Spaces:
Running
on
CPU Upgrade
Running
on
CPU Upgrade
First commit
Browse files- app.py +352 -3
- release_date_mapping.json +857 -0
- requirements.txt +7 -0
- utils.py +234 -0
app.py
CHANGED
@@ -1,7 +1,356 @@
|
|
|
|
|
|
|
|
|
|
|
|
1 |
import gradio as gr
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
2 |
|
3 |
-
|
4 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
5 |
|
6 |
-
demo = gr.Interface(fn=greet, inputs="text", outputs="text")
|
7 |
demo.launch()
|
|
|
1 |
+
import os
|
2 |
+
import pickle
|
3 |
+
|
4 |
+
import pandas as pd
|
5 |
+
import numpy as np
|
6 |
import gradio as gr
|
7 |
+
from datetime import datetime
|
8 |
+
from huggingface_hub import HfApi
|
9 |
+
from apscheduler.schedulers.background import BackgroundScheduler
|
10 |
+
import plotly.graph_objects as go
|
11 |
+
|
12 |
+
from utils import (
|
13 |
+
KEY_TO_CATEGORY_NAME,
|
14 |
+
CAT_NAME_TO_EXPLANATION,
|
15 |
+
download_latest_data_from_space,
|
16 |
+
get_constants,
|
17 |
+
update_release_date_mapping,
|
18 |
+
format_data,
|
19 |
+
get_trendlines,
|
20 |
+
find_crossover_point,
|
21 |
+
sigmoid_transition
|
22 |
+
)
|
23 |
+
|
24 |
+
###################
|
25 |
+
### Initialize scheduler
|
26 |
+
###################
|
27 |
+
|
28 |
+
|
29 |
+
def restart_space():
|
30 |
+
HfApi(token=os.getenv("HF_TOKEN", None)).restart_space(
|
31 |
+
repo_id="andrewrreed/closed-vs-open-arena-elo"
|
32 |
+
)
|
33 |
+
print(f"Space restarted on {datetime.now()}")
|
34 |
+
|
35 |
+
|
36 |
+
# restart the space every day at 9am
|
37 |
+
scheduler = BackgroundScheduler()
|
38 |
+
scheduler.add_job(restart_space, "cron", day_of_week="mon-sun", hour=7, minute=0)
|
39 |
+
scheduler.start()
|
40 |
+
|
41 |
+
###################
|
42 |
+
### Load Data
|
43 |
+
###################
|
44 |
+
|
45 |
+
# gather ELO data
|
46 |
+
latest_elo_file_local = download_latest_data_from_space(
|
47 |
+
repo_id="lmsys/chatbot-arena-leaderboard", file_type="pkl"
|
48 |
+
)
|
49 |
+
|
50 |
+
with open(latest_elo_file_local, "rb") as fin:
|
51 |
+
elo_results = pickle.load(fin)
|
52 |
+
|
53 |
+
# TO-DO: need to also include vision
|
54 |
+
elo_results = elo_results["text"]
|
55 |
+
|
56 |
+
arena_dfs = {}
|
57 |
+
for k in KEY_TO_CATEGORY_NAME.keys():
|
58 |
+
if k not in elo_results:
|
59 |
+
continue
|
60 |
+
arena_dfs[KEY_TO_CATEGORY_NAME[k]] = elo_results[k]["leaderboard_table_df"]
|
61 |
+
|
62 |
+
# gather open llm leaderboard data
|
63 |
+
latest_leaderboard_file_local = download_latest_data_from_space(
|
64 |
+
repo_id="lmsys/chatbot-arena-leaderboard", file_type="csv"
|
65 |
+
)
|
66 |
+
leaderboard_df = pd.read_csv(latest_leaderboard_file_local)
|
67 |
+
|
68 |
+
# load release date mapping data
|
69 |
+
release_date_mapping = pd.read_json("release_date_mapping.json", orient="records")
|
70 |
+
|
71 |
+
###################
|
72 |
+
### Prepare Data
|
73 |
+
###################
|
74 |
+
|
75 |
+
# update release date mapping with new models
|
76 |
+
# check for new models in ELO data
|
77 |
+
new_model_keys_to_add = [
|
78 |
+
model
|
79 |
+
for model in arena_dfs["Overall"].index.to_list()
|
80 |
+
if model not in release_date_mapping["key"].to_list()
|
81 |
+
]
|
82 |
+
if new_model_keys_to_add:
|
83 |
+
release_date_mapping = update_release_date_mapping(
|
84 |
+
new_model_keys_to_add, leaderboard_df, release_date_mapping
|
85 |
+
)
|
86 |
+
|
87 |
+
# merge leaderboard data with ELO data
|
88 |
+
merged_dfs = {}
|
89 |
+
for k, v in arena_dfs.items():
|
90 |
+
merged_dfs[k] = (
|
91 |
+
pd.merge(arena_dfs[k], leaderboard_df, left_index=True, right_on="key")
|
92 |
+
.sort_values("rating", ascending=False)
|
93 |
+
.reset_index(drop=True)
|
94 |
+
)
|
95 |
+
|
96 |
+
# add release dates into the merged data
|
97 |
+
for k, v in merged_dfs.items():
|
98 |
+
merged_dfs[k] = pd.merge(
|
99 |
+
merged_dfs[k], release_date_mapping[["key", "Release Date"]], on="key"
|
100 |
+
)
|
101 |
+
|
102 |
+
# format dataframes
|
103 |
+
merged_dfs = {k: format_data(v) for k, v in merged_dfs.items()}
|
104 |
+
|
105 |
+
# get constants
|
106 |
+
min_elo_score, max_elo_score, _ = get_constants(merged_dfs)
|
107 |
+
date_updated = elo_results["full"]["last_updated_datetime"].split(" ")[0]
|
108 |
+
orgs = merged_dfs["Overall"].Organization.unique().tolist()
|
109 |
+
|
110 |
+
###################
|
111 |
+
### Build and Plot Data
|
112 |
+
###################
|
113 |
+
|
114 |
+
|
115 |
+
df = merged_dfs["Overall"]
|
116 |
+
top_orgs = df.groupby("Organization")["rating"].max().nlargest(11).index.tolist()
|
117 |
+
|
118 |
+
df = df.loc[(df["Organization"].isin(top_orgs)) & (df["rating"] > 1000)]
|
119 |
+
print(df)
|
120 |
+
|
121 |
+
df = df.loc[~df["Release Date"].isna()]
|
122 |
+
|
123 |
+
def get_data_split(dfs, set_name):
|
124 |
+
df = dfs[set_name].copy(deep=True)
|
125 |
+
return df.reset_index(drop=True)
|
126 |
+
|
127 |
+
|
128 |
+
def clean_df_for_display(df):
|
129 |
+
df = df.loc[
|
130 |
+
:,
|
131 |
+
[
|
132 |
+
"Model",
|
133 |
+
"rating",
|
134 |
+
"MMLU",
|
135 |
+
"MT-bench (score)",
|
136 |
+
"Release Date",
|
137 |
+
"Organization",
|
138 |
+
"License",
|
139 |
+
"Link",
|
140 |
+
],
|
141 |
+
].rename(columns={"rating": "ELO Score", "MT-bench (score)": "MT-Bench"})
|
142 |
+
|
143 |
+
df["Release Date"] = df["Release Date"].astype(str)
|
144 |
+
df.sort_values("ELO Score", ascending=False, inplace=True)
|
145 |
+
df.reset_index(drop=True, inplace=True)
|
146 |
+
return df
|
147 |
+
|
148 |
+
def format_data(df):
|
149 |
+
"""
|
150 |
+
Formats the given DataFrame by performing the following operations:
|
151 |
+
- Converts the 'License' column values to 'Proprietary LLM' if they are in PROPRIETARY_LICENSES, otherwise 'Open LLM'.
|
152 |
+
- Converts the 'Release Date' column to datetime format.
