Spaces:
Sleeping
Sleeping
Update app.py
Browse files
app.py
CHANGED
|
@@ -1,9 +1,9 @@
|
|
|
|
|
|
|
|
| 1 |
import pandas as pd
|
| 2 |
import matplotlib.pyplot as plt
|
| 3 |
import seaborn as sns
|
| 4 |
import gradio as gr
|
| 5 |
-
import requests
|
| 6 |
-
from bs4 import BeautifulSoup
|
| 7 |
import io
|
| 8 |
import os
|
| 9 |
import base64
|
|
@@ -13,11 +13,12 @@ from io import BytesIO
|
|
| 13 |
import tempfile
|
| 14 |
import sys
|
| 15 |
|
| 16 |
-
|
| 17 |
# --------------------------------------------------------------------
|
| 18 |
-
# PART 1:
|
| 19 |
# --------------------------------------------------------------------
|
| 20 |
|
|
|
|
|
|
|
| 21 |
data_full = [
|
| 22 |
['CultriX/Qwen2.5-14B-SLERPv7', 'https://huggingface.co/CultriX/Qwen2.5-14B-SLERPv7', 0.7205, 0.8272, 0.7541, 0.6581, 0.5, 0.729],
|
| 23 |
['djuna/Q2.5-Veltha-14B-0.5', 'https://huggingface.co/djuna/Q2.5-Veltha-14B-0.5', 0.7492, 0.8386, 0.7305, 0.598, 0.43, 0.7817],
|
|
@@ -44,10 +45,12 @@ data_full = [
|
|
| 44 |
['CultriX/Qwen2.5-14B-Wernickev6', 'https://huggingface.co/CultriX/Qwen2.5-14B-Wernickev6', 0.6994, 0.7549, 0.5816, 0.6991, 0.58, 0.7267],
|
| 45 |
['CultriX/Qwen2.5-14B-Wernickev7', 'https://huggingface.co/CultriX/Qwen2.5-14B-Wernickev7', 0.7147, 0.7599, 0.6097, 0.7056, 0.57, 0.7164],
|
| 46 |
['CultriX/Qwen2.5-14B-FinalMerge-tmp2', 'https://huggingface.co/CultriX/Qwen2.5-14B-FinalMerge-tmp2', 0.7255, 0.8192, 0.7535, 0.6671, 0.5, 0.7612],
|
| 47 |
-
['CultriX/Qwen2.5-14B-BrocaV8', 'https://huggingface.co/CultriX/Qwen2.5-14B-BrocaV8', 0.7415, 0.8396, 0.7334, 0.5785, 0.
|
|
|
|
|
|
|
|
|
|
|
|
|
| 48 |
]
|
| 49 |
-
columns = ["Model Configuration", "Model Link", "tinyArc", "tinyHellaswag",
|
| 50 |
-
"tinyMMLU", "tinyTruthfulQA", "tinyTruthfulQA_mc1", "tinyWinogrande"]
|
| 51 |
df_full = pd.DataFrame(data_full, columns=columns)
|
| 52 |
|
| 53 |
def plot_average_scores():
|
|
@@ -75,8 +78,10 @@ def plot_average_scores():
|
|
| 75 |
return pil_image, temp_image_file.name
|
| 76 |
|
| 77 |
def plot_task_performance():
|
| 78 |
-
df_full_melted = df_full.melt(
|
| 79 |
-
|
|
|
|
|
|
|
| 80 |
|
| 81 |
plt.figure(figsize=(16, 12))
|
| 82 |
for model in df_full["Model Configuration"]:
|
|
@@ -127,8 +132,13 @@ def plot_task_specific_top_models():
|
|
| 127 |
|
| 128 |
def plot_heatmap():
|
| 129 |
plt.figure(figsize=(14, 10))
|
| 130 |
-
sns.heatmap(
|
| 131 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 132 |
plt.title("Performance Heatmap", fontsize=16)
|
| 133 |
plt.tight_layout()
|
| 134 |
|
|
@@ -143,7 +153,15 @@ def plot_heatmap():
|
|
| 143 |
return pil_image, temp_image_file.name
|
| 144 |
|
| 145 |
def scrape_mergekit_config(model_name):
|
| 146 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 147 |
response = requests.get(model_link)
|
| 148 |
if response.status_code != 200:
|
| 149 |
return f"Failed to fetch model page for {model_name}. Please check the link."
|
|
@@ -155,12 +173,18 @@ def scrape_mergekit_config(model_name):
|
|
| 155 |
return f"No YAML configuration found for {model_name}."
|
| 156 |
|
| 157 |
def download_yaml(yaml_content, model_name):
|
|
|
|
|
|
|
|
|
|
| 158 |
if "No YAML configuration found" in yaml_content or "Failed to fetch model page" in yaml_content:
|
| 159 |
return None
|
| 160 |
filename = f"{model_name.replace('/', '_')}_config.yaml"
|
| 161 |
return gr.File(value=yaml_content.encode(), filename=filename)
|
| 162 |
|
| 163 |
def scrape_model_page(model_url):
|
|
|
|
|
|
|
|
|
|
| 164 |
try:
|
| 165 |
response = requests.get(model_url)
|
| 166 |
if response.status_code != 200:
|
|
@@ -176,9 +200,18 @@ def scrape_model_page(model_url):
|
|
| 176 |
return f"Error: {str(e)}"
|
| 177 |
|
| 178 |
def display_scraped_model_data(model_url):
|
|
|
|
|
|
|
|
|
|
| 179 |
return scrape_model_page(model_url)
|
| 180 |
|
| 181 |
def download_all_data():
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 182 |
import io
|
| 183 |
csv_buffer = io.StringIO()
|
| 184 |
df_full.to_csv(csv_buffer, index=False)
|
|
@@ -200,28 +233,27 @@ def download_all_data():
|
|
| 200 |
with zipfile.ZipFile(zip_buffer, 'w') as zf:
|
| 201 |
zf.writestr("model_scores.csv", csv_data)
|
| 202 |
|
|
|
|
| 203 |
for name, (pil_image, filename) in plot_dict.items():
|
| 204 |
image_bytes = io.BytesIO()
|
| 205 |
pil_image.save(image_bytes, format='PNG')
|
| 206 |
image_bytes.seek(0)
|
| 207 |
zf.writestr(filename, image_bytes.read())
|
| 208 |
|
| 209 |
-
# Also try scraping each model for a YAML config
|
| 210 |
for model_name in df_full["Model Configuration"].to_list():
|
| 211 |
yaml_content = scrape_mergekit_config(model_name)
|
| 212 |
if ("No YAML configuration found" not in yaml_content) and ("Failed to fetch model page" not in yaml_content):
|
| 213 |
-
|
| 214 |
|
| 215 |
zip_buffer.seek(0)
|
| 216 |
return zip_buffer, "analysis_data.zip"
|
| 217 |
|
| 218 |
-
|
| 219 |
# --------------------------------------------------------------------
|
| 220 |
-
# PART 2:
|
| 221 |
# --------------------------------------------------------------------
|
|
|
|
| 222 |
benchmark_data = [
