1.105 0 2-.895 2-2s-.895-2-2-2-2 .895-2 2 .895 2 2 2zM35 41c1.105 0 2-.895 2-2s-.895-2-2-2-2 .895-2 2 .895 2 2 2zM12 60c1.105 0 2-.895 2-2s-.895-2-2-2-2 .895-2 2 .895 2 2 2z' fill='%230E1E3A' fill-opacity='0.05' fill-rule='evenodd'/%3E%3C/svg%3E");
|
| pointer-events: none;
|
| }
|
|
|
| /* Main Content Area */
|
| .stApp > header:first-child,
|
| .stApp > div:has(div[data-testid="stSidebarContent"]) + div,
|
| .stApp > div:first-child > div:has(div[data-testid="stVerticalBlock"]) > div:first-child {
|
| padding: 2rem 3rem 3.5rem 3rem !important; /* Expanded padding */
|
| position: relative !important;
|
| z-index: 1 !important; /* Above background pattern */
|
| backdrop-filter: blur(3px) !important; /* Subtle blur for depth */
|
| }
|
|
|
| /* Hide Streamlit Header */
|
| .stAppHeader {
|
| display: none !important;
|
| }
|
|
|
| /* Futuristic Sidebar */
|
| .stSidebar {
|
| background: linear-gradient(180deg, #0D0D15 0%, #101020 50%, #131325 100%) !important; /* Darker, more intense gradient */
|
| padding: 2.5rem 1.75rem !important;
|
| border-right: 1px solid #1F1F33 !important; /* Stronger border */
|
| position: relative !important;
|
| color: #E9EDF0 !important;
|
| box-shadow: 5px 0 30px rgba(0, 0, 0, 0.8) !important; /* Deeper shadow */
|
| z-index: 10 !important; /* Ensure sidebar is on top */
|
| }
|
| /* Glowing edge for sidebar */
|
| .stSidebar::before {
|
| content: "";
|
| position: absolute;
|
| top: 0;
|
| bottom: 0;
|
| right: 0;
|
| width: 2px; /* Wider glow edge */
|
| background: linear-gradient(to bottom,
|
| transparent,
|
| rgba(0, 180, 255, 0.4) 20%, /* Brighter blue glow */
|
| rgba(0, 180, 255, 0.7) 50%,
|
| rgba(0, 180, 255, 0.4) 80%,
|
| transparent) !important;
|
| filter: blur(1px); /* Soften the glow */
|
| }
|
| /* Subtle radial highlight in sidebar */
|
| .stSidebar::after {
|
| content: "";
|
| position: absolute;
|
| top: 0;
|
| left: 0;
|
| width: 100%;
|
| height: 100%;
|
| background: radial-gradient(circle at top left, rgba(0, 180, 255, 0.04), transparent 70%) !important;
|
| pointer-events: none;
|
| }
|
| /* Sidebar Headers */
|
| .stSidebar h1, .stSidebar h2, .stSidebar h3, .stSidebar h4 {
|
| color: #00BFFF !important; /* Brighter, more vibrant blue */
|
| margin-top: 0.75rem !important;
|
| margin-bottom: 1.25rem !important;
|
| position: relative !important;
|
| padding-bottom: 0.75rem !important;
|
| text-shadow: 0px 0px 5px rgba(0, 180, 255, 0.3) !important; /* More prominent text glow */
|
| }
|
| /* Sidebar Header Underlines */
|
| .stSidebar h1::after, .stSidebar h2::after {
|
| content: "";
|
| position: absolute;
|
| left: 0;
|
| bottom: 0;
|
| width: 60%; /* Wider underline */
|
| height: 3px; /* Thicker underline */
|
| background: linear-gradient(90deg, #00BFFF, transparent) !important;
|
| box-shadow: 0px 0px 6px rgba(0, 180, 255, 0.5) !important; /* Stronger glowing underline */
|
| }
|
| .stSidebar h3::after, .stSidebar h4::after {
|
| content: "";
|
| position: absolute;
|
| left: 0;
|
| bottom: 0;
|
| width: 40%; /* Wider underline */
|
| height: 2px; /* Thicker underline */
|
| background: linear-gradient(90deg, #00BFFF, transparent) !important;
|
| box-shadow: 0px 0px 4px rgba(0, 180, 255, 0.3) !important; /* Stronger glowing underline */
|
| }
|
| /* Sidebar Labels */
|
| .stSidebar label {
|
| color: #8BE9FD !important; /* Cyan-like color */
|
| font-weight: normal !important;
|
| margin-bottom: 0.75rem !important;
|
| display: block !important;
|
| font-size: 0.95rem !important;
|
| letter-spacing: 0.02rem !important;
|
| }
|
| .stSidebar .stSelectbox > label,
|
| .stSidebar .stSlider > label,
|
| .stSidebar .stCheckbox > label,
|
| .stSidebar .stTextInput > label,
|
| .stSidebar .stFileUploader > label {
|
| font-weight: bold !important;
|
| color: #50FA7B !important; /* Green */
|
| text-shadow: 0px 0px 3px rgba(80, 250, 123, 0.4) !important; /* Enhanced glow */
|
| position: relative !important;
|
| padding-left: 0.75rem !important;
|
| margin-top: 1.5rem !important; /* Space above controls */
|
| margin-bottom: 0.5rem !important;
|
| }
|
| /* Vertical accent bar for sidebar labels */
|
| .stSidebar .stSelectbox > label::before,
|
| .stSidebar .stSlider > label::before,
|
| .stSidebar .stCheckbox > label::before,
|
| .stSidebar .stTextInput > label::before,
|
| .stSidebar .stFileUploader > label::before {
|
| content: "";
|
| position: absolute;
|
| left: 0;
|
| top: 0.1rem;
|
| bottom: 0.1rem;
|
| width: 4px; /* Thicker bar */
|
| border-radius: 2px;
|
| background: linear-gradient(to bottom, #50FA7B, rgba(80, 250, 123, 0.4)) !important; /* Gradient vertical bar */
|
| box-shadow: 0px 0px 4px rgba(80, 250, 123, 0.4) !important; /* Glowing accent */
|
| }
|
|
|
|
|
| /* Futuristic Headers */
|
| h1, h2, h3, h4, h5, h6 {
|
| color: #8BE9FD !important; /* Cyan-like color */
|
| margin-top: 2rem !important;
|
| margin-bottom: 1.25rem !important;
|
| padding-bottom: 0.5rem !important;
|
| border-bottom: none !important;
|
| letter-spacing: -0.015rem !important;
|
| position: relative !important;
|
| text-shadow: 0px 0px 3px rgba(139, 233, 253, 0.2) !important; /* Subtle text glow */
|
| }
|
| h1 {
|
| font-size: 3rem !important; /* Larger main title */
|
| color: #FF79C6 !important; /* Pink */
|
| padding-bottom: 1rem !important;
|
| margin-bottom: 2.5rem !important;
|
| text-shadow: 0px 1px 6px rgba(255, 121, 198, 0.3) !important; /* Enhanced text glow */
|
| }
|
| /* Main Title Underline/Accent */
|
| h1::before {
|
| content: "";
|
| position: absolute;
|
| bottom: 0;
|
| left: 0;
|
| width: 100%;
|
| height: 4px; /* Thicker line */
|
| background: linear-gradient(to right, #FF79C6, #BD93F9, transparent 95%) !important; /* Pink to purple gradient */
|
| border-radius: 3px !important;
|
| box-shadow: 0px 0px 8px rgba(255, 121, 198, 0.5) !important; /* Glowing border */
|
| }
|
| h1::after {
|
| content: "";
|
| position: absolute;
|
| bottom: -10px; /* Further below */
|
| left: 0;
|
| width: 50%; /* Wider secondary line */
|
| height: 2px; /* Thicker secondary line */
|
| background: linear-gradient(to right, rgba(255, 121, 198, 0.6), transparent) !important;
|
| border-radius: 1px !important;
|
| }
|
| h2 {
|
| font-size: 2.4rem !important; /* Larger secondary header */
|
| background: linear-gradient(90deg, #50FA7B, #8BE9FD) !important; /* Green to Cyan gradient */
|
| -webkit-background-clip: text !important;
|
| background-clip: text !important;
|
| color: transparent !important;
|
| margin-top: 3.5rem !important; /* More space above */
|
| border-bottom: 1px solid #3A3A4C !important; /* Darker separator */
|
| padding-bottom: 1rem !important; /* More padding */
|
| margin-bottom: 2.25rem !important; /* More space below */
|
| }
|
| /* H2 Underline/Accent */
|
| h2::before {
|
| content: "";
|
| position: absolute;
|
| bottom: -1px;
|
| left: 0;
|
| width: 160px; /* Wider accent */
|
| height: 3px; /* Thicker accent */
|
| background: linear-gradient(90deg, #50FA7B, rgba(80, 250, 123, 0.3)) !important;
|
| border-radius: 2px !important;
|
| box-shadow: 0px 0px 5px rgba(80, 250, 123, 0.4) !important; /* Glowing accent */
|
| }
|
| h2::after {
|
| content: "";
|
| position: absolute;
|
| bottom: -5px; /* Further below */
|
| left: 0;
|
| width: 80px; /* Wider secondary */
|
| height: 1px;
|
| background: linear-gradient(90deg, rgba(80, 250, 123, 0.5), transparent) !important;
|
| }
|
| h3 {
|
| font-size: 2rem !important; /* Larger h3 */
|
| color: #BD93F9 !important; /* Purple */
|
| margin-top: 2.5rem !important;
|
| margin-bottom: 1.75rem !important;
|
| position: relative !important;
|
| padding-left: 1.5rem !important; /* More padding */
|
| text-shadow: 0px 0px 3px rgba(189, 147, 249, 0.3) !important; /* Enhanced glow */
|
| }
|
| /* H3 Vertical Bar Accent */
|
| h3::before {
|
| content: "";
|
| position: absolute;
|
| left: 0;
|
| top: 0.1rem;
|
| bottom: 0.1rem;
|
| width: 5px; /* Thicker bar */
|
| background: linear-gradient(to bottom, #BD93F9, rgba(189, 147, 249, 0.4)) !important; /* Gradient vertical bar */
|
| border-radius: 3px !important;
|
| box-shadow: 0px 0px 5px rgba(189, 147, 249, 0.4) !important; /* Glowing accent */
|
| }
|
| h4 {
|
| font-size: 1.6rem !important; /* Larger h4 */
|
| background: linear-gradient(90deg, #FFB86C, #FF79C6) !important; /* Orange to Pink gradient */
|
| -webkit-background-clip: text !important;
|
| background-clip: text !important;
|
| color: transparent !important;
|
| margin-top: 2rem !important;
|
| margin-bottom: 1.25rem !important;
|
| position: relative !important;
|
| padding-left: 1rem !important; /* More padding */
|
| }
|
| /* H4 Vertical Bar Accent */
|
| h4::before {
|
| content: "";
|
| position: absolute;
|
| left: 0;
|
| top: 0.2rem;
|
| bottom: 0.2rem;
|
| width: 4px; /* Thicker bar */
|
| background: linear-gradient(to bottom, #FFB86C, rgba(255, 184, 108, 0.4)) !important; /* Gradient vertical bar */
|
| border-radius: 2px !important;
|
| }
|
| h5 {
|
| font-size: 1.4rem !important;
|
| color: #D4D4E8 !important;
|
| margin-top: 1.8rem !important;
|
| margin-bottom: 1rem !important;
|
| border-bottom: 1px dashed #4F4F68 !important; /* Stronger dashed border */
|
| padding-bottom: 0.5rem !important;
|
| position: relative !important;
|
| padding-left: 0.5rem !important;
|
| }
|
| h5::before {
|
| content: "◊"; /* Diamond accent */
|
| position: absolute;
|
| left: -0.5rem;
|
| top: 0.5rem;
|
| color: #BD93F9 !important; /* Purple accent */
|
| text-shadow: 0px 0px 3px rgba(189, 147, 249, 0.3) !important;
|
| font-size: 1em;
|
| }
|
| h6 {
|
| font-size: 1.2rem !important;
|
| color: #B0B0C0 !important;
|
| margin-top: 1.2rem !important;
|
| margin-bottom: 0.6rem !important;
|
| position: relative !important;
|
| padding-left: 0.3rem !important;
|
| }
|
| h6::before {
|
| content: "▪"; /* Square accent */
|
| position: absolute;
|
| left: -0.3rem;
|
| top: 0.4rem;
|
| color: #FFB86C !important; /* Orange accent */
|
| text-shadow: 0px 0px 2px rgba(255, 184, 108, 0.3) !important;
|
| font-size: 0.8em;
|
| }
|
|
|
|
|
| /* Futuristic Button */
|
| .stButton>button {
|
| background: linear-gradient(180deg, #50FA7B 0%, #48E772 50%, #40D067 100%) !important; /* Vibrant green gradient */
|
| color: #0A0A12 !important; /* Very dark text */
|
| border: 1px solid #3ACD5A !important;
|
| border-radius: 0.6rem !important; /* Slightly larger border radius */
|
| padding: 1rem 2.5rem !important; /* More padding */
|
| font-size: 1.1rem !important; /* Slightly larger font */
|
| font-weight: 700 !important; /* Bolder */
|
| transition: all 0.3s cubic-bezier(0.165, 0.84, 0.44, 1) !important; /* Smoother easing */
|
| box-shadow:
|
| 0px 3px 0px #2F994A, /* Deeper bottom shadow */
|
| 0px 5px 15px rgba(0, 0, 0, 0.6), /* Stronger outer shadow */
|
| inset 0px 1px 0px rgba(255, 255, 255, 0.2) !important; /* Enhanced inner highlight */
|
| margin-top: 2.5rem !important;
|
| letter-spacing: 0.8px !important; /* More spacing */
|
| position: relative !important;
|
| overflow: hidden !important;
|
| text-shadow: 0px 1px 3px rgba(0, 0, 0, 0.3) !important; /* Deeper text shadow */
|
| }
|