|
153 |
+
- Adds a new 'Month-Year' column by extracting the month and year from the 'Release Date' column.
|
154 |
+
- Rounds the 'rating' column to the nearest integer.
|
155 |
+
- Resets the index of the DataFrame.
|
156 |
+
Args:
|
157 |
+
df (pandas.DataFrame): The DataFrame to be formatted.
|
158 |
+
Returns:
|
159 |
+
pandas.DataFrame: The formatted DataFrame.
|
160 |
+
"""
|
161 |
+
|
162 |
+
PROPRIETARY_LICENSES = ["Proprietary", "Proprietory"]
|
163 |
+
|
164 |
+
df["License"] = df["License"].apply(
|
165 |
+
lambda x: "Proprietary LLM" if x in PROPRIETARY_LICENSES else "Open LLM"
|
166 |
+
)
|
167 |
+
df["Release Date"] = pd.to_datetime(df["Release Date"])
|
168 |
+
df["Month-Year"] = df["Release Date"].dt.to_period("M")
|
169 |
+
df["rating"] = df["rating"].round()
|
170 |
+
return df.reset_index(drop=True)
|
171 |
+
|
172 |
+
|
173 |
+
# Define organization to country mapping and colors
|
174 |
+
org_info = {
|
175 |
+
"OpenAI": ("#00A67E", "🇺🇸"), # Teal
|
176 |
+
"Google": ("#4285F4", "🇺🇸"), # Google Blue
|
177 |
+
"xAI": ("black", "🇺🇸"), # Bright Orange
|
178 |
+
"Anthropic": ("#cc785c", "🇺🇸"), # Brown (as requested)
|
179 |
+
"Meta": ("#0064E0", "🇺🇸"), # Facebook Blue
|
180 |
+
"Alibaba": ("#6958cf", "🇨🇳"),
|
181 |
+
"DeepSeek": ("#C70039", "🇨🇳"),
|
182 |
+
"01 AI": ("#11871e", "🇨🇳"), # Bright Green
|
183 |
+
"DeepSeek AI": ("#9900CC", "🇨🇳"), # Purple
|
184 |
+
"Mistral": ("#ff7000", "🇫🇷"), # Mistral Orange (as requested)
|
185 |
+
"AI21 Labs": ("#1E90FF", "🇮🇱"), # Dodger Blue,
|
186 |
+
"Reka AI": ("#FFC300", "🇺🇸"),
|
187 |
+
"Zhipu AI": ("#FFC300", "🇨🇳"),
|
188 |
+
}
|
189 |
+
|
190 |
+
def make_figure(df):
|
191 |
+
fig = go.Figure()
|
192 |
+
|
193 |
+
for i, org in enumerate(
|
194 |
+
df.groupby("Organization")["rating"]
|
195 |
+
.max()
|
196 |
+
.sort_values(ascending=False)
|
197 |
+
.index.tolist()
|
198 |
+
):
|
199 |
+
org_data = df[df["Organization"] == org]
|
200 |
+
|
201 |
+
if len(org_data) > 0:
|
202 |
+
x_values = []
|
203 |
+
y_values = []
|
204 |
+
current_best = -np.inf
|
205 |
+
best_models = []
|
206 |
+
|
207 |
+
# Group by date and get the best model for each date
|
208 |
+
daily_best = org_data.groupby("Release Date").first().reset_index()
|
209 |
+
|
210 |
+
for _, row in daily_best.iterrows():
|
211 |
+
if row["rating"] > current_best:
|
212 |
+
if len(x_values) > 0:
|
213 |
+
# Create smooth transition
|
214 |
+
transition_days = (row["Release Date"] - x_values[-1]).days
|
215 |
+
transition_points = pd.date_range(
|
216 |
+
x_values[-1],
|
217 |
+
row["Release Date"],
|
218 |
+
periods=max(100, transition_days),
|
219 |
+
)
|
220 |
+
x_values.extend(transition_points)
|
221 |
+
|
222 |
+
transition_y = current_best + (
|
223 |
+
row["rating"] - current_best
|
224 |
+
) * sigmoid_transition(
|
225 |
+
np.linspace(-6, 6, len(transition_points)), 0, k=1
|
226 |
+
)
|
227 |
+
y_values.extend(transition_y)
|
228 |
+
|
229 |
+
x_values.append(row["Release Date"])
|
230 |
+
y_values.append(row["rating"])
|
231 |
+
current_best = row["rating"]
|
232 |
+
best_models.append(row)
|
233 |
+
|
234 |
+
# Extend the line to the current date
|
235 |
+
if x_values[-1] < current_date:
|
236 |
+
x_values.append(current_date)
|
237 |
+
y_values.append(current_best)
|
238 |
+
|
239 |
+
# Get org color and flag
|
240 |
+
color, flag = org_info.get(org, ("#808080", ""))
|
241 |
+
|
242 |
+
# Add line plot
|
243 |
+
fig.add_trace(
|
244 |
+
go.Scatter(
|
245 |
+
x=x_values,
|
246 |
+
y=y_values,
|
247 |
+
mode="lines",
|
248 |
+
name=f"{i+1}. {org} {flag}",
|
249 |
+
line=dict(color=color, width=2),
|
250 |
+
hoverinfo="skip",
|
251 |
+
)
|
252 |
+
)
|
253 |
+
|
254 |
+
# Add scatter plot for best model points
|
255 |
+
best_models_df = pd.DataFrame(best_models)
|
256 |
+
fig.add_trace(
|
257 |
+
go.Scatter(
|
258 |
+
x=best_models_df["Release Date"],
|
259 |
+
y=best_models_df["rating"],
|
260 |
+
mode="markers",
|
261 |
+
name=org,
|
262 |
+
showlegend=False,
|
263 |
+
marker=dict(color=color, size=8, symbol="circle"),
|
264 |
+
text=best_models_df["Model"],
|
265 |
+
hovertemplate="<b>%{text}</b><br>Date: %{x}<br>ELO Score: %{y:.2f}<extra></extra>",
|
266 |
+
)
|
267 |
+
)
|
268 |
+
|
269 |
+
|
270 |
+
# Update layout
|
271 |
+
fig.update_layout(
|
272 |
+
xaxis_title="Date",
|
273 |
+
title="La course au classement",
|
274 |
+
yaxis_title="Score ELO",
|
275 |
+
legend_title="Classement en Novembre 2024",
|
276 |
+
xaxis_range=[pd.Timestamp("2024-01-01"), current_date], # Extend x-axis for labels
|
277 |
+
yaxis_range=[1103, 1350],
|
278 |
+
)
|
279 |
+
# apply_template(fig)
|
280 |
+
|
281 |
+
fig.update_xaxes(
|
282 |
+
tickformat="%m-%Y",
|
283 |
+
)
|
284 |
+
|
285 |
+
return fig, df
|
286 |
+
|
287 |
+
def filter_df():
|
288 |
+
return df
|
289 |
+
|
290 |
+
|
291 |
+
set_dark_mode = """
|
292 |
+
function refresh() {
|
293 |
+
const url = new URL(window.location);
|
294 |
+
|
295 |
+
if (url.searchParams.get('__theme') !== 'dark') {
|
296 |
+
url.searchParams.set('__theme', 'dark');
|
297 |
+
window.location.href = url.href;
|
298 |
+
}
|
299 |
+
}
|
300 |
+
"""
|
301 |
+
|
302 |
+
with gr.Blocks(
|
303 |
+
theme=gr.themes.Soft(
|
304 |
+
primary_hue=gr.themes.colors.sky,
|
305 |
+
secondary_hue=gr.themes.colors.green,
|
306 |
+
# spacing_size=gr.themes.sizes.spacing_sm,
|
307 |
+
text_size=gr.themes.sizes.text_sm,
|
308 |
+
font=[
|
309 |
+
gr.themes.GoogleFont("Open Sans"),
|
310 |
+
"ui-sans-serif",
|
311 |
+
"system-ui",
|
312 |
+
"sans-serif",
|
313 |
+
],
|
314 |
+
),
|
315 |
+
js=set_dark_mode,
|
316 |
+
) as demo:
|
317 |
+
gr.Markdown(
|
318 |
+
"""
|
319 |
+
<div style="text-align: center; max-width: 650px; margin: auto;">
|
320 |
+
<h1 style="font-weight: 900; margin-top: 5px;">🚀 The race for the best LLM 🚀</h1>
|
321 |
+
<p style="text-align: left; margin-top: 30px; margin-bottom: 30px; line-height: 20px;">
|
322 |
+
This app visualizes the progress of LLMs over time as scored by the <a href="https://leaderboard.lmsys.org/">LMSYS Chatbot Arena</a>.