|
| 223 |
-
# The entire dataset from your "DATA START", rank 44..105
|
| 224 |
-
# (the code you posted with "knowledge of config" or scraping logic)
|
| 225 |
{
|
| 226 |
"rank": 44,
|
| 227 |
"name": "sometimesanotion/Qwen2.5-14B-Vimarckoso-v3",
|
|
@@ -248,21 +280,533 @@ benchmark_data = [
|
|
| 248 |
}
|
| 249 |
}
|
| 250 |
},
|
| 251 |
-
|
| 252 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 253 |
]
|
| 254 |
|
| 255 |
-
|
| 256 |
def snippet_scrape_model_page(url):
|
| 257 |
"""
|
| 258 |
-
|
|
|
|
| 259 |
"""
|
| 260 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 261 |
|
| 262 |
def snippet_print_benchmark_and_config_info(model_info):
|
| 263 |
"""
|
| 264 |
-
Prints an overview for each model
|
| 265 |
-
|
| 266 |
"""
|
| 267 |
print(f"---\nModel Rank: {model_info['rank']}")
|
| 268 |
print(f"Model Name: {model_info['name']}")
|
|
@@ -274,7 +818,7 @@ def snippet_print_benchmark_and_config_info(model_info):
|
|
| 274 |
print(f"Models average score in MUSR benchmarks in %: {model_info['scores']['MUSR']}")
|
| 275 |
print(f"Models average score in MMLU-PRO benchmarks in %: {model_info['scores']['MMLU-PRO']}")
|
| 276 |
|
| 277 |
-
# If there's a known_config, print it
|
| 278 |
if model_info["known_config"] is not None:
|
| 279 |
print("###")
|
| 280 |
print("models:")
|
|
@@ -284,25 +828,22 @@ def snippet_print_benchmark_and_config_info(model_info):
|
|
| 284 |
print(f"base_model: {model_info['known_config']['base_model']}")
|
| 285 |
print(f"dtype: {model_info['known_config']['dtype']}")
|
| 286 |
print("parameters:")
|
| 287 |
-
|
|
|
|
| 288 |
print("###")
|
| 289 |
return
|
| 290 |
|
| 291 |
-
# Otherwise,
|
| 292 |
scraped = snippet_scrape_model_page(model_info["hf_url"])
|
| 293 |
if isinstance(scraped, str):
|
| 294 |
# Means it's an error string or something
|
| 295 |
-
|
| 296 |
-
|
| 297 |
-
# optionally print snippet
|
| 298 |
-
else:
|
| 299 |
-
print(scraped)
|
| 300 |
return
|
| 301 |
else:
|
| 302 |
-
# It's presumably a dict
|
| 303 |
-
if
|
| 304 |
print("(No MergeKit configuration found.)\n")
|
| 305 |
-
# Print your snippet code
|
| 306 |
print("You can try the following Python script to scrape the model page:\n")
|
| 307 |
print("#" * 70)
|
| 308 |
print(f'''import requests
|
|
@@ -336,14 +877,15 @@ if __name__ == "__main__":
|
|
| 336 |
print(result)''')
|
| 337 |
print("#" * 70)
|
| 338 |
else:
|
|
|
|
| 339 |
print("###")
|
| 340 |
print(scraped["yaml_configuration"])
|
| 341 |
print("###")
|
| 342 |
|
| 343 |
def run_non_tiny_benchmarks():
|
| 344 |
"""
|
| 345 |
-
Captures the stdout from printing each model in benchmark_data
|
| 346 |
-
|
| 347 |
"""
|
| 348 |
old_stdout = sys.stdout
|
| 349 |
buffer = io.StringIO()
|
|
@@ -355,14 +897,13 @@ def run_non_tiny_benchmarks():
|
|
| 355 |
sys.stdout = old_stdout
|
| 356 |
return buffer.getvalue()
|
| 357 |
|
| 358 |
-
|
| 359 |
# --------------------------------------------------------------------
|
| 360 |
-
# PART 3:
|
| 361 |
# --------------------------------------------------------------------
|
| 362 |
with gr.Blocks() as demo:
|
| 363 |
gr.Markdown("# Comprehensive Model Performance Analysis with Hugging Face Links")
|
| 364 |
|
| 365 |
-
# The existing UI
|
| 366 |
with gr.Row():
|
| 367 |
btn1 = gr.Button("Show Average Performance")
|
| 368 |
img1 = gr.Image(type="pil", label="Average Performance Plot")
|
|
@@ -387,6 +928,7 @@ with gr.Blocks() as demo:
|
|
| 387 |
heatmap_download = gr.File(label="Download Heatmap")
|
| 388 |
btn4.click(plot_heatmap, outputs=[heatmap_img, heatmap_download])
|
| 389 |
|
|
|
|
| 390 |
with gr.Row():
|
| 391 |
model_selector = gr.Dropdown(choices=df_full["Model Configuration"].tolist(), label="Select a Model")
|
| 392 |
with gr.Column():
|
|
@@ -398,12 +940,13 @@ with gr.Blocks() as demo:
|
|
| 398 |
yaml_download = gr.File(label="Download MergeKit Configuration")
|
| 399 |
save_yaml_btn.click(download_yaml, inputs=[yaml_output, model_selector], outputs=yaml_download)
|
| 400 |
|
|
|
|
| 401 |
with gr.Row():
|
| 402 |
download_all_btn = gr.Button("Download Everything")
|
| 403 |
all_downloads = gr.File(label="Download All Data")
|
| 404 |
download_all_btn.click(download_all_data, outputs=all_downloads)
|
| 405 |
|
| 406 |
-
# Live
|
| 407 |
gr.Markdown("## Live Scraping Features")
|
| 408 |
with gr.Row():
|
| 409 |
url_input = gr.Textbox(label="Enter Hugging Face Model URL", placeholder="https://huggingface.co/<model>")
|
|
@@ -411,11 +954,11 @@ with gr.Blocks() as demo:
|
|
| 411 |
live_scrape_output = gr.Textbox(label="Scraped Data", lines=15)
|
| 412 |
live_scrape_btn.click(display_scraped_model_data, inputs=url_input, outputs=live_scrape_output)
|
| 413 |
|
| 414 |
-
#
|
| 415 |
gr.Markdown("## Non-Tiny Benchmark Parser (Ranks 44–105)")
|
| 416 |
with gr.Row():
|
| 417 |
parse_non_tiny_btn = gr.Button("Parse Non-Tiny Benchmarks")
|
| 418 |
parse_non_tiny_output = gr.Textbox(label="Non-Tiny Benchmark Output", lines=30)
|
| 419 |
parse_non_tiny_btn.click(fn=run_non_tiny_benchmarks, outputs=parse_non_tiny_output)
|
| 420 |
|
| 421 |
-
demo.launch()