| /* Button Top Highlight */
|
| .stButton>button::before {
|
| content: "";
|
| position: absolute;
|
| top: 0;
|
| left: 0;
|
| width: 100%;
|
| height: 100%;
|
| background: linear-gradient(to bottom, rgba(255, 255, 255, 0.15), transparent) !important;
|
| opacity: 0.8;
|
| }
|
| /* Button Hover State */
|
| .stButton>button:hover {
|
| background: linear-gradient(180deg, #69FF94 0%, #50FA7B 50%, #48E772 100%) !important; /* Brighter gradient */
|
| border-color: #50FA7B !important;
|
| transform: translateY(-3px) scale(1.02) !important; /* More pronounced lift */
|
| box-shadow:
|
| 0px 5px 0px #3ACD5A, /* Match hover lift */
|
| 0px 8px 18px rgba(0, 0, 0, 0.7), /* Stronger hover shadow */
|
| inset 0px 1px 0px rgba(255, 255, 255, 0.3) !important; /* Brighter inner highlight */
|
| }
|
| /* Button Active State */
|
| .stButton>button:active {
|
| background: linear-gradient(180deg, #3A8E51 0%, #40D067 50%, #40D067 100%) !important;
|
| border-color: #40D067 !important;
|
| transform: translateY(1px) scale(0.98) !important;
|
| box-shadow:
|
| 0px 1px 0px #2F994A,
|
| 0px 3px 8px rgba(0, 0, 0, 0.6),
|
| inset 0px 1px 4px rgba(0, 0, 0, 0.4) !important; /* Deeper inner shadow */
|
| }
|
| /* Button Shine Effect */
|
| .stButton>button::after {
|
| content: "";
|
| position: absolute;
|
| top: -50%;
|
| left: -50%;
|
| width: 200%;
|
| height: 200%;
|
| opacity: 0;
|
| transform: rotate(30deg);
|
| background: linear-gradient(
|
| to right,
|
| rgba(255, 255, 255, 0) 0%,
|
| rgba(255, 255, 255, 0.4) 50%, /* Brighter shine */
|
| rgba(255, 255, 255, 0) 100%
|
| );
|
| transition: opacity 0.8s ease, transform 0.8s ease !important; /* Slower, smoother transition */
|
| }
|
| .stButton>button:hover::after {
|
| opacity: 1;
|
| transform: rotate(30deg) translate(60%, 60%) !important; /* More pronounced shine movement */
|
| }
|
|
|
|
|
| /* Futuristic Text Areas and Inputs */
|
| div[data-baseweb="textarea"] textarea,
|
| div[data-baseweb="input"] input,
|
| .stTextInput>div>div>input,
|
| .stSelectbox>div>div[data-baseweb="select"]>div,
|
| .stFileUploader>div>div {
|
| border: 1px solid #3F3F55 !important; /* Stronger border */
|
| border-radius: 0.6rem !important; /* Slightly larger radius */
|
| padding: 1.1rem 1.35rem !important; /* More padding */
|
| font-size: 1.05rem !important; /* Slightly larger font */
|
| background-color: #1A1A28 !important; /* Dark bluish background */
|
| color: #E9EDF0 !important;
|
| box-shadow:
|
| inset 1px 1px 10px rgba(0, 0, 0, 0.4), /* Deeper inner shadow */
|
| 0px 1px 2px rgba(255, 255, 255, 0.05),
|
| 0px 0px 0px 2px rgba(63, 63, 85, 0.6) !important; /* Wider outer glow */
|
| width: 100% !important;
|
| line-height: 1.6 !important;
|
| transition: all 0.3s cubic-bezier(0.165, 0.84, 0.44, 1) !important;
|
| position: relative !important;
|
| }
|
| .stSelectbox>div>div[data-baseweb="select"]>div {
|
| padding: 0.8rem 1.35rem !important; /* Adjust padding for select */
|
| }
|
| .stFileUploader>div>div {
|
| padding: 1rem !important; /* Adjust padding for file uploader drop area */
|
| border: 1px dashed #556270 !important; /* Dashed border for uploader */
|
| background-color: #1A1A28 !important;
|
| box-shadow: inset 1px 1px 10px rgba(0, 0, 0, 0.4), 0px 2px 8px rgba(0, 0, 0, 0.3);
|
| }
|
| .stFileUploader>div>div>button {
|
| font-size: 0.95rem !important;
|
| padding: 0.6rem 1rem !important;
|
| margin-top: 1rem !important;
|
| background: linear-gradient(180deg, #BD93F9 0%, #B08FE5 100%) !important; /* Purple gradient */
|
| color: #0A0A12 !important;
|
| border: 1px solid #A989D2 !important;
|
| border-radius: 0.4rem !important;
|
| box-shadow: 0px 2px 0px #7A639B, 0px 4px 8px rgba(0, 0, 0, 0.4);
|
| }
|
| .stFileUploader>div>div>button:hover {
|
| background: linear-gradient(180deg, #D7AEFB 0%, #BD93F9 100%) !important;
|
| border-color: #BD93F9 !important;
|
| transform: translateY(-1px);
|
| box-shadow: 0px 3px 0px #7A639B, 0px 5px 10px rgba(0, 0, 0, 0.5);
|
| }
|
| .stFileUploader>div>div>button:active {
|
| transform: translateY(1px);
|
| box-shadow: 0px 1px 0px #7A639B, 0px 2px 5px rgba(0, 0, 0, 0.4);
|
| }
|
|
|
|
|
| /* Input/Select Top Edge Highlight */
|
| div[data-baseweb="textarea"]::after,
|
| div[data-baseweb="input"]::after,
|
| .stTextInput>div>div::after,
|
| .stSelectbox>div>div[data-baseweb="select"]::after {
|
| content: "";
|
| position: absolute;
|
| top: 0;
|
| left: 0;
|
| width: 100%;
|
| height: 2px; /* Thicker highlight */
|
| background: linear-gradient(90deg, transparent, rgba(255, 255, 255, 0.1), transparent) !important;
|
| pointer-events: none;
|
| }
|
| /* Futuristic focus state with glow */
|
| div[data-baseweb="textarea"] textarea:focus,
|
| div[data-baseweb="input"] input:focus,
|
| .stTextInput>div>div>input:focus,
|
| .stSelectbox>div>div[data-baseweb="select"]:focus-within { /* Use :focus-within for selectbox */
|
| border-color: #00BFFF !important; /* Brighter blue glow color */
|
| background-color: #1F1F30 !important; /* Lighter when focused */
|
| box-shadow:
|
| inset 1px 1px 6px rgba(0, 0, 0, 0.3), /* Deeper inner shadow */
|
| 0 0 0 3px rgba(0, 180, 255, 0.35), /* Wider, more intense outer glow */
|
| 0 0 12px rgba(0, 180, 255, 0.2) !important; /* Ambient glow */
|
| outline: none !important;
|
| }
|
|
|
| /* Futuristic Status Box */
|
| [data-testid="stStatus"] {
|
| background: linear-gradient(135deg, #1A1A28, #28283D) !important; /* Richer gradient */
|
| border: 1px solid #40405A !important; /* Stronger border */
|
| border-radius: 0.7rem !important; /* Larger radius */
|
| padding: 1.8rem !important; /* More padding */
|
| margin-bottom: 2.5rem !important; /* More space below */
|
| box-shadow:
|
| 3px 4px 20px rgba(0, 0, 0, 0.45), /* Stronger shadow */
|
| 0px 1px 0px rgba(255, 255, 255, 0.08) inset,
|
| 0px 0px 0px 2px rgba(64, 64, 90, 0.8) !important; /* Wider outer glow */
|
| position: relative !important;
|
| overflow: hidden !important;
|
| }
|
| /* Status Box Top Accent */
|
| [data-testid="stStatus"]::before {
|
| content: "";
|
| position: absolute;
|
| top: 0;
|
| left: 0;
|
| width: 100%;
|
| height: 4px; /* Thicker accent */
|
| background: linear-gradient(90deg, #00BFFF, #FF79C6) !important; /* Blue to Pink gradient */
|
| border-top-left-radius: 0.7rem !important;
|
| border-top-right-radius: 0.7rem !important;
|
| box-shadow: 0px 0px 10px rgba(0, 180, 255, 0.6) !important; /* Stronger glow */
|
| }
|
| /* Status Box Internal Lighting */
|
| [data-testid="stStatus"]::after {
|
| content: "";
|
| position: absolute;
|
| top: 0;
|
| left: 0;
|
| width: 100%;
|
| height: 100%;
|
| background:
|
| radial-gradient(circle at top right, rgba(255, 121, 198, 0.06), transparent 70%), /* Pink highlight */
|
| linear-gradient(135deg, rgba(255, 255, 255, 0.04), transparent) !important; /* Glass effect */
|
| pointer-events: none;
|
| }
|
| [data-testid="stStatus"] .stMarkdown p,
|
| [data-testid="stStatus"] .stAlert {
|
| color: #D4D4E8 !important;
|
| background-color: transparent !important;
|
| border: none !important;
|
| padding: 0.3rem 0 !important; /* More padding */
|
| position: relative !important;
|
| z-index: 1 !important;
|
| }
|
| [data-testid="stStatus"] .stProgress {
|
| margin-top: 1.5rem !important; /* More space */
|
| margin-bottom: 1rem !important;
|
| position: relative !important;
|
| }
|
| [data-testid="stStatus"] .stProgress > div > div {
|
| background: linear-gradient(90deg, #00BFFF, #50FA7B) !important; /* Blue to Green gradient */
|
| box-shadow: 0px 0px 8px rgba(0, 180, 255, 0.5) !important; /* Glowing progress bar */
|
| height: 8px !important; /* Taller */
|
| border-radius: 4px !important;
|
| }
|
| [data-testid="stStatus"] .stProgress::before {
|
| content: "";
|
| position: absolute;
|
| top: 0;
|
| left: 0;
|
| width: 100%;
|
| height: 100%;
|
| background-color: rgba(20, 20, 30, 0.4) !important; /* Darker track */
|
| border-radius: 4px !important;
|
| z-index: -1;
|
| }
|
|
|
| /* Futuristic Alert Boxes */
|
| .stAlert {
|
| border-radius: 0.8rem !important; /* Larger radius */
|
| margin-bottom: 2.5rem !important; /* More space */
|
| padding: 1.6rem 1.8rem !important; /* More padding */
|
| font-size: 1.05rem !important;
|
| border-left: 6px solid transparent !important; /* Thicker border */
|
| color: #E9EDF0 !important;
|
| box-shadow:
|
| 3px 4px 20px rgba(0, 0, 0, 0.5),
|
| 0px 0px 0px 1px rgba(64, 64, 90, 0.6) !important; /* Subtle outer glow */
|
| position: relative !important;
|
| overflow: hidden !important;
|
| backdrop-filter: blur(5px) !important; /* More blur for glass */
|
| background-color: rgba(30, 30, 45, 0.6) !important; /* Semi-transparent background */
|
| }
|
| /* Alert Glass Effect */
|
| .stAlert::before {
|
| content: "";
|
| position: absolute;
|
| top: 0;
|
| left: 0;
|
| width: 100%;
|
| height: 100%;
|
| background:
|
| linear-gradient(135deg, rgba(255, 255, 255, 0.08), transparent), /* Stronger highlight */
|
| radial-gradient(circle at top left, rgba(255, 255, 255, 0.05), transparent 70%) !important;
|
| pointer-events: none;
|
| }
|
| /* Alert Top Accent Line */
|
| .stAlert::after {
|
| content: "";
|
| position: absolute;
|
| top: 0;
|
| left: 6px; /* Account for border-left */
|
| width: calc(100% - 6px);
|
| height: 2px; /* Thicker line */
|
| background: linear-gradient(90deg, rgba(255, 255, 255, 0.15), transparent) !important;
|
| pointer-events: none;
|
| }
|
| .stAlert.stAlert-info {
|
| border-left-color: #00BFFF !important; /* Brighter blue */
|
| background-color: rgba(15, 30, 45, 0.6) !important; /* Match border color */
|
| }
|
| .stAlert.stAlert-success {
|
| border-left-color: #50FA7B !important; /* Green */
|
| background-color: rgba(15, 40, 30, 0.6) !important; /* Match border color */
|
| }
|
| .stAlert.stAlert-warning {
|
| border-left-color: #FFB86C !important; /* Orange */
|
| background-color: rgba(40, 35, 15, 0.6) !important; /* Match border color */
|
| }
|
| .stAlert.stAlert-error {
|
| border-left-color: #F44747 !important; /* Red */
|
| background-color: rgba(45, 20, 25, 0.6) !important; /* Match border color */
|
| }
|
|
|
|
|
| /* Futuristic Expander */
|
| .streamlit-expanderHeader {
|
| background: linear-gradient(90deg, #1A1A28, #28283D) !important; /* Richer gradient */
|
| color: #E0E0EA !important;
|
| border-radius: 0.7rem !important; /* Larger radius */
|
| padding: 1.4rem 1.8rem !important; /* More padding */
|
| margin-top: 2rem !important; /* More space above */
|
| margin-bottom: 0.8rem !important; /* More space below */
|
| font-weight: 700 !important; /* Bolder */
|
| font-size: 1.2rem !important; /* Larger font */
|
| cursor: pointer !important;
|
| transition: all 0.3s cubic-bezier(0.165, 0.84, 0.44, 1) !important;
|
| box-shadow:
|
| 0px 3px 10px rgba(0, 0, 0, 0.4),
|
| 0px 0px 0px 2px rgba(64, 64, 90, 0.7) !important; /* Wider outer glow */
|
| position: relative !important;
|
| border: 1px solid #3F3F55 !important; /* Stronger border */
|
| overflow: hidden !important;
|
| backdrop-filter: blur(4px) !important; /* Glass effect */
|
| }
|
| /* Expander Header Glass Effect */
|