|
323 |
+
The app is adapted from <a href="https://huggingface.co/spaces/andrewrreed/closed-vs-open-arena-elo"> this app</a> by Andew Reed,
|
324 |
+
and is intended to stay up-to-date as new models are released and evaluated.
|
325 |
+
<div style="text-align: left;">
|
326 |
+
<strong>Plot info:</strong>
|
327 |
+
<br>
|
328 |
+
<ul style="padding-left: 20px;">
|
329 |
+
<li> The ELO score (y-axis) is a measure of the relative strength of a model based on its performance against other models in the arena. </li>
|
330 |
+
<li> The Release Date (x-axis) corresponds to when the model was first publicly released or when its ELO results were first reported (for ease of automated updates). </li>
|
331 |
+
<li> Trend lines are based on Ordinary Least Squares (OLS) regression and adjust based on the filter criteria. </li>
|
332 |
+
<ul>
|
333 |
+
</div>
|
334 |
+
</p>
|
335 |
+
</div>
|
336 |
+
"""
|
337 |
+
)
|
338 |
+
filtered_df = gr.State()
|
339 |
+
with gr.Group():
|
340 |
+
with gr.Tab("Plot"):
|
341 |
+
plot = gr.Plot(show_label=False)
|
342 |
+
with gr.Tab("Raw Data"):
|
343 |
+
display_df = gr.DataFrame()
|
344 |
+
|
345 |
|
346 |
+
demo.load(
|
347 |
+
fn=filter_df,
|
348 |
+
inputs=[],
|
349 |
+
outputs=filtered_df,
|
350 |
+
).then(
|
351 |
+
fn=make_figure,
|
352 |
+
inputs=[filtered_df],
|
353 |
+
outputs=[plot, display_df],
|
354 |
+
)
|
355 |
|
|
|
356 |
demo.launch()
|
release_date_mapping.json
ADDED
@@ -0,0 +1,857 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
[
|
2 |
+
{
|
3 |
+
"key": "gpt-4-turbo-2024-04-09",
|
4 |
+
"Model": "GPT-4-Turbo-2024-04-09",
|
5 |
+
"Release Date": "2024-04-09"
|
6 |
+
},
|
7 |
+
{
|
8 |
+
"key": "gpt-4-1106-preview",
|
9 |
+
"Model": "GPT-4-1106-preview",
|
10 |
+
"Release Date": "2023-11-06"
|
11 |
+
},
|
12 |
+
{
|
13 |
+
"key": "claude-3-opus-20240229",
|
14 |
+
"Model": "Claude 3 Opus",
|
15 |
+
"Release Date": "2024-02-29"
|
16 |
+
},
|
17 |
+
{
|
18 |
+
"key": "gemini-1.5-pro-api-0409-preview",
|
19 |
+
"Model": "Gemini 1.5 Pro API-0409-Preview",
|
20 |
+
"Release Date": "2024-04-09"
|
21 |
+
},
|
22 |
+
{
|
23 |
+
"key": "gpt-4-0125-preview",
|
24 |
+
"Model": "GPT-4-0125-preview",
|
25 |
+
"Release Date": "2024-01-25"
|
26 |
+
},
|
27 |
+
{
|
28 |
+
"key": "bard-jan-24-gemini-pro",
|
29 |
+
"Model": "Bard (Gemini Pro)",
|
30 |
+
"Release Date": "2024-02-01"
|
31 |
+
},
|
32 |
+
{
|
33 |
+
"key": "llama-3-70b-instruct",
|
34 |
+
"Model": "Llama-3-70b-Instruct",
|
35 |
+
"Release Date": "2024-04-18"
|
36 |
+
},
|
37 |
+
{
|
38 |
+
"key": "claude-3-sonnet-20240229",
|
39 |
+
"Model": "Claude 3 Sonnet",
|
40 |
+
"Release Date": "2024-02-29"
|
41 |
+
},
|
42 |
+
{
|
43 |
+
"key": "command-r-plus",
|
44 |
+
"Model": "Command R+",
|
45 |
+
"Release Date": "2024-04-04"
|
46 |
+
},
|
47 |
+
{
|
48 |
+
"key": "gpt-4-0314",
|
49 |
+
"Model": "GPT-4-0314",
|
50 |
+
"Release Date": "2023-03-14"
|
51 |
+
},
|
52 |
+
{
|
53 |
+
"key": "claude-3-haiku-20240307",
|
54 |
+
"Model": "Claude 3 Haiku",
|
55 |
+
"Release Date": "2024-03-07"
|
56 |
+
},
|
57 |
+
{
|
58 |
+
"key": "gpt-4-0613",
|
59 |
+
"Model": "GPT-4-0613",
|
60 |
+
"Release Date": "2023-06-13"
|
61 |
+
},
|
62 |
+
{
|
63 |
+
"key": "mistral-large-2402",
|
64 |
+
"Model": "Mistral-Large-2402",
|
65 |
+
"Release Date": "2024-02-24"
|
66 |
+
},
|
67 |
+
{
|
68 |
+
"key": "qwen1.5-72b-chat",
|
69 |
+
"Model": "Qwen1.5-72B-Chat",
|
70 |
+
"Release Date": "2024-02-04"
|
71 |
+
},
|
72 |
+
{
|
73 |
+
"key": "reka-flash-21b-20240226-online",
|
74 |
+
"Model": "Reka-Flash-21B-online",
|
75 |
+
"Release Date": "2024-02-26"
|
76 |
+
},
|
77 |
+
{
|
78 |
+
"key": "claude-1",
|
79 |
+
"Model": "Claude-1",
|
80 |
+
"Release Date": "2023-03-14"
|
81 |
+
},
|
82 |
+
{
|
83 |
+
"key": "reka-flash-21b-20240226",
|
84 |
+
"Model": "Reka-Flash-21B",
|
85 |
+
"Release Date": "2024-02-26"
|
86 |
+
},
|
87 |
+
{
|
88 |
+
"key": "command-r",
|
89 |
+
"Model": "Command R",
|
90 |
+
"Release Date": "2024-03-11"
|
91 |
+
},
|
92 |
+
{
|
93 |
+
"key": "mistral-medium",
|
94 |
+
"Model": "Mistral Medium",
|
95 |
+
"Release Date": "2023-12-11"
|
96 |
+
},
|
97 |
+
{
|
98 |
+
"key": "mixtral-8x22b-instruct-v0.1",
|
99 |
+
"Model": "Mixtral-8x22b-Instruct-v0.1",
|
100 |
+
"Release Date": "2024-04-17"
|
101 |
+
},
|
102 |
+
{
|
103 |
+
"key": "llama-3-8b-instruct",
|
104 |
+
"Model": "Llama-3-8b-Instruct",
|
105 |
+
"Release Date": "2024-04-18"
|
106 |
+
},
|
107 |
+
{
|
108 |
+
"key": "gemini-pro-dev-api",
|
109 |
+
"Model": "Gemini Pro (Dev API)",
|
110 |
+
"Release Date": "2023-12-13"
|
111 |
+
},
|
112 |
+
{
|
113 |
+
"key": "qwen1.5-32b-chat",
|
114 |
+
"Model": "Qwen1.5-32B-Chat",
|
115 |
+
"Release Date": "2024-02-04"
|
116 |
+
},
|
117 |
+
{
|
118 |
+
"key": "claude-2.0",
|
119 |
+
"Model": "Claude-2.0",
|
120 |
+
"Release Date": "2023-07-11"
