Update app.py
Browse files
app.py
CHANGED
@@ -1,21 +1,32 @@
|
|
1 |
#!/usr/bin/env python3
|
2 |
import os
|
3 |
import glob
|
|
|
4 |
import time
|
|
|
5 |
import streamlit as st
|
|
|
|
|
|
|
|
|
|
|
|
|
6 |
import fitz # PyMuPDF
|
7 |
import requests
|
8 |
from PIL import Image
|
9 |
-
from transformers import AutoTokenizer, AutoModel
|
10 |
-
from diffusers import StableDiffusionPipeline
|
11 |
import cv2
|
12 |
import numpy as np
|
13 |
import logging
|
14 |
import asyncio
|
15 |
import aiofiles
|
16 |
from io import BytesIO
|
|
|
|
|
|
|
|
|
|
|
17 |
|
18 |
-
# Logging setup
|
19 |
logging.basicConfig(level=logging.INFO, format="%(asctime)s - %(levelname)s - %(message)s")
|
20 |
logger = logging.getLogger(__name__)
|
21 |
log_records = []
|
@@ -28,40 +39,212 @@ logger.addHandler(LogCaptureHandler())
|
|
28 |
|
29 |
# Page Configuration
|
30 |
st.set_page_config(
|
31 |
-
page_title="AI Vision Titans 🚀",
|
32 |
page_icon="🤖",
|
33 |
layout="wide",
|
34 |
initial_sidebar_state="expanded",
|
35 |
-
menu_items={
|
|
|
|
|
|
|
|
|
36 |
)
|
37 |
|
38 |
# Initialize st.session_state
|
39 |
if 'captured_files' not in st.session_state:
|
40 |
st.session_state['captured_files'] = []
|
|
|
|
|
|
|
|
|
41 |
if 'processing' not in st.session_state:
|
42 |
st.session_state['processing'] = {}
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
43 |
|
44 |
# Utility Functions
|
45 |
-
def
|
46 |
-
|
47 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
48 |
|
49 |
def get_gallery_files(file_types):
|
50 |
return sorted([f for ext in file_types for f in glob.glob(f"*.{ext}")])
|
51 |
|
52 |
-
def update_gallery():
|
53 |
-
media_files = get_gallery_files(["png", "txt"])
|
54 |
-
if media_files:
|
55 |
-
cols = st.sidebar.columns(2)
|
56 |
-
for idx, file in enumerate(media_files[:gallery_size * 2]):
|
57 |
-
with cols[idx % 2]:
|
58 |
-
if file.endswith(".png"):
|
59 |
-
st.image(Image.open(file), caption=file, use_container_width=True)
|
60 |
-
elif file.endswith(".txt"):
|
61 |
-
with open(file, "r") as f:
|
62 |
-
content = f.read()
|
63 |
-
st.text(content[:50] + "..." if len(content) > 50 else content, help=file)
|
64 |
-
|
65 |
def download_pdf(url, output_path):
|
66 |
try:
|
67 |
response = requests.get(url, stream=True, timeout=10)
|
@@ -74,28 +257,86 @@ def download_pdf(url, output_path):
|
|
74 |
logger.error(f"Failed to download {url}: {e}")
|
75 |
return False
|
76 |
|
77 |
-
#
|
78 |
-
def
|
79 |
-
|
80 |
-
|
81 |
-
|
82 |
-
|
83 |
-
|
84 |
-
|
85 |
-
|
86 |
-
|
87 |
-
|
88 |
-
|
89 |
-
|
90 |
-
|
91 |
-
|
92 |
-
|
93 |
-
|
94 |
-
|
95 |
-
|
96 |
-
|
97 |
-
|
98 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
99 |
|
100 |
# Async Processing Functions
|
101 |
async def process_pdf_snapshot(pdf_path, mode="thumbnail"):
|
@@ -104,21 +345,19 @@ async def process_pdf_snapshot(pdf_path, mode="thumbnail"):
|
|
104 |
status.text(f"Processing PDF Snapshot ({mode})... (0s)")
|
105 |
doc = fitz.open(pdf_path)
|
106 |
output_files = []
|
107 |
-
|
108 |
if mode == "thumbnail":
|
109 |
page = doc[0]
|
110 |
-
pix = page.get_pixmap(matrix=fitz.Matrix(0.5, 0.5))
|
111 |
output_file = generate_filename("thumbnail", "png")
|
112 |
pix.save(output_file)
|
113 |
output_files.append(output_file)
|
114 |
elif mode == "twopage":
|
115 |
for i in range(min(2, len(doc))):
|
116 |
page = doc[i]
|
117 |
-
pix = page.get_pixmap(matrix=fitz.Matrix(1.0, 1.0))
|
118 |
output_file = generate_filename(f"twopage_{i}", "png")
|
119 |
pix.save(output_file)
|
120 |
output_files.append(output_file)
|
121 |
-
|
122 |
doc.close()
|
123 |
elapsed = int(time.time() - start_time)
|
124 |
status.text(f"PDF Snapshot ({mode}) completed in {elapsed}s!")
|
@@ -143,50 +382,43 @@ async def process_ocr(image, output_file):
|
|
143 |
update_gallery()
|
144 |
return result
|
145 |
|
146 |
-
async def process_image_gen(prompt, output_file):
|
147 |
-
start_time = time.time()
|
148 |
-
status = st.empty()
|
149 |
-
status.text("Processing Image Gen... (0s)")
|
150 |
-
pipeline = load_image_gen()
|
151 |
-
gen_image = pipeline(prompt, num_inference_steps=20).images[0]
|
152 |
-
elapsed = int(time.time() - start_time)
|
153 |
-
status.text(f"Image Gen completed in {elapsed}s!")
|
154 |
-
gen_image.save(output_file)
|
155 |
-
if output_file not in st.session_state['captured_files']:
|
156 |
-
st.session_state['captured_files'].append(output_file)
|
157 |
-
update_gallery()
|
158 |
-
return gen_image
|
159 |
-
|
160 |
-
async def process_line_drawing(image, style, output_file):
|
161 |
-
start_time = time.time()
|
162 |
-
status = st.empty()
|
163 |
-
status.text(f"Processing Line Drawing ({style})... (0s)")
|
164 |
-
edge_fn = load_line_drawer()
|
165 |
-
line_drawing = edge_fn(image, style=style)
|
166 |
-
elapsed = int(time.time() - start_time)
|
167 |
-
status.text(f"Line Drawing ({style}) completed in {elapsed}s!")
|
168 |
-
line_drawing.save(output_file)
|
169 |
-
if output_file not in st.session_state['captured_files']:
|
170 |
-
st.session_state['captured_files'].append(output_file)
|
171 |
-
update_gallery()
|
172 |
-
return line_drawing
|
173 |
-
|
174 |
# Main App
|
175 |
-
st.title("AI Vision Titans 🚀")
|
176 |
|
177 |
-
# Sidebar
|
178 |
st.sidebar.header("Captured Files 📜")
|
179 |
gallery_size = st.sidebar.slider("Gallery Size", 1, 10, 4)
|
180 |
update_gallery()
|
181 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
182 |
st.sidebar.subheader("Action Logs 📜")
|
183 |
log_container = st.sidebar.empty()
|
184 |
with log_container:
|
185 |
for record in log_records:
|
186 |
st.write(f"{record.asctime} - {record.levelname} - {record.message}")
|
187 |
|
|
|
|
|
|
|
|
|
|
|
|
|
188 |
# Tabs
|
189 |
-
tab1, tab2, tab3, tab4, tab5 = st.tabs([
|
|
|
|
|
|
|
190 |
|
191 |
with tab1:
|
192 |
st.header("Camera Snap 📷")
|
@@ -202,6 +434,7 @@ with tab1:
|
|
202 |
st.image(Image.open(filename), caption=filename, use_container_width=True)
|
203 |
logger.info(f"Saved snapshot from Camera 0: {filename}")
|
204 |
st.session_state['captured_files'].append(filename)
|
|
|
205 |
update_gallery()
|
206 |
with cols[1]:
|
207 |
cam1_img = st.camera_input("Take a picture - Cam 1", key="cam1")
|
@@ -213,6 +446,7 @@ with tab1:
|
|
213 |
st.image(Image.open(filename), caption=filename, use_container_width=True)
|
214 |
logger.info(f"Saved snapshot from Camera 1: {filename}")
|
215 |
st.session_state['captured_files'].append(filename)
|
|
|
216 |
update_gallery()
|
217 |
|
218 |
st.subheader("Burst Capture")
|
@@ -231,6 +465,7 @@ with tab1:
|
|
231 |
f.write(img.getvalue())
|
232 |
st.session_state['burst_frames'].append(filename)
|
233 |
logger.info(f"Saved burst frame {i}: {filename}")
|
|
|
234 |
st.image(Image.open(filename), caption=filename, use_container_width=True)
|
235 |
time.sleep(0.5)
|
236 |
st.session_state['captured_files'].extend([f for f in st.session_state['burst_frames'] if f not in st.session_state['captured_files']])
|
@@ -248,6 +483,7 @@ with tab2:
|
|
248 |
pdf_path = generate_filename("downloaded", "pdf")
|
249 |
if download_pdf(url, pdf_path):
|
250 |
logger.info(f"Downloaded PDF from {url} to {pdf_path}")
|
|
|
251 |
snapshots = asyncio.run(process_pdf_snapshot(pdf_path, mode.lower().replace(" ", "")))
|
252 |
for snapshot in snapshots:
|
253 |
st.image(Image.open(snapshot), caption=snapshot, use_container_width=True)
|
@@ -255,6 +491,147 @@ with tab2:
|
|
255 |
st.error(f"Failed to download {url}")
|
256 |
|
257 |
with tab3:
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
258 |
st.header("Test OCR 🔍")
|
259 |
captured_files = get_gallery_files(["png"])
|
260 |
if captured_files:
|
@@ -265,13 +642,14 @@ with tab3:
|
|
265 |
output_file = generate_filename("ocr_output", "txt")
|
266 |
st.session_state['processing']['ocr'] = True
|
267 |
result = asyncio.run(process_ocr(image, output_file))
|
|
|
268 |
st.text_area("OCR Result", result, height=200, key="ocr_result")
|
269 |
st.success(f"OCR output saved to {output_file}")
|
270 |
st.session_state['processing']['ocr'] = False
|
271 |
else:
|
272 |
st.warning("No images captured yet. Use Camera Snap or Download PDFs first!")
|
273 |
|
274 |
-
with
|
275 |
st.header("Test Image Gen 🎨")
|
276 |
captured_files = get_gallery_files(["png"])
|
277 |
if captured_files:
|
@@ -283,29 +661,12 @@ with tab4:
|
|
283 |
output_file = generate_filename("gen_output", "png")
|
284 |
st.session_state['processing']['gen'] = True
|
285 |
result = asyncio.run(process_image_gen(prompt, output_file))
|
|
|
286 |
st.image(result, caption="Generated Image", use_container_width=True)
|
287 |
st.success(f"Image saved to {output_file}")
|
288 |
st.session_state['processing']['gen'] = False
|
289 |
else:
|
290 |
st.warning("No images captured yet. Use Camera Snap or Download PDFs first!")
|
291 |
|
292 |
-
with tab5:
|
293 |
-
st.header("Test Line Drawings ✏️")
|
294 |
-
captured_files = get_gallery_files(["png"])
|
295 |
-
if captured_files:
|
296 |
-
selected_file = st.selectbox("Select Image", captured_files, key="line_select")
|
297 |
-
image = Image.open(selected_file)
|
298 |
-
st.image(image, caption="Input Image", use_container_width=True)
|
299 |
-
style = st.selectbox("Line Style", ["Fine", "Bold"], key="line_style")
|
300 |
-
if st.button("Run Line Drawing 🚀", key="line_run"):
|
301 |
-
output_file = generate_filename(f"line_{style.lower()}", "png")
|
302 |
-
st.session_state['processing']['line'] = True
|
303 |
-
result = asyncio.run(process_line_drawing(image, style.lower(), output_file))
|
304 |
-
st.image(result, caption=f"{style} Line Drawing", use_container_width=True)
|
305 |
-
st.success(f"Line drawing saved to {output_file}")
|
306 |
-
st.session_state['processing']['line'] = False
|
307 |
-
else:
|
308 |
-
st.warning("No images captured yet. Use Camera Snap or Download PDFs first!")
|
309 |
-
|
310 |
# Initial Gallery Update
|
311 |
update_gallery()
|
|
|
1 |
#!/usr/bin/env python3
|
2 |
import os
|
3 |
import glob
|
4 |
+
import base64
|
5 |
import time
|
6 |
+
import shutil
|
7 |
import streamlit as st
|
8 |
+
import pandas as pd
|
9 |
+
import torch
|
10 |
+
from transformers import AutoModelForCausalLM, AutoTokenizer, AutoModel
|
11 |
+
from diffusers import StableDiffusionPipeline
|
12 |
+
from torch.utils.data import Dataset, DataLoader
|
13 |
+
import csv
|
14 |
import fitz # PyMuPDF
|
15 |
import requests
|
16 |
from PIL import Image
|
|
|
|
|
17 |
import cv2
|
18 |
import numpy as np
|
19 |
import logging
|
20 |
import asyncio
|
21 |
import aiofiles
|
22 |
from io import BytesIO
|
23 |
+
from dataclasses import dataclass
|
24 |
+
from typing import Optional, Tuple
|
25 |
+
import zipfile
|
26 |
+
import math
|
27 |
+
import random
|
28 |
|
29 |
+
# Logging setup with custom buffer
|
30 |
logging.basicConfig(level=logging.INFO, format="%(asctime)s - %(levelname)s - %(message)s")
|
31 |
logger = logging.getLogger(__name__)
|
32 |
log_records = []
|
|
|
39 |
|
40 |
# Page Configuration
|
41 |
st.set_page_config(
|
42 |
+
page_title="AI Vision & SFT Titans 🚀",
|
43 |
page_icon="🤖",
|
44 |
layout="wide",
|
45 |
initial_sidebar_state="expanded",
|
46 |
+
menu_items={
|
47 |
+
'Get Help': 'https://huggingface.co/awacke1',
|
48 |
+
'Report a Bug': 'https://huggingface.co/spaces/awacke1',
|
49 |
+
'About': "AI Vision & SFT Titans: PDFs, OCR, Image Gen, Line Drawings, and SFT on CPU! 🌌"
|
50 |
+
}
|
51 |
)
|
52 |
|
53 |
# Initialize st.session_state
|
54 |
if 'captured_files' not in st.session_state:
|
55 |
st.session_state['captured_files'] = []
|
56 |
+
if 'builder' not in st.session_state:
|
57 |
+
st.session_state['builder'] = None
|
58 |
+
if 'model_loaded' not in st.session_state:
|
59 |
+
st.session_state['model_loaded'] = False
|
60 |
if 'processing' not in st.session_state:
|
61 |
st.session_state['processing'] = {}
|
62 |
+
if 'history' not in st.session_state:
|
63 |
+
st.session_state['history'] = []
|
64 |
+
|
65 |
+
# Model Configuration Classes
|
66 |
+
@dataclass
|
67 |
+
class ModelConfig:
|
68 |
+
name: str
|
69 |
+
base_model: str
|
70 |
+
size: str
|
71 |
+
domain: Optional[str] = None
|
72 |
+
model_type: str = "causal_lm"
|
73 |
+
@property
|
74 |
+
def model_path(self):
|
75 |
+
return f"models/{self.name}"
|
76 |
+
|
77 |
+
@dataclass
|
78 |
+
class DiffusionConfig:
|
79 |
+
name: str
|
80 |
+
base_model: str
|
81 |
+
size: str
|
82 |
+
@property
|
83 |
+
def model_path(self):
|
84 |
+
return f"diffusion_models/{self.name}"
|
85 |
+
|
86 |
+
# Datasets
|
87 |
+
class SFTDataset(Dataset):
|
88 |
+
def __init__(self, data, tokenizer, max_length=128):
|
89 |
+
self.data = data
|
90 |
+
self.tokenizer = tokenizer
|
91 |
+
self.max_length = max_length
|
92 |
+
def __len__(self):
|
93 |
+
return len(self.data)
|
94 |
+
def __getitem__(self, idx):
|
95 |
+
prompt = self.data[idx]["prompt"]
|
96 |
+
response = self.data[idx]["response"]
|
97 |
+
full_text = f"{prompt} {response}"
|
98 |
+
full_encoding = self.tokenizer(full_text, max_length=self.max_length, padding="max_length", truncation=True, return_tensors="pt")
|
99 |
+
prompt_encoding = self.tokenizer(prompt, max_length=self.max_length, padding=False, truncation=True, return_tensors="pt")
|
100 |
+
input_ids = full_encoding["input_ids"].squeeze()
|
101 |
+
attention_mask = full_encoding["attention_mask"].squeeze()
|
102 |
+
labels = input_ids.clone()
|
103 |
+
prompt_len = prompt_encoding["input_ids"].shape[1]
|
104 |
+
if prompt_len < self.max_length:
|
105 |
+
labels[:prompt_len] = -100
|
106 |
+
return {"input_ids": input_ids, "attention_mask": attention_mask, "labels": labels}
|
107 |
+
|
108 |
+
class DiffusionDataset(Dataset):
|
109 |
+
def __init__(self, images, texts):
|
110 |
+
self.images = images
|
111 |
+
self.texts = texts
|
112 |
+
def __len__(self):
|
113 |
+
return len(self.images)
|
114 |
+
def __getitem__(self, idx):
|
115 |
+
return {"image": self.images[idx], "text": self.texts[idx]}
|
116 |
+
|
117 |
+
# Model Builders
|
118 |
+
class ModelBuilder:
|
119 |
+
def __init__(self):
|
120 |
+
self.config = None
|
121 |
+
self.model = None
|
122 |
+
self.tokenizer = None
|
123 |
+
self.sft_data = None
|
124 |
+
self.jokes = ["Why did the AI go to therapy? Too many layers to unpack! 😂", "Training complete! Time for a binary coffee break. ☕"]
|
125 |
+
def load_model(self, model_path: str, config: Optional[ModelConfig] = None):
|
126 |
+
with st.spinner(f"Loading {model_path}... ⏳"):
|
127 |
+
self.model = AutoModelForCausalLM.from_pretrained(model_path)
|
128 |
+
self.tokenizer = AutoTokenizer.from_pretrained(model_path)
|
129 |
+
if self.tokenizer.pad_token is None:
|
130 |
+
self.tokenizer.pad_token = self.tokenizer.eos_token
|
131 |
+
if config:
|
132 |
+
self.config = config
|
133 |
+
self.model.to("cuda" if torch.cuda.is_available() else "cpu")
|
134 |
+
st.success(f"Model loaded! 🎉 {random.choice(self.jokes)}")
|
135 |
+
return self
|
136 |
+
def fine_tune_sft(self, csv_path: str, epochs: int = 3, batch_size: int = 4):
|
137 |
+
self.sft_data = []
|
138 |
+
with open(csv_path, "r") as f:
|
139 |
+
reader = csv.DictReader(f)
|
140 |
+
for row in reader:
|
141 |
+
self.sft_data.append({"prompt": row["prompt"], "response": row["response"]})
|
142 |
+
dataset = SFTDataset(self.sft_data, self.tokenizer)
|
143 |
+
dataloader = DataLoader(dataset, batch_size=batch_size, shuffle=True)
|
144 |
+
optimizer = torch.optim.AdamW(self.model.parameters(), lr=2e-5)
|
145 |
+
self.model.train()
|
146 |
+
device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
|
147 |
+
self.model.to(device)
|
148 |
+
for epoch in range(epochs):
|
149 |
+
with st.spinner(f"Training epoch {epoch + 1}/{epochs}... ⚙️"):
|
150 |
+
total_loss = 0
|
151 |
+
for batch in dataloader:
|
152 |
+
optimizer.zero_grad()
|
153 |
+
input_ids = batch["input_ids"].to(device)
|
154 |
+
attention_mask = batch["attention_mask"].to(device)
|
155 |
+
labels = batch["labels"].to(device)
|
156 |
+
outputs = self.model(input_ids=input_ids, attention_mask=attention_mask, labels=labels)
|
157 |
+
loss = outputs.loss
|
158 |
+
loss.backward()
|
159 |
+
optimizer.step()
|
160 |
+
total_loss += loss.item()
|
161 |
+
st.write(f"Epoch {epoch + 1} completed. Average loss: {total_loss / len(dataloader):.4f}")
|
162 |
+
st.success(f"SFT Fine-tuning completed! 🎉 {random.choice(self.jokes)}")
|
163 |
+
return self
|
164 |
+
def save_model(self, path: str):
|
165 |
+
with st.spinner("Saving model... 💾"):
|
166 |
+
os.makedirs(os.path.dirname(path), exist_ok=True)
|
167 |
+
self.model.save_pretrained(path)
|
168 |
+
self.tokenizer.save_pretrained(path)
|
169 |
+
st.success(f"Model saved at {path}! ✅")
|
170 |
+
def evaluate(self, prompt: str, status_container=None):
|
171 |
+
self.model.eval()
|
172 |
+
if status_container:
|
173 |
+
status_container.write("Preparing to evaluate... 🧠")
|
174 |
+
try:
|
175 |
+
with torch.no_grad():
|
176 |
+
inputs = self.tokenizer(prompt, return_tensors="pt", max_length=128, truncation=True).to(self.model.device)
|
177 |
+
outputs = self.model.generate(**inputs, max_new_tokens=50, do_sample=True, top_p=0.95, temperature=0.7)
|
178 |
+
return self.tokenizer.decode(outputs[0], skip_special_tokens=True)
|
179 |
+
except Exception as e:
|
180 |
+
if status_container:
|
181 |
+
status_container.error(f"Oops! Something broke: {str(e)} 💥")
|
182 |
+
return f"Error: {str(e)}"
|
183 |
+
|
184 |
+
class DiffusionBuilder:
|
185 |
+
def __init__(self):
|
186 |
+
self.config = None
|
187 |
+
self.pipeline = None
|
188 |
+
def load_model(self, model_path: str, config: Optional[DiffusionConfig] = None):
|
189 |
+
with st.spinner(f"Loading diffusion model {model_path}... ⏳"):
|
190 |
+
self.pipeline = StableDiffusionPipeline.from_pretrained(model_path, torch_dtype=torch.float32).to("cpu")
|
191 |
+
if config:
|
192 |
+
self.config = config
|
193 |
+
st.success(f"Diffusion model loaded! 🎨")
|
194 |
+
return self
|
195 |
+
def fine_tune_sft(self, images, texts, epochs=3):
|
196 |
+
dataset = DiffusionDataset(images, texts)
|
197 |
+
dataloader = DataLoader(dataset, batch_size=1, shuffle=True)
|
198 |
+
optimizer = torch.optim.AdamW(self.pipeline.unet.parameters(), lr=1e-5)
|
199 |
+
self.pipeline.unet.train()
|
200 |
+
for epoch in range(epochs):
|
201 |
+
with—for st.spinner(f"Training diffusion epoch {epoch + 1}/{epochs}... ⚙️"):
|
202 |
+
total_loss = 0
|
203 |
+
for batch in dataloader:
|
204 |
+
optimizer.zero_grad()
|
205 |
+
image = batch["image"][0].to(self.pipeline.device)
|
206 |
+
text = batch["text"][0]
|
207 |
+
latents = self.pipeline.vae.encode(torch.tensor(np.array(image)).permute(2, 0, 1).unsqueeze(0).float().to(self.pipeline.device)).latent_dist.sample()
|
208 |
+
noise = torch.randn_like(latents)
|
209 |
+
timesteps = torch.randint(0, self.pipeline.scheduler.num_train_timesteps, (latents.shape[0],), device=latents.device)
|
210 |
+
noisy_latents = self.pipeline.scheduler.add_noise(latents, noise, timesteps)
|
211 |
+
text_embeddings = self.pipeline.text_encoder(self.pipeline.tokenizer(text, return_tensors="pt").input_ids.to(self.pipeline.device))[0]
|
212 |
+
pred_noise = self.pipeline.unet(noisy_latents, timesteps, encoder_hidden_states=text_embeddings).sample
|
213 |
+
loss = torch.nn.functional.mse_loss(pred_noise, noise)
|
214 |
+
loss.backward()
|
215 |
+
optimizer.step()
|
216 |
+
total_loss += loss.item()
|
217 |
+
st.write(f"Epoch {epoch + 1} completed. Average loss: {total_loss / len(dataloader):.4f}")
|
218 |
+
st.success("Diffusion SFT Fine-tuning completed! 🎨")
|
219 |
+
return self
|
220 |
+
def save_model(self, path: str):
|
221 |
+
with st.spinner("Saving diffusion model... 💾"):
|
222 |
+
os.makedirs(os.path.dirname(path), exist_ok=True)
|
223 |
+
self.pipeline.save_pretrained(path)
|
224 |
+
st.success(f"Diffusion model saved at {path}! ✅")
|
225 |
+
def generate(self, prompt: str):
|
226 |
+
return self.pipeline(prompt, num_inference_steps=20).images[0]
|
227 |
|
228 |
# Utility Functions
|
229 |
+
def get_download_link(file_path, mime_type="text/plain", label="Download"):
|
230 |
+
with open(file_path, 'rb') as f:
|
231 |
+
data = f.read()
|
232 |
+
b64 = base64.b64encode(data).decode()
|
233 |
+
return f'<a href="data:{mime_type};base64,{b64}" download="{os.path.basename(file_path)}">{label} 📥</a>'
|
234 |
+
|
235 |
+
def zip_directory(directory_path, zip_path):
|
236 |
+
with zipfile.ZipFile(zip_path, 'w', zipfile.ZIP_DEFLATED) as zipf:
|
237 |
+
for root, _, files in os.walk(directory_path):
|
238 |
+
for file in files:
|
239 |
+
zipf.write(os.path.join(root, file), os.path.relpath(os.path.join(root, file), os.path.dirname(directory_path)))
|
240 |
+
|
241 |
+
def get_model_files(model_type="causal_lm"):
|
242 |
+
path = "models/*" if model_type == "causal_lm" else "diffusion_models/*"
|
243 |
+
return [d for d in glob.glob(path) if os.path.isdir(d)]
|
244 |
|
245 |
def get_gallery_files(file_types):
|
246 |
return sorted([f for ext in file_types for f in glob.glob(f"*.{ext}")])
|
247 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
248 |
def download_pdf(url, output_path):
|
249 |
try:
|
250 |
response = requests.get(url, stream=True, timeout=10)
|
|
|
257 |
logger.error(f"Failed to download {url}: {e}")
|
258 |
return False
|
259 |
|
260 |
+
# Mock Search Tool for RAG
|
261 |
+
def mock_search(query: str) -> str:
|
262 |
+
if "superhero" in query.lower():
|
263 |
+
return "Latest trends: Gold-plated Batman statues, VR superhero battles."
|
264 |
+
return "No relevant results found."
|
265 |
+
|
266 |
+
def mock_duckduckgo_search(query: str) -> str:
|
267 |
+
if "superhero party trends" in query.lower():
|
268 |
+
return """
|
269 |
+
Latest trends for 2025:
|
270 |
+
- Luxury decorations: Gold-plated Batman statues, holographic Avengers displays.
|
271 |
+
- Entertainment: Live stunt shows with Iron Man suits, VR superhero battles.
|
272 |
+
- Catering: Gourmet kryptonite-green cocktails, Thor’s hammer-shaped appetizers.
|
273 |
+
"""
|
274 |
+
return "No relevant results found."
|
275 |
+
|
276 |
+
# Agent Classes
|
277 |
+
class PartyPlannerAgent:
|
278 |
+
def __init__(self, model, tokenizer):
|
279 |
+
self.model = model
|
280 |
+
self.tokenizer = tokenizer
|
281 |
+
self.device = torch.device("cuda" if torch.cuda.is_available() else "cpu")
|
282 |
+
self.model.to(self.device)
|
283 |
+
def generate(self, prompt: str) -> str:
|
284 |
+
self.model.eval()
|
285 |
+
with torch.no_grad():
|
286 |
+
inputs = self.tokenizer(prompt, return_tensors="pt", max_length=128, truncation=True).to(self.device)
|
287 |
+
outputs = self.model.generate(**inputs, max_new_tokens=100, do_sample=True, top_p=0.95, temperature=0.7)
|
288 |
+
return self.tokenizer.decode(outputs[0], skip_special_tokens=True)
|
289 |
+
def plan_party(self, task: str) -> pd.DataFrame:
|
290 |
+
search_result = mock_duckduckgo_search("latest superhero party trends")
|
291 |
+
prompt = f"Given this context: '{search_result}'\n{task}"
|
292 |
+
plan_text = self.generate(prompt)
|
293 |
+
locations = {
|
294 |
+
"Wayne Manor": (42.3601, -71.0589),
|
295 |
+
"New York": (40.7128, -74.0060),
|
296 |
+
"Los Angeles": (34.0522, -118.2437),
|
297 |
+
"London": (51.5074, -0.1278)
|
298 |
+
}
|
299 |
+
wayne_coords = locations["Wayne Manor"]
|
300 |
+
travel_times = {loc: calculate_cargo_travel_time(coords, wayne_coords) for loc, coords in locations.items() if loc != "Wayne Manor"}
|
301 |
+
catchphrases = ["To the Batmobile!", "Avengers, assemble!", "I am Iron Man!", "By the power of Grayskull!"]
|
302 |
+
data = [
|
303 |
+
{"Location": "New York", "Travel Time (hrs)": travel_times["New York"], "Luxury Idea": "Gold-plated Batman statues", "Catchphrase": random.choice(catchphrases)},
|
304 |
+
{"Location": "Los Angeles", "Travel Time (hrs)": travel_times["Los Angeles"], "Luxury Idea": "Holographic Avengers displays", "Catchphrase": random.choice(catchphrases)},
|
305 |
+
{"Location": "London", "Travel Time (hrs)": travel_times["London"], "Luxury Idea": "Live stunt shows with Iron Man suits", "Catchphrase": random.choice(catchphrases)},
|
306 |
+
{"Location": "Wayne Manor", "Travel Time (hrs)": 0.0, "Luxury Idea": "VR superhero battles", "Catchphrase": random.choice(catchphrases)},
|
307 |
+
{"Location": "New York", "Travel Time (hrs)": travel_times["New York"], "Luxury Idea": "Gourmet kryptonite-green cocktails", "Catchphrase": random.choice(catchphrases)},
|
308 |
+
{"Location": "Los Angeles", "Travel Time (hrs)": travel_times["Los Angeles"], "Luxury Idea": "Thor’s hammer-shaped appetizers", "Catchphrase": random.choice(catchphrases)},
|
309 |
+
]
|
310 |
+
return pd.DataFrame(data)
|
311 |
+
|
312 |
+
class CVPartyPlannerAgent:
|
313 |
+
def __init__(self, pipeline):
|
314 |
+
self.pipeline = pipeline
|
315 |
+
def generate(self, prompt: str) -> Image.Image:
|
316 |
+
return self.pipeline(prompt, num_inference_steps=20).images[0]
|
317 |
+
def plan_party(self, task: str) -> pd.DataFrame:
|
318 |
+
search_result = mock_search("superhero party trends")
|
319 |
+
prompt = f"Given this context: '{search_result}'\n{task}"
|
320 |
+
data = [
|
321 |
+
{"Theme": "Batman", "Image Idea": "Gold-plated Batman statue"},
|
322 |
+
{"Theme": "Avengers", "Image Idea": "VR superhero battle scene"}
|
323 |
+
]
|
324 |
+
return pd.DataFrame(data)
|
325 |
+
|
326 |
+
def calculate_cargo_travel_time(origin_coords: Tuple[float, float], destination_coords: Tuple[float, float], cruising_speed_kmh: float = 750.0) -> float:
|
327 |
+
def to_radians(degrees: float) -> float:
|
328 |
+
return degrees * (math.pi / 180)
|
329 |
+
lat1, lon1 = map(to_radians, origin_coords)
|
330 |
+
lat2, lon2 = map(to_radians, destination_coords)
|
331 |
+
EARTH_RADIUS_KM = 6371.0
|
332 |
+
dlon = lon2 - lon1
|
333 |
+
dlat = lat2 - lat1
|
334 |
+
a = (math.sin(dlat / 2) ** 2 + math.cos(lat1) * math.cos(lat2) * math.sin(dlon / 2) ** 2)
|
335 |
+
c = 2 * math.asin(math.sqrt(a))
|
336 |
+
distance = EARTH_RADIUS_KM * c
|
337 |
+
actual_distance = distance * 1.1
|
338 |
+
flight_time = (actual_distance / cruising_speed_kmh) + 1.0
|
339 |
+
return round(flight_time, 2)
|
340 |
|
341 |
# Async Processing Functions
|
342 |
async def process_pdf_snapshot(pdf_path, mode="thumbnail"):
|
|
|
345 |
status.text(f"Processing PDF Snapshot ({mode})... (0s)")
|
346 |
doc = fitz.open(pdf_path)
|
347 |
output_files = []
|
|
|
348 |
if mode == "thumbnail":
|
349 |
page = doc[0]
|
350 |
+
pix = page.get_pixmap(matrix=fitz.Matrix(0.5, 0.5))
|
351 |
output_file = generate_filename("thumbnail", "png")
|
352 |
pix.save(output_file)
|
353 |
output_files.append(output_file)
|
354 |
elif mode == "twopage":
|
355 |
for i in range(min(2, len(doc))):
|
356 |
page = doc[i]
|
357 |
+
pix = page.get_pixmap(matrix=fitz.Matrix(1.0, 1.0))
|
358 |
output_file = generate_filename(f"twopage_{i}", "png")
|
359 |
pix.save(output_file)
|
360 |
output_files.append(output_file)
|
|
|
361 |
doc.close()
|
362 |
elapsed = int(time.time() - start_time)
|
363 |
status.text(f"PDF Snapshot ({mode}) completed in {elapsed}s!")
|
|
|
382 |
update_gallery()
|
383 |
return result
|
384 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
385 |
# Main App
|
386 |
+
st.title("AI Vision & SFT Titans 🚀")
|
387 |
|
388 |
+
# Sidebar
|
389 |
st.sidebar.header("Captured Files 📜")
|
390 |
gallery_size = st.sidebar.slider("Gallery Size", 1, 10, 4)
|
391 |
update_gallery()
|
392 |
|
393 |
+
st.sidebar.subheader("Model Management 🗂️")
|
394 |
+
model_type = st.sidebar.selectbox("Model Type", ["Causal LM", "Diffusion"], key="sidebar_model_type")
|
395 |
+
model_dirs = get_model_files(model_type)
|
396 |
+
selected_model = st.sidebar.selectbox("Select Saved Model", ["None"] + model_dirs, key="sidebar_model_select")
|
397 |
+
if selected_model != "None" and st.sidebar.button("Load Model 📂"):
|
398 |
+
builder = ModelBuilder() if model_type == "Causal LM" else DiffusionBuilder()
|
399 |
+
config = (ModelConfig if model_type == "Causal LM" else DiffusionConfig)(name=os.path.basename(selected_model), base_model="unknown", size="small")
|
400 |
+
builder.load_model(selected_model, config)
|
401 |
+
st.session_state['builder'] = builder
|
402 |
+
st.session_state['model_loaded'] = True
|
403 |
+
st.rerun()
|
404 |
+
|
405 |
st.sidebar.subheader("Action Logs 📜")
|
406 |
log_container = st.sidebar.empty()
|
407 |
with log_container:
|
408 |
for record in log_records:
|
409 |
st.write(f"{record.asctime} - {record.levelname} - {record.message}")
|
410 |
|
411 |
+
st.sidebar.subheader("History 📜")
|
412 |
+
history_container = st.sidebar.empty()
|
413 |
+
with history_container:
|
414 |
+
for entry in st.session_state['history'][-5:]:
|
415 |
+
st.write(entry)
|
416 |
+
|
417 |
# Tabs
|
418 |
+
tab1, tab2, tab3, tab4, tab5, tab6, tab7, tab8 = st.tabs([
|
419 |
+
"Camera Snap 📷", "Download PDFs 📥", "Build Titan 🌱", "Fine-Tune Titan 🔧",
|
420 |
+
"Test Titan 🧪", "Agentic RAG Party 🌐", "Test OCR 🔍", "Test Image Gen 🎨"
|
421 |
+
])
|
422 |
|
423 |
with tab1:
|
424 |
st.header("Camera Snap 📷")
|
|
|
434 |
st.image(Image.open(filename), caption=filename, use_container_width=True)
|
435 |
logger.info(f"Saved snapshot from Camera 0: {filename}")
|
436 |
st.session_state['captured_files'].append(filename)
|
437 |
+
st.session_state['history'].append(f"Snapshot from Cam 0: {filename}")
|
438 |
update_gallery()
|
439 |
with cols[1]:
|
440 |
cam1_img = st.camera_input("Take a picture - Cam 1", key="cam1")
|
|
|
446 |
st.image(Image.open(filename), caption=filename, use_container_width=True)
|
447 |
logger.info(f"Saved snapshot from Camera 1: {filename}")
|
448 |
st.session_state['captured_files'].append(filename)
|
449 |
+
st.session_state['history'].append(f"Snapshot from Cam 1: {filename}")
|
450 |
update_gallery()
|
451 |
|
452 |
st.subheader("Burst Capture")
|
|
|
465 |
f.write(img.getvalue())
|
466 |
st.session_state['burst_frames'].append(filename)
|
467 |
logger.info(f"Saved burst frame {i}: {filename}")
|
468 |
+
st.session_state['history'].append(f"Burst frame {i}: {filename}")
|
469 |
st.image(Image.open(filename), caption=filename, use_container_width=True)
|
470 |
time.sleep(0.5)
|
471 |
st.session_state['captured_files'].extend([f for f in st.session_state['burst_frames'] if f not in st.session_state['captured_files']])
|
|
|
483 |
pdf_path = generate_filename("downloaded", "pdf")
|
484 |
if download_pdf(url, pdf_path):
|
485 |
logger.info(f"Downloaded PDF from {url} to {pdf_path}")
|
486 |
+
st.session_state['history'].append(f"Downloaded PDF: {pdf_path}")
|
487 |
snapshots = asyncio.run(process_pdf_snapshot(pdf_path, mode.lower().replace(" ", "")))
|
488 |
for snapshot in snapshots:
|
489 |
st.image(Image.open(snapshot), caption=snapshot, use_container_width=True)
|
|
|
491 |
st.error(f"Failed to download {url}")
|
492 |
|
493 |
with tab3:
|
494 |
+
st.header("Build Titan 🌱")
|
495 |
+
model_type = st.selectbox("Model Type", ["Causal LM", "Diffusion"], key="build_type")
|
496 |
+
base_model = st.selectbox("Select Tiny Model",
|
497 |
+
["HuggingFaceTB/SmolLM-135M", "Qwen/Qwen1.5-0.5B-Chat"] if model_type == "Causal LM" else
|
498 |
+
["OFA-Sys/small-stable-diffusion-v0", "stabilityai/stable-diffusion-2-base"])
|
499 |
+
model_name = st.text_input("Model Name", f"tiny-titan-{int(time.time())}")
|
500 |
+
domain = st.text_input("Target Domain", "general")
|
501 |
+
if st.button("Download Model ⬇️"):
|
502 |
+
config = (ModelConfig if model_type == "Causal LM" else DiffusionConfig)(name=model_name, base_model=base_model, size="small", domain=domain)
|
503 |
+
builder = ModelBuilder() if model_type == "Causal LM" else DiffusionBuilder()
|
504 |
+
builder.load_model(base_model, config)
|
505 |
+
builder.save_model(config.model_path)
|
506 |
+
st.session_state['builder'] = builder
|
507 |
+
st.session_state['model_loaded'] = True
|
508 |
+
st.session_state['history'].append(f"Built {model_type} model: {model_name}")
|
509 |
+
st.success(f"Model downloaded and saved to {config.model_path}! 🎉")
|
510 |
+
st.rerun()
|
511 |
+
|
512 |
+
with tab4:
|
513 |
+
st.header("Fine-Tune Titan 🔧")
|
514 |
+
if 'builder' not in st.session_state or not st.session_state.get('model_loaded', False):
|
515 |
+
st.warning("Please build or load a Titan first! ⚠️")
|
516 |
+
else:
|
517 |
+
if isinstance(st.session_state['builder'], ModelBuilder):
|
518 |
+
if st.button("Generate Sample CSV 📝"):
|
519 |
+
sample_data = [
|
520 |
+
{"prompt": "What is AI?", "response": "AI is artificial intelligence, simulating human smarts in machines."},
|
521 |
+
{"prompt": "Explain machine learning", "response": "Machine learning is AI’s gym where models bulk up on data."},
|
522 |
+
]
|
523 |
+
csv_path = f"sft_data_{int(time.time())}.csv"
|
524 |
+
with open(csv_path, "w", newline="") as f:
|
525 |
+
writer = csv.DictWriter(f, fieldnames=["prompt", "response"])
|
526 |
+
writer.writeheader()
|
527 |
+
writer.writerows(sample_data)
|
528 |
+
st.markdown(get_download_link(csv_path, "text/csv", "Download Sample CSV"), unsafe_allow_html=True)
|
529 |
+
st.success(f"Sample CSV generated as {csv_path}! ✅")
|
530 |
+
|
531 |
+
uploaded_csv = st.file_uploader("Upload CSV for SFT", type="csv")
|
532 |
+
if uploaded_csv and st.button("Fine-Tune with Uploaded CSV 🔄"):
|
533 |
+
csv_path = f"uploaded_sft_data_{int(time.time())}.csv"
|
534 |
+
with open(csv_path, "wb") as f:
|
535 |
+
f.write(uploaded_csv.read())
|
536 |
+
new_model_name = f"{st.session_state['builder'].config.name}-sft-{int(time.time())}"
|
537 |
+
new_config = ModelConfig(name=new_model_name, base_model=st.session_state['builder'].config.base_model, size="small", domain=st.session_state['builder'].config.domain)
|
538 |
+
st.session_state['builder'].config = new_config
|
539 |
+
st.session_state['builder'].fine_tune_sft(csv_path)
|
540 |
+
st.session_state['builder'].save_model(new_config.model_path)
|
541 |
+
zip_path = f"{new_config.model_path}.zip"
|
542 |
+
zip_directory(new_config.model_path, zip_path)
|
543 |
+
st.session_state['history'].append(f"Fine-tuned Causal LM: {new_model_name}")
|
544 |
+
st.markdown(get_download_link(zip_path, "application/zip", "Download Fine-Tuned Titan"), unsafe_allow_html=True)
|
545 |
+
st.rerun()
|
546 |
+
elif isinstance(st.session_state['builder'], DiffusionBuilder):
|
547 |
+
captured_files = get_gallery_files(["png"])
|
548 |
+
if len(captured_files) >= 2:
|
549 |
+
demo_data = [{"image": img, "text": f"Superhero {os.path.basename(img).split('.')[0]}"} for img in captured_files[:min(len(captured_files), slice_count)]]
|
550 |
+
edited_data = st.data_editor(pd.DataFrame(demo_data), num_rows="dynamic")
|
551 |
+
if st.button("Fine-Tune with Dataset 🔄"):
|
552 |
+
images = [Image.open(row["image"]) for _, row in edited_data.iterrows()]
|
553 |
+
texts = [row["text"] for _, row in edited_data.iterrows()]
|
554 |
+
new_model_name = f"{st.session_state['builder'].config.name}-sft-{int(time.time())}"
|
555 |
+
new_config = DiffusionConfig(name=new_model_name, base_model=st.session_state['builder'].config.base_model, size="small")
|
556 |
+
st.session_state['builder'].config = new_config
|
557 |
+
st.session_state['builder'].fine_tune_sft(images, texts)
|
558 |
+
st.session_state['builder'].save_model(new_config.model_path)
|
559 |
+
zip_path = f"{new_config.model_path}.zip"
|
560 |
+
zip_directory(new_config.model_path, zip_path)
|
561 |
+
st.session_state['history'].append(f"Fine-tuned Diffusion: {new_model_name}")
|
562 |
+
st.markdown(get_download_link(zip_path, "application/zip", "Download Fine-Tuned Diffusion Model"), unsafe_allow_html=True)
|
563 |
+
csv_path = f"sft_dataset_{int(time.time())}.csv"
|
564 |
+
with open(csv_path, "w", newline="") as f:
|
565 |
+
writer = csv.writer(f)
|
566 |
+
writer.writerow(["image", "text"])
|
567 |
+
for _, row in edited_data.iterrows():
|
568 |
+
writer.writerow([row["image"], row["text"]])
|
569 |
+
st.markdown(get_download_link(csv_path, "text/csv", "Download SFT Dataset CSV"), unsafe_allow_html=True)
|
570 |
+
|
571 |
+
with tab5:
|
572 |
+
st.header("Test Titan 🧪")
|
573 |
+
if 'builder' not in st.session_state or not st.session_state.get('model_loaded', False):
|
574 |
+
st.warning("Please build or load a Titan first! ⚠️")
|
575 |
+
else:
|
576 |
+
if isinstance(st.session_state['builder'], ModelBuilder):
|
577 |
+
if st.session_state['builder'].sft_data:
|
578 |
+
st.write("Testing with SFT Data:")
|
579 |
+
for item in st.session_state['builder'].sft_data[:3]:
|
580 |
+
prompt = item["prompt"]
|
581 |
+
expected = item["response"]
|
582 |
+
status_container = st.empty()
|
583 |
+
generated = st.session_state['builder'].evaluate(prompt, status_container)
|
584 |
+
st.write(f"**Prompt**: {prompt}")
|
585 |
+
st.write(f"**Expected**: {expected}")
|
586 |
+
st.write(f"**Generated**: {generated}")
|
587 |
+
st.write("---")
|
588 |
+
status_container.empty()
|
589 |
+
test_prompt = st.text_area("Enter Test Prompt", "What is AI?")
|
590 |
+
if st.button("Run Test ▶️"):
|
591 |
+
status_container = st.empty()
|
592 |
+
result = st.session_state['builder'].evaluate(test_prompt, status_container)
|
593 |
+
st.session_state['history'].append(f"Causal LM Test: {test_prompt} -> {result}")
|
594 |
+
st.write(f"**Generated Response**: {result}")
|
595 |
+
status_container.empty()
|
596 |
+
elif isinstance(st.session_state['builder'], DiffusionBuilder):
|
597 |
+
test_prompt = st.text_area("Enter Test Prompt", "Neon Batman")
|
598 |
+
if st.button("Run Test ▶️"):
|
599 |
+
image = st.session_state['builder'].generate(test_prompt)
|
600 |
+
output_file = generate_filename("diffusion_test", "png")
|
601 |
+
image.save(output_file)
|
602 |
+
st.session_state['captured_files'].append(output_file)
|
603 |
+
st.session_state['history'].append(f"Diffusion Test: {test_prompt} -> {output_file}")
|
604 |
+
st.image(image, caption="Generated Image")
|
605 |
+
update_gallery()
|
606 |
+
|
607 |
+
with tab6:
|
608 |
+
st.header("Agentic RAG Party 🌐")
|
609 |
+
if 'builder' not in st.session_state or not st.session_state.get('model_loaded', False):
|
610 |
+
st.warning("Please build or load a Titan first! ⚠️")
|
611 |
+
else:
|
612 |
+
if isinstance(st.session_state['builder'], ModelBuilder):
|
613 |
+
if st.button("Run NLP RAG Demo 🎉"):
|
614 |
+
agent = PartyPlannerAgent(st.session_state['builder'].model, st.session_state['builder'].tokenizer)
|
615 |
+
task = "Plan a luxury superhero-themed party at Wayne Manor."
|
616 |
+
plan_df = agent.plan_party(task)
|
617 |
+
st.session_state['history'].append(f"NLP RAG Demo: Planned party at Wayne Manor")
|
618 |
+
st.dataframe(plan_df)
|
619 |
+
elif isinstance(st.session_state['builder'], DiffusionBuilder):
|
620 |
+
if st.button("Run CV RAG Demo 🎉"):
|
621 |
+
agent = CVPartyPlannerAgent(st.session_state['builder'].pipeline)
|
622 |
+
task = "Generate images for a luxury superhero-themed party."
|
623 |
+
plan_df = agent.plan_party(task)
|
624 |
+
st.session_state['history'].append(f"CV RAG Demo: Generated party images")
|
625 |
+
st.dataframe(plan_df)
|
626 |
+
for _, row in plan_df.iterrows():
|
627 |
+
image = agent.generate(row["Image Idea"])
|
628 |
+
output_file = generate_filename(f"cv_rag_{row['Theme'].lower()}", "png")
|
629 |
+
image.save(output_file)
|
630 |
+
st.session_state['captured_files'].append(output_file)
|
631 |
+
st.image(image, caption=f"{row['Theme']} - {row['Image Idea']}")
|
632 |
+
update_gallery()
|
633 |
+
|
634 |
+
with tab7:
|
635 |
st.header("Test OCR 🔍")
|
636 |
captured_files = get_gallery_files(["png"])
|
637 |
if captured_files:
|
|
|
642 |
output_file = generate_filename("ocr_output", "txt")
|
643 |
st.session_state['processing']['ocr'] = True
|
644 |
result = asyncio.run(process_ocr(image, output_file))
|
645 |
+
st.session_state['history'].append(f"OCR Test: {selected_file} -> {output_file}")
|
646 |
st.text_area("OCR Result", result, height=200, key="ocr_result")
|
647 |
st.success(f"OCR output saved to {output_file}")
|
648 |
st.session_state['processing']['ocr'] = False
|
649 |
else:
|
650 |
st.warning("No images captured yet. Use Camera Snap or Download PDFs first!")
|
651 |
|
652 |
+
with tab8:
|
653 |
st.header("Test Image Gen 🎨")
|
654 |
captured_files = get_gallery_files(["png"])
|
655 |
if captured_files:
|
|
|
661 |
output_file = generate_filename("gen_output", "png")
|
662 |
st.session_state['processing']['gen'] = True
|
663 |
result = asyncio.run(process_image_gen(prompt, output_file))
|
664 |
+
st.session_state['history'].append(f"Image Gen Test: {prompt} -> {output_file}")
|
665 |
st.image(result, caption="Generated Image", use_container_width=True)
|
666 |
st.success(f"Image saved to {output_file}")
|
667 |
st.session_state['processing']['gen'] = False
|
668 |
else:
|
669 |
st.warning("No images captured yet. Use Camera Snap or Download PDFs first!")
|
670 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
671 |
# Initial Gallery Update
|
672 |
update_gallery()
|