|
|
|
|
| 1 |
+
import requests
|
| 2 |
+
from bs4 import BeautifulSoup
|
| 3 |
import pandas as pd
|
| 4 |
import matplotlib.pyplot as plt
|
| 5 |
import seaborn as sns
|
| 6 |
import gradio as gr
|
|
|
|
|
|
|
| 7 |
import io
|
| 8 |
import os
|
| 9 |
import base64
|
|
|
|
| 13 |
import tempfile
|
| 14 |
import sys
|
| 15 |
|
|
|
|
| 16 |
# --------------------------------------------------------------------
|
| 17 |
+
# PART 1: TINY DATA + PLOTS
|
| 18 |
# --------------------------------------------------------------------
|
| 19 |
|
| 20 |
+
# This dataframe is your “tiny” version of model performance data.
|
| 21 |
+
# Used for plotting & demonstration in the Gradio app.
|
| 22 |
data_full = [
|
| 23 |
['CultriX/Qwen2.5-14B-SLERPv7', 'https://huggingface.co/CultriX/Qwen2.5-14B-SLERPv7', 0.7205, 0.8272, 0.7541, 0.6581, 0.5, 0.729],
|
| 24 |
['djuna/Q2.5-Veltha-14B-0.5', 'https://huggingface.co/djuna/Q2.5-Veltha-14B-0.5', 0.7492, 0.8386, 0.7305, 0.598, 0.43, 0.7817],
|
|
|
|
| 45 |
['CultriX/Qwen2.5-14B-Wernickev6', 'https://huggingface.co/CultriX/Qwen2.5-14B-Wernickev6', 0.6994, 0.7549, 0.5816, 0.6991, 0.58, 0.7267],
|
| 46 |
['CultriX/Qwen2.5-14B-Wernickev7', 'https://huggingface.co/CultriX/Qwen2.5-14B-Wernickev7', 0.7147, 0.7599, 0.6097, 0.7056, 0.57, 0.7164],
|
| 47 |
['CultriX/Qwen2.5-14B-FinalMerge-tmp2', 'https://huggingface.co/CultriX/Qwen2.5-14B-FinalMerge-tmp2', 0.7255, 0.8192, 0.7535, 0.6671, 0.5, 0.7612],
|
| 48 |
+
['CultriX/Qwen2.5-14B-BrocaV8', 'https://huggingface.co/CultriX/Qwen2.5-14B-BrocaV8', 0.7415, 0.8396, 0.7334, 0.5785, 0.43, 0.7646],
|
| 49 |
+
]
|
| 50 |
+
columns = [
|
| 51 |
+
"Model Configuration", "Model Link", "tinyArc", "tinyHellaswag",
|
| 52 |
+
"tinyMMLU", "tinyTruthfulQA", "tinyTruthfulQA_mc1", "tinyWinogrande"
|
| 53 |
]
|
|
|
|
|
|
|
| 54 |
df_full = pd.DataFrame(data_full, columns=columns)
|
| 55 |
|
| 56 |
def plot_average_scores():
|
|
|
|
| 78 |
return pil_image, temp_image_file.name
|
| 79 |
|
| 80 |
def plot_task_performance():
|
| 81 |
+
df_full_melted = df_full.melt(
|
| 82 |
+
id_vars=["Model Configuration", "Model Link"],
|
| 83 |
+
var_name="Task", value_name="Score"
|
| 84 |
+
)
|
| 85 |
|
| 86 |
plt.figure(figsize=(16, 12))
|
| 87 |
for model in df_full["Model Configuration"]:
|
|
|
|
| 132 |
|
| 133 |
def plot_heatmap():
|
| 134 |
plt.figure(figsize=(14, 10))
|
| 135 |
+
sns.heatmap(
|
| 136 |
+
df_full.iloc[:, 2:],
|
| 137 |
+
annot=True,
|
| 138 |
+
cmap="YlGnBu",
|
| 139 |
+
xticklabels=columns[2:],
|
| 140 |
+
yticklabels=df_full["Model Configuration"]
|
| 141 |
+
)
|
| 142 |
plt.title("Performance Heatmap", fontsize=16)
|
| 143 |
plt.tight_layout()
|
| 144 |
|
|
|
|
| 153 |
return pil_image, temp_image_file.name
|
| 154 |
|
| 155 |
def scrape_mergekit_config(model_name):
|
| 156 |
+
"""
|
| 157 |
+
For the *tiny* table’s model links.
|
| 158 |
+
Scrapes <pre> tags on the huggingface model page to find a YAML config.
|
| 159 |
+
"""
|
| 160 |
+
df_row = df_full.loc[df_full["Model Configuration"] == model_name]
|
| 161 |
+
if df_row.empty:
|
| 162 |
+
return f"No data found for model {model_name}."
|
| 163 |
+
|
| 164 |
+
model_link = df_row["Model Link"].values[0]
|
| 165 |
response = requests.get(model_link)
|
| 166 |
if response.status_code != 200:
|
| 167 |
return f"Failed to fetch model page for {model_name}. Please check the link."
|
|
|
|
| 173 |
return f"No YAML configuration found for {model_name}."
|
| 174 |
|
| 175 |
def download_yaml(yaml_content, model_name):
|
| 176 |
+
"""
|
| 177 |
+
Let users download the scraped YAML if it exists.
|
| 178 |
+
"""
|
| 179 |
if "No YAML configuration found" in yaml_content or "Failed to fetch model page" in yaml_content:
|
| 180 |
return None
|
| 181 |
filename = f"{model_name.replace('/', '_')}_config.yaml"
|
| 182 |
return gr.File(value=yaml_content.encode(), filename=filename)
|
| 183 |
|
| 184 |
def scrape_model_page(model_url):
|
| 185 |
+
"""
|
| 186 |
+
Used for the "Live Scraping" text box in the Gradio UI.
|
| 187 |
+
"""
|
| 188 |
try:
|
| 189 |
response = requests.get(model_url)
|
| 190 |
if response.status_code != 200:
|
|
|
|
| 200 |
return f"Error: {str(e)}"
|
| 201 |
|
| 202 |
def display_scraped_model_data(model_url):
|
| 203 |
+
"""
|
| 204 |
+
Helper for the "Live Scraping Features" section of the Gradio app.
|
| 205 |
+
"""
|
| 206 |
return scrape_model_page(model_url)
|
| 207 |
|
| 208 |
def download_all_data():
|
| 209 |
+
"""
|
| 210 |
+
Builds and returns a zip of:
|
| 211 |
+
- the CSV of your 'tiny' data,
|
| 212 |
+
- four plots (average performance, task performance, top models, heatmap),
|
| 213 |
+
- any YAML configurations for the 'tiny' table's models (if found).