|
121 |
+
},
|
122 |
+
{
|
123 |
+
"key": "mistral-next",
|
124 |
+
"Model": "Mistral-Next",
|
125 |
+
"Release Date": "2024-02-17"
|
126 |
+
},
|
127 |
+
{
|
128 |
+
"key": "zephyr-orpo-141b-A35b-v0.1",
|
129 |
+
"Model": "Zephyr-ORPO-141b-A35b-v0.1",
|
130 |
+
"Release Date": "2024-04-12"
|
131 |
+
},
|
132 |
+
{
|
133 |
+
"key": "gpt-3.5-turbo-0613",
|
134 |
+
"Model": "GPT-3.5-Turbo-0613",
|
135 |
+
"Release Date": "2023-06-13"
|
136 |
+
},
|
137 |
+
{
|
138 |
+
"key": "claude-2.1",
|
139 |
+
"Model": "Claude-2.1",
|
140 |
+
"Release Date": "2023-11-21"
|
141 |
+
},
|
142 |
+
{
|
143 |
+
"key": "qwen1.5-14b-chat",
|
144 |
+
"Model": "Qwen1.5-14B-Chat",
|
145 |
+
"Release Date": "2024-02-04"
|
146 |
+
},
|
147 |
+
{
|
148 |
+
"key": "starling-lm-7b-beta",
|
149 |
+
"Model": "Starling-LM-7B-beta",
|
150 |
+
"Release Date": "2024-03-20"
|
151 |
+
},
|
152 |
+
{
|
153 |
+
"key": "gemini-pro",
|
154 |
+
"Model": "Gemini Pro",
|
155 |
+
"Release Date": "2023-12-13"
|
156 |
+
},
|
157 |
+
{
|
158 |
+
"key": "mixtral-8x7b-instruct-v0.1",
|
159 |
+
"Model": "Mixtral-8x7b-Instruct-v0.1",
|
160 |
+
"Release Date": "2023-12-11"
|
161 |
+
},
|
162 |
+
{
|
163 |
+
"key": "claude-instant-1",
|
164 |
+
"Model": "Claude-Instant-1",
|
165 |
+
"Release Date": "2023-03-14"
|
166 |
+
},
|
167 |
+
{
|
168 |
+
"key": "yi-34b-chat",
|
169 |
+
"Model": "Yi-34B-Chat",
|
170 |
+
"Release Date": "2024-01-23"
|
171 |
+
},
|
172 |
+
{
|
173 |
+
"key": "gpt-3.5-turbo-0314",
|
174 |
+
"Model": "GPT-3.5-Turbo-0314",
|
175 |
+
"Release Date": "2023-03-14"
|
176 |
+
},
|
177 |
+
{
|
178 |
+
"key": "wizardlm-70b",
|
179 |
+
"Model": "WizardLM-70B-v1.0",
|
180 |
+
"Release Date": "2023-08-09"
|
181 |
+
},
|
182 |
+
{
|
183 |
+
"key": "gpt-3.5-turbo-0125",
|
184 |
+
"Model": "GPT-3.5-Turbo-0125",
|
185 |
+
"Release Date": "2024-01-25"
|
186 |
+
},
|
187 |
+
{
|
188 |
+
"key": "tulu-2-dpo-70b",
|
189 |
+
"Model": "Tulu-2-DPO-70B",
|
190 |
+
"Release Date": "2023-11-12"
|
191 |
+
},
|
192 |
+
{
|
193 |
+
"key": "dbrx-instruct-preview",
|
194 |
+
"Model": "DBRX-Instruct-Preview",
|
195 |
+
"Release Date": "2024-03-27"
|
196 |
+
},
|
197 |
+
{
|
198 |
+
"key": "openchat-3.5-0106",
|
199 |
+
"Model": "OpenChat-3.5-0106",
|
200 |
+
"Release Date": "2024-01-06"
|
201 |
+
},
|
202 |
+
{
|
203 |
+
"key": "vicuna-33b",
|
204 |
+
"Model": "Vicuna-33B",
|
205 |
+
"Release Date": "2023-06-21"
|
206 |
+
},
|
207 |
+
{
|
208 |
+
"key": "starling-lm-7b-alpha",
|
209 |
+
"Model": "Starling-LM-7B-alpha",
|
210 |
+
"Release Date": "2023-11-25"
|
211 |
+
},
|
212 |
+
{
|
213 |
+
"key": "llama-2-70b-chat",
|
214 |
+
"Model": "Llama-2-70b-chat",
|
215 |
+
"Release Date": "2023-07-18"
|
216 |
+
},
|
217 |
+
{
|
218 |
+
"key": "nous-hermes-2-mixtral-8x7b-dpo",
|
219 |
+
"Model": "Nous-Hermes-2-Mixtral-8x7B-DPO",
|
220 |
+
"Release Date": "2024-01-13"
|
221 |
+
},
|
222 |
+
{
|
223 |
+
"key": "gemma-1.1-7b-it",
|
224 |
+
"Model": "Gemma-1.1-7B-it",
|
225 |
+
"Release Date": "2024-04-09"
|
226 |
+
},
|
227 |
+
{
|
228 |
+
"key": "llama2-70b-steerlm-chat",
|
229 |
+
"Model": "NV-Llama2-70B-SteerLM-Chat",
|
230 |
+
"Release Date": "2023-11-24"
|
231 |
+
},
|
232 |
+
{
|
233 |
+
"key": "deepseek-llm-67b-chat",
|
234 |
+
"Model": "DeepSeek-LLM-67B-Chat",
|
235 |
+
"Release Date": "2023-11-29"
|
236 |
+
},
|
237 |
+
{
|
238 |
+
"key": "openhermes-2.5-mistral-7b",
|
239 |
+
"Model": "OpenHermes-2.5-Mistral-7b",
|
240 |
+
"Release Date": "2023-10-29"
|
241 |
+
},
|
242 |
+
{
|
243 |
+
"key": "openchat-3.5",
|
244 |
+
"Model": "OpenChat-3.5",
|
245 |
+
"Release Date": "2023-11-16"
|
246 |
+
},
|
247 |
+
{
|
248 |
+
"key": "pplx-70b-online",
|
249 |
+
"Model": "pplx-70b-online",
|
250 |
+
"Release Date": "2023-11-29"
|
251 |
+
},
|
252 |
+
{
|
253 |
+
"key": "mistral-7b-instruct-v0.2",
|
254 |
+
"Model": "Mistral-7B-Instruct-v0.2",
|
255 |
+
"Release Date": "2023-12-11"
|
256 |
+
},
|
257 |
+
{
|
258 |
+
"key": "qwen1.5-7b-chat",
|
259 |
+
"Model": "Qwen1.5-7B-Chat",
|
260 |
+
"Release Date": "2024-02-04"
|
261 |
+
},
|
262 |
+
{
|
263 |
+
"key": "gpt-3.5-turbo-1106",
|
264 |
+
"Model": "GPT-3.5-Turbo-1106",
|
265 |
+
"Release Date": "2023-11-06"
|
266 |
+
},
|
267 |
+
{
|
268 |
+
"key": "dolphin-2.2.1-mistral-7b",
|
269 |
+
"Model": "Dolphin-2.2.1-Mistral-7B",
|
270 |
+
"Release Date": "2023-10-30"
|
271 |
+
},
|
272 |
+
{
|
273 |
+
"key": "solar-10.7b-instruct-v1.0",
|
274 |
+
"Model": "SOLAR-10.7B-Instruct-v1.0",
|
275 |
+
"Release Date": "2023-12-13"
|
276 |
+
},
|
277 |
+
{
|
278 |
+
"key": "phi-3-mini-128k-instruct",
|
279 |
+
"Model": "Phi-3-Mini-128k-Instruct",
|
280 |
+
"Release Date": "2024-04-23"
|
281 |
+
},
|
282 |
+
{
|
283 |
+
"key": "wizardlm-13b",
|
284 |
+
"Model": "WizardLM-13b-v1.2",
|
285 |
+
"Release Date": "2023-07-25"
|
286 |
+
},
|
287 |
+
{
|
288 |
+
"key": "llama-2-13b-chat",
|
289 |
+
"Model": "Llama-2-13b-chat",
|
290 |
+
"Release Date": "2023-07-18"
|
291 |
+
},
|
292 |
+
{
|
293 |
+
"key": "zephyr-7b-beta",
|
294 |
+
"Model": "Zephyr-7b-beta",
|
295 |
+
"Release Date": "2023-10-26"
|
296 |
+
},
|
297 |
+
{
|
298 |
+
"key": "codellama-70b-instruct",
|
299 |
+
"Model": "CodeLlama-70B-instruct",
|
300 |
+
"Release Date": "2024-01-29"
|
301 |
+
},
|
302 |
+
{
|
303 |
+
"key": "mpt-30b-chat",
|
304 |
+
"Model": "MPT-30B-chat",
|
305 |
+
"Release Date": "2023-06-09"
|
306 |
+
},
|
307 |
+
{
|
308 |
+
"key": "vicuna-13b",
|
309 |
+
"Model": "Vicuna-13B",
|
310 |
+
"Release Date": "2023-07-23"
|
311 |
+
},
|
312 |
+
{
|
313 |
+
"key": "codellama-34b-instruct",
|
314 |
+
"Model": "CodeLlama-34B-instruct",
|
315 |
+
"Release Date": "2023-08-24"
|
316 |
+
},
|
317 |
+
{
|
318 |
+
"key": "gemma-7b-it",
|
319 |
+
"Model": "Gemma-7B-it",
|
320 |
+
"Release Date": "2024-02-21"
|
321 |
+
},
|
322 |
+
{
|
323 |
+
"key": "pplx-7b-online",
|
324 |
+
"Model": "pplx-7b-online",
|
325 |
+
"Release Date": "2023-11-29"
|
326 |
+
},
|
327 |
+
{
|
328 |
+
"key": "zephyr-7b-alpha",
|
329 |
+
"Model": "Zephyr-7b-alpha",
|
330 |
+
"Release Date": "2023-10-09"
|
331 |
+
},
|
332 |
+
{
|
333 |
+
"key": "llama-2-7b-chat",
|
334 |
+
"Model": "Llama-2-7b-chat",
|
335 |
+
"Release Date": "2023-07-18"
|
336 |
+
},
|
337 |
+
{
|
338 |
+
"key": "qwen-14b-chat",
|
339 |
+
"Model": "Qwen-14B-Chat",
|
340 |
+
"Release Date": "2023-09-24"
|
341 |
+
},
|
342 |
+
{
|
343 |
+
"key": "falcon-180b-chat",
|
344 |
+
"Model": "falcon-180b-chat",
|
345 |
+
"Release Date": "2023-09-05"
|
346 |
+
},
|
347 |
+
{
|
348 |
+
"key": "guanaco-33b",
|
349 |
+
"Model": "Guanaco-33B",
|
350 |
+
"Release Date": "2023-05-22"
|
351 |
+
},
|
352 |
+
{
|
353 |
+
"key": "stripedhyena-nous-7b",
|
354 |
+
"Model": "StripedHyena-Nous-7B",
|
355 |
+
"Release Date": "2023-12-07"
|
356 |
+
},
|
357 |
+
{
|
358 |
+
"key": "olmo-7b-instruct",
|
359 |
+
"Model": "OLMo-7B-instruct",
|
360 |
+
"Release Date": "2024-02-23"
|
361 |
+
},
|
362 |
+
{
|
363 |
+
"key": "gemma-1.1-2b-it",
|
364 |
+
"Model": "Gemma-1.1-2B-it",
|
365 |
+
"Release Date": "2024-04-09"
|
366 |
+
},
|
367 |
+
{
|
368 |
+
"key": "mistral-7b-instruct",
|
369 |
+
"Model": "Mistral-7B-Instruct-v0.1",
|
370 |
+
"Release Date": "2023-09-27"
|
371 |
+
},
|
372 |
+
{
|
373 |
+
"key": "palm-2",
|
374 |
+
"Model": "PaLM-Chat-Bison-001",
|
375 |
+
"Release Date": "2023-07-10"
|
376 |
+
},
|
377 |
+
{
|
378 |
+
"key": "vicuna-7b",
|
379 |
+
"Model": "Vicuna-7B",
|
380 |
+
"Release Date": "2023-07-29"
|
381 |
+
},
|
382 |
+
{
|
383 |
+
"key": "qwen1.5-4b-chat",
|
384 |
+
"Model": "Qwen1.5-4B-Chat",
|
385 |
+
"Release Date": "2024-02-04"
|
386 |
+
},
|
387 |
+
{
|
388 |
+
"key": "gemma-2b-it",
|
389 |
+
"Model": "Gemma-2B-it",
|
390 |
+
"Release Date": "2024-02-21"
|
391 |
+
},
|
392 |
+
{
|
393 |
+
"key": "koala-13b",
|
394 |
+
"Model": "Koala-13B",
|
395 |
+
"Release Date": "2023-04-03"
|
396 |
+
},
|
397 |
+
{
|
398 |
+
"key": "chatglm3-6b",
|
399 |
+
"Model": "ChatGLM3-6B",
|
400 |
+
"Release Date": "2023-10-25"
|
401 |
+
},
|
402 |
+
{
|
403 |
+
"key": "gpt4all-13b-snoozy",
|
404 |
+
"Model": "GPT4All-13B-Snoozy",
|
405 |
+
"Release Date": "2023-04-24"
|
406 |
+
},
|
407 |
+
{
|
408 |
+
"key": "chatglm2-6b",
|
409 |
+
"Model": "ChatGLM2-6B",
|
410 |
+
"Release Date": "2023-06-25"
|
411 |
+
},
|
412 |
+
{
|
413 |
+
"key": "mpt-7b-chat",
|
414 |
+
"Model": "MPT-7B-Chat",
|
415 |
+
"Release Date": "2023-05-04"
|
416 |
+
},
|
417 |
+
{
|
418 |
+
"key": "RWKV-4-Raven-14B",
|
419 |
+
"Model": "RWKV-4-Raven-14B",
|
420 |
+
"Release Date": "2023-05-22"
|
421 |
+
},
|
422 |
+
{
|
423 |
+
"key": "alpaca-13b",
|
424 |
+
"Model": "Alpaca-13B",
|
425 |
+
"Release Date": "2023-03-13"
|
426 |
+
},
|
427 |
+
{
|
428 |
+
"key": "oasst-pythia-12b",
|
429 |
+
"Model": "OpenAssistant-Pythia-12B",
|
430 |
+
"Release Date": "2023-04-03"
|
431 |
+
},
|
432 |
+
{
|
433 |
+
"key": "chatglm-6b",
|
434 |
+
"Model": "ChatGLM-6B",
|
435 |
+
"Release Date": "2023-03-13"
|
436 |
+
},
|
437 |
+
{
|
438 |
+
"key": "fastchat-t5-3b",
|
439 |
+
"Model": "FastChat-T5-3B",
|
440 |
+
"Release Date": "2023-04-27"
|
441 |
+
},
|
442 |
+
{
|
443 |
+
"key": "stablelm-tuned-alpha-7b",
|
444 |
+
"Model": "StableLM-Tuned-Alpha-7B",
|
445 |
+
"Release Date": "2023-04-19"
|
446 |
+
},
|
447 |
+
{
|
448 |
+
"key": "dolly-v2-12b",
|
449 |
+
"Model": "Dolly-V2-12B",
|
450 |
+
"Release Date": "2023-04-12"
|
451 |
+
},
|
452 |
+
{
|
453 |
+
"key": "llama-13b",
|
454 |
+
"Model": "LLaMA-13B",
|
455 |
+
"Release Date": "2023-02-27"
|
456 |
+
},
|
457 |
+
{
|
458 |
+
"key": "snowflake-arctic-instruct",
|
459 |
+
"Model": "Snowflake Arctic Instruct",
|
460 |
+
"Release Date": "2024-04-24"
|
461 |
+
},
|
462 |
+
{
|
463 |
+
"key": "gpt-4o-2024-05-13",
|
464 |
+
"Model": "GPT-4o-2024-05-13",
|
465 |
+
"Release Date": "2024-05-16"
|
466 |
+
},
|
467 |
+
{
|
468 |
+
"key": "qwen-max-0428",
|
469 |
+
"Model": "Qwen-Max-0428",
|
470 |
+
"Release Date": "2024-05-16"
|
471 |
+
},
|
472 |
+
{
|
473 |
+
"key": "qwen1.5-110b-chat",
|
474 |
+
"Model": "Qwen1.5-110B-Chat",
|
475 |
+
"Release Date": "2024-05-16"
|
476 |
+
},
|
477 |
+
{
|
478 |
+
"key": "reka-core-20240501",
|
479 |
+
"Model": "Reka-Core-20240501",
|
480 |
+
"Release Date": "2024-05-16"
|
481 |
+
},
|
482 |
+
{
|
483 |
+
"key": "glm-4-0116",
|
484 |
+
"Model": "GLM-4-0116",
|
485 |
+
"Release Date": "2024-05-23"
|
486 |
+
},
|
487 |
+
{
|
488 |
+
"key": "phi-3-mini-4k-instruct",
|
489 |
+
"Model": "Phi-3-Mini-4k-Instruct",
|
490 |
+
"Release Date": "2024-05-23"
|
491 |
+
},
|
492 |
+
{
|
493 |
+
"key": "yi-large-preview",
|
494 |
+
"Model": "Yi-Large-preview",
|
495 |
+
"Release Date": "2024-05-23"
|
496 |
+
},
|
497 |
+
{
|
498 |
+
"key": "claude-3-5-sonnet-20240620",
|
499 |
+
"Model": "Claude 3.5 Sonnet",
|
500 |
+
"Release Date": "2024-06-20"
|
501 |
+
},
|
502 |
+
{
|
503 |
+
"key": "deepseek-coder-v2",
|
504 |
+
"Model": "DeepSeek-Coder-V2-Instruct",
|
505 |
+
"Release Date": "2024-06-17"
|
506 |
+
},
|
507 |
+
{
|
508 |
+
"key": "gemini-1.5-flash-api-0514",
|
509 |
+
"Model": "Gemini-1.5-Flash-API-0514",
|
510 |
+
"Release Date": "2024-05-24"
|
511 |
+
},
|
512 |
+
{
|
513 |
+
"key": "gemini-1.5-pro-api-0514",
|
514 |
+
"Model": "Gemini-1.5-Pro-API-0514",
|
515 |
+
"Release Date": "2024-05-24"
|
516 |
+
},
|
517 |
+
{
|
518 |
+
"key": "gemini-advanced-0514",
|
519 |
+
"Model": "Gemini-Advanced-0514",
|
520 |
+
"Release Date": "2024-05-24"
|
521 |
+
},
|
522 |
+
{
|
523 |
+
"key": "gemma-2-27b-it",
|
524 |
+
"Model": "Gemma-2-27B-it",
|
525 |
+
"Release Date": "2024-07-01"
|
526 |
+
},
|
527 |
+
{
|
528 |
+
"key": "gemma-2-9b-it",
|
529 |
+
"Model": "Gemma-2-9B-it",
|
530 |
+
"Release Date": "2024-06-27"
|
531 |
+
},
|
532 |
+
{
|
533 |
+
"key": "glm-4-0520",
|
534 |
+
"Model": "GLM-4-0520",
|
535 |
+
"Release Date": "2024-05-20"
|
536 |
+
},
|
537 |
+
{
|
538 |
+
"key": "nemotron-4-340b-instruct",
|
539 |
+
"Model": "Nemotron-4-340B-Instruct",
|
540 |
+
"Release Date": "2024-06-14"
|
541 |
+
},
|
542 |
+
{
|
543 |
+
"key": "phi-3-medium-4k-instruct",
|
544 |
+
"Model": "Phi-3-Medium-4k-Instruct",
|
545 |
+
"Release Date": "2024-05-21"
|
546 |
+
},
|
547 |
+
{
|
548 |
+
"key": "phi-3-small-8k-instruct",
|
549 |
+
"Model": "Phi-3-Small-8k-Instruct",
|
550 |
+
"Release Date": "2024-05-21"
|
551 |
+
},
|
552 |
+
{
|
553 |
+
"key": "qwen2-72b-instruct",
|
554 |
+
"Model": "Qwen2-72B-Instruct",
|
555 |
+
"Release Date": "2024-06-06"
|
556 |
+
},
|
557 |
+
{
|
558 |
+
"key": "reka-flash-preview-20240611",
|
559 |
+
"Model": "Reka-Flash-Preview-20240611",
|
560 |
+
"Release Date": "2024-06-11"
|
561 |
+
},
|
562 |
+
{
|
563 |
+
"key": "yi-1.5-34b-chat",
|
564 |
+
"Model": "Yi-1.5-34B-Chat",
|
565 |
+
"Release Date": "2024-05-13"
|
566 |
+
},
|
567 |
+
{
|
568 |
+
"key": "yi-large",
|
569 |
+
"Model": "Yi-Large",
|
570 |
+
"Release Date": "2024-05-13"
|
571 |
+
},
|
572 |
+
{
|
573 |
+
"key": "phi-3-mini-4k-instruct-june-2024",
|
574 |
+
"Model": "Phi-3-Mini-4k-Instruct-June-24",
|
575 |
+
"Release Date": "2024-06-24"
|
576 |
+
},
|
577 |
+
{
|
578 |
+
"key": "athene-70b-0725",
|
579 |
+
"Model": "athene-70b-0725",
|
580 |
+
"Release Date": "2024-07-25"
|
581 |
+
},
|
582 |
+
{
|
583 |
+
"key": "athene-70b-0725",
|
584 |
+
"Model": "athene-70b-0725",
|
585 |
+
"Release Date": "2024-07-25"
|
586 |
+
},
|
587 |
+
{
|
588 |
+
"key": "deepseek-coder-v2-0724",
|
589 |
+
"Model": "Deepseek-Coder-v2-0724",
|
590 |
+
"Release Date": "2024-07-24"
|
591 |
+
},
|
592 |
+
{
|
593 |
+
"key": "deepseek-v2-api-0628",
|
594 |
+
"Model": "Deepseek-v2-API-0628",
|
595 |
+
"Release Date": "2024-06-28"
|
596 |
+
},
|
597 |
+
{
|
598 |
+
"key": "gemini-1.5-pro-exp-0801",
|
599 |
+
"Model": "Gemini-1.5-Pro-Exp-0801",
|
600 |
+
"Release Date": "2024-08-01"
|
601 |
+
},
|
602 |
+
{
|
603 |
+
"key": "gemma-2-2b-it",
|
604 |
+
"Model": "Gemma-2-2b-it",
|
605 |
+
"Release Date": "2024-07-31"
|
606 |
+
},
|
607 |
+
{
|
608 |
+
"key": "gpt-4o-mini-2024-07-18",
|
609 |
+
"Model": "GPT-4o-mini-2024-07-18",
|
610 |
+
"Release Date": "2024-07-18"
|
611 |
+
},
|
612 |
+
{
|
613 |
+
"key": "llama-3.1-405b-instruct",
|
614 |
+
"Model": "Meta-Llama-3.1-405b-Instruct",
|
615 |
+
"Release Date": "2024-07-23"
|
616 |
+
},
|
617 |
+
{
|
618 |
+
"key": "llama-3.1-70b-instruct",
|
619 |
+
"Model": "Meta-Llama-3.1-70b-Instruct",
|
620 |
+
"Release Date": "2024-07-23"
|
621 |
+
},
|
622 |
+
{
|
623 |
+
"key": "llama-3.1-8b-instruct",
|
624 |
+
"Model": "Meta-Llama-3.1-8b-Instruct",
|
625 |
+
"Release Date": "2024-07-23"
|
626 |
+
},
|
627 |
+
{
|
628 |
+
"key": "mistral-large-2407",
|
629 |
+
"Model": "Mistral-Large-2407",
|
630 |
+
"Release Date": "2024-07-24"
|
631 |
+
},
|
632 |
+
{
|
633 |
+
"key": "reka-core-20240722",
|
634 |
+
"Model": "Reka-Core-20240722",
|
635 |
+
"Release Date": "2024-07-22"
|
636 |
+
},
|
637 |
+
{
|
638 |
+
"key": "reka-flash-20240722",
|
639 |
+
"Model": "Reka-Flash-20240722",
|
640 |
+
"Release Date": "2024-07-22"
|
641 |
+
},
|
642 |
+
{
|
643 |
+
"key": "chatgpt-4o-latest-20240808",
|
644 |
+
"Model": "ChatGPT-4o-latest (2024-08-08)",
|
645 |
+
"Release Date": "2024-08-08"
|
646 |
+
},
|
647 |
+
{
|
648 |
+
"key": "chatgpt-4o-latest-20240903",
|
649 |
+
"Model": "ChatGPT-4o-latest (2024-09-03)",
|
650 |
+
"Release Date": "2024-09-03"
|
651 |
+
},
|
652 |
+
{
|
653 |
+
"key": "command-r-08-2024",
|
654 |
+
"Model": "Command R (08-2024)",
|
655 |
+
"Release Date": "2024-08-08"
|
656 |
+
},
|
657 |
+
{
|
658 |
+
"key": "command-r-plus-08-2024",
|
659 |
+
"Model": "Command R+ (08-2024)",
|
660 |
+
"Release Date": "2024-08-08"
|
661 |
+
},
|
662 |
+
{
|
663 |
+
"key": "deepseek-v2.5",
|
664 |
+
"Model": "Deepseek-v2.5",
|
665 |
+
"Release Date": "2024-09-05"
|
666 |
+
},
|
667 |
+
{
|
668 |
+
"key": "gemini-1.5-flash-8b-exp-0827",
|
669 |
+
"Model": "Gemini-1.5-Flash-8b-Exp-0827",
|
670 |
+
"Release Date": "2024-08-27"
|
671 |
+
},
|
672 |
+
{
|
673 |
+
"key": "gemini-1.5-flash-exp-0827",
|
674 |
+
"Model": "Gemini-1.5-Flash-Exp-0827",
|
675 |
+
"Release Date": "2024-08-27"
|
676 |
+
},
|
677 |
+
{
|
678 |
+
"key": "gemini-1.5-pro-exp-0827",
|
679 |
+
"Model": "Gemini-1.5-Pro-Exp-0827",
|
680 |
+
"Release Date": "2024-08-27"
|
681 |
+
},
|
682 |
+
{
|
683 |
+
"key": "gemma-2-9b-it-simpo",
|
684 |
+
"Model": "Gemma-2-9b-it-SimPO",
|
685 |
+
"Release Date": "2024-09-07"
|
686 |
+
},
|
687 |
+
{
|
688 |
+
"key": "gpt-4o-2024-08-06",
|
689 |
+
"Model": "GPT-4o-2024-08-06",
|
690 |
+
"Release Date": "2024-08-06"
|
691 |
+
},
|
692 |
+
{
|
693 |
+
"key": "grok-2-2024-08-13",
|
694 |
+
"Model": "Grok-2-08-13",
|
695 |
+
"Release Date": "2024-08-13"
|
696 |
+
},
|
697 |
+
{
|
698 |
+
"key": "grok-2-mini-2024-08-13",
|
699 |
+
"Model": "Grok-2-Mini-08-13",
|
700 |
+
"Release Date": "2024-08-13"
|
701 |
+
},
|
702 |
+
{
|
703 |
+
"key": "internlm2_5-20b-chat",
|
704 |
+
"Model": "InternLM2.5-20b-chat",
|
705 |
+
"Release Date": "2024-07-30"
|
706 |
+
},
|
707 |
+
{
|
708 |
+
"key": "jamba-1.5-large",
|
709 |
+
"Model": "Jamba-1.5-Large",
|
710 |
+
"Release Date": "2024-08-22"
|
711 |
+
},
|
712 |
+
{
|
713 |
+
"key": "jamba-1.5-mini",
|
714 |
+
"Model": "Jamba-1.5-Mini",
|
715 |
+
"Release Date": "2024-08-22"
|
716 |
+
},
|
717 |
+
{
|
718 |
+
"key": "llama-3.1-405b-instruct-bf16",
|
719 |
+
"Model": "Meta-Llama-3.1-405b-Instruct-bf16",
|
720 |
+
"Release Date": "2024-07-23"
|
721 |
+
},
|
722 |
+
{
|
723 |
+
"key": "llama-3.1-405b-instruct-fp8",
|
724 |
+
"Model": "Meta-Llama-3.1-405b-Instruct-fp8",
|
725 |
+
"Release Date": "2024-07-23"
|
726 |
+
},
|
727 |
+
{
|
728 |
+
"key": "llama-3.2-1b-instruct",
|
729 |
+
"Model": "Meta-Llama-3.2-1b-Instruct",
|
730 |
+
"Release Date": "2024-09-25"
|
731 |
+
},
|
732 |
+
{
|
733 |
+
"key": "llama-3.2-3b-instruct",
|
734 |
+
"Model": "Meta-Llama-3.2-3b-Instruct",
|
735 |
+
"Release Date": "2024-09-25"
|
736 |
+
},
|
737 |
+
{
|
738 |
+
"key": "o1-mini",
|
739 |
+
"Model": "o1-mini",
|
740 |
+
"Release Date": "2024-09-12"
|
741 |
+
},
|
742 |
+
{
|
743 |
+
"key": "o1-preview",
|
744 |
+
"Model": "o1-preview",
|
745 |
+
"Release Date": "2024-09-12"
|
746 |
+
},
|
747 |
+
{
|
748 |
+
"key": "qwen-plus-0828",
|
749 |
+
"Model": "Qwen-Plus-0828",
|
750 |
+
"Release Date": "2024-08-28"
|
751 |
+
},
|
752 |
+
{
|
753 |
+
"key": "qwen2.5-72b-instruct",
|
754 |
+
"Model": "Qwen2.5-72b-Instruct",
|
755 |
+
"Release Date": "2024-09-19"
|
756 |
+
},
|
757 |
+
{
|
758 |
+
"key": "chatgpt-4o-latest-20241120",
|
759 |
+
"Model": "ChatGPT-4o-latest (2024-11-20)",
|
760 |
+
"Release Date": "2024-11-20"
|
761 |
+
},
|
762 |
+
{
|
763 |
+
"key": "claude-3-5-sonnet-20241022",
|
764 |
+
"Model": "Claude 3.5 Sonnet (20241022)",
|
765 |
+
"Release Date": "2024-10-22"
|
766 |
+
},
|
767 |
+
{
|
768 |
+
"key": "gemini-1.5-flash-001",
|
769 |
+
"Model": "Gemini-1.5-Flash-001",
|
770 |
+
"Release Date": "2024-05-24"
|
771 |
+
},
|
772 |
+
{
|
773 |
+
"key": "gemini-1.5-flash-002",
|
774 |
+
"Model": "Gemini-1.5-Flash-002",
|
775 |
+
"Release Date": "2024-09-24"
|
776 |
+
},
|
777 |
+
{
|
778 |
+
"key": "gemini-1.5-flash-8b-001",
|
779 |
+
"Model": "Gemini-1.5-Flash-8B-001",
|
780 |
+
"Release Date": "2024-10-03"
|
781 |
+
},
|
782 |
+
{
|
783 |
+
"key": "gemini-1.5-pro-001",
|
784 |
+
"Model": "Gemini-1.5-Pro-001",
|
785 |
+
"Release Date": "2024-05-24"
|
786 |
+
},
|
787 |
+
{
|
788 |
+
"key": "gemini-1.5-pro-002",
|
789 |
+
"Model": "Gemini-1.5-Pro-002",
|
790 |
+
"Release Date": "2024-09-24"
|
791 |
+
},
|
792 |
+
{
|
793 |
+
"key": "gemini-exp-1114",
|
794 |
+
"Model": "Gemini-Exp-1114",
|
795 |
+
"Release Date": "2024-11-14"
|
796 |
+
},
|
797 |
+
{
|
798 |
+
"key": "gemini-exp-1121",
|
799 |
+
"Model": "Gemini-Exp-1121",
|
800 |
+
"Release Date": "2024-11-21"
|
801 |
+
},
|
802 |
+
{
|
803 |
+
"key": "glm-4-plus",
|
804 |
+
"Model": "GLM-4-Plus",
|
805 |
+
"Release Date": "2024-08-30"
|
806 |
+
},
|
807 |
+
{
|
808 |
+
"key": "granite-3.0-2b-instruct",
|
809 |
+
"Model": "Granite-3.0-2B-Instruct",
|
810 |
+
"Release Date": "2024-10-21"
|
811 |
+
},
|
812 |
+
{
|
813 |
+
"key": "granite-3.0-8b-instruct",
|
814 |
+
"Model": "Granite-3.0-8B-Instruct",
|
815 |
+
"Release Date": "2024-10-21"
|
816 |
+
},
|
817 |
+
{
|
818 |
+
"key": "hunyuan-standard-256k",
|
819 |
+
"Model": "Hunyuan-Standard-256K",
|
820 |
+
"Release Date": "2024-05-28"
|
821 |
+
},
|
822 |
+
{
|
823 |
+
"key": "llama-3.1-nemotron-51b-instruct",
|
824 |
+
"Model": "Llama-3.1-Nemotron-51B-Instruct",
|
825 |
+
"Release Date": "2024-09-23"
|
826 |
+
},
|
827 |
+
{
|
828 |
+
"key": "llama-3.1-nemotron-70b-instruct",
|
829 |
+
"Model": "Llama-3.1-Nemotron-70B-Instruct",
|
830 |
+
"Release Date": "2024-10-11"
|
831 |
+
},
|
832 |
+
{
|
833 |
+
"key": "ministral-8b-2410",
|
834 |
+
"Model": "Ministral-8B-2410",
|
835 |
+
"Release Date": "2024-10-24"
|
836 |
+
},
|
837 |
+
{
|
838 |
+
"key": "qwen-max-0919",
|
839 |
+
"Model": "Qwen-Max-0919",
|
840 |
+
"Release Date": "2024-09-19"
|
841 |
+
},
|
842 |
+
{
|
843 |
+
"key": "qwen2.5-coder-32b-instruct",
|
844 |
+
"Model": "Qwen2.5-Coder-32B-Instruct",
|
845 |
+
"Release Date": "2024-11-05"
|
846 |
+
},
|
847 |
+
{
|
848 |
+
"key": "yi-lightning",
|
849 |
+
"Model": "Yi-Lightning",
|
850 |
+
"Release Date": "2024-10-16"
|
851 |
+
},
|
852 |
+
{
|
853 |
+
"key": "yi-lightning-lite",
|
854 |
+
"Model": "Yi-Lightning-lite",
|
855 |
+
"Release Date": "2024-10-16"
|
856 |
+
}
|
857 |
+
]
|
requirements.txt
ADDED
@@ -0,0 +1,7 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
huggingface_hub
|
2 |
+
pandas
|
3 |
+
numpy
|
4 |
+
plotly
|
5 |
+
gradio
|
6 |
+
statsmodels
|
7 |
+
apscheduler
|
utils.py
ADDED
@@ -0,0 +1,234 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import json
|
2 |
+
from datetime import datetime
|
3 |
+
|
4 |
+
from typing import Literal, List
|
5 |
+
|
6 |
+
import pandas as pd
|
7 |
+
import plotly.express as px
|
8 |
+
from huggingface_hub import HfFileSystem, hf_hub_download
|
9 |
+
|
10 |
+
# from: https://github.com/lm-sys/FastChat/blob/main/fastchat/serve/monitor/monitor.py#L389
|
11 |
+
KEY_TO_CATEGORY_NAME = {
|
12 |
+
"full": "Overall",
|
13 |
+
"dedup": "De-duplicate Top Redundant Queries (soon to be default)",
|
14 |
+
"math": "Math",
|
15 |
+
"if": "Instruction Following",
|
16 |
+
"multiturn": "Multi-Turn",
|
17 |
+
"coding": "Coding",
|
18 |
+
"hard_6": "Hard Prompts (Overall)",
|
19 |
+
"hard_english_6": "Hard Prompts (English)",
|
20 |
+
"long_user": "Longer Query",
|
21 |
+
"english": "English",
|
22 |
+
"chinese": "Chinese",
|
23 |
+
"french": "French",
|
24 |
+
"german": "German",
|
25 |
+
"spanish": "Spanish",
|
26 |
+
"russian": "Russian",
|
27 |
+
"japanese": "Japanese",
|
28 |
+
"korean": "Korean",
|
29 |
+
"no_tie": "Exclude Ties",
|
30 |
+
"no_short": "Exclude Short Query (< 5 tokens)",
|
31 |
+
"no_refusal": "Exclude Refusal",
|
32 |
+
"overall_limit_5_user_vote": "overall_limit_5_user_vote",
|
33 |
+
"full_old": "Overall (Deprecated)",
|
34 |
+
}
|
35 |
+
|
36 |
+
CAT_NAME_TO_EXPLANATION = {
|
37 |
+
"Overall": "Overall Questions",
|
38 |
+
"De-duplicate Top Redundant Queries (soon to be default)": "De-duplicate top redundant queries (top 0.1%). See details in [blog post](https://lmsys.org/blog/2024-05-17-category-hard/#note-enhancing-quality-through-de-duplication).",
|
39 |
+
"Math": "Math",
|
40 |
+
"Instruction Following": "Instruction Following",
|
41 |
+
"Multi-Turn": "Multi-Turn Conversation (>= 2 turns)",
|
42 |
+
"Coding": "Coding: whether conversation contains code snippets",
|
43 |
+
"Hard Prompts (Overall)": "Hard Prompts (Overall): details in [blog post](https://lmsys.org/blog/2024-05-17-category-hard/)",
|
44 |
+
"Hard Prompts (English)": "Hard Prompts (English), note: the delta is to English Category. details in [blog post](https://lmsys.org/blog/2024-05-17-category-hard/)",
|
45 |
+
"Longer Query": "Longer Query (>= 500 tokens)",
|
46 |
+
"English": "English Prompts",
|
47 |
+
"Chinese": "Chinese Prompts",
|
48 |
+
"French": "French Prompts",
|
49 |
+
"German": "German Prompts",
|
50 |
+
"Spanish": "Spanish Prompts",
|
51 |
+
"Russian": "Russian Prompts",
|
52 |
+
"Japanese": "Japanese Prompts",
|
53 |
+
"Korean": "Korