Update app.py
Browse files
app.py
CHANGED
@@ -1,134 +1,8 @@
|
|
1 |
-
|
2 |
-
st.title("π Cutting-Edge ML Outline Generator")
|
3 |
-
|
4 |
-
col1, col2 = st.columns(2)
|
5 |
-
|
6 |
-
with col1:
|
7 |
-
st.header("π Markdown Outline")
|
8 |
-
|
9 |
-
# Display the markdown content
|
10 |
-
st.markdown(ml_markdown)
|
11 |
-
|
12 |
-
# Create a download button for the markdown file
|
13 |
-
st.download_button(
|
14 |
-
label="Download Markdown",
|
15 |
-
data=ml_markdown,
|
16 |
-
file_name="ml_outline.md",
|
17 |
-
mime="text/markdown"
|
18 |
-
)
|
19 |
-
|
20 |
-
# Show the markdown source code in an expandable section
|
21 |
-
with st.expander("View Markdown Source"):
|
22 |
-
st.code(ml_markdown, language="markdown")
|
23 |
-
|
24 |
-
with col2:
|
25 |
-
st.header("π PDF Preview & Demos")
|
26 |
-
|
27 |
-
# Library Demos
|
28 |
-
st.subheader("Library Demos")
|
29 |
-
if st.button("Run PDF Library Demos"):
|
30 |
-
with st.spinner("Running demos..."):
|
31 |
-
# Create tabs for each demo
|
32 |
-
demo_tabs = st.tabs(["PikePDF", "FPDF", "PyMuPDF", "Image Demo"])
|
33 |
-
|
34 |
-
with demo_tabs[0]:
|
35 |
-
# pikepdf demo
|
36 |
-
pike_pdf = demo_pikepdf()
|
37 |
-
st.download_button("Download pikepdf Demo", pike_pdf, "pikepdf_demo.pdf")
|
38 |
-
st.write("PikePDF demo created successfully!")
|
39 |
-
st.info("This PDF contains the multilevel markdown outline in a two-column layout.")
|
40 |
-
|
41 |
-
with demo_tabs[1]:
|
42 |
-
# fpdf demo
|
43 |
-
fpdf_pdf = demo_fpdf()
|
44 |
-
st.download_button("Download fpdf Demo", fpdf_pdf, "fpdf_demo.pdf")
|
45 |
-
st.write("FPDF demo created successfully!")
|
46 |
-
st.info("This PDF contains the multilevel markdown outline in a two-column layout.")
|
47 |
-
|
48 |
-
with demo_tabs[2]:
|
49 |
-
# pymupdf demo
|
50 |
-
pymupdf_pdf = demo_pymupdf()
|
51 |
-
st.download_button("Download pymupdf Demo", pymupdf_pdf, "pymupdf_demo.pdf")
|
52 |
-
st.write("PyMuPDF demo created successfully!")
|
53 |
-
st.info("This PDF contains the multilevel markdown outline in a two-column layout.")
|
54 |
-
|
55 |
-
with demo_tabs[3]:
|
56 |
-
# Image demo
|
57 |
-
img_data = demo_image_capture()
|
58 |
-
st.image(img_data, caption="Demo Image (Camera simulation)")
|
59 |
-
|
60 |
-
# Main PDF Generation
|
61 |
-
st.subheader("Main Outline PDF")
|
62 |
-
if st.button("Generate Main PDF"):
|
63 |
-
with st.spinner("Generating PDF..."):
|
64 |
-
try:
|
65 |
-
pdf_bytes = create_main_pdf(ml_markdown)
|
66 |
-
|
67 |
-
st.download_button(
|
68 |
-
label="Download Main PDF",
|
69 |
-
data=pdf_bytes,
|
70 |
-
file_name="ml_outline.pdf",
|
71 |
-
mime="application/pdf"
|
72 |
-
)
|
73 |
-
|
74 |
-
# Display the PDF in the app
|
75 |
-
base64_pdf = base64.b64encode(pdf_bytes).decode('utf-8')
|
76 |
-
pdf_display = f'''
|
77 |
-
<embed
|
78 |
-
src="data:application/pdf;base64,{base64_pdf}"
|
79 |
-
width="100%"
|
80 |
-
height="400px"
|
81 |
-
type="application/pdf">
|
82 |
-
'''
|
83 |
-
st.markdown(pdf_display, unsafe_allow_html=True)
|
84 |
-
|
85 |
-
st.success("PDF generated successfully! The PDF displays the multilevel markdown outline in a two-column layout.")
|
86 |
-
except Exception as e:
|
87 |
-
st.error(f"Error generating PDF: {str(e)}")
|
88 |
-
|
89 |
-
# Show the PDF rendering code in an expandable section
|
90 |
-
with st.expander("View PDF Rendering Code"):
|
91 |
-
st.code("""
|
92 |
-
# Process multilevel markdown for PDF output
|
93 |
-
def markdown_to_pdf_content(markdown_text):
|
94 |
-
# Convert markdown headers to styled text for PDF
|
95 |
-
lines = markdown_text.strip().split('\\n')
|
96 |
-
pdf_content = []
|
97 |
-
|
98 |
-
for line in lines:
|
99 |
-
if line.startswith('# '):
|
100 |
-
# Main header - will be handled separately
|
101 |
-
pass
|
102 |
-
elif line.startswith('## '):
|
103 |
-
# Section header - add as a bold item
|
104 |
-
section = line.replace('## ', '').strip()
|
105 |
-
pdf_content.append(f"<b>{section}</b>")
|
106 |
-
elif line.startswith('- '):
|
107 |
-
# List item - add as a normal item
|
108 |
-
item = line.replace('- ', '').strip()
|
109 |
-
pdf_content.append(item)
|
110 |
-
|
111 |
-
# Split the content for two columns
|
112 |
-
mid_point = len(pdf_content) // 2
|
113 |
-
left_column = pdf_content[:mid_point]
|
114 |
-
right_column = pdf_content[mid_point:]
|
115 |
-
|
116 |
-
return left_column, right_column
|
117 |
-
""", language="python")
|
118 |
-
|
119 |
-
# Add custom CSS for better appearance
|
120 |
-
st.markdown("""
|
121 |
-
<style>
|
122 |
-
.stButton>button {
|
123 |
-
background-color: #4CAF50;
|
124 |
-
color: white;
|
125 |
-
font-weight: bold;
|
126 |
-
}
|
127 |
-
.stTabsimport streamlit as st
|
128 |
import base64
|
129 |
from reportlab.lib.pagesizes import A4
|
130 |
-
from reportlab.platypus import SimpleDocTemplate, Paragraph, Spacer
|
131 |
-
from reportlab.lib.styles import getSampleStyleSheet
|
132 |
from reportlab.lib import colors
|
133 |
import pikepdf
|
134 |
import fpdf
|
@@ -138,91 +12,76 @@ import numpy as np
|
|
138 |
from PIL import Image
|
139 |
import io
|
140 |
import os
|
|
|
141 |
|
142 |
# Define the ML outline as a markdown string for multilevel content
|
143 |
ml_markdown = """# Cutting-Edge ML Outline
|
144 |
|
145 |
## Core ML Techniques
|
146 |
-
|
147 |
-
|
148 |
-
|
149 |
-
|
150 |
|
151 |
-
|
152 |
-
|
153 |
-
|
154 |
-
|
155 |
|
156 |
-
|
157 |
-
|
158 |
-
|
159 |
-
|
160 |
|
161 |
## Training Methods
|
162 |
-
|
163 |
-
|
164 |
-
|
165 |
-
|
166 |
|
167 |
-
|
168 |
-
|
169 |
-
|
170 |
-
|
171 |
|
172 |
-
|
173 |
-
|
174 |
-
|
175 |
-
|
176 |
|
177 |
## Optimization & Deployment
|
178 |
-
|
179 |
-
|
180 |
-
|
181 |
-
|
182 |
|
183 |
-
|
184 |
-
|
185 |
-
|
186 |
-
|
187 |
|
188 |
-
|
189 |
-
|
190 |
-
|
191 |
-
|
192 |
|
193 |
## Knowledge Representation
|
194 |
-
|
195 |
-
|
196 |
-
|
197 |
-
|
198 |
-
|
199 |
-
|
200 |
-
|
201 |
-
|
202 |
-
|
203 |
-
|
204 |
-
|
205 |
-
|
206 |
-
|
207 |
-
|
208 |
"""
|
209 |
|
210 |
-
# For compatibility with previous code, also maintain the list version
|
211 |
-
ml_outline = [
|
212 |
-
"π 1. Mixture of Experts (MoE)",
|
213 |
-
"π₯ 2. Supervised Fine-Tuning (SFT) using PyTorch",
|
214 |
-
"π€ 3. Large Language Models (LLM) using Transformers",
|
215 |
-
"π 4. Self-Rewarding Learning using NPS 0-10 and Verbatims",
|
216 |
-
"π 5. Reinforcement Learning from Human Feedback (RLHF)",
|
217 |
-
"π 6. MergeKit: Merging Models to Same Embedding Space",
|
218 |
-
"π 7. DistillKit: Model Size Reduction with Spectrum Analysis",
|
219 |
-
"π§ 8. Agentic RAG Agents using Document Inputs",
|
220 |
-
"β³ 9. Longitudinal Data Summarization from Multiple Docs",
|
221 |
-
"π 10. Knowledge Extraction using Markdown Knowledge Graphs",
|
222 |
-
"πΊοΈ 11. Knowledge Mapping with Mermaid Diagrams",
|
223 |
-
"π» 12. ML Code Generation with Streamlit/Gradio/HTML5+JS"
|
224 |
-
]
|
225 |
-
|
226 |
# Process multilevel markdown for PDF output
|
227 |
def markdown_to_pdf_content(markdown_text):
|
228 |
"""Convert markdown text to a format suitable for PDF generation"""
|
@@ -231,25 +90,49 @@ def markdown_to_pdf_content(markdown_text):
|
|
231 |
# Convert markdown headers to styled text for PDF
|
232 |
lines = markdown_text.strip().split('\n')
|
233 |
pdf_content = []
|
|
|
|
|
|
|
234 |
|
235 |
for line in lines:
|
|
|
|
|
|
|
|
|
236 |
if line.startswith('# '):
|
237 |
# Main header - will be handled separately in the PDF generation
|
238 |
pass
|
239 |
elif line.startswith('## '):
|
240 |
# Section header - add as a bold item
|
|
|
|
|
|
|
|
|
|
|
|
|
241 |
section = line.replace('## ', '').strip()
|
242 |
pdf_content.append(f"<b>{section}</b>")
|
243 |
-
|
244 |
-
|
245 |
-
|
246 |
-
|
247 |
-
|
248 |
-
|
249 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
250 |
|
251 |
-
#
|
252 |
-
|
|
|
253 |
|
254 |
# Split the content for two columns
|
255 |
mid_point = len(pdf_content) // 2
|
@@ -271,7 +154,8 @@ def demo_pikepdf():
|
|
271 |
# Set up fonts and colors
|
272 |
title_font = "helv-b"
|
273 |
section_font = "helv-b"
|
274 |
-
|
|
|
275 |
blue_color = (0, 0, 0.8)
|
276 |
black_color = (0, 0, 0)
|
277 |
|
@@ -280,25 +164,65 @@ def demo_pikepdf():
|
|
280 |
|
281 |
# First column
|
282 |
x1, y1 = 50, 80
|
283 |
-
|
284 |
-
|
|
|
|
|
285 |
# Section header
|
|
|
|
|
|
|
|
|
286 |
text = item.replace('<b>', '').replace('</b>', '')
|
287 |
-
page.insert_text((x1,
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
288 |
else:
|
289 |
-
#
|
290 |
-
page.insert_text((x1,
|
|
|
291 |
|
292 |
# Second column
|
293 |
x2, y2 = 450, 80
|
294 |
-
|
295 |
-
|
|
|
|
|
296 |
# Section header
|
|
|
|
|
|
|
|
|
297 |
text = item.replace('<b>', '').replace('</b>', '')
|
298 |
-
page.insert_text((x2,
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
299 |
else:
|
300 |
-
#
|
301 |
-
page.insert_text((x2,
|
|
|
302 |
|
303 |
# Draw a dividing line
|
304 |
page.draw_line((421, 70), (421, 550))
|
@@ -330,45 +254,57 @@ def demo_fpdf():
|
|
330 |
pdf.cell(0, 10, txt="Cutting-Edge ML Outline (FPDF Demo)", ln=True, align='C')
|
331 |
pdf.ln(10)
|
332 |
|
333 |
-
#
|
334 |
x_col1 = 20
|
335 |
-
y_start = pdf.get_y()
|
336 |
-
|
337 |
-
for item in left_column:
|
338 |
-
if item.startswith('<b>'):
|
339 |
-
# Section header
|
340 |
-
text = item.replace('<b>', '').replace('</b>', '')
|
341 |
-
pdf.set_font("Arial", 'B', size=14)
|
342 |
-
pdf.set_text_color(0, 0, 128) # Dark blue
|
343 |
-
else:
|
344 |
-
# Normal item
|
345 |
-
text = item
|
346 |
-
pdf.set_font("Arial", size=11)
|
347 |
-
pdf.set_text_color(0, 0, 0) # Black
|
348 |
-
|
349 |
-
pdf.set_x(x_col1)
|
350 |
-
pdf.multi_cell(180, 10, txt=text, align='L')
|
351 |
-
pdf.ln(2)
|
352 |
-
|
353 |
-
# Second column
|
354 |
x_col2 = pdf.w / 2 + 10
|
355 |
-
pdf.
|
356 |
|
357 |
-
|
358 |
-
|
359 |
-
|
360 |
-
text = item.replace('<b>', '').replace('</b>', '')
|
361 |
-
pdf.set_font("Arial", 'B', size=14)
|
362 |
-
pdf.set_text_color(0, 0, 128) # Dark blue
|
363 |
-
else:
|
364 |
-
# Normal item
|
365 |
-
text = item
|
366 |
-
pdf.set_font("Arial", size=11)
|
367 |
-
pdf.set_text_color(0, 0, 0) # Black
|
368 |
|
369 |
-
|
370 |
-
|
371 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
372 |
|
373 |
# Draw a dividing line
|
374 |
pdf.line(pdf.w/2, 30, pdf.w/2, 280)
|
@@ -389,7 +325,8 @@ def demo_pymupdf():
|
|
389 |
# Set up fonts and colors
|
390 |
title_font = "helv-b"
|
391 |
section_font = "helv-b"
|
392 |
-
|
|
|
393 |
blue_color = (0, 0, 0.8)
|
394 |
black_color = (0, 0, 0)
|
395 |
|
@@ -398,42 +335,65 @@ def demo_pymupdf():
|
|
398 |
|
399 |
# First column
|
400 |
x1, y1 = 50, 80
|
401 |
-
line_height = 25
|
402 |
current_y = y1
|
403 |
|
404 |
for item in left_column:
|
405 |
-
if item.startswith('<b>'):
|
|
|
406 |
# Add extra space before sections (except the first one)
|
407 |
if current_y > y1:
|
408 |
current_y += 10
|
409 |
|
410 |
-
# Section header
|
411 |
text = item.replace('<b>', '').replace('</b>', '')
|
412 |
page.insert_text((x1, current_y), text, fontname=section_font, fontsize=14, color=blue_color)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
413 |
else:
|
414 |
-
#
|
415 |
-
page.insert_text((x1
|
416 |
-
|
417 |
-
current_y += line_height
|
418 |
|
419 |
# Second column
|
420 |
x2, y2 = 450, 80
|
421 |
current_y = y2
|
422 |
|
423 |
for item in right_column:
|
424 |
-
if item.startswith('<b>'):
|
|
|
425 |
# Add extra space before sections (except the first one)
|
426 |
if current_y > y2:
|
427 |
current_y += 10
|
428 |
|
429 |
-
# Section header
|
430 |
text = item.replace('<b>', '').replace('</b>', '')
|
431 |
page.insert_text((x2, current_y), text, fontname=section_font, fontsize=14, color=blue_color)
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
432 |
else:
|
433 |
-
#
|
434 |
-
page.insert_text((x2
|
435 |
-
|
436 |
-
current_y += line_height
|
437 |
|
438 |
# Draw a dividing line
|
439 |
page.draw_line((421, 70), (421, 550))
|
@@ -518,7 +478,15 @@ def create_main_pdf(markdown_text):
|
|
518 |
parent=styles['Normal'],
|
519 |
fontSize=11,
|
520 |
leading=14,
|
521 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
522 |
)
|
523 |
|
524 |
# Add title
|
@@ -526,31 +494,55 @@ def create_main_pdf(markdown_text):
|
|
526 |
story.append(Spacer(1, 20))
|
527 |
|
528 |
# Prepare data for table
|
529 |
-
|
530 |
-
max_rows = max(len(left_column), len(right_column))
|
531 |
-
|
532 |
-
# Process left and right columns into paragraphs
|
533 |
-
left_paragraphs = []
|
534 |
for item in left_column:
|
535 |
-
if item.startswith('<b>'):
|
|
|
536 |
text = item.replace('<b>', '').replace('</b>', '')
|
537 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
538 |
else:
|
539 |
-
|
|
|
540 |
|
541 |
-
|
542 |
for item in right_column:
|
543 |
-
if item.startswith('<b>'):
|
|
|
544 |
text = item.replace('<b>', '').replace('</b>', '')
|
545 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
546 |
else:
|
547 |
-
|
548 |
-
|
549 |
-
|
550 |
-
|
551 |
-
|
552 |
-
|
553 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
554 |
|
555 |
# Calculate column widths
|
556 |
col_width = (A4[1] - 120) / 2.0 # Page width minus margins divided by 2
|
@@ -587,16 +579,21 @@ col1, col2 = st.columns(2)
|
|
587 |
|
588 |
with col1:
|
589 |
st.header("π Markdown Outline")
|
590 |
-
|
591 |
-
|
|
|
592 |
|
593 |
# Create a download button for the markdown file
|
594 |
st.download_button(
|
595 |
label="Download Markdown",
|
596 |
-
data=
|
597 |
file_name="ml_outline.md",
|
598 |
mime="text/markdown"
|
599 |
)
|
|
|
|
|
|
|
|
|
600 |
|
601 |
with col2:
|
602 |
st.header("π PDF Preview & Demos")
|
@@ -613,18 +610,21 @@ with col2:
|
|
613 |
pike_pdf = demo_pikepdf()
|
614 |
st.download_button("Download pikepdf Demo", pike_pdf, "pikepdf_demo.pdf")
|
615 |
st.write("PikePDF demo created successfully!")
|
|
|
616 |
|
617 |
with demo_tabs[1]:
|
618 |
# fpdf demo
|
619 |
fpdf_pdf = demo_fpdf()
|
620 |
st.download_button("Download fpdf Demo", fpdf_pdf, "fpdf_demo.pdf")
|
621 |
st.write("FPDF demo created successfully!")
|
|
|
622 |
|
623 |
with demo_tabs[2]:
|
624 |
# pymupdf demo
|
625 |
pymupdf_pdf = demo_pymupdf()
|
626 |
st.download_button("Download pymupdf Demo", pymupdf_pdf, "pymupdf_demo.pdf")
|
627 |
st.write("PyMuPDF demo created successfully!")
|
|
|
628 |
|
629 |
with demo_tabs[3]:
|
630 |
# Image demo
|
@@ -636,7 +636,7 @@ with col2:
|
|
636 |
if st.button("Generate Main PDF"):
|
637 |
with st.spinner("Generating PDF..."):
|
638 |
try:
|
639 |
-
pdf_bytes = create_main_pdf(
|
640 |
|
641 |
st.download_button(
|
642 |
label="Download Main PDF",
|
@@ -656,10 +656,44 @@ with col2:
|
|
656 |
'''
|
657 |
st.markdown(pdf_display, unsafe_allow_html=True)
|
658 |
|
659 |
-
st.success("PDF generated successfully!
|
660 |
except Exception as e:
|
661 |
st.error(f"Error generating PDF: {str(e)}")
|
662 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
663 |
# Add custom CSS for better appearance
|
664 |
st.markdown("""
|
665 |
<style>
|
|
|
1 |
+
import streamlit as st
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2 |
import base64
|
3 |
from reportlab.lib.pagesizes import A4
|
4 |
+
from reportlab.platypus import SimpleDocTemplate, Paragraph, Spacer, Table, TableStyle
|
5 |
+
from reportlab.lib.styles import getSampleStyleSheet, ParagraphStyle
|
6 |
from reportlab.lib import colors
|
7 |
import pikepdf
|
8 |
import fpdf
|
|
|
12 |
from PIL import Image
|
13 |
import io
|
14 |
import os
|
15 |
+
import re
|
16 |
|
17 |
# Define the ML outline as a markdown string for multilevel content
|
18 |
ml_markdown = """# Cutting-Edge ML Outline
|
19 |
|
20 |
## Core ML Techniques
|
21 |
+
1. π **Mixture of Experts (MoE)**
|
22 |
+
- Conditional computation techniques
|
23 |
+
- Sparse gating mechanisms
|
24 |
+
- Training specialized sub-models
|
25 |
|
26 |
+
2. π₯ **Supervised Fine-Tuning (SFT) using PyTorch**
|
27 |
+
- Loss function customization
|
28 |
+
- Gradient accumulation strategies
|
29 |
+
- Learning rate schedulers
|
30 |
|
31 |
+
3. π€ **Large Language Models (LLM) using Transformers**
|
32 |
+
- Attention mechanisms
|
33 |
+
- Tokenization strategies
|
34 |
+
- Position encodings
|
35 |
|
36 |
## Training Methods
|
37 |
+
4. π **Self-Rewarding Learning using NPS 0-10 and Verbatims**
|
38 |
+
- Custom reward functions
|
39 |
+
- Feedback categorization
|
40 |
+
- Signal extraction from text
|
41 |
|
42 |
+
5. π **Reinforcement Learning from Human Feedback (RLHF)**
|
43 |
+
- Preference datasets
|
44 |
+
- PPO implementation
|
45 |
+
- KL divergence constraints
|
46 |
|
47 |
+
6. π **MergeKit: Merging Models to Same Embedding Space**
|
48 |
+
- TIES merging
|
49 |
+
- Task arithmetic
|
50 |
+
- SLERP interpolation
|
51 |
|
52 |
## Optimization & Deployment
|
53 |
+
7. π **DistillKit: Model Size Reduction with Spectrum Analysis**
|
54 |
+
- Knowledge distillation
|
55 |
+
- Quantization techniques
|
56 |
+
- Model pruning strategies
|
57 |
|
58 |
+
8. π§ **Agentic RAG Agents using Document Inputs**
|
59 |
+
- Vector database integration
|
60 |
+
- Query planning
|
61 |
+
- Self-reflection mechanisms
|
62 |
|
63 |
+
9. β³ **Longitudinal Data Summarization from Multiple Docs**
|
64 |
+
- Multi-document compression
|
65 |
+
- Timeline extraction
|
66 |
+
- Entity tracking
|
67 |
|
68 |
## Knowledge Representation
|
69 |
+
10. π **Knowledge Extraction using Markdown Knowledge Graphs**
|
70 |
+
- Entity recognition
|
71 |
+
- Relationship mapping
|
72 |
+
- Hierarchical structuring
|
73 |
+
|
74 |
+
11. πΊοΈ **Knowledge Mapping with Mermaid Diagrams**
|
75 |
+
- Flowchart generation
|
76 |
+
- Sequence diagram creation
|
77 |
+
- State diagrams
|
78 |
+
|
79 |
+
12. π» **ML Code Generation with Streamlit/Gradio/HTML5+JS**
|
80 |
+
- Code completion
|
81 |
+
- Unit test generation
|
82 |
+
- Documentation synthesis
|
83 |
"""
|
84 |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
85 |
# Process multilevel markdown for PDF output
|
86 |
def markdown_to_pdf_content(markdown_text):
|
87 |
"""Convert markdown text to a format suitable for PDF generation"""
|
|
|
90 |
# Convert markdown headers to styled text for PDF
|
91 |
lines = markdown_text.strip().split('\n')
|
92 |
pdf_content = []
|
93 |
+
in_list_item = False
|
94 |
+
current_item = None
|
95 |
+
sub_items = []
|
96 |
|
97 |
for line in lines:
|
98 |
+
line = line.strip()
|
99 |
+
if not line:
|
100 |
+
continue
|
101 |
+
|
102 |
if line.startswith('# '):
|
103 |
# Main header - will be handled separately in the PDF generation
|
104 |
pass
|
105 |
elif line.startswith('## '):
|
106 |
# Section header - add as a bold item
|
107 |
+
if current_item and sub_items:
|
108 |
+
# Store the previous item with its sub-items
|
109 |
+
pdf_content.append([current_item, sub_items])
|
110 |
+
sub_items = []
|
111 |
+
current_item = None
|
112 |
+
|
113 |
section = line.replace('## ', '').strip()
|
114 |
pdf_content.append(f"<b>{section}</b>")
|
115 |
+
in_list_item = False
|
116 |
+
elif re.match(r'^\d+\.', line):
|
117 |
+
# Numbered list item
|
118 |
+
if current_item and sub_items:
|
119 |
+
# Store the previous item with its sub-items
|
120 |
+
pdf_content.append([current_item, sub_items])
|
121 |
+
sub_items = []
|
122 |
+
|
123 |
+
current_item = line.strip()
|
124 |
+
in_list_item = True
|
125 |
+
elif line.startswith('- ') and in_list_item:
|
126 |
+
# Sub-item under a numbered list item
|
127 |
+
sub_items.append(line.strip())
|
128 |
+
else:
|
129 |
+
# Regular line
|
130 |
+
if not in_list_item:
|
131 |
+
pdf_content.append(line.strip())
|
132 |
|
133 |
+
# Add the last item if there is one
|
134 |
+
if current_item and sub_items:
|
135 |
+
pdf_content.append([current_item, sub_items])
|
136 |
|
137 |
# Split the content for two columns
|
138 |
mid_point = len(pdf_content) // 2
|
|
|
154 |
# Set up fonts and colors
|
155 |
title_font = "helv-b"
|
156 |
section_font = "helv-b"
|
157 |
+
item_font = "helv-b"
|
158 |
+
subitem_font = "helv"
|
159 |
blue_color = (0, 0, 0.8)
|
160 |
black_color = (0, 0, 0)
|
161 |
|
|
|
164 |
|
165 |
# First column
|
166 |
x1, y1 = 50, 80
|
167 |
+
current_y = y1
|
168 |
+
|
169 |
+
for item in left_column:
|
170 |
+
if isinstance(item, str) and item.startswith('<b>'):
|
171 |
# Section header
|
172 |
+
# Add extra space before sections (except the first one)
|
173 |
+
if current_y > y1:
|
174 |
+
current_y += 10
|
175 |
+
|
176 |
text = item.replace('<b>', '').replace('</b>', '')
|
177 |
+
page.insert_text((x1, current_y), text, fontname=section_font, fontsize=14, color=blue_color)
|
178 |
+
current_y += 25
|
179 |
+
elif isinstance(item, list):
|
180 |
+
# Main item with sub-items
|
181 |
+
main_item, sub_items = item
|
182 |
+
page.insert_text((x1, current_y), main_item, fontname=item_font, fontsize=12, color=black_color)
|
183 |
+
current_y += 20
|
184 |
+
|
185 |
+
# Add sub-items
|
186 |
+
for sub_item in sub_items:
|
187 |
+
page.insert_text((x1 + 20, current_y), sub_item, fontname=subitem_font, fontsize=10, color=black_color)
|
188 |
+
current_y += 15
|
189 |
+
|
190 |
+
current_y += 5 # Extra space after a group
|
191 |
else:
|
192 |
+
# Regular item
|
193 |
+
page.insert_text((x1, current_y), item, fontname=item_font, fontsize=12, color=black_color)
|
194 |
+
current_y += 20
|
195 |
|
196 |
# Second column
|
197 |
x2, y2 = 450, 80
|
198 |
+
current_y = y2
|
199 |
+
|
200 |
+
for item in right_column:
|
201 |
+
if isinstance(item, str) and item.startswith('<b>'):
|
202 |
# Section header
|
203 |
+
# Add extra space before sections (except the first one)
|
204 |
+
if current_y > y2:
|
205 |
+
current_y += 10
|
206 |
+
|
207 |
text = item.replace('<b>', '').replace('</b>', '')
|
208 |
+
page.insert_text((x2, current_y), text, fontname=section_font, fontsize=14, color=blue_color)
|
209 |
+
current_y += 25
|
210 |
+
elif isinstance(item, list):
|
211 |
+
# Main item with sub-items
|
212 |
+
main_item, sub_items = item
|
213 |
+
page.insert_text((x2, current_y), main_item, fontname=item_font, fontsize=12, color=black_color)
|
214 |
+
current_y += 20
|
215 |
+
|
216 |
+
# Add sub-items
|
217 |
+
for sub_item in sub_items:
|
218 |
+
page.insert_text((x2 + 20, current_y), sub_item, fontname=subitem_font, fontsize=10, color=black_color)
|
219 |
+
current_y += 15
|
220 |
+
|
221 |
+
current_y += 5 # Extra space after a group
|
222 |
else:
|
223 |
+
# Regular item
|
224 |
+
page.insert_text((x2, current_y), item, fontname=item_font, fontsize=12, color=black_color)
|
225 |
+
current_y += 20
|
226 |
|
227 |
# Draw a dividing line
|
228 |
page.draw_line((421, 70), (421, 550))
|
|
|
254 |
pdf.cell(0, 10, txt="Cutting-Edge ML Outline (FPDF Demo)", ln=True, align='C')
|
255 |
pdf.ln(10)
|
256 |
|
257 |
+
# Define coordinates for columns
|
258 |
x_col1 = 20
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
259 |
x_col2 = pdf.w / 2 + 10
|
260 |
+
y_start = pdf.get_y()
|
261 |
|
262 |
+
# Function to render a column
|
263 |
+
def render_column(items, x_start, y_start):
|
264 |
+
y_pos = y_start
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
265 |
|
266 |
+
for item in items:
|
267 |
+
if isinstance(item, str) and item.startswith('<b>'):
|
268 |
+
# Section header
|
269 |
+
text = item.replace('<b>', '').replace('</b>', '')
|
270 |
+
pdf.set_font("Arial", 'B', size=14)
|
271 |
+
pdf.set_text_color(0, 0, 128) # Dark blue
|
272 |
+
pdf.set_xy(x_start, y_pos)
|
273 |
+
pdf.cell(0, 10, txt=text, ln=True)
|
274 |
+
y_pos += 10
|
275 |
+
elif isinstance(item, list):
|
276 |
+
# Main item with sub-items
|
277 |
+
main_item, sub_items = item
|
278 |
+
|
279 |
+
# Main item
|
280 |
+
pdf.set_font("Arial", 'B', size=11)
|
281 |
+
pdf.set_text_color(0, 0, 0) # Black
|
282 |
+
pdf.set_xy(x_start, y_pos)
|
283 |
+
pdf.multi_cell(180, 6, txt=main_item, align='L')
|
284 |
+
main_height = pdf.get_y() - y_pos
|
285 |
+
y_pos += main_height + 2
|
286 |
+
|
287 |
+
# Sub-items
|
288 |
+
pdf.set_font("Arial", size=10)
|
289 |
+
for sub_item in sub_items:
|
290 |
+
pdf.set_xy(x_start + 10, y_pos)
|
291 |
+
pdf.multi_cell(170, 5, txt=sub_item, align='L')
|
292 |
+
sub_height = pdf.get_y() - y_pos
|
293 |
+
y_pos += sub_height + 1
|
294 |
+
|
295 |
+
y_pos += 2 # Extra space after a group
|
296 |
+
else:
|
297 |
+
# Regular item
|
298 |
+
pdf.set_font("Arial", 'B', size=11)
|
299 |
+
pdf.set_text_color(0, 0, 0) # Black
|
300 |
+
pdf.set_xy(x_start, y_pos)
|
301 |
+
pdf.multi_cell(180, 6, txt=item, align='L')
|
302 |
+
item_height = pdf.get_y() - y_pos
|
303 |
+
y_pos += item_height + 4
|
304 |
+
|
305 |
+
# Render both columns
|
306 |
+
render_column(left_column, x_col1, y_start)
|
307 |
+
render_column(right_column, x_col2, y_start)
|
308 |
|
309 |
# Draw a dividing line
|
310 |
pdf.line(pdf.w/2, 30, pdf.w/2, 280)
|
|
|
325 |
# Set up fonts and colors
|
326 |
title_font = "helv-b"
|
327 |
section_font = "helv-b"
|
328 |
+
item_font = "helv-b"
|
329 |
+
subitem_font = "helv"
|
330 |
blue_color = (0, 0, 0.8)
|
331 |
black_color = (0, 0, 0)
|
332 |
|
|
|
335 |
|
336 |
# First column
|
337 |
x1, y1 = 50, 80
|
|
|
338 |
current_y = y1
|
339 |
|
340 |
for item in left_column:
|
341 |
+
if isinstance(item, str) and item.startswith('<b>'):
|
342 |
+
# Section header
|
343 |
# Add extra space before sections (except the first one)
|
344 |
if current_y > y1:
|
345 |
current_y += 10
|
346 |
|
|
|
347 |
text = item.replace('<b>', '').replace('</b>', '')
|
348 |
page.insert_text((x1, current_y), text, fontname=section_font, fontsize=14, color=blue_color)
|
349 |
+
current_y += 25
|
350 |
+
elif isinstance(item, list):
|
351 |
+
# Main item with sub-items
|
352 |
+
main_item, sub_items = item
|
353 |
+
page.insert_text((x1, current_y), main_item, fontname=item_font, fontsize=12, color=black_color)
|
354 |
+
current_y += 20
|
355 |
+
|
356 |
+
# Add sub-items
|
357 |
+
for sub_item in sub_items:
|
358 |
+
page.insert_text((x1 + 20, current_y), sub_item, fontname=subitem_font, fontsize=10, color=black_color)
|
359 |
+
current_y += 15
|
360 |
+
|
361 |
+
current_y += 5 # Extra space after a group
|
362 |
else:
|
363 |
+
# Regular item
|
364 |
+
page.insert_text((x1, current_y), item, fontname=item_font, fontsize=12, color=black_color)
|
365 |
+
current_y += 20
|
|
|
366 |
|
367 |
# Second column
|
368 |
x2, y2 = 450, 80
|
369 |
current_y = y2
|
370 |
|
371 |
for item in right_column:
|
372 |
+
if isinstance(item, str) and item.startswith('<b>'):
|
373 |
+
# Section header
|
374 |
# Add extra space before sections (except the first one)
|
375 |
if current_y > y2:
|
376 |
current_y += 10
|
377 |
|
|
|
378 |
text = item.replace('<b>', '').replace('</b>', '')
|
379 |
page.insert_text((x2, current_y), text, fontname=section_font, fontsize=14, color=blue_color)
|
380 |
+
current_y += 25
|
381 |
+
elif isinstance(item, list):
|
382 |
+
# Main item with sub-items
|
383 |
+
main_item, sub_items = item
|
384 |
+
page.insert_text((x2, current_y), main_item, fontname=item_font, fontsize=12, color=black_color)
|
385 |
+
current_y += 20
|
386 |
+
|
387 |
+
# Add sub-items
|
388 |
+
for sub_item in sub_items:
|
389 |
+
page.insert_text((x2 + 20, current_y), sub_item, fontname=subitem_font, fontsize=10, color=black_color)
|
390 |
+
current_y += 15
|
391 |
+
|
392 |
+
current_y += 5 # Extra space after a group
|
393 |
else:
|
394 |
+
# Regular item
|
395 |
+
page.insert_text((x2, current_y), item, fontname=item_font, fontsize=12, color=black_color)
|
396 |
+
current_y += 20
|
|
|
397 |
|
398 |
# Draw a dividing line
|
399 |
page.draw_line((421, 70), (421, 550))
|
|
|
478 |
parent=styles['Normal'],
|
479 |
fontSize=11,
|
480 |
leading=14,
|
481 |
+
fontName='Helvetica-Bold'
|
482 |
+
)
|
483 |
+
|
484 |
+
subitem_style = ParagraphStyle(
|
485 |
+
'SubItemStyle',
|
486 |
+
parent=styles['Normal'],
|
487 |
+
fontSize=10,
|
488 |
+
leading=12,
|
489 |
+
leftIndent=20
|
490 |
)
|
491 |
|
492 |
# Add title
|
|
|
494 |
story.append(Spacer(1, 20))
|
495 |
|
496 |
# Prepare data for table
|
497 |
+
left_cells = []
|
|
|
|
|
|
|
|
|
498 |
for item in left_column:
|
499 |
+
if isinstance(item, str) and item.startswith('<b>'):
|
500 |
+
# Section header
|
501 |
text = item.replace('<b>', '').replace('</b>', '')
|
502 |
+
left_cells.append(Paragraph(text, section_style))
|
503 |
+
elif isinstance(item, list):
|
504 |
+
# Main item with sub-items
|
505 |
+
main_item, sub_items = item
|
506 |
+
left_cells.append(Paragraph(main_item, item_style))
|
507 |
+
|
508 |
+
# Sub items
|
509 |
+
for sub_item in sub_items:
|
510 |
+
left_cells.append(Paragraph(sub_item, subitem_style))
|
511 |
else:
|
512 |
+
# Regular item
|
513 |
+
left_cells.append(Paragraph(item, item_style))
|
514 |
|
515 |
+
right_cells = []
|
516 |
for item in right_column:
|
517 |
+
if isinstance(item, str) and item.startswith('<b>'):
|
518 |
+
# Section header
|
519 |
text = item.replace('<b>', '').replace('</b>', '')
|
520 |
+
right_cells.append(Paragraph(text, section_style))
|
521 |
+
elif isinstance(item, list):
|
522 |
+
# Main item with sub-items
|
523 |
+
main_item, sub_items = item
|
524 |
+
right_cells.append(Paragraph(main_item, item_style))
|
525 |
+
|
526 |
+
# Sub items
|
527 |
+
for sub_item in sub_items:
|
528 |
+
right_cells.append(Paragraph(sub_item, subitem_style))
|
529 |
else:
|
530 |
+
# Regular item
|
531 |
+
right_cells.append(Paragraph(item, item_style))
|
532 |
+
|
533 |
+
# Make sure both columns have the same number of rows by adding empty cells
|
534 |
+
max_cells = max(len(left_cells), len(right_cells))
|
535 |
+
if len(left_cells) < max_cells:
|
536 |
+
for i in range(max_cells - len(left_cells)):
|
537 |
+
left_cells.append("")
|
538 |
+
if len(right_cells) < max_cells:
|
539 |
+
for i in range(max_cells - len(right_cells)):
|
540 |
+
right_cells.append("")
|
541 |
+
|
542 |
+
# Create table data (one row per cell)
|
543 |
+
table_data = []
|
544 |
+
for i in range(max_cells):
|
545 |
+
table_data.append([left_cells[i], right_cells[i]])
|
546 |
|
547 |
# Calculate column widths
|
548 |
col_width = (A4[1] - 120) / 2.0 # Page width minus margins divided by 2
|
|
|
579 |
|
580 |
with col1:
|
581 |
st.header("π Markdown Outline")
|
582 |
+
|
583 |
+
# Display the markdown content
|
584 |
+
st.markdown(ml_markdown)
|
585 |
|
586 |
# Create a download button for the markdown file
|
587 |
st.download_button(
|
588 |
label="Download Markdown",
|
589 |
+
data=ml_markdown,
|
590 |
file_name="ml_outline.md",
|
591 |
mime="text/markdown"
|
592 |
)
|
593 |
+
|
594 |
+
# Show the markdown source code in an expandable section
|
595 |
+
with st.expander("View Markdown Source"):
|
596 |
+
st.code(ml_markdown, language="markdown")
|
597 |
|
598 |
with col2:
|
599 |
st.header("π PDF Preview & Demos")
|
|
|
610 |
pike_pdf = demo_pikepdf()
|
611 |
st.download_button("Download pikepdf Demo", pike_pdf, "pikepdf_demo.pdf")
|
612 |
st.write("PikePDF demo created successfully!")
|
613 |
+
st.info("This PDF contains the multilevel markdown outline in a two-column layout.")
|
614 |
|
615 |
with demo_tabs[1]:
|
616 |
# fpdf demo
|
617 |
fpdf_pdf = demo_fpdf()
|
618 |
st.download_button("Download fpdf Demo", fpdf_pdf, "fpdf_demo.pdf")
|
619 |
st.write("FPDF demo created successfully!")
|
620 |
+
st.info("This PDF contains the multilevel markdown outline in a two-column layout.")
|
621 |
|
622 |
with demo_tabs[2]:
|
623 |
# pymupdf demo
|
624 |
pymupdf_pdf = demo_pymupdf()
|
625 |
st.download_button("Download pymupdf Demo", pymupdf_pdf, "pymupdf_demo.pdf")
|
626 |
st.write("PyMuPDF demo created successfully!")
|
627 |
+
st.info("This PDF contains the multilevel markdown outline in a two-column layout.")
|
628 |
|
629 |
with demo_tabs[3]:
|
630 |
# Image demo
|
|
|
636 |
if st.button("Generate Main PDF"):
|
637 |
with st.spinner("Generating PDF..."):
|
638 |
try:
|
639 |
+
pdf_bytes = create_main_pdf(ml_markdown)
|
640 |
|
641 |
st.download_button(
|
642 |
label="Download Main PDF",
|
|
|
656 |
'''
|
657 |
st.markdown(pdf_display, unsafe_allow_html=True)
|
658 |
|
659 |
+
st.success("PDF generated successfully! The PDF displays the multilevel markdown outline in a two-column layout.")
|
660 |
except Exception as e:
|
661 |
st.error(f"Error generating PDF: {str(e)}")
|
662 |
|
663 |
+
# Show the PDF rendering code in an expandable section
|
664 |
+
with st.expander("View PDF Rendering Code"):
|
665 |
+
st.code("""
|
666 |
+
# Process multilevel markdown for PDF output
|
667 |
+
def markdown_to_pdf_content(markdown_text):
|
668 |
+
# Convert markdown headers to styled text for PDF
|
669 |
+
lines = markdown_text.strip().split('\\n')
|
670 |
+
pdf_content = []
|
671 |
+
|
672 |
+
for line in lines:
|
673 |
+
if line.startswith('# '):
|
674 |
+
# Main header - will be handled separately
|
675 |
+
pass
|
676 |
+
elif line.startswith('## '):
|
677 |
+
# Section header - add as a bold item
|
678 |
+
section = line.replace('## ', '').strip()
|
679 |
+
pdf_content.append(f"<b>{section}</b>")
|
680 |
+
elif re.match(r'^\\d+\\.', line):
|
681 |
+
# Numbered list item
|
682 |
+
item = line.strip()
|
683 |
+
pdf_content.append(item)
|
684 |
+
elif line.startswith('- '):
|
685 |
+
# Sub-item under a numbered list item
|
686 |
+
sub_item = line.strip()
|
687 |
+
pdf_content.append(" " + sub_item)
|
688 |
+
|
689 |
+
# Split the content for two columns
|
690 |
+
mid_point = len(pdf_content) // 2
|
691 |
+
left_column = pdf_content[:mid_point]
|
692 |
+
right_column = pdf_content[mid_point:]
|
693 |
+
|
694 |
+
return left_column, right_column
|
695 |
+
""", language="python")
|
696 |
+
|
697 |
# Add custom CSS for better appearance
|
698 |
st.markdown("""
|
699 |
<style>
|