|
| 214 |
+
"""
|
| 215 |
import io
|
| 216 |
csv_buffer = io.StringIO()
|
| 217 |
df_full.to_csv(csv_buffer, index=False)
|
|
|
|
| 233 |
with zipfile.ZipFile(zip_buffer, 'w') as zf:
|
| 234 |
zf.writestr("model_scores.csv", csv_data)
|
| 235 |
|
| 236 |
+
# Add the images
|
| 237 |
for name, (pil_image, filename) in plot_dict.items():
|
| 238 |
image_bytes = io.BytesIO()
|
| 239 |
pil_image.save(image_bytes, format='PNG')
|
| 240 |
image_bytes.seek(0)
|
| 241 |
zf.writestr(filename, image_bytes.read())
|
| 242 |
|
| 243 |
+
# Also try scraping each model in the *tiny* dataset for a YAML config
|
| 244 |
for model_name in df_full["Model Configuration"].to_list():
|
| 245 |
yaml_content = scrape_mergekit_config(model_name)
|
| 246 |
if ("No YAML configuration found" not in yaml_content) and ("Failed to fetch model page" not in yaml_content):
|
| 247 |
+
zf.writestr(f"{model_name.replace('/', '_')}_config.yaml", yaml_content.encode())
|
| 248 |
|
| 249 |
zip_buffer.seek(0)
|
| 250 |
return zip_buffer, "analysis_data.zip"
|
| 251 |
|
|
|
|
| 252 |
# --------------------------------------------------------------------
|
| 253 |
+
# PART 2: THE "DATA START" SNIPPET (RANKS 44–105) + Parser
|
| 254 |
# --------------------------------------------------------------------
|
| 255 |
+
# This is your larger dataset, rank = 44..105
|
| 256 |
benchmark_data = [
|
|
|
|
|
|
|
| 257 |
{
|
| 258 |
"rank": 44,
|
| 259 |
"name": "sometimesanotion/Qwen2.5-14B-Vimarckoso-v3",
|
|
|
|
| 280 |
}
|
| 281 |
}
|
| 282 |
},
|
| 283 |
+
{
|
| 284 |
+
"rank": 45,
|
| 285 |
+
"name": "sthenno-com/miscii-14b-1225",
|
| 286 |
+
"scores": {
|
| 287 |
+
"average": 40.08,
|
| 288 |
+
"IFEval": 78.78,
|
| 289 |
+
"BBH": 50.91,
|
| 290 |
+
"MATH": 31.57,
|
| 291 |
+
"GPQA": 17.00,
|
| 292 |
+
"MUSR": 14.77,
|
| 293 |
+
"MMLU-PRO": 47.46
|
| 294 |
+
},
|
| 295 |
+
"hf_url": "https://huggingface.co/sthenno-com/miscii-14b-1225",
|
| 296 |
+
"known_config": None
|
| 297 |
+
},
|
| 298 |
+
{
|
| 299 |
+
"rank": 46,
|
| 300 |
+
"name": "djuna/Q2.5-Veltha-14B-0.5",
|
| 301 |
+
"scores": {
|
| 302 |
+
"average": 39.96,
|
| 303 |
+
"IFEval": 77.96,
|
| 304 |
+
"BBH": 50.32,
|
| 305 |
+
"MATH": 33.84,
|
| 306 |
+
"GPQA": 15.77,
|
| 307 |
+
"MUSR": 14.17,
|
| 308 |
+
"MMLU-PRO": 47.72
|
| 309 |
+
},
|
| 310 |
+
"hf_url": "https://huggingface.co/djuna/Q2.5-Veltha-14B-0.5",
|
| 311 |
+
"known_config": None
|
| 312 |
+
},
|
| 313 |
+
{
|
| 314 |
+
"rank": 48,
|
| 315 |
+
"name": "sometimesanotion/Qwen2.5-14B-Vimarckoso-v3-model_stock",
|
| 316 |
+
"scores": {
|
| 317 |
+
"average": 39.81,
|
| 318 |
+
"IFEval": 71.62,
|
| 319 |
+
"BBH": 48.76,
|
| 320 |
+
"MATH": 33.99,
|
| 321 |
+
"GPQA": 17.34,
|
| 322 |
+
"MUSR": 19.23,
|
| 323 |
+
"MMLU-PRO": 47.95
|
| 324 |
+
},
|
| 325 |
+
"hf_url": "https://huggingface.co/sometimesanotion/Qwen2.5-14B-Vimarckoso-v3-model_stock",
|
| 326 |
+
"known_config": None
|
| 327 |
+
},
|
| 328 |
+
{
|
| 329 |
+
"rank": 50,
|
| 330 |
+
"name": "sometimesanotion/Qwen2.5-14B-Vimarckoso-v3-Prose01",
|
| 331 |
+
"scores": {
|
| 332 |
+
"average": 39.46,
|
| 333 |
+
"IFEval": 68.72,
|
| 334 |
+
"BBH": 47.71,
|
| 335 |
+
"MATH": 35.05,
|
| 336 |
+
"GPQA": 18.23,
|
| 337 |
+
"MUSR": 19.56,
|
| 338 |
+
"MMLU-PRO": 47.50
|
| 339 |
+
},
|
| 340 |
+
"hf_url": "https://huggingface.co/sometimesanotion/Qwen2.5-14B-Vimarckoso-v3-Prose01",
|
| 341 |
+
"known_config": None
|
| 342 |
+
},
|
| 343 |
+
{
|
| 344 |
+
"rank": 52,
|
| 345 |
+
"name": "arcee-ai/Virtuoso-Small",
|
| 346 |
+
"scores": {
|
| 347 |
+
"average": 39.43,
|
| 348 |
+
"IFEval": 79.35,
|
| 349 |
+
"BBH": 50.40,
|
| 350 |
+
"MATH": 34.29,
|
| 351 |
+
"GPQA": 11.52,
|
| 352 |
+
"MUSR": 14.44,
|
| 353 |
+
"MMLU-PRO": 46.57
|
| 354 |
+
},
|
| 355 |
+
"hf_url": "https://huggingface.co/arcee-ai/Virtuoso-Small",
|
| 356 |
+
"known_config": None
|
| 357 |
+
},
|
| 358 |
+
{
|
| 359 |
+
"rank": 54,
|
| 360 |
+
"name": "sometimesanotion/Qwentinuum-14B-v6",
|
| 361 |
+
"scores": {
|
| 362 |
+
"average": 39.23,
|
| 363 |
+
"IFEval": 63.04,
|
| 364 |
+
"BBH": 50.23,
|
| 365 |
+
"MATH": 33.84,
|
| 366 |
+
"GPQA": 18.23,
|
| 367 |
+
"MUSR": 21.18,
|
| 368 |
+
"MMLU-PRO": 48.89
|
| 369 |
+
},
|
| 370 |
+
"hf_url": "https://huggingface.co/sometimesanotion/Qwentinuum-14B-v6",
|
| 371 |
+
"known_config": None
|
| 372 |
+
},
|
| 373 |
+
{
|
| 374 |
+
"rank": 55,
|
| 375 |
+
"name": "djuna/Q2.5-Veltha-14B",
|
| 376 |
+
"scores": {
|
| 377 |
+
"average": 39.21,
|
| 378 |
+
"IFEval": 82.92,
|
| 379 |
+
"BBH": 49.75,
|
| 380 |
+
"MATH": 28.02,
|
| 381 |
+
"GPQA": 14.54,
|
| 382 |
+
"MUSR": 12.26,
|
| 383 |
+
"MMLU-PRO": 47.76
|
| 384 |
+
},
|
| 385 |
+
"hf_url": "https://huggingface.co/djuna/Q2.5-Veltha-14B",
|
| 386 |
+
"known_config": None
|
| 387 |
+
},
|
| 388 |
+
{
|
| 389 |
+
"rank": 57,
|
| 390 |
+
"name": "allknowingroger/QwenSlerp6-14B",
|
| 391 |
+
"scores": {
|
| 392 |
+
"average": 39.02,
|
| 393 |
+
"IFEval": 68.67,
|
| 394 |
+
"BBH": 47.59,
|
| 395 |
+
"MATH": 34.14,
|
| 396 |
+
"GPQA": 16.44,
|
| 397 |
+
"MUSR": 18.32,
|
| 398 |
+
"MMLU-PRO": 48.95
|
| 399 |
+
},
|
| 400 |
+
"hf_url": "https://huggingface.co/allknowingroger/QwenSlerp6-14B",
|
| 401 |
+
"known_config": None
|
| 402 |
+
},
|
| 403 |
+
{
|
| 404 |
+
"rank": 58,
|
| 405 |
+
"name": "allknowingroger/QwenSlerp5-14B",
|
| 406 |
+
"scores": {
|
| 407 |
+
"average": 38.94,
|
| 408 |
+
"IFEval": 71.19,
|
| 409 |
+
"BBH": 47.39,
|
| 410 |
+
"MATH": 33.16,
|
| 411 |
+
"GPQA": 15.32,
|
| 412 |
+
"MUSR": 17.81,
|
| 413 |
+
"MMLU-PRO": 48.78
|
| 414 |
+
},
|
| 415 |
+
"hf_url": "https://huggingface.co/allknowingroger/QwenSlerp5-14B",
|
| 416 |
+
"known_config": None
|
| 417 |
+
},
|
| 418 |
+
{
|
| 419 |
+
"rank": 59,
|
| 420 |
+
"name": "sometimesanotion/Qwentinuum-14B-v5",
|
| 421 |
+
"scores": {
|
| 422 |
+
"average": 38.87,
|
| 423 |
+
"IFEval": 62.86,
|
| 424 |
+
"BBH": 50.28,
|
| 425 |
+
"MATH": 31.57,
|
| 426 |
+
"GPQA": 18.34,
|
| 427 |
+
"MUSR": 21.09,
|
| 428 |
+
"MMLU-PRO": 49.09
|
| 429 |
+
},
|
| 430 |
+
"hf_url": "https://huggingface.co/sometimesanotion/Qwentinuum-14B-v5",
|
| 431 |
+
"known_config": None
|
| 432 |
+
},
|
| 433 |
+
{
|
| 434 |
+
"rank": 60,
|
| 435 |
+
"name": "sometimesanotion/Qwenvergence-14B-v6-Prose",
|
| 436 |
+
"scores": {
|
| 437 |
+
"average": 38.82,
|
| 438 |
+
"IFEval": 59.90,
|
| 439 |
+
"BBH": 50.12,
|
| 440 |
+
"MATH": 34.89,
|
| 441 |
+
"GPQA": 18.46,
|
| 442 |
+
"MUSR": 21.02,
|
| 443 |
+
"MMLU-PRO": 48.56
|
| 444 |
+
},
|
| 445 |
+
"hf_url": "https://huggingface.co/sometimesanotion/Qwenvergence-14B-v6-Prose",
|
| 446 |
+
"known_config": None
|
| 447 |
+
},
|
| 448 |
+
{
|
| 449 |
+
"rank": 61,
|
| 450 |
+
"name": "CultriX/Qwen2.5-14B-Brocav3",
|
| 451 |
+
"scores": {
|
| 452 |
+
"average": 38.76,
|
| 453 |
+
"IFEval": 69.52,
|
| 454 |
+
"BBH": 49.05,
|
| 455 |
+
"MATH": 32.25,
|
| 456 |
+
"GPQA": 14.54,
|
| 457 |
+
"MUSR": 19.25,
|
| 458 |
+
"MMLU-PRO": 47.97
|
| 459 |
+
},
|
| 460 |
+
"hf_url": "https://huggingface.co/CultriX/Qwen2.5-14B-Brocav3",
|
| 461 |
+
"known_config": None
|
| 462 |
+
},
|
| 463 |
+
{
|
| 464 |
+
"rank": 62,
|
| 465 |
+
"name": "sometimesanotion/Qwentinuum-14B-v7",
|
| 466 |
+
"scores": {
|
| 467 |
+
"average": 38.76,
|
| 468 |
+
"IFEval": 61.09,
|
| 469 |
+
"BBH": 50.35,
|
| 470 |
+
"MATH": 33.38,
|
| 471 |
+
"GPQA": 18.79,
|
| 472 |
+
"MUSR": 19.95,
|
| 473 |
+
"MMLU-PRO": 49.00
|
| 474 |
+
},
|
| 475 |
+
"hf_url": "https://huggingface.co/sometimesanotion/Qwentinuum-14B-v7",
|
| 476 |
+
"known_config": None
|
| 477 |
+
},
|
| 478 |
+
{
|
| 479 |
+
"rank": 64,
|
| 480 |
+
"name": "sometimesanotion/Qwentinuum-14B-v3",
|
| 481 |
+
"scores": {
|
| 482 |
+
"average": 38.74,
|
| 483 |
+
"IFEval": 61.58,
|
| 484 |
+
"BBH": 50.04,
|
| 485 |
+
"MATH": 32.85,
|
| 486 |
+
"GPQA": 18.34,
|
| 487 |
+
"MUSR": 20.62,
|
| 488 |
+
"MMLU-PRO": 49.03
|
| 489 |
+
},
|
| 490 |
+
"hf_url": "https://huggingface.co/sometimesanotion/Qwentinuum-14B-v3",
|
| 491 |
+
"known_config": None
|
| 492 |
+
},
|
| 493 |
+
{
|
| 494 |
+
"rank": 65,
|
| 495 |
+
"name": "allura-org/TQ2.5-14B-Aletheia-v1",
|
| 496 |
+
"scores": {
|
| 497 |
+
"average": 38.74,
|
| 498 |
+
"IFEval": 75.30,
|
| 499 |
+
"BBH": 50.88,
|
| 500 |
+
"MATH": 29.53,
|
| 501 |
+
"GPQA": 14.99,
|
| 502 |
+
"MUSR": 14.61,
|
| 503 |
+
"MMLU-PRO": 47.12
|
| 504 |
+
},
|
| 505 |
+
"hf_url": "https://huggingface.co/allura-org/TQ2.5-14B-Aletheia-v1",
|
| 506 |
+
"known_config": None
|
| 507 |
+
},
|
| 508 |
+
{
|
| 509 |
+
"rank": 66,
|
| 510 |
+
"name": "qingy2024/Fusion4-14B-Instruct",
|
| 511 |
+
"scores": {
|
| 512 |
+
"average": 38.73,
|
| 513 |
+
"IFEval": 76.49,
|
| 514 |
+
"BBH": 50.70,
|
| 515 |
+
"MATH": 33.91,
|
| 516 |
+
"GPQA": 10.74,
|
| 517 |
+
"MUSR": 13.97,
|
| 518 |
+
"MMLU-PRO": 46.60
|
| 519 |
+
},
|
| 520 |
+
"hf_url": "https://huggingface.co/qingy2024/Fusion4-14B-Instruct",
|
| 521 |
+
"known_config": None
|
| 522 |
+
},
|
| 523 |
+
{
|
| 524 |
+
"rank": 68,
|
| 525 |
+
"name": "CultriX/Qwen2.5-14B-Brocav7",
|
| 526 |
+
"scores": {
|
| 527 |
+
"average": 38.52,
|
| 528 |
+
"IFEval": 67.24,
|
| 529 |
+
"BBH": 48.91,
|
| 530 |
+
"MATH": 31.87,
|
| 531 |
+
"GPQA": 15.66,
|
| 532 |
+
"MUSR": 20.15,
|
| 533 |
+
"MMLU-PRO": 47.31
|
| 534 |
+
},
|
| 535 |
+
"hf_url": "https://huggingface.co/CultriX/Qwen2.5-14B-Brocav7",
|
| 536 |
+
"known_config": None
|
| 537 |
+
},
|
| 538 |
+
{
|
| 539 |
+
"rank": 71,
|
| 540 |
+
"name": "sometimesanotion/Qwentinuum-14B-v6-Prose",
|
| 541 |
+
"scores": {
|
| 542 |
+
"average": 38.46,
|
| 543 |
+
"IFEval": 56.43,
|
| 544 |
+
"BBH": 50.14,
|
| 545 |
+
"MATH": 35.57,
|
| 546 |
+
"GPQA": 18.46,
|
| 547 |
+
"MUSR": 21.34,
|
| 548 |
+
"MMLU-PRO": 48.80
|
| 549 |
+
},
|
| 550 |
+
"hf_url": "https://huggingface.co/sometimesanotion/Qwentinuum-14B-v6-Prose",
|
| 551 |
+
"known_config": None
|
| 552 |
+
},
|
| 553 |
+
{
|
| 554 |
+
"rank": 76,
|
| 555 |
+
"name": "CultriX/Qwen2.5-14B-Brocav6",
|
| 556 |
+
"scores": {
|
| 557 |
+
"average": 38.32,
|
| 558 |
+
"IFEval": 69.95,
|
| 559 |
+
"BBH": 47.82,
|
| 560 |
+
"MATH": 29.61,
|
| 561 |
+
"GPQA": 15.66,
|
| 562 |
+
"MUSR": 18.88,
|
| 563 |
+
"MMLU-PRO": 47.99
|
| 564 |
+
},
|
| 565 |
+
"hf_url": "https://huggingface.co/CultriX/Qwen2.5-14B-Brocav6",
|
| 566 |
+
"known_config": None
|
| 567 |
+
},
|
| 568 |
+
{
|
| 569 |
+
"rank": 80,
|
| 570 |
+
"name": "CultriX/SeQwence-14Bv1",
|
| 571 |
+
"scores": {
|
| 572 |
+
"average": 38.20,
|
| 573 |
+
"IFEval": 66.78,
|
| 574 |
+
"BBH": 47.19,
|
| 575 |
+
"MATH": 33.53,
|
| 576 |
+
"GPQA": 14.88,
|
| 577 |
+
"MUSR": 18.80,
|
| 578 |
+
"MMLU-PRO": 48.00
|
| 579 |
+
},
|
| 580 |
+
"hf_url": "https://huggingface.co/CultriX/SeQwence-14Bv1",
|
| 581 |
+
"known_config": None
|
| 582 |
+
},
|
| 583 |
+
{
|
| 584 |
+
"rank": 85,
|
| 585 |
+
"name": "sometimesanotion/Qwentinuum-14B-v013",
|
| 586 |
+
"scores": {
|
| 587 |
+
"average": 37.96,
|
| 588 |
+
"IFEval": 67.11,
|
| 589 |
+
"BBH": 43.97,
|
| 590 |
+
"MATH": 33.01,
|
| 591 |
+
"GPQA": 14.32,
|
| 592 |
+
"MUSR": 24.99,
|
| 593 |
+
"MMLU-PRO": 44.34
|
| 594 |
+
},
|
| 595 |
+
"hf_url": "https://huggingface.co/sometimesanotion/Qwentinuum-14B-v013",
|
| 596 |
+
"known_config": None
|
| 597 |
+
},
|
| 598 |
+
{
|
| 599 |
+
"rank": 86,
|
| 600 |
+
"name": "CultriX/Qwen2.5-14B-Wernickev3",
|
| 601 |
+
"scores": {
|
| 602 |
+
"average": 37.94,
|
| 603 |
+
"IFEval": 70.48,
|
| 604 |
+
"BBH": 44.58,
|
| 605 |
+
"MATH": 32.78,
|
| 606 |
+
"GPQA": 14.99,
|
| 607 |
+
"MUSR": 18.69,
|
| 608 |
+
"MMLU-PRO": 46.13
|
| 609 |
+
},
|
| 610 |
+
"hf_url": "https://huggingface.co/CultriX/Qwen2.5-14B-Wernickev3",
|
| 611 |
+
"known_config": None
|
| 612 |
+
},
|
| 613 |
+
{
|
| 614 |
+
"rank": 88,
|
| 615 |
+
"name": "allknowingroger/QwenSlerp4-14B",
|
| 616 |
+
"scores": {
|
| 617 |
+
"average": 37.80,
|
| 618 |
+
"IFEval": 63.28,
|
| 619 |
+
"BBH": 49.38,
|
| 620 |
+
"MATH": 30.97,
|
| 621 |
+
"GPQA": 16.33,
|
| 622 |
+
"MUSR": 17.59,
|
| 623 |
+
"MMLU-PRO": 49.28
|
| 624 |
+
},
|
| 625 |
+
"hf_url": "https://huggingface.co/allknowingroger/QwenSlerp4-14B",
|
| 626 |
+
"known_config": None
|
| 627 |
+
},
|
| 628 |
+
{
|
| 629 |
+
"rank": 89,
|
| 630 |
+
"name": "CultriX/Qwen2.5-14B-Broca",
|
| 631 |
+
"scores": {
|
| 632 |
+
"average": 37.72,
|
| 633 |
+
"IFEval": 56.04,
|
| 634 |
+
"BBH": 50.03,
|
| 635 |
+
"MATH": 34.59,
|
| 636 |
+
"GPQA": 18.23,
|
| 637 |
+
"MUSR": 18.95,
|
| 638 |
+
"MMLU-PRO": 48.49
|
| 639 |
+
},
|
| 640 |
+
"hf_url": "https://huggingface.co/CultriX/Qwen2.5-14B-Broca",
|
| 641 |
+
"known_config": None
|
| 642 |
+
},
|
| 643 |
+
{
|
| 644 |
+
"rank": 90,
|
| 645 |
+
"name": "CultriX/Qwen2.5-14B-Emerged",
|
| 646 |
+
"scores": {
|
| 647 |
+
"average": 37.66,
|
| 648 |
+
"IFEval": 70.00,
|
| 649 |
+
"BBH": 45.93,
|
| 650 |
+
"MATH": 30.74,
|
| 651 |
+
"GPQA": 14.32,
|
| 652 |
+
"MUSR": 18.47,
|
| 653 |
+
"MMLU-PRO": 46.51
|
| 654 |
+
},
|
| 655 |
+
"hf_url": "https://huggingface.co/CultriX/Qwen2.5-14B-Emerged",
|
| 656 |
+
"known_config": None
|
| 657 |
+
},
|
| 658 |
+
{
|
| 659 |
+
"rank": 91,
|
| 660 |
+
"name": "sometimesanotion/Qwentinuum-14B-v8",
|
| 661 |
+
"scores": {
|
| 662 |
+
"average": 37.65,
|
| 663 |
+
"IFEval": 54.12,
|
| 664 |
+
"BBH": 50.11,
|
| 665 |
+
"MATH": 34.14,
|
| 666 |
+
"GPQA": 17.79,
|
| 667 |
+
"MUSR": 20.75,
|
| 668 |
+
"MMLU-PRO": 49.02
|
| 669 |
+
},
|
| 670 |
+
"hf_url": "https://huggingface.co/sometimesanotion/Qwentinuum-14B-v8",
|
| 671 |
+
"known_config": None
|
| 672 |
+
},
|
| 673 |
+
{
|
| 674 |
+
"rank": 92,
|
| 675 |
+
"name": "qingy2024/Fusion-14B-Instruct",
|
| 676 |
+
"scores": {
|
| 677 |
+
"average": 37.64,
|
| 678 |
+
"IFEval": 72.60,
|
| 679 |
+
"BBH": 48.58,
|
| 680 |
+
"MATH": 30.97,
|
| 681 |
+
"GPQA": 13.98,
|
| 682 |
+
"MUSR": 14.81,
|
| 683 |
+
"MMLU-PRO": 44.93
|
| 684 |
+
},
|
| 685 |
+
"hf_url": "https://huggingface.co/qingy2024/Fusion-14B-Instruct",
|
| 686 |
+
"known_config": None
|
| 687 |
+
},
|
| 688 |
+
{
|
| 689 |
+
"rank": 94,
|
| 690 |
+
"name": "CultriX/Qwestion-14B",
|
| 691 |
+
"scores": {
|
| 692 |
+
"average": 37.63,
|
| 693 |
+
"IFEval": 63.18,
|
| 694 |
+
"BBH": 48.76,
|
| 695 |
+
"MATH": 31.72,
|
| 696 |
+
"GPQA": 15.77,
|
| 697 |
+
"MUSR": 17.22,
|
| 698 |
+
"MMLU-PRO": 49.14
|
| 699 |
+
},
|
| 700 |
+
"hf_url": "https://huggingface.co/CultriX/Qwestion-14B",
|
| 701 |
+
"known_config": None
|
| 702 |
+
},
|
| 703 |
+
{
|
| 704 |
+
"rank": 99,
|
| 705 |
+
"name": "sometimesanotion/Qwenvergence-14B-v3-Prose",
|
| 706 |
+
"scores": {
|
| 707 |
+
"average": 37.37,
|
| 708 |
+
"IFEval": 49.18,
|
| 709 |
+
"BBH": 49.80,
|
| 710 |
+
"MATH": 35.57,
|
| 711 |
+
"GPQA": 19.35,
|
| 712 |
+
"MUSR": 21.77,
|
| 713 |
+
"MMLU-PRO": 48.55
|
| 714 |
+
},
|
| 715 |
+
"hf_url": "https://huggingface.co/sometimesanotion/Qwenvergence-14B-v3-Prose",
|
| 716 |
+
"known_config": None
|
| 717 |
+
},
|
| 718 |
+
{
|
| 719 |
+
"rank": 102,
|
| 720 |
+
"name": "CultriX/SeQwence-14B-v5",
|
| 721 |
+
"scores": {
|
| 722 |
+
"average": 37.27,
|
| 723 |
+
"IFEval": 59.20,
|
| 724 |
+
"BBH": 50.00,
|
| 725 |
+
"MATH": 31.04,
|
| 726 |
+
"GPQA": 16.00,
|
| 727 |
+
"MUSR": 18.33,
|
| 728 |
+
"MMLU-PRO": 49.05
|
| 729 |
+
},
|
| 730 |
+
"hf_url": "https://huggingface.co/CultriX/SeQwence-14B-v5",
|
| 731 |
+
"known_config": None
|
| 732 |
+
},
|
| 733 |
+
{
|
| 734 |
+
"rank": 103,
|
| 735 |
+
"name": "sometimesanotion/Qwen-14B-ProseStock-v4",
|
| 736 |
+
"scores": {
|
| 737 |
+
"average": 37.23,
|
| 738 |
+
"IFEval": 49.42,
|
| 739 |
+
"BBH": 49.54,
|
| 740 |
+
"MATH": 35.50,
|
| 741 |
+
"GPQA": 18.46,
|
| 742 |
+
"MUSR": 21.70,
|
| 743 |
+
"MMLU-PRO": 48.74
|
| 744 |
+
},
|
| 745 |
+
"hf_url": "https://huggingface.co/sometimesanotion/Qwen-14B-ProseStock-v4",
|
| 746 |
+
"known_config": None
|
| 747 |
+
},
|
| 748 |
+
{
|
| 749 |
+
"rank": 104,
|
| 750 |
+
"name": "sometimesanotion/IF-reasoning-experiment-40",
|
| 751 |
+
"scores": {
|
| 752 |
+
"average": 37.21,
|
| 753 |
+
"IFEval": 63.30,
|
| 754 |
+
"BBH": 44.31,
|
| 755 |
+
"MATH": 27.72,
|
| 756 |
+
"GPQA": 17.34,
|
| 757 |
+
"MUSR": 25.86,
|
| 758 |
+
"MMLU-PRO": 44.72
|
| 759 |
+
},
|
| 760 |
+
"hf_url": "https://huggingface.co/sometimesanotion/IF-reasoning-experiment-40",
|
| 761 |
+
"known_config": None
|
| 762 |
+
},
|
| 763 |
+
{
|
| 764 |
+
"rank": 105,
|
| 765 |
+
"name": "CultriX/SeQwence-14B-EvolMerge",
|
| 766 |
+
"scores": {
|
| 767 |
+
"average": 37.20,
|
| 768 |
+
"IFEval": 53.82,
|
| 769 |
+
"BBH": 50.78,
|
| 770 |
+
"MATH": 31.80,
|
| 771 |
+
"GPQA": 17.45,
|
| 772 |
+
"MUSR": 20.26,
|
| 773 |
+
"MMLU-PRO": 49.10
|
| 774 |
+
},
|
| 775 |
+
"hf_url": "https://huggingface.co/CultriX/SeQwence-14B-EvolMerge",
|
| 776 |
+
"known_config": None
|
| 777 |
+
}
|
| 778 |
]
|
| 779 |
|
|
|
|
| 780 |
def snippet_scrape_model_page(url):
|
| 781 |
"""
|
| 782 |
+
Equivalent scraping function for the larger dataset
|
| 783 |
+
to look for <pre> YAML and a .metadata section.
|
| 784 |
"""
|
| 785 |
+
try:
|
| 786 |
+
response = requests.get(url)
|
| 787 |
+
if response.status_code != 200:
|
| 788 |
+
return f"Error: Unable to fetch the page (Status Code: {response.status_code})"
|
| 789 |
+
|
| 790 |
+
soup = BeautifulSoup(response.text, "html.parser")
|
| 791 |
+
|
| 792 |
+
yaml_config = soup.find("pre")
|
| 793 |
+
yaml_text = yaml_config.text.strip() if yaml_config else "No YAML configuration found."
|
| 794 |
+
|
| 795 |
+
metadata_section = soup.find("div", class_="metadata")
|
| 796 |
+
metadata_text = metadata_section.text.strip() if metadata_section else "No metadata found."
|
| 797 |
+
|
| 798 |
+
return {
|
| 799 |
+
"yaml_configuration": yaml_text,
|
| 800 |
+
"metadata": metadata_text
|
| 801 |
+
}
|
| 802 |
+
|
| 803 |
+
except Exception as e:
|
| 804 |
+
return f"Error: {str(e)}"
|
| 805 |
|
| 806 |
def snippet_print_benchmark_and_config_info(model_info):
|
| 807 |
"""
|
| 808 |
+
Prints an overview for each model in the rank=44..105 dataset.
|
| 809 |
+
If known_config is not None, prints it. Otherwise attempts to scrape.
|
| 810 |
"""
|
| 811 |
print(f"---\nModel Rank: {model_info['rank']}")
|
| 812 |
print(f"Model Name: {model_info['name']}")
|
|
|
|
| 818 |
print(f"Models average score in MUSR benchmarks in %: {model_info['scores']['MUSR']}")
|
| 819 |
print(f"Models average score in MMLU-PRO benchmarks in %: {model_info['scores']['MMLU-PRO']}")
|
| 820 |
|
| 821 |
+
# If there's a known_config, print it in YAML form and stop.
|
| 822 |
if model_info["known_config"] is not None:
|
| 823 |
print("###")
|
| 824 |
print("models:")
|
|
|
|
| 828 |
print(f"base_model: {model_info['known_config']['base_model']}")
|
| 829 |
print(f"dtype: {model_info['known_config']['dtype']}")
|
| 830 |
print("parameters:")
|
| 831 |
+
t_vals = model_info["known_config"]["parameters"]["t"]
|
| 832 |
+
print(f" t: {t_vals} # V shaped curve: Hermes for input & output, WizardMath in the middle layers")
|
| 833 |
print("###")
|
| 834 |
return
|
| 835 |
|
| 836 |
+
# Otherwise, do scraping:
|
| 837 |
scraped = snippet_scrape_model_page(model_info["hf_url"])
|
| 838 |
if isinstance(scraped, str):
|
| 839 |
# Means it's an error string or something
|
| 840 |
+
print("(No MergeKit configuration found or scraping error.)")
|
| 841 |
+
print(scraped)
|
|
|
|
|
|
|
|
|
|
| 842 |
return
|
| 843 |
else:
|
| 844 |
+
# It's presumably a dict
|
| 845 |
+
if "No YAML configuration found." in scraped["yaml_configuration"]:
|
| 846 |
print("(No MergeKit configuration found.)\n")
|
|
|
|
| 847 |
print("You can try the following Python script to scrape the model page:\n")
|
| 848 |
print("#" * 70)
|
| 849 |
print(f'''import requests
|
|
|
|
| 877 |
print(result)''')
|
| 878 |
print("#" * 70)
|
| 879 |
else:
|
| 880 |
+
# Found some YAML
|
| 881 |
print("###")
|
| 882 |
print(scraped["yaml_configuration"])
|
| 883 |
print("###")
|
| 884 |
|
| 885 |
def run_non_tiny_benchmarks():
|
| 886 |
"""
|
| 887 |
+
Captures the stdout from printing each model in benchmark_data (ranks 44..105),
|
| 888 |
+
returning the entire output as a single string for Gradio to display.
|
| 889 |
"""
|
| 890 |
old_stdout = sys.stdout
|
| 891 |
buffer = io.StringIO()
|
|
|
|
| 897 |
sys.stdout = old_stdout
|
| 898 |
return buffer.getvalue()
|
| 899 |
|
|
|
|
| 900 |
# --------------------------------------------------------------------
|
| 901 |
+
# PART 3: The Gradio App
|
| 902 |
# --------------------------------------------------------------------
|
| 903 |
with gr.Blocks() as demo:
|
| 904 |
gr.Markdown("# Comprehensive Model Performance Analysis with Hugging Face Links")
|
| 905 |
|
| 906 |
+
# The existing UI for the “tiny” data
|
| 907 |
with gr.Row():
|
| 908 |
btn1 = gr.Button("Show Average Performance")
|
| 909 |
img1 = gr.Image(type="pil", label="Average Performance Plot")
|
|
|
|
| 928 |
heatmap_download = gr.File(label="Download Heatmap")
|
| 929 |
btn4.click(plot_heatmap, outputs=[heatmap_img, heatmap_download])
|
| 930 |
|
| 931 |
+
# Scraping & YAML handling for the *tiny* table
|
| 932 |
with gr.Row():
|
| 933 |
model_selector = gr.Dropdown(choices=df_full["Model Configuration"].tolist(), label="Select a Model")
|
| 934 |
with gr.Column():
|
|
|
|
| 940 |
yaml_download = gr.File(label="Download MergeKit Configuration")
|
| 941 |
save_yaml_btn.click(download_yaml, inputs=[yaml_output, model_selector], outputs=yaml_download)
|
| 942 |
|
| 943 |
+
# Download everything (CSV, plots, any found YAML)
|
| 944 |
with gr.Row():
|
| 945 |
download_all_btn = gr.Button("Download Everything")
|
| 946 |
all_downloads = gr.File(label="Download All Data")
|
| 947 |
download_all_btn.click(download_all_data, outputs=all_downloads)
|
| 948 |
|
| 949 |
+
# Live Scraping
|
| 950 |
gr.Markdown("## Live Scraping Features")
|
| 951 |
with gr.Row():
|
| 952 |
url_input = gr.Textbox(label="Enter Hugging Face Model URL", placeholder="https://huggingface.co/<model>")
|
|
|
|
| 954 |
live_scrape_output = gr.Textbox(label="Scraped Data", lines=15)
|
| 955 |
live_scrape_btn.click(display_scraped_model_data, inputs=url_input, outputs=live_scrape_output)
|
| 956 |
|
| 957 |
+
# Non-Tiny Benchmarks
|
| 958 |
gr.Markdown("## Non-Tiny Benchmark Parser (Ranks 44–105)")
|
| 959 |
with gr.Row():
|
| 960 |
parse_non_tiny_btn = gr.Button("Parse Non-Tiny Benchmarks")
|
| 961 |
parse_non_tiny_output = gr.Textbox(label="Non-Tiny Benchmark Output", lines=30)
|
| 962 |
parse_non_tiny_btn.click(fn=run_non_tiny_benchmarks, outputs=parse_non_tiny_output)
|
| 963 |
|
| 964 |
+
demo.launch()
|