| .streamlit-expanderHeader::before {
|
| content: "";
|
| position: absolute;
|
| top: 0;
|
| left: 0;
|
| width: 100%;
|
| height: 100%;
|
| background:
|
| linear-gradient(135deg, rgba(255, 255, 255, 0.05), transparent),
|
| radial-gradient(circle at top right, rgba(189, 147, 249, 0.03), transparent 70%) !important; /* Purple highlight */
|
| pointer-events: none;
|
| opacity: 0.8;
|
| transition: opacity 0.3s ease !important;
|
| }
|
| /* Expander Header Hover State */
|
| .streamlit-expanderHeader:hover {
|
| background: linear-gradient(90deg, #28283D, #3A3A4C) !important;
|
| color: #F0F0F5 !important;
|
| box-shadow:
|
| 0px 4px 12px rgba(0, 0, 0, 0.5),
|
| 0px 0px 0px 2px rgba(189, 147, 249, 0.3) !important; /* Purple glow on hover */
|
| transform: translateY(-2px) !important;
|
| }
|
| .streamlit-expanderHeader:hover::before {
|
| opacity: 1 !important;
|
| }
|
| /* Expander Header Arrow */
|
| .streamlit-expanderHeader::after {
|
| content: "";
|
| position: absolute;
|
| right: 1.8rem; /* More space */
|
| top: 50%;
|
| transform: translateY(-50%);
|
| width: 14px; /* Larger arrow */
|
| height: 14px; /* Larger arrow */
|
| border-right: 3px solid #BD93F9 !important; /* Thicker, purple arrow */
|
| border-bottom: 3px solid #BD93F9 !important; /* Thicker, purple arrow */
|
| transform-origin: 75% 75%;
|
| transition: transform 0.4s cubic-bezier(0.68, -0.55, 0.27, 1.55) !important; /* Bouncier animation */
|
| box-shadow: 2px 2px 5px rgba(0, 0, 0, 0.3) !important; /* Stronger shadow on arrow */
|
| }
|
| .streamlit-expanderHeader[aria-expanded="true"]::after {
|
| transform: translateY(-50%) rotate(45deg) scale(1.15) !important; /* Larger scale */
|
| }
|
| .streamlit-expanderHeader[aria-expanded="false"]::after {
|
| transform: translateY(-50%) rotate(-45deg) scale(1.15) !important;
|
| }
|
|
|
| /* Futuristic Expander Content */
|
| .streamlit-expanderContent {
|
| background: linear-gradient(135deg, #1A1A28, #232335) !important; /* Subtle gradient */
|
| border: 1px solid #3F3F55 !important;
|
| border-top: none !important;
|
| border-bottom-left-radius: 0.7rem !important;
|
| border-bottom-right-radius: 0.7rem !important;
|
| padding: 2.5rem !important; /* More padding */
|
| margin-top: 0 !important;
|
| color: #E9EDF0 !important;
|
| box-shadow:
|
| inset 0 3px 12px rgba(0, 0, 0, 0.3), /* Deeper inner shadow */
|
| 0px 4px 12px rgba(0, 0, 0, 0.3) !important; /* Outer shadow */
|
| position: relative !important;
|
| overflow: hidden !important;
|
| backdrop-filter: blur(4px) !important; /* Match header blur */
|
| }
|
| /* Expander Content Top Separator */
|
| .streamlit-expanderContent::before {
|
| content: "";
|
| position: absolute;
|
| top: 0;
|
| left: 5%;
|
| width: 90%;
|
| height: 2px; /* Thicker separator */
|
| background: linear-gradient(90deg, transparent, rgba(189, 147, 249, 0.4), transparent) !important; /* Purple gradient separator */
|
| box-shadow: 0px 0px 6px rgba(189, 147, 249, 0.3) !important; /* Glowing separator */
|
| }
|
| /* Expander Content Internal Lighting */
|
| .streamlit-expanderContent::after {
|
| content: "";
|
| position: absolute;
|
| top: 0;
|
| left: 0;
|
| width: 100%;
|
| height: 100%;
|
| background:
|
| radial-gradient(circle at bottom right, rgba(80, 250, 123, 0.04), transparent 70%), /* Green highlight */
|
| linear-gradient(135deg, rgba(255, 255, 255, 0.03), transparent) !important; /* Glass effect */
|
| pointer-events: none;
|
| }
|
| /* Text within expander content */
|
| .streamlit-expanderContent .stMarkdown p,
|
| .streamlit-expanderContent .stMarkdown li {
|
| color: #E9EDF0 !important;
|
| margin-bottom: 1rem !important; /* More space */
|
| line-height: 1.8 !important; /* Increased line height */
|
| letter-spacing: 0.01rem !important;
|
| }
|
| /* Futuristic code blocks inside expanders */
|
| .streamlit-expanderContent .stMarkdown pre code {
|
| background: linear-gradient(135deg, #0D0D15, #101020) !important; /* Darker gradient */
|
| border: 1px solid #3F3F55 !important;
|
| border-radius: 0.7rem !important; /* Larger radius */
|
| padding: 1.8rem !important; /* More padding */
|
| margin: 1.2rem 0 !important; /* More margin */
|
| overflow-x: auto !important;
|
| color: #BD93F9 !important; /* Purple code text */
|
| font-family: 'Fira Code', Consolas, Monaco, 'Andale Mono', 'Ubuntu Mono', monospace !important; /* Use Fira Code if available */
|
| line-height: 1.7 !important;
|
| font-size: 0.95rem !important;
|
| display: block !important;
|
| word-wrap: normal !important;
|
| white-space: pre-wrap !important;
|
| box-shadow:
|
| inset 0 0 20px rgba(0, 0, 0, 0.4) !important, /* Deeper inner shadow */
|
| 0px 4px 15px rgba(0, 0, 0, 0.3) !important; /* Enhanced outer shadow */
|
| position: relative !important;
|
| backdrop-filter: blur(2px); /* Slight blur for code block */
|
| }
|
| .streamlit-expanderContent .stMarkdown pre code::before {
|
| content: "";
|
| position: absolute;
|
| top: 0;
|
| left: 0;
|
| width: 100%;
|
| height: 100%;
|
| background:
|
| linear-gradient(135deg, rgba(255, 255, 255, 0.02), transparent) !important,
|
| radial-gradient(circle at top right, rgba(189, 147, 249, 0.03), transparent 70%) !important; /* Purple highlight */
|
| pointer-events: none;
|
| border-radius: 0.7rem !important;
|
| }
|
| /* Futuristic inline code inside expanders */
|
| .streamlit-expanderContent .stMarkdown code {
|
| background: linear-gradient(135deg, #28283D, #3F3F55) !important; /* Gradient background */
|
| border-radius: 0.5rem !important; /* Larger radius */
|
| padding: 0.3em 0.6em !important; /* More padding */
|
| color: #FFB86C !important; /* Orange inline code */
|
| font-family: 'Fira Code', Consolas, Monaco, 'Andale Mono', 'Ubuntu Mono', monospace !important;
|
| font-size: 0.95rem !important;
|
| border: 1px solid rgba(255, 184, 108, 0.3) !important; /* Subtle orange border */
|
| box-shadow: 0px 1px 4px rgba(0, 0, 0, 0.3) !important; /* Subtle shadow */
|
| }
|
|
|
| /* Ultra Premium Telemetry Box */
|
| .telemetry-box {
|
| background: linear-gradient(135deg, #1F1F30, #2A2A40) !important; /* Richer gradient */
|
| border: 1px solid #4F4F68 !important; /* Stronger border */
|
| border-radius: 0.8rem !important; /* Larger radius */
|
| padding: 1.5rem !important; /* More padding */
|
| margin-top: 2.5rem !important; /* More space */
|
| font-size: 0.9rem !important;
|
| color: #B0B0C0 !important;
|
| text-align: center !important;
|
| box-shadow:
|
| 3px 4px 20px rgba(0, 0, 0, 0.45),
|
| 0px 0px 0px 2px rgba(79, 79, 104, 0.8) !important; /* Wider outer glow */
|
| position: relative !important;
|
| overflow: hidden !important;
|
| backdrop-filter: blur(3px) !important; /* Glass effect */
|
| }
|
| /* Telemetry Box Glass Effect */
|
| .telemetry-box::before {
|
| content: "";
|
| position: absolute;
|
| top: 0;
|
| left: 0;
|
| width: 100%;
|
| height: 100%;
|
| background:
|
| linear-gradient(135deg, rgba(255, 255, 255, 0.04), transparent),
|
| radial-gradient(circle at bottom right, rgba(139, 233, 253, 0.03), transparent 70%) !important; /* Cyan highlight */
|
| pointer-events: none;
|
| }
|
| /* Telemetry Box Top Accent Line */
|
| .telemetry-box::after {
|
| content: "";
|
| position: absolute;
|
| top: 0;
|
| left: 0;
|
| width: 100%;
|
| height: 2px; /* Thicker line */
|
| background: linear-gradient(90deg, transparent, rgba(255, 255, 255, 0.15), transparent) !important;
|
| pointer-events: none;
|
| }
|
| .telemetry-box strong {
|
| color: #F8F8F2 !important; /* Dracular foreground */
|
| font-weight: 700 !important; /* Bolder */
|
| letter-spacing: 0.03rem !important; /* More spacing */
|
| text-shadow: 0px 0px 2px rgba(248, 248, 242, 0.3) !important; /* Stronger text glow */
|
| }
|
|
|
| /* Ultra Premium Self-Consistency Consensus Styling */
|
| .consensus-answer { /* Use hyphenated class name */
|
| background: linear-gradient(135deg, #153323, #1A482C, #206038) !important; /* Richer green gradient */
|
| color: #A3E6CE !important;
|
| border: 1px solid #3E8E56 !important; /* Stronger border */
|
| border-radius: 1rem !important; /* Very large radius */
|
| padding: 2rem !important; /* More padding */
|
| margin-top: 3rem !important; /* More space */
|
| margin-bottom: 2.5rem !important;
|
| font-size: 1.4rem !important; /* Larger font */
|
| font-weight: 700 !important; /* Bolder */
|
| box-shadow:
|
| 0 6px 25px rgba(0, 0, 0, 0.6), /* Stronger shadow */
|
| 0 1px 0 rgba(255, 255, 255, 0.08) inset,
|
| 0px 0px 0px 3px rgba(62, 142, 86, 0.8) !important; /* Wide green outer glow */
|
| display: flex !important;
|
| align-items: center !important;
|
| position: relative !important;
|
| overflow: hidden !important;
|
| backdrop-filter: blur(6px) !important; /* More blur for glass */
|
| background-color: rgba(21, 51, 35, 0.7) !important; /* Semi-transparent green */
|
| }
|
| /* Consensus Answer Left Accent */
|
| .consensus-answer::before {
|
| content: "";
|
| position: absolute;
|
| top: 0;
|
| left: 0;
|
| width: 8px; /* Wider accent */
|
| height: 100%;
|
| background: linear-gradient(to bottom, #69FF94, #3E8E56) !important; /* Brighter green gradient */
|
| box-shadow: 0px 0px 10px rgba(105, 255, 148, 0.6) !important; /* Stronger glowing accent */
|
| border-top-left-radius: 1rem !important;
|
| border-bottom-left-radius: 1rem !important;
|
| }
|
| /* Consensus Answer Glass Effect */
|
| .consensus-answer::after {
|
| content: "";
|
| position: absolute;
|
| top: 0;
|
| left: 0;
|
| width: 100%;
|
| height: 100%;
|
| background:
|
| linear-gradient(135deg, rgba(255, 255, 255, 0.1), transparent), /* Stronger highlight */
|
| radial-gradient(circle at bottom right, rgba(105, 255, 148, 0.05), transparent 70%) !important;
|
| pointer-events: none;
|
| }
|
| .consensus-answer strong {
|
| color: #8BE9FD !important; /* Cyan label */
|
| margin-right: 1.5rem !important; /* More space */
|
| flex-shrink: 0 !important;
|
| position: relative !important;
|
| padding-right: 1.5rem !important;
|
| text-shadow: 0px 0px 4px rgba(139, 233, 253, 0.4) !important; /* Stronger text glow */
|
| }
|
| /* Consensus Answer Label Separator */
|
| .consensus-answer strong::after {
|
| content: "";
|
| position: absolute;
|
| right: 0;
|
| top: 15%; /* Adjust position */
|
| height: 70%; /* Adjust height */
|
| width: 2px; /* Thicker separator */
|
| background: linear-gradient(to bottom, transparent, rgba(139, 233, 253, 0.8), transparent) !important;
|
| box-shadow: 0px 0px 4px rgba(139, 233, 253, 0.5) !important;
|
| }
|
| .consensus-answer p {
|
| color: #F8F8F2 !important; /* Dracular foreground */
|
| margin: 0 !important;
|
| padding: 0 !important;
|
| flex-grow: 1 !important;
|
| word-break: break-word !important;
|
| line-height: 1.8 !important;
|
| letter-spacing: 0.01rem !important;
|
| }
|
|
|
|
|
| /* Ultra Premium Vote Counts Styling */
|
| .vote-counts {
|
| background: linear-gradient(135deg, #1F1F30, #2A2A40) !important; /* Richer gradient */
|
| border: 1px solid #4F4F68 !important;
|
| border-radius: 0.8rem !important; /* Larger radius */
|
| padding: 1.8rem !important; /* More padding */
|
| margin-top: 2.5rem !important; /* More space */
|
| margin-bottom: 2.5rem !important;
|
| font-size: 1rem !important;
|
| box-shadow:
|
| 3px 4px 20px rgba(0, 0, 0, 0.45),
|
| 0px 0px 0px 2px rgba(79, 79, 104, 0.7) !important; /* Wider outer glow */
|
| position: relative !important;
|
| overflow: hidden !important;
|
| backdrop-filter: blur(5px) !important; /* Glass effect */
|
| background-color: rgba(30, 30, 45, 0.6) !important; /* Semi-transparent */
|
| }
|
| /* Vote Counts Top Accent */
|
| .vote-counts::before {
|
| content: "";
|
| position: absolute;
|
| top: 0;
|
| left: 0;
|
| width: 100%;
|
| height: 4px; /* Thicker accent */
|
| background: linear-gradient(90deg, #8BE9FD, transparent) !important; /* Cyan gradient */
|
| box-shadow: 0px 0px 8px rgba(139, 233, 253, 0.5) !important; /* Glowing top border */
|
| }
|
| /* Vote Counts Glass Effect */
|
| .vote-counts::after {
|
| content: "";
|
| position: absolute;
|
| top: 0;
|
| left: 0;
|
| width: 100%;
|
| height: 100%;
|
| background:
|
| linear-gradient(135deg, rgba(255, 255, 255, 0.04), transparent),
|
| radial-gradient(circle at top right, rgba(139, 233, 253, 0.03), transparent 70%) !important; /* Cyan highlight */
|
| pointer-events: none;
|
| }
|
| .vote-counts strong {
|
| color: #8BE9FD !important; /* Cyan label */
|
| display: block !important;
|
| margin-bottom: 1.2rem !important; /* More space */
|
| font-weight: 700 !important; /* Bolder */
|
| letter-spacing: 0.02rem !important;
|
| position: relative !important;
|
| padding-bottom: 0.6rem !important;
|
| text-shadow: 0px 0px 3px rgba(139, 233, 253, 0.3) !important; /* Text glow */
|
| }
|
| /* Vote Counts Label Underline */
|
| .vote-counts strong::after {
|
| content: "";
|
| position: absolute;
|
| bottom: 0;
|
| left: 0;
|
| width: 70px; /* Wider underline */
|
| height: 2px; /* Thicker underline */
|
| background: linear-gradient(90deg, rgba(139, 233, 253, 0.7), transparent) !important;
|
| }
|
| .vote-counts ul {
|
| padding-left: 2.5rem !important; /* More padding */
|
| margin-bottom: 1rem !important;
|
| list-style-type: none !important;
|
| position: relative !important;
|
| }
|
| .vote-counts li {
|
| margin-bottom: 0.8rem !important; /* More space */
|
| color: #D4D4E8 !important;
|
| position: relative !important;
|
| padding-left: 0.6rem !important; /* More padding */
|
| line-height: 1.7 !important;
|
| }
|
| /* Vote Counts Custom Bullets */
|
| .vote-counts li::before {
|
| content: "•"; /* Custom bullet */
|
| position: absolute;
|
| left: -1.4rem; /* Adjust position */
|
| color: #BD93F9 !important; /* Purple bullets */
|
| text-shadow: 0px 0px 3px rgba(189, 147, 249, 0.4) !important; /* Glowing bullets */
|
| font-size: 1em;
|
| }
|
|
|
|
|
| /* Ultra Premium Chat Message Styling */
|
| [data-testid="chatContainer"] [data-testid="stChatMessage"] {
|
| background: linear-gradient(135deg, #1A1A28, #232335) !important; /* Richer gradient */
|
| border-radius: 1rem !important; /* Larger radius */
|
| padding: 1.8rem 2.2rem !important; /* More padding */
|
| margin-bottom: 2rem !important; /* More space */
|
| border: 1px solid #3F3F55 !important; /* Stronger border */
|
| box-shadow:
|
| 0 6px 22px rgba(0, 0, 0, 0.5), /* Stronger shadow */
|
| 0px 0px 0px 2px rgba(63, 63, 85, 0.8) !important; /* Wider outer glow */
|
| position: relative !important;
|
| transition: all 0.3s cubic-bezier(0.165, 0.84, 0.44, 1) !important;
|
| overflow: hidden !important;
|
| backdrop-filter: blur(5px) !important; /* Glass effect */
|
| background-color: rgba(26, 26, 40, 0.6) !important; /* Semi-transparent */
|
| }
|
| /* Chat Message Glass Effect */
|
| [data-testid="chatContainer"] [data-testid="stChatMessage"]::before {
|
| content: "";
|
| position: absolute;
|
| top: 0;
|
| left: 0;
|
| width: 100%;
|
| height: 100%;
|
| background:
|
| linear-gradient(135deg, rgba(255, 255, 255, 0.05), transparent), /* Stronger highlight */
|
| radial-gradient(circle at top right, rgba(255, 255, 255, 0.02), transparent 70%) !important;
|
| pointer-events: none;
|
| opacity: 0.9; /* More visible glass */
|
| transition: opacity 0.3s ease !important;
|
| }
|
| /* Chat Message Hover State */
|
| [data-testid="chatContainer"] [data-testid="stChatMessage"]:hover {
|
| box-shadow:
|
| 0 8px 25px rgba(0, 0, 0, 0.6),
|
| 0px 0px 0px 2px rgba(80, 250, 123, 0.4) !important; /* Green outer glow on hover */
|
| transform: translateY(-3px) !important; /* More pronounced lift */
|
| }
|
| [data-testid="chatContainer"] [data-testid="stChatMessage"]:hover::before {
|
| opacity: 1 !important; /* Full glass effect on hover */
|
| }
|
| /* User vs Assistant message differentiation with enhanced styling */
|
| [data-testid="chatContainer"] [data-testid="stChatMessage"][data-testid="user-message"] {
|
| border-left: 5px solid #BD93F9 !important; /* Thicker purple accent */
|
| }
|
| /* User Message Left Accent */
|
| [data-testid="chatContainer"] [data-testid="stChatMessage"][data-testid="user-message"]::after {
|
| content: "";
|
| position: absolute;
|
| top: 0;
|
| left: -5px; /* Match border width */
|
| height: 100%;
|
| width: 5px; /* Match border width */
|
| background: linear-gradient(to bottom, #BD93F9, rgba(189, 147, 249, 0.4)) !important; /* Gradient border */
|
| box-shadow: 0px 0px 10px rgba(189, 147, 249, 0.6) !important; /* Glowing border */
|
| border-top-left-radius: 1rem !important;
|
| border-bottom-left-radius: 1rem !important;
|
| pointer-events: none;
|
| }
|
| [data-testid="chatContainer"] [data-testid="stChatMessage"]:not([data-testid="user-message"]) {
|
| border-left: 5px solid #69FF94 !important; /* Thicker green accent */
|
| }
|
| /* Assistant Message Left Accent */
|
| [data-testid="chatContainer"] [data-testid="stChatMessage"]:not([data-testid="user-message"])::after {
|
| content: "";
|
| position: absolute;
|
| top: 0;
|
| left: -5px; /* Match border width */
|
| height: 100%;
|
| width: 5px; /* Match border width */
|
| background: linear-gradient(to bottom, #69FF94, rgba(105, 255, 148, 0.4)) !important; /* Gradient border */
|
| box-shadow: 0px 0px 10px rgba(105, 255, 148, 0.6) !important; /* Glowing border */
|
| border-top-left-radius: 1rem !important;
|
| border-bottom-left-radius: 1rem !important;
|
| pointer-events: none;
|
| }
|
| /* Enhanced content within the chat message */
|
| [data-testid="chatContainer"] [data-testid="stChatMessage"] .stMarkdown p {
|
| color: #F8F8F2 !important; /* Dracular foreground */
|
| margin-bottom: 1rem !important;
|
| line-height: 1.8 !important;
|
| letter-spacing: 0.01rem !important;
|
| }
|
| /* Futuristic code blocks inside chat messages */
|
| [data-testid="chatContainer"] [data-testid="stChatMessage"] .stMarkdown pre code {
|
| background: linear-gradient(135deg, #0D0D15, #101020) !important; /* Darker gradient */
|
| border: 1px solid #3F3F55 !important;
|
| border-radius: 0.7rem !important;
|
| padding: 1.8rem !important;
|
| margin: 1.2rem 0 !important;
|
| overflow-x: auto !important;
|
| color: #BD93F9 !important; /* Purple code text */
|
| font-family: 'Fira Code', Consolas, Monaco, 'Andale Mono', 'Ubuntu Mono', monospace !important;
|
| line-height: 1.7 !important;
|
| font-size: 0.95rem !important;
|
| display: block !important;
|
| word-wrap: normal !important;
|
| white-space: pre-wrap !important;
|
| box-shadow:
|
| inset 0 0 20px rgba(0, 0, 0, 0.4) !important,
|
| 0px 4px 15px rgba(0, 0, 0, 0.3) !important;
|
| position: relative !important;
|
| backdrop-filter: blur(2px); /* Slight blur */
|
| }
|
| [data-testid="chatContainer"] [data-testid="stChatMessage"] .stMarkdown pre code::before {
|
| content: "";
|
| position: absolute;
|
| top: 0;
|
| left: 0;
|
| width: 100%;
|
| height: 100%;
|
| background:
|
| linear-gradient(135deg, rgba(255, 255, 255, 0.02), transparent) !important,
|
| radial-gradient(circle at top right, rgba(189, 147, 249, 0.03), transparent 70%) !important; /* Purple highlight */
|
| pointer-events: none;
|
| border-radius: 0.7rem !important;
|
| }
|
| /* Futuristic inline code inside chat messages */
|
| [data-testid="chatContainer"] [data-testid="stChatMessage"] code {
|
| background: linear-gradient(135deg, #28283D, #3F3F55) !important; /* Gradient background */
|
| border-radius: 0.5rem !important;
|
| padding: 0.35em 0.7em !important; /* More padding */
|
| color: #FFB86C !important; /* Orange inline code */
|
| font-family: 'Fira Code', Consolas, Monaco, 'Andale Mono', 'Ubuntu Mono', monospace !important;
|
| font-size: 0.95rem !important;
|
| border: 1px solid rgba(255, 184, 108, 0.3) !important; /* Subtle orange border */
|
| box-shadow: 0px 1px 4px rgba(0, 0, 0, 0.3) !important;
|
| }
|
|
|
| /* Ultra Premium File Uploader Styling */
|
| [data-testid="stFileUploader"] label {
|
| color: #FFB86C !important; /* Orange label */
|
| font-weight: 700 !important;
|
| margin-bottom: 1rem !important;
|
| display: block !important;
|
| position: relative !important;
|
| letter-spacing: 0.01rem !important;
|
| text-shadow: 0px 0px 2px rgba(255, 184, 108, 0.3) !important; /* Subtle text glow */
|
| }
|
| /* File Uploader Label Underline */
|
| [data-testid="stFileUploader"] label::after {
|
| content: "";
|
| position: absolute;
|
| left: 0;
|
| bottom: -4px;
|
| width: 50px; /* Wider underline */
|
| height: 2px;
|
| background: linear-gradient(90deg, #FFB86C, transparent) !important; /* Orange gradient */
|
| border-radius: 1px !important;
|
| box-shadow: 0px 0px 5px rgba(255, 184, 108, 0.4) !important; /* Glowing underline */
|
| }
|
| [data-testid="stFileUploader"] div[data-baseweb="file-uploader"] {
|
| background: linear-gradient(135deg, #1A1A28, #232335) !important; /* Gradient background */
|
| border: 2px dashed #556270 !important; /* Thicker dashed border */
|
| border-radius: 0.8rem !important; /* Larger radius */
|
| padding: 2.5rem !important; /* More padding */
|
| text-align: center !important;
|
| color: #B0B0C0 !important;
|
| margin-bottom: 2rem !important;
|
| transition: all 0.3s cubic-bezier(0.165, 0.84, 0.44, 1) !important;
|
| position: relative !important;
|
| overflow: hidden !important;
|
| box-shadow:
|
| inset 0 3px 12px rgba(0, 0, 0, 0.3),
|
| 0px 3px 10px rgba(0, 0, 0, 0.3) !important;
|
| backdrop-filter: blur(4px) !important; /* Glass effect */
|
| }
|
| /* File Uploader Drop Area Glow */
|
| [data-testid="stFileUploader"] div[data-baseweb="file-uploader"]::before {
|
| content: "";
|
| position: absolute;
|
| top: -50%;
|
| left: -50%;
|
| width: 200%;
|
| height: 200%;
|
| background:
|
| radial-gradient(circle, rgba(189, 147, 249, 0.2) 0%, transparent 70%), /* Purple radial glow */
|
| linear-gradient(135deg, rgba(255, 255, 255, 0.04), transparent) !important;
|
| opacity: 0;
|
| transition: opacity 0.6s ease, transform 0.6s ease !important;
|
| transform: scale(0.9);
|
| transform-origin: center;
|
| }
|
| [data-testid="stFileUploader"] div[data-baseweb="file-uploader"]:hover {
|
| border-color: #BD93F9 !important; /* Purple border on hover */
|
| box-shadow:
|
| inset 0 3px 12px rgba(0, 0, 0, 0.4),
|
| 0 0 0 2px rgba(189, 147, 249, 0.4), /* Purple outer glow */
|
| 0px 4px 15px rgba(0, 0, 0, 0.4) !important;
|
| transform: translateY(-2px) !important;
|
| }
|
| [data-testid="stFileUploader"] div[data-baseweb="file-uploader"]:hover::before {
|
| opacity: 1 !important;
|
| transform: scale(1) !important;
|
| }
|
| /* Futuristic uploaded file chip styling */
|
| [data-testid="stFileUploader"] div[data-testid="FileThumb"] {
|
| background: linear-gradient(135deg, #28283D, #3F3F55) !important; /* Gradient background */
|
| border: 1px solid #556270 !important;
|
| border-radius: 0.7rem !important;
|
| padding: 1.2rem !important; /* More padding */
|
| margin-top: 1.5rem !important;
|
| color: #E9EDF0 !important;
|
| box-shadow:
|
| 3px 4px 15px rgba(0, 0, 0, 0.35),
|
| 0px 0px 0px 1px rgba(79, 79, 104, 0.7) !important;
|
| position: relative !important;
|
| overflow: hidden !important;
|
| backdrop-filter: blur(3px); /* Glass effect */
|
| transition: all 0.3s cubic-bezier(0.165, 0.84, 0.44, 1);
|
| }
|
| /* File Chip Glass Effect */
|
| [data-testid="stFileUploader"] div[data-testid="FileThumb"]::before {
|
| content: "";
|
| position: absolute;
|
| top: 0;
|
| left: 0;
|
| width: 100%;
|
| height: 100%;
|
| background: linear-gradient(135deg, rgba(255, 255, 255, 0.05), transparent) !important;
|
| pointer-events: none;
|
| }
|
| [data-testid="stFileUploader"] div[data-testid="FileThumb"] p {
|
| color: #F8F8F2 !important; /* Dracular foreground */
|
| font-size: 0.95rem !important;
|
| position: relative !important;
|
| }
|
| [data-testid="stFileUploader"] div[data-testid="FileThumb"]:hover {
|
| box-shadow:
|
| 3px 4px 18px rgba(0, 0, 0, 0.45),
|
| 0px 0px 0px 1px rgba(80, 250, 123, 0.3) !important; /* Green outer glow */
|
| transform: translateY(-2px) !important;
|
| border-color: #69FF94 !important; /* Brighter green */
|
| }
|
| /* Futuristic image display in chat */
|
| .chat-image {
|
| max-width: 100% !important;
|
| border-radius: 0.8rem !important;
|
| margin: 1.5rem 0 !important; /* More margin */
|
| box-shadow:
|
| 0 8px 25px rgba(0, 0, 0, 0.6),
|
| 0px 0px 0px 2px rgba(63, 63, 85, 0.8) !important; /* Premium outer glow */
|
| border: 1px solid #3F3F55 !important;
|
| transition: all 0.3s cubic-bezier(0.165, 0.84, 0.44, 1) !important;
|
| position: relative !important;
|
| overflow: hidden !important;
|
| }
|
| /* Chat Image Top Accent */
|
| .chat-image::after {
|
| content: "";
|
| position: absolute;
|
| top: 0;
|
| left: 0;
|
| width: 100%;
|
| height: 3px; /* Thicker line */
|
| background: linear-gradient(90deg, rgba(255, 255, 255, 0.15), transparent) !important;
|
| border-top-left-radius: 0.8rem !important;
|
| border-top-right-radius: 0.8rem !important;
|
| pointer-events: none;
|
| }
|
| .chat-image:hover {
|
| transform: scale(1.02) !important;
|
| box-shadow:
|
| 0 12px 30px rgba(0, 0, 0, 0.7),
|
| 0px 0px 0px 2px rgba(105, 255, 148, 0.4) !important; /* Green outer glow on hover */
|
| border-color: #69FF94 !important; /* Brighter green */
|
| }
|
| /* Style for the thumbnail images in chat */
|
| [data-testid="chatContainer"] [data-testid="stChatMessage"] img {
|
| border-radius: 0.6rem !important;
|
| border: 1px solid #3F3F55 !important;
|
| box-shadow: 0 3px 10px rgba(0, 0, 0, 0.3) !important;
|
| transition: all 0.2s ease !important;
|
| }
|
| [data-testid="chatContainer"] [data-testid="stChatMessage"] img:hover {
|
| transform: scale(1.05);
|
| box-shadow: 0 5px 15px rgba(0, 0, 0, 0.4) !important;
|
| border-color: #BD93F9 !important;
|
| }
|
| /* Remove default Streamlit spacing for columns inside chat message */
|
| [data-testid="chatContainer"] [data-testid="stChatMessage"] [data-testid="stHorizontalBlock"] {
|
| gap: 1rem !important;
|
| }
|
|
|
|
|
| </style>
|
| """, unsafe_allow_html=True)
|
|
|
|
|
|
|
|
|
| try:
|
| pynvml.nvmlInit()
|
| GPU_AVAILABLE = True
|
|
|
| GPU_COUNT = pynvml.nvmlDeviceGetCount()
|
| if GPU_COUNT == 0:
|
| GPU_AVAILABLE = False
|
| logger.info("No NVML-compatible GPU devices found.")
|
| else:
|
|
|
| selected_device = st.session_state.get("device_select", "cpu")
|
| if selected_device.startswith("cuda"):
|
| try:
|
|
|
| device_index = int(selected_device.split(":")[-1]) if ":" in selected_device else 0
|
|
|
| if device_index >= GPU_COUNT:
|
| GPU_AVAILABLE = False
|
| logger.warning(f"Selected CUDA device index {device_index} is out of range (Max index: {GPU_COUNT - 1}). Telemetry disabled for this selection.")
|
| else:
|
| pynvml.nvmlDeviceGetHandleByIndex(device_index)
|
| logger.info(f"NVML initialized and ready for device {device_index}. Total GPU(s) found: {GPU_COUNT}")
|
| except Exception as e:
|
| GPU_AVAILABLE = False
|
| logger.error(f"Error accessing NVML handle for device {selected_device}: {e}. Telemetry disabled.")
|
| else:
|
| GPU_AVAILABLE = False
|
| logger.info(f"Selected device '{selected_device}' is not a CUDA device. Telemetry disabled.")
|
|
|
| except Exception as e:
|
| GPU_AVAILABLE = False
|
| logger.error(f"pynvml initialization failed: {e}. GPU Telemetry disabled.")
|
|
|
|
|
|
|
|
|
| telemetry_placeholder = st.sidebar.empty()
|
|
|
|
|
| def update_telemetry():
|
| """Updates the telemetry display in the dedicated placeholder."""
|
| telemetry_text = "📊 System Status: [Initializing...]"
|
| selected_device = st.session_state.get("device_select", "cpu")
|
|
|
| if not GPU_AVAILABLE or not selected_device.startswith("cuda"):
|
| telemetry_text = "📊 System Status: [No Compatible GPU Available or Selected for Telemetry]"
|
| else:
|
| try:
|
|
|
| device_index = int(selected_device.split(":")[-1]) if ":" in selected_device else 0
|
| handle = pynvml.nvmlDeviceGetHandleByIndex(device_index)
|
|
|
|
|
| gpu_name_bytes = pynvml.nvmlDeviceGetName(handle)
|
| try:
|
| gpu_name = gpu_name_bytes.decode('utf-8')
|
| except (UnicodeDecodeError, AttributeError):
|
|
|
| gpu_name = str(gpu_name_bytes)
|
|
|
|
|
| utilization = pynvml.nvmlDeviceGetUtilizationRates(handle)
|
| memory = pynvml.nvmlDeviceGetMemoryInfo(handle)
|
| mem_used_mb = memory.used // 1024**2
|
| mem_total_mb = memory.total // 1024**2
|
| telemetry_text = f"📊 <strong>GPU {device_index} ({gpu_name}):</strong> {utilization.gpu}% Util | {mem_used_mb}/{mem_total_mb} MB VRAM"
|
| except pynvml.NVMLError as e:
|
| telemetry_text = f"📊 System Status: [Telemetry Error: {e}]"
|
| logger.error(f"pynvml error during telemetry update: {e}")
|
| except Exception as e:
|
| telemetry_text = "📊 System Status: [Telemetry Error]"
|
| logger.error(f"Unexpected error during telemetry update: {e}")
|
|
|
|
|
|
|
| telemetry_placeholder.markdown(f'<div class="telemetry-box">{telemetry_text}</div>', unsafe_allow_html=True)
|
|
|
|
|
| update_telemetry()
|
|
|
|
|
|
|
| @st.cache_resource(show_spinner=False)
|
| def _load_model_and_processor_cached(model_name: str, device: str):
|
| """
|
| Loads the model and its associated processor (tokenizer + image processor if multimodal)
|
| using a prioritized approach (CausalLM, Seq2SeqLM, then AutoModel).
|
| This function is cached by Streamlit.
|
| """
|
| logger.info(f"Attempting to load model '{model_name}' and processor on '{device}'...")
|
|
|
| model = None
|
| processor = None
|
| model_config = None
|
| model_loaded_successfully = False
|
|
|
| status_box = st.empty()
|
|
|
| try:
|
|
|
| status_box.write("Fetching model configuration...")
|
| try:
|
| model_config = AutoConfig.from_pretrained(model_name, trust_remote_code=True)
|
| logger.info(f"Model config loaded. Architecture: {model_config.architectures}")
|
| status_box.write("Configuration loaded. Loading processor...")
|
| except Exception as config_e:
|
| logger.warning(f"Could not load model config: {config_e}. Proceeding without config.")
|
| status_box.warning(f"Could not load model config: {config_e}. Proceeding without config.")
|
| model_config = None
|
|
|
|
|
|
|
|
|
| status_box.write("Loading processor...")
|
| try:
|
| processor = AutoProcessor.from_pretrained(model_name, trust_remote_code=True)
|
| logger.info(f"Processor loaded successfully. Processor type: {type(processor)}")
|
| status_box.write("Processor loaded. Loading model weights...")
|
| except Exception as proc_e:
|
| logger.error(f"Failed to load processor: {proc_e}")
|
| status_box.error(f"Failed to load processor for '{model_name}'. Details: {proc_e}")
|
|
|
| if model_config is not None: del model_config
|
| if torch.cuda.is_available(): torch.cuda.empty_cache()
|
| gc.collect()
|
| return None, None, False
|
|
|
|
|
|
|
| torch_dtype = torch.float32
|
| if torch.cuda.is_available() and device.startswith("cuda"):
|
|
|
|
|
|
|
| try:
|
| gpu_major_version = torch.cuda.get_device_properties(device).major
|
| if gpu_major_version >= 8:
|
|
|
| if model_config is not None and hasattr(model_config, 'torch_dtype') and model_config.torch_dtype == torch.bfloat16:
|
| torch_dtype = torch.bfloat16
|
| logger.info("GPU supports bfloat16 and model config specifies bfloat16. Loading with bfloat16.")
|
| elif model_config is not None and model_config.architectures and any(arch.lower() in ["llama", "mistral", "gemma", "qwen", "smol", "idefics", "paligemma"] for arch in model_config.architectures):
|
|
|
| torch_dtype = torch.bfloat16
|
| logger.info("GPU supports bfloat16 and model architecture suggests bfloat16 compatibility. Loading with bfloat16.")
|
| else:
|
| logger.info("GPU supports bfloat16 but model config/architecture doesn't explicitly suggest it. Loading with float32.")
|
| except Exception as dtype_e:
|
| logger.warning(f"Could not determine GPU bfloat16 support or model preference ({dtype_e}). Loading with float32.")
|
| torch_dtype = torch.float32
|
| else:
|
| logger.info("Not on CUDA device. Loading with float32.")
|
|
|
|
|
|
|
|
|
| loading_attempts = [
|
| (AutoModelForCausalLM, "CausalLM"),
|
| (AutoModelForSeq2SeqLM, "Seq2SeqLM"),
|
| (AutoModel, "General AutoModel")
|
| ]
|
|
|
| for auto_model_class, class_name in loading_attempts:
|
| status_box.write(f"Attempting to load model as {class_name}...")
|
| try:
|
| model = auto_model_class.from_pretrained(
|
| model_name,
|
| trust_remote_code=True,
|
| low_cpu_mem_usage=True,
|
| torch_dtype=torch_dtype
|
| )
|
| model_loaded_successfully = True
|
| logger.info(f"Model loaded successfully as {class_name}.")
|
|
|
| if auto_model_class == AutoModel:
|
| if not hasattr(model, 'generate') or not callable(getattr(model, 'generate', None)):
|
| logger.error(f"Model loaded with {class_name} does not have a compatible .generate() method.")
|
| status_box.error(f"Loaded model ({type(model).__name__}) does not have a compatible .generate() method required for text generation. Please try a different model.")
|
| model = None
|
| model_loaded_successfully = False
|
| else:
|
| logger.info(f"Model loaded with {class_name} has a compatible .generate() method.")
|
|
|
| break
|
|
|
| except Exception as e:
|
| logger.warning(f"Failed to load as {class_name}: {e}")
|
| model = None
|
|
|
|
|
|
|
| if not model_loaded_successfully:
|
| logger.error(f"All model loading attempts failed for '{model_name}'.")
|
| status_box.error(f"Failed to load model '{model_name}' using any method. Please check the model ID/path and its compatibility.")
|
|
|
| if model is not None: del model
|
| if processor is not None: del processor
|
| if model_config is not None: del model_config
|
| if torch.cuda.is_available(): torch.cuda.empty_cache()
|
| gc.collect()
|
| return None, processor, False
|
|
|
|
|
|
|
| model.to(device)
|
| model.eval()
|
| logger.info(f"Model moved to {device} and set to eval mode. Model type: {type(model)}")
|
|
|
|
|
|
|
|
|
|
|
| tokenizer = getattr(processor, 'tokenizer', None)
|
| if tokenizer and tokenizer.pad_token_id is not None:
|
|
|
|
|
| try:
|
| if hasattr(model, 'resize_token_embeddings') and hasattr(model, 'get_input_embeddings') and model.get_input_embeddings() is not None and len(tokenizer) > model.get_input_embeddings().weight.size(0):
|
| logger.info(f"Tokenizer size ({len(tokenizer)}) is larger than model embedding size ({model.get_input_embeddings().weight.size(0)}). Resizing model embeddings...")
|
| model.resize_token_embeddings(len(tokenizer))
|
| logger.info("Model embeddings resized successfully.")
|
| elif hasattr(model, 'get_input_embeddings') and model.get_input_embeddings() is not None:
|
| logger.debug("Tokenizer size matches or is smaller than model embedding size. No resize needed.")
|
| else:
|
| logger.warning("Model does not have get_input_embeddings method or embeddings are None. Cannot check or resize embeddings.")
|
| except Exception as resize_e:
|
| logger.warning(f"Failed to check or resize model embeddings: {resize_e}")
|
| status_box.warning(f"Failed to check or resize model embeddings: {resize_e}")
|
| elif tokenizer is None:
|
| logger.warning("Processor does not have a tokenizer. Skipping embedding resize check.")
|
|
|
|
|
|
|
|
|
|
|
| if hasattr(model, 'config'):
|
| model.config.return_dict_in_generate = True
|
|
|
| if hasattr(model.config, 'output_scores'):
|
| model.config.output_scores = True
|
|
|
|
|
|
|
|
|
| has_image_processor = hasattr(processor, 'image_processor') and processor.image_processor is not None
|
| model_architecture_name = type(model).__name__ if model else "Unknown"
|
| processor_architecture_name = type(processor).__name__ if processor else "Unknown"
|
| logger.info(f"Loaded Model Class: {model_architecture_name}")
|
| logger.info(f"Loaded Processor Class: {processor_architecture_name}")
|
| logger.info(f"Image Processor available (Multimodal Capable for Input): {has_image_processor}")
|
|
|
| status_box.empty()
|
|
|
|
|
| return model, processor, has_image_processor
|
|
|
|
|
| except Exception as e:
|
| logger.error(f"❌ Model or Processor loading failed for '{model_name}'. Details: {e}")
|
| status_box = st.empty()
|
| status_box.error(f"❌ Model or Processor loading failed for '{model_name}'. Details: {e}")
|
|
|
| if model is not None: del model
|
| if processor is not None: del processor
|
| if model_config is not None: del model_config
|
| if torch.cuda.is_available(): torch.cuda.empty_cache()
|
| gc.collect()
|
|
|
|
|
| return None, None, False
|
|
|
|
|
|
|
| def safe_load_model_with_status(model_name: str, device: str):
|
| """
|
| Calls the cached loading function and handles Streamlit status updates
|
| and error reporting. Now loads processor and checks multimodal capability.
|
| """
|
|
|
| status_text = f"🌐 Loading model and processor for '{model_name}'..."
|
|
|
| main_loading_status = st.status(status_text, expanded=True)
|
|
|
| with main_loading_status:
|
| main_loading_status.write(f"Initiating model and processor load on device '{device}'...")
|
| update_telemetry()
|
|
|
| try:
|
|
|
|
|
| model, processor, has_image_processor = _load_model_and_processor_cached(
|
| model_name=model_name,
|
| device=device
|
| )
|
|
|
|
|
| if model and processor:
|
|
|
| tokenizer = getattr(processor, 'tokenizer', None)
|
| if tokenizer:
|
| if tokenizer.pad_token_id is None:
|
| main_loading_status.warning(f"Tokenizer has no pad_token_id. Batch generation (Self-Consistency) might be unstable.")
|
| else:
|
| main_loading_status.write(f"Tokenizer pad_token_id: {tokenizer.pad_token_id}.")
|
|
|
| if tokenizer.eos_token_id is not None:
|
| main_loading_status.write(f"Tokenizer eos_token_id: {tokenizer.eos_token_id}.")
|
| else:
|
| main_loading_status.warning(f"Tokenizer has no eos_token_id.")
|
| else:
|
| main_loading_status.warning("Processor does not have a tokenizer attribute. Cannot check pad/eos token IDs.")
|
|
|
|
|
|
|
| if has_image_processor:
|
| st.session_state.multimodal_capable_model_loaded = True
|
| main_loading_status.info("🖼️ Loaded model supports image inputs (multimodal).")
|
| else:
|
| st.session_state.multimodal_capable_model_loaded = False
|
| main_loading_status.warning("ℹ️ Loaded model appears text-only (no image processor found). Multimodal features will be disabled.")
|
|
|
| update_telemetry()
|
|
|
|
|
| main_loading_status.update(label=f"✅ Model Loaded: {model_name}", state="complete")
|
|
|
|
|
|
|
| return model, processor, has_image_processor
|
|
|
| else:
|
|
|
|
|
| st.session_state.multimodal_capable_model_loaded = False
|
| update_telemetry()
|
|
|
|
|
|
|
| return None, None, False
|
|
|
| except Exception as e:
|
|
|
| logger.error(f"An unexpected error occurred during model loading status wrapping: {e}")
|
| st.session_state.multimodal_capable_model_loaded = False
|
| main_loading_status.error(f"❌ An unexpected error occurred during loading status: {e}")
|
| update_telemetry()
|
| st.exception(e)
|
| main_loading_status.update(label="🔴 Model Loading Failed", state="complete")
|
|
|
| return None, None, False
|
|
|
|
|
|
|
|
|
| def normalize_answer(answer: str) -> str:
|
| """
|
| Normalizes a string answer for robust comparison during voting.
|
| - Converts to lowercase.
|
| - Strips leading/trailing whitespace.
|
| - Removes common punctuation and articles.
|
| - Handles simple cases of number words (e.g., "two" -> "2").
|
| - Removes extra internal whitespace.
|
| """
|
| if not isinstance(answer, str):
|
| return ""
|
|
|
| normalized = answer.lower().strip()
|
|
|
|
|
| normalized = re.sub(r'[.,!?;:]+$', '', normalized).strip()
|
|
|
|
|
| normalized = re.sub(r'^\s*(?:the answer is|result|output)\s*[:\-]?\s*', '', normalized, flags=re.IGNORECASE).strip()
|
|
|
|
|
| normalized = re.sub(r'^\s*(a|an|the)\s+', '', normalized, flags=re.IGNORECASE).strip()
|
|
|
|
|
| num_word_map = {
|
| 'zero': '0', 'one': '1', 'two': '2', 'three': '3', 'four': '4',
|
| 'five': '5', 'six': '6', 'seven': '7', 'eight': '8', 'nine': '9',
|
| 'ten': '10', 'eleven': '11', 'twelve': '12', 'thirteen': '13',
|
| 'fourteen': '14', 'fifteen': '15', 'sixteen': '16', 'seventeen': '17',
|
| 'eighteen': '18', 'nineteen': '19', 'twenty': '20', 'thirty': '30',
|
| 'forty': '40', 'fifty': '50', 'sixty': '60', 'seventy': '70',
|
| 'eighty': '80', 'ninety': '90', 'hundred': '100', 'thousand': '1000',
|
| 'million': '1000000', 'billion': '1000000000'
|
| }
|
|
|
|
|
| words = normalized.split()
|
| normalized_words = [num_word_map.get(word, word) for word in words]
|
| normalized = " ".join(normalized_words)
|
|
|
|
|
|
|
| normalized = re.sub(r'\s+', ' ', normalized).strip()
|
|
|
|
|
| normalized = normalized.strip()
|
|
|
|
|
| return normalized
|
|
|
| def perform_self_consistency_voting(final_answers: List[str]) -> Tuple[Optional[str], Dict[str, int], List[str]]:
|
| """
|
| Performs simple majority voting on a list of final answers after normalization.
|
|
|
| Args:
|
| final_answers (List[str]): A list of raw final answer strings from the wrapper.
|
|
|
| Returns:
|
| Tuple[Optional[str], Dict[str, int], List[str]]: A tuple containing:
|
| - The winning (most common) normalized answer, or None if no valid answers.
|
| - A dictionary mapping normalized answers to their vote counts.
|
| - A list of the raw answers that matched the consensus normalized answer.
|
| """
|
| if not final_answers:
|
| return None, {}, []
|
|
|
|
|
|
|
| normalized_answer_pairs = [(normalize_answer(str(ans)), str(ans)) for ans in final_answers if ans is not None and str(ans).strip()]
|
|
|
|
|
| normalized_answers_only = [pair[0] for pair in normalized_answer_pairs]
|
| if not normalized_answers_only:
|
| return None, {}, []
|
|
|
| answer_distribution = Counter(normalized_answers_only)
|
|
|
|
|
|
|
|
|
| if not answer_distribution:
|
| return None, {}, []
|
|
|
|
|
|
|
| most_common_list = answer_distribution.most_common(1)
|
| if not most_common_list:
|
| return None, dict(answer_distribution), []
|
| consensus_normalized_answer = most_common_list[0][0]
|
|
|
|
|
|
|
| consensus_raw_answers = [
|
| original_ans for original_ans in final_answers
|
| if original_ans is not None and normalize_answer(str(original_ans)) == consensus_normalized_answer
|
| ]
|
|
|
|
|
| return consensus_normalized_answer, dict(answer_distribution), consensus_raw_answers
|
|
|
|
|
|
|
|
|
|
|
| if "chat_history" not in st.session_state:
|
| st.session_state.chat_history = []
|
|
|
| st.session_state.chat_history.append({
|
| "role": "assistant",
|
|
|
| "content": "✨ Greetings. I am your AI companion, ready to explore the frontiers of thought with you. Present your question or problem below, and sculpt the analytical process using the parameters in the sidebar.",
|
| "type": "text",
|
| "id": 0
|
| })
|
|
|
|
|
|
|
| if "model" not in st.session_state:
|
| st.session_state.model = None
|
| if "processor" not in st.session_state:
|
| st.session_state.processor = None
|
| if "cot_wrapper" not in st.session_state:
|
| st.session_state.cot_wrapper = None
|
|
|
| if "multimodal_capable_model_loaded" not in st.session_state:
|
| st.session_state.multimodal_capable_model_loaded = False
|
|
|
|
|
| st.title("🧠 NeuroReasoner CoT (Multimodal)")
|
|
|
|
|
| with st.sidebar:
|
| st.header("⚙️ Controls")
|
|
|
|
|
| st.subheader("🌐 Model Configuration")
|
|
|
|
|
| default_model = "gpt2"
|
|
|
|
|
| model_name_input = st.text_input(
|
| "Hugging Face Model ID or Local Path",
|
| value=st.session_state.get("model_name_select", default_model),
|
| key="model_name_select"
|
| )
|
|
|
|
|
| available_devices = ["cpu"]
|
| if torch.cuda.is_available():
|
| for i in range(torch.cuda.device_count()):
|
| available_devices.append(f"cuda:{i}")
|
|
|
| default_device_index = 0
|
| if "device_select" in st.session_state and st.session_state.device_select in available_devices:
|
| default_device_index = available_devices.index(st.session_state.device_select)
|
| elif len(available_devices) > 0:
|
| default_device_index = 0
|
|
|
|
|
| device_select = st.selectbox(
|
| "Device",
|
| available_devices,
|
| index=default_device_index,
|
| key="device_select",
|
| help="Select the device (CPU or GPU) to load the model onto. Larger or multimodal models often require GPU."
|
| )
|
|
|
|
|
| if st.button("Load Model"):
|
| logger.info("Load Model button clicked. Initiating model loading process.")
|
|
|
|
|
| st.cache_resource.clear()
|
| logger.info("Streamlit resource cache cleared.")
|
|
|
|
|
| st.session_state.model = None
|
| st.session_state.processor = None
|
| st.session_state.cot_wrapper = None
|
| st.session_state.multimodal_capable_model_loaded = False
|
|
|
|
|
| model, processor, has_image_processor = safe_load_model_with_status(model_name_input, device_select)
|
|
|
|
|
| st.session_state.model = model
|
| st.session_state.processor = processor
|
| st.session_state.multimodal_capable_model_loaded = has_image_processor
|
|
|
|
|
|
|
| if model is not None and processor is not None and (hasattr(model, 'generate') and callable(getattr(model, 'generate', None)) or isinstance(model, GenerationMixin)):
|
| try:
|
|
|
| st.session_state.cot_wrapper = ChainOfThoughtWrapper(
|
| model=model,
|
| processor=processor,
|
| device=device_select,
|
|
|
|
|
|
|
|
|
| )
|
| st.sidebar.success("🚀 CoT Wrapper initialized!")
|
| logger.info("CoT Wrapper initialized successfully in GUI.")
|
| except Exception as e:
|
| st.session_state.cot_wrapper = None
|
| st.sidebar.error(f"❌ Failed to initialize CoT Wrapper: {e}")
|
| logger.error(f"Failed to initialize CoT Wrapper in GUI: {e}")
|
| st.exception(e)
|
| elif model is not None:
|
|
|
|
|
| logger.warning("Model loaded but not generation compatible, skipping wrapper initialization.")
|
| st.sidebar.warning("Wrapper initialization skipped: Loaded model is not compatible with text generation.")
|
| else:
|
| logger.warning("Model or Processor loading failed, skipping wrapper initialization.")
|
| st.sidebar.warning("Wrapper initialization skipped due to model loading failure.")
|
|
|
|
|
|
|
| st.subheader("✨ Generation Parameters")
|
|
|
|
|
| st.checkbox(
|
| "Enable Self-Consistency Voting",
|
| value=st.session_state.get("self_consistency_checkbox", True),
|
| key="self_consistency_checkbox",
|
| help="Generate multiple reasoning chains and vote for the most common answer."
|
| )
|
|
|
|
|
|
|
| st.slider(
|
| "Number of Chains (N)",
|
| min_value=1,
|
| max_value=10,
|
| step=1,
|
| key="num_chains",
|
| help="The number of independent reasoning chains to generate for self-consistency."
|
| "Requires a model that supports batch generation and a tokenizer with a pad_token_id."
|
| )
|
|
|
|
|
|
|
| effective_num_chains_display = st.session_state.get("num_chains", 1) if st.session_state.get("self_consistency_checkbox", False) else 1
|
| st.info(f"Effective Chains Generated: {effective_num_chains_display}")
|
|
|
|
|
|
|
|
|
| st.slider(
|
| "Max New Tokens",
|
| min_value=50,
|
| max_value=2048,
|
| value=st.session_state.get("max_new_tokens", 512),
|
| step=10,
|
| key="max_new_tokens",
|
| help="Maximum number of tokens to generate in each chain."
|
| )
|
|
|
|
|
|
|
| st.slider(
|
| "Temperature",
|
| min_value=0.0,
|
| max_value=2.0,
|
| value=st.session_state.get("temperature", 0.7),
|
| step=0.05,
|
| key="temperature",
|
| help="Controls the randomness of generation. Lower = more deterministic, Higher = more creative."
|
| "Set to 0 for greedy decoding (if do_sample is False)."
|
| )
|
|
|
|
|
|
|
| st.slider(
|
| "Top K",
|
| min_value=0,
|
| max_value=100,
|
| value=st.session_state.get("top_k", 50),
|
| step=1,
|
| key="top_k",
|
| help="Limits the sampling pool to the top K most likely tokens."
|
| "Set to 0 to disable Top-K sampling (when do_sample is True)."
|
| "Recommended: Use either Top-K or Top-P, not both (e.g., Top-P 1.0 with Top-K > 0, or Top-K 0 with Top-P < 1.0)."
|
| )
|
|
|
|
|
|
|
| st.slider(
|
| "Top P",
|
| min_value=0.0,
|
| max_value=1.0,
|
| value=st.session_state.get("top_p", 1.0),
|
| step=0.01,
|
| key="top_p",
|
| help="Limits the sampling pool to the smallest set of tokens whose cumulative probability exceeds Top P."
|
| "Set to 1.0 to disable Top-P sampling (when do_sample is True)."
|
| "Recommended: Use either Top-K or Top-P, not both (e.g., Top-P 1.0 with Top-K > 0, or Top-K 0 with Top-P < 1.0)."
|
| )
|
|
|
|
|
|
|
| st.checkbox(
|
| "Do Sample",
|
| value=st.session_state.get("do_sample", True),
|
| key="do_sample",
|
| help="If True, use sampling (with Temperature, Top K, Top P). Otherwise, use greedy decoding (Temperature, Top K, Top P are ignored)."
|
| )
|
|
|
|
|
|
|
| st.slider(
|
| "Repetition Penalty",
|
| min_value=1.0,
|
| max_value=2.0,
|
| value=st.session_state.get("repetition_penalty", 1.1),
|
| step=0.01,
|
| key="repetition_penalty",
|
| help="Penalizes repeated tokens based on the number of repetitions. Value > 1.0 encourages diversity."
|
| )
|
|
|
|
|
|
|
| st.slider(
|
| "No Repeat Ngram Size",
|
| min_value=0,
|
| max_value=10,
|
| value=st.session_state.get("no_repeat_ngram_size", 0),
|
| step=1,
|
| key="no_repeat_ngram_size",
|
| help="All ngrams of this size appearing in the generation are penalized so that they do not appear more than once in the generation."
|
| "Set to 0 to disable."
|
| )
|
|
|
|
|
|
|
|
|
| st.subheader("💾 Memory Management")
|
| if st.button("Clear Chat History"):
|
| logger.info("Clear Chat History button clicked.")
|
| st.session_state.chat_history = []
|
|
|
| st.session_state.chat_history.append({
|
| "role": "assistant",
|
|
|
| "content": "✨ Greetings. I am your AI companion, ready to explore the frontiers of thought with you. Present your question or problem below, and sculpt the analytical process using the parameters in the sidebar.",
|
| "type": "text",
|
| "id": 0
|
| })
|
| logger.info("Chat history cleared and welcome message added.")
|
| st.rerun()
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| if st.session_state.multimodal_capable_model_loaded:
|
| st.info("🖼️ Multimodal model loaded. You can now upload images alongside your text input.")
|
| else:
|
| st.warning("ℹ️ Text-only model loaded or no model loaded or loaded model is not multimodal compatible. Multimodal input (images) will not be processed for generation. Load a multimodal model (like `HuggingFaceTB/SmolVLM-Instruct`) compatible with Hugging Face Auto classes for generation to enable this.")
|
|
|
|
|
|
|
|
|
| for message in st.session_state.chat_history:
|
| role = message["role"]
|
|
|
| content_data = message["content"]
|
| message_type = message.get("type", "text")
|
| message_id = message.get("id", None)
|
|
|
|
|
|
|
| with st.chat_message(role):
|
| if role == "user":
|
|
|
|
|
| user_text = content_data.get("text", "") if isinstance(content_data, dict) else str(content_data)
|
| image_data_list = content_data.get("image_data", []) if isinstance(content_data, dict) else []
|
|
|
|
|
| if user_text:
|
|
|
| st.markdown(str(user_text), unsafe_allow_html=True)
|
|
|
|
|
| if image_data_list:
|
| if user_text:
|
| st.write("---")
|
|
|
| st.write(f"Uploaded Image{'s' if len(image_data_list) > 1 else ''}:")
|
|
|
|
|
| cols = st.columns(len(image_data_list)) if len(image_data_list) > 1 else [st.container()]
|
| for i, img_bytes in enumerate(image_data_list):
|
| try:
|
| img = Image.open(io.BytesIO(img_bytes))
|
|
|
| with (cols[i] if len(image_data_list) > 1 else cols[0]):
|
|
|
|
|
| st.image(img, caption=f"Image {i+1}", use_column_width=False, width=150)
|
| except Exception as e:
|
| logger.error(f"Could not display uploaded image {i+1} for message ID {message_id}: {e}")
|
| with (cols[i] if len(image_data_list) > 1 else cols[0]):
|
|
|
| st.warning(f"Could not display image {i+1}")
|
|
|
|
|
| elif role == "assistant":
|
|
|
| if message_type == "text":
|
|
|
|
|
| st.markdown(str(content_data), unsafe_allow_html=True)
|
| elif message_type == "error_message":
|
|
|
|
|
| st.error(str(content_data))
|
| elif message_type == "cot_response":
|
|
|
| if not isinstance(content_data, dict):
|
| logger.error(f"Assistant message ID {message_id} content is not in expected dictionary format for cot_response.")
|
| st.error("Error: Assistant CoT response content is not in expected dictionary format.")
|
|
|
| st.markdown(str(content_data), unsafe_allow_html=True)
|
| continue
|
|
|
|
|
|
|
|
|
|
|
|
|
| main_output_display = str(content_data.get("main_output_display", "Error: Could not retrieve response."))
|
|
|
| main_output_label = str(content_data.get("main_output_label", "Output"))
|
|
|
|
|
| details = content_data.get("details", {})
|
|
|
|
|
| generated_results = details.get("generated_results", [])
|
| consensus_answer = details.get("consensus_answer")
|
| vote_counts = details.get("vote_counts")
|
| consensus_raw_answers = details.get("consensus_raw_answers", [])
|
| self_consistency_was_enabled = details.get("self_consistency_enabled", False)
|
| requested_chains = details.get("requested_chains", len(generated_results))
|
|
|
|
|
| generation_params = content_data.get("generation_params", {})
|
|
|
| model_name = str(content_data.get('model_name', 'N/A'))
|
| device_name = str(content_data.get('device', 'N/A'))
|
| model_type_name = str(content_data.get('model_type', 'Unknown'))
|
| is_multimodal = content_data.get('is_multimodal_compatible', False)
|
| model_info = f"Model: {model_name} | Device: {device_name} | Loaded Class: {model_type_name} | Capability: {'Multimodal' if is_multimodal else 'Text-Only'}"
|
|
|
|
|
|
|
|
|
| st.markdown(f'**{str(main_output_label)}:**', unsafe_allow_html=True)
|
|
|
|
|
|
|
| st.code(str(main_output_display), language='markdown')
|
|
|
|
|
|
|
|
|
| should_show_details_expander = len(generated_results) > 0 or (self_consistency_was_enabled and requested_chains > 1 and (consensus_answer is not None or vote_counts is not None))
|
|
|
| if should_show_details_expander:
|
|
|
| with st.expander("View Reasoning Details"):
|
|
|
|
|
|
|
|
|
| if self_consistency_was_enabled and requested_chains > 1:
|
| st.markdown("#### Consensus Details", unsafe_allow_html=True)
|
|
|
|
|
|
|
| if consensus_answer is not None and str(consensus_answer).strip():
|
|
|
|
|
| st.markdown(f'<div class="consensus-answer"><strong>Consensus Answer:</strong> <p>{str(consensus_answer)}</p></div>', unsafe_allow_html=True)
|
| else:
|
| st.info("No consensus answer found among the generated chains (all unique or invalid).")
|
|
|
|
|
| if vote_counts and isinstance(vote_counts, dict):
|
|
|
| sorted_votes = sorted(vote_counts.items(), key=lambda item: item[1], reverse=True)
|
|
|
|
|
| vote_list_items = "".join([f"<li><code>{str(ans)[:150]}{'...' if len(str(ans)) > 150 else ''}</code>: {str(count)} vote(s)</li>" for ans, count in sorted_votes])
|
| st.markdown(f'<div class="vote-counts"><strong>Self-Consistency Vote Counts ({len(generated_results)} Chains Analyzed):</strong><ul style="list-style-type: disc;">{vote_list_items}</ul></div>', unsafe_allow_html=True)
|
|
|
| elif self_consistency_was_enabled and requested_chains > 1 and len(generated_results) > 0:
|
| st.info("Self-Consistency voting skipped: Fewer than 2 valid answers extracted from generated chains for voting.")
|
|
|
|
|
|
|
| if (consensus_answer is not None and str(consensus_answer).strip()) or (vote_counts and isinstance(vote_counts, dict)):
|
| st.markdown("---")
|
|
|
|
|
|
|
| st.markdown("#### Individual Reasoning Chains", unsafe_allow_html=True)
|
|
|
| if not generated_results:
|
| st.warning("No reasoning chains were generated successfully by the wrapper.")
|
| elif not isinstance(generated_results, list):
|
| st.error("Error: generated_results content is not in expected list format.")
|
| st.markdown(str(generated_results), unsafe_allow_html=True)
|
| else:
|
|
|
| for i, chain_data in enumerate(generated_results):
|
|
|
| if not isinstance(chain_data, dict):
|
| st.error(f"Error: Content for Chain {i+1} is not in expected dictionary format.")
|
| st.markdown(str(chain_data), unsafe_allow_html=True)
|
| continue
|
|
|
| chain_number = i + 1
|
|
|
|
|
| st.markdown(f"##### Chain {chain_number}", unsafe_allow_html=True)
|
|
|
|
|
|
|
| chain_answer = chain_data.get("final_answer", None)
|
|
|
|
|
| chain_answer_display = str(chain_answer) if chain_answer is not None and str(chain_answer).strip() else "N/A (Extraction Failed)"
|
|
|
|
|
|
|
| reasoning_steps = chain_data.get("reasoning_steps", [])
|
|
|
|
|
| if reasoning_steps and isinstance(reasoning_steps, list) and any(str(step).strip() for step in reasoning_steps):
|
| st.markdown("###### Reasoning Steps", unsafe_allow_html=True)
|
|
|
|
|
| steps_markdown = "\n".join([f"- {str(step).strip()}" for step in reasoning_steps if str(step).strip()])
|
| st.markdown(steps_markdown, unsafe_allow_html=True)
|
| else:
|
| st.info("No reasoning steps parsed for this chain.")
|
|
|
|
|
|
|
| st.markdown("###### Final Answer (Chain Only)", unsafe_allow_html=True)
|
|
|
|
|
| st.markdown(f'**{str(chain_answer_display)}**', unsafe_allow_html=True)
|
|
|
|
|
|
|
| st.markdown("###### Full Generated Output (Chain Only)", unsafe_allow_html=True)
|
|
|
| full_text = str(chain_data.get("full_text", "N/A (No output returned)"))
|
|
|
|
|
| st.code(str(full_text), language='markdown')
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| if i < len(generated_results) - 1:
|
| st.markdown("---", unsafe_allow_html=True)
|
|
|
|
|
|
|
|
|
| if generated_results or (self_consistency_was_enabled and requested_chains > 1 and (consensus_answer is not None or (vote_counts and isinstance(vote_counts, dict)))):
|
| st.markdown("---", unsafe_allow_html=True)
|
|
|
|
|
|
|
| st.markdown("#### Generation Parameters Used", unsafe_allow_html=True)
|
|
|
| st.markdown("###### Model Info", unsafe_allow_html=True)
|
|
|
| st.info(str(model_info))
|
|
|
| st.markdown("###### Parameters", unsafe_allow_html=True)
|
| if generation_params and isinstance(generation_params, dict):
|
|
|
| param_keys_to_exclude_from_display = ['self_consistency_enabled', 'pad_token_id', 'eos_token_id', 'requested_chains']
|
|
|
| filtered_gen_params = {k: v for k, v in generation_params.items() if k not in param_keys_to_exclude_from_display}
|
|
|
| if filtered_gen_params:
|
|
|
| sorted_display_params = sorted(filtered_gen_params.items())
|
|
|
| params_markdown = "\n".join([f"- **{str(key)}:** {str(value)}" for key, value in sorted_display_params])
|
| st.markdown(params_markdown, unsafe_allow_html=True)
|
| else:
|
| st.info("No specific generation parameters recorded for this turn.")
|
| else:
|
| st.info("No generation parameters recorded for this turn.")
|
|
|
|
|
|
|
|
|
| st.markdown("<br>", unsafe_allow_html=True)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| with st.form("chat_form", clear_on_submit=True):
|
|
|
| col1, col2 = st.columns([4, 1])
|
|
|
| with col1:
|
| user_input = st.text_area(
|
| "Your Message",
|
| key="user_input_text_area",
|
| label_visibility="collapsed",
|
| placeholder="Enter your question or problem here...",
|
| height=100
|
| )
|
|
|
| with col2:
|
|
|
|
|
| st.markdown("###### Upload Image")
|
|
|
| if st.session_state.multimodal_capable_model_loaded:
|
| uploaded_files = st.file_uploader(
|
| "Upload Image(s)",
|
| type=["png", "jpg", "jpeg", "gif", "webp"],
|
| accept_multiple_files=True,
|
| label_visibility="collapsed",
|
|
|
| )
|
|
|
| uploaded_image_data = []
|
| if uploaded_files:
|
| for uploaded_file in uploaded_files:
|
|
|
| try:
|
| file_bytes = uploaded_file.getvalue()
|
| uploaded_image_data.append(file_bytes)
|
| except Exception as e:
|
| logger.error(f"Error reading uploaded file {uploaded_file.name}: {e}")
|
| st.warning(f"Could not read file: {uploaded_file.name}")
|
|
|
|
|
|
|
| else:
|
|
|
| st.info("Upload disabled (Text-only or incompatible model loaded)")
|
| uploaded_files = None
|
| uploaded_image_data = []
|
|
|
|
|
|
|
|
|
|
|
| st.markdown("###### Send")
|
| send_button = st.form_submit_button("Send")
|
|
|
|
|
|
|
|
|
| thinking_message_placeholder = st.empty()
|
|
|
|
|
|
|
|
|
| if send_button and (user_input or uploaded_image_data):
|
|
|
| is_model_generation_compatible = (
|
| st.session_state.model is not None and
|
| (hasattr(st.session_state.model, 'generate') and callable(getattr(st.session_state.model, 'generate', None)) or isinstance(st.session_state.model, GenerationMixin))
|
| )
|
|
|
| if st.session_state.model is None or st.session_state.processor is None or st.session_state.cot_wrapper is None or not is_model_generation_compatible:
|
| logger.warning("Attempted generation without loaded model/wrapper or with incompatible model.")
|
| if st.session_state.model is not None and not is_model_generation_compatible:
|
| st.warning("Loaded model is not compatible with text generation. Please load a different model.")
|
| assistant_msg_content = "🔴 Loaded model is not compatible with text generation. Please load a different model."
|
| else:
|
| st.warning("Please load a model first.")
|
| assistant_msg_content = "Please load a model first to start the conversation."
|
|
|
|
|
|
|
|
|
| st.session_state.chat_history.append({
|
| "role": "assistant",
|
| "content": assistant_msg_content,
|
| "type": "error_message",
|
| "id": len(st.session_state.chat_history)
|
| })
|
| st.rerun()
|
|
|
| else:
|
|
|
|
|
| user_message_content = {
|
| "text": user_input,
|
| "image_data": uploaded_image_data
|
| }
|
|
|
| if user_input.strip() or uploaded_image_data:
|
| st.session_state.chat_history.append({"role": "user", "content": user_message_content, "type": "user_input", "id": len(st.session_state.chat_history)})
|
| logger.info(f"User input received (text: {len(user_input.strip()) > 0}, images: {len(uploaded_image_data)}). Proceeding with generation.")
|
| else:
|
|
|
| logger.info("Empty user input and no image uploaded. Skipping generation.")
|
|
|
| thinking_message_placeholder.empty()
|
| st.stop()
|
|
|
|
|
|
|
|
|
| thinking_message_placeholder.info("💭 NeuroReasoner is thinking...")
|
|
|
|
|
|
|
|
|
| self_consistency_enabled = st.session_state.get("self_consistency_checkbox", False)
|
| num_chains_requested = st.session_state.get("num_chains", 1)
|
| effective_num_chains = num_chains_requested if self_consistency_enabled else 1
|
|
|
|
|
|
|
| current_gen_params = {
|
| "max_new_tokens": st.session_state.get("max_new_tokens", 512),
|
| "temperature": st.session_state.get("temperature", 0.7),
|
| "top_k": st.session_state.get("top_k", 50),
|
| "top_p": st.session_state.get("top_p", 1.0),
|
| "do_sample": st.session_state.get("do_sample", True),
|
| "repetition_penalty": st.session_state.get("repetition_penalty", 1.1),
|
| "no_repeat_ngram_size": st.session_state.get("no_repeat_ngram_size", 0),
|
| "num_return_sequences": effective_num_chains,
|
|
|
|
|
|
|
| "pad_token_id": getattr(getattr(st.session_state.processor, 'tokenizer', None), 'pad_token_id', None),
|
| "eos_token_id": getattr(getattr(st.session_state.processor, 'tokenizer', None), 'eos_token_id', None),
|
| "self_consistency_enabled": self_consistency_enabled,
|
| "requested_chains": num_chains_requested
|
|
|
| }
|
| model_name_current = st.session_state.get("model_name_select", "N/A")
|
| device_current = st.session_state.get("device_select", "N/A")
|
|
|
|
|
| actual_model_class_name = type(st.session_state.model).__name__ if st.session_state.model else "Unknown"
|
| is_multimodal_compatible_loaded = st.session_state.multimodal_capable_model_loaded
|
|
|
|
|
| generation_status_text = f"✨ Generating {effective_num_chains} reasoning chain(s)..."
|
| with st.status(generation_status_text, expanded=True) as status_box:
|
| start_time = time.time()
|
| status_box.write("Starting generation process...")
|
|
|
| if uploaded_image_data and is_multimodal_compatible_loaded:
|
| status_box.write(f"Processing {len(uploaded_image_data)} uploaded image(s)...")
|
|
|
| update_telemetry()
|
|
|
| outputs = None
|
| try:
|
|
|
|
|
| outputs = st.session_state.cot_wrapper.generate(
|
| input_text=user_input,
|
| image_data=uploaded_image_data,
|
| generation_params=current_gen_params,
|
| chat_history=st.session_state.chat_history
|
| )
|
| end_time = time.time()
|
| duration = end_time - start_time
|
| status_box.write(f"Generation complete in {duration:.2f} seconds.")
|
| update_telemetry()
|
|
|
|
|
| generated_results_list = []
|
| all_final_answers = []
|
|
|
|
|
|
|
| if outputs and isinstance(outputs, dict) and \
|
| isinstance(outputs.get("full_texts"), list) and \
|
| isinstance(outputs.get("reasoning_steps"), list) and \
|
| isinstance(outputs.get("final_answers"), list):
|
|
|
| num_returned = len(outputs.get("full_texts", []))
|
| logger.info(f"Wrapper returned {num_returned} sequences.")
|
|
|
|
|
| if num_returned == 0:
|
| status_box.warning("Wrapper returned no generated sequences.")
|
| elif num_returned != effective_num_chains:
|
| status_box.warning(f"Wrapper returned {num_returned} sequences, expected {effective_num_chains}. Displaying returned sequences.")
|
|
|
|
|
| for i in range(num_returned):
|
|
|
|
|
| full_text = outputs.get("full_texts", [])[i] if i < len(outputs.get("full_texts", [])) else "N/A (No output)"
|
|
|
| reasoning_steps_data = outputs.get("reasoning_steps", [])[i] if i < len(outputs.get("reasoning_steps", [])) else []
|
|
|
| reasoning_steps_list = reasoning_steps_data if isinstance(reasoning_steps_data, list) else [str(reasoning_steps_data)]
|
|
|
| final_answer = outputs.get("final_answers", [])[i] if i < len(outputs.get("final_answers", [])) else None
|
|
|
|
|
|
|
| chain_result = {
|
| "full_text": full_text,
|
| "reasoning_steps": reasoning_steps_list,
|
| "final_answer": final_answer
|
|
|
|
|
| }
|
| generated_results_list.append(chain_result)
|
|
|
| all_final_answers.append(final_answer)
|
|
|
|
|
| else:
|
| logger.error("Wrapper returned unexpected output structure or empty lists.")
|
| status_box.error("❌ Wrapper did not return expected output structure or empty lists.")
|
|
|
|
|
|
|
|
|
| consensus_answer = None
|
| vote_counts = None
|
| consensus_raw_answers = []
|
|
|
| if self_consistency_enabled and len(generated_results_list) > 1 and any(ans is not None and str(ans).strip() for ans in all_final_answers):
|
| status_box.write("Performing Self-Consistency voting...")
|
|
|
| consensus_answer, vote_counts, consensus_raw_answers = perform_self_consistency_voting(all_final_answers)
|
| if consensus_answer is not None:
|
| status_box.success(f"Self-Consistency consensus reached: '{consensus_answer}'")
|
| else:
|
|
|
| status_box.warning("Self-Consistency voting found no common answers or no valid answers among the generated chains.")
|
|
|
| elif self_consistency_enabled and len(generated_results_list) > 0 and num_chains_requested > 1:
|
|
|
|
|
| if len(generated_results_list) > 0:
|
| status_box.info("Self-Consistency voting skipped: Fewer than 2 valid answers extracted from generated chains for voting.")
|
| else:
|
|
|
| status_box.warning("Self-Consistency voting skipped: No chains were generated.")
|
|
|
|
|
|
|
|
|
| main_output_to_show = None
|
| main_output_label = "Output"
|
|
|
|
|
| if generated_results_list and len(generated_results_list) > 0:
|
|
|
| first_chain_full_text = generated_results_list[0].get("full_text", "N/A (No full output returned for Chain 1)")
|
| main_output_to_show = str(first_chain_full_text)
|
| main_output_label = "Full Output (Chain 1)"
|
|
|
|
|
| if self_consistency_enabled and consensus_answer is not None and str(consensus_answer).strip():
|
| main_output_label = "Full Output (Chain 1) | Consensus Answer Available Below"
|
| else:
|
| main_output_label = "Full Output (Chain 1)"
|
|
|
|
|
| else:
|
|
|
| main_output_to_show = "Generation completed, but no output was returned or could be processed."
|
| main_output_label = "Status"
|
|
|
|
|
|
|
| assistant_response_data = {
|
| "prompt": user_input,
|
|
|
|
|
| "main_output_display": main_output_to_show,
|
| "main_output_label": main_output_label,
|
|
|
|
|
| "details": {
|
| "generated_results": generated_results_list,
|
| "consensus_answer": consensus_answer,
|
| "vote_counts": vote_counts,
|
| "consensus_raw_answers": consensus_raw_answers,
|
|
|
| "self_consistency_enabled": self_consistency_enabled,
|
| "requested_chains": num_chains_requested
|
| },
|
|
|
|
|
| "generation_params": current_gen_params,
|
| "model_name": model_name_current,
|
| "device": device_current,
|
| "model_type": actual_model_class_name,
|
| "is_multimodal_compatible": is_multimodal_compatible_loaded,
|
|
|
| "type": "cot_response",
|
| "id": len(st.session_state.chat_history)
|
| }
|
| st.session_state.chat_history.append({"role": "assistant", "content": assistant_response_data, "type": "cot_response", "id": len(st.session_state.chat_history)})
|
|
|
|
|
|
|
| status_box.update(label="✅ Generation Complete", state="complete")
|
|
|
| except Exception as e:
|
| end_time = time.time()
|
| duration = end_time - start_time
|
| status_box.error(f"❌ An error occurred during generation after {duration:.2f} seconds.")
|
| logger.error(f"Error during wrapper generation call: {e}")
|
| update_telemetry()
|
| st.exception(e)
|
|
|
|
|
|
|
| st.session_state.chat_history.append({
|
| "role": "assistant",
|
| "content": f"🔴 An unexpected error occurred during reasoning. Details: {e}",
|
| "type": "error_message",
|
| "id": len(st.session_state.chat_history)
|
| })
|
| status_box.update(label="🔴 Generation Failed", state="complete")
|
|
|
|
|
|
|
| thinking_message_placeholder.empty()
|
|
|
|
|
| if torch.cuda.is_available():
|
| try:
|
| torch.cuda.empty_cache()
|
| logger.debug("GPU memory cache cleared after generation attempt.")
|
| except Exception as cleanup_e:
|
| logger.warning(f"Error during cuda empty_cache after generation attempt: {cleanup_e}")
|
| pass
|
| gc.collect()
|
| logger.debug("Garbage collection performed after generation attempt.")
|
|
|
|
|
| st.rerun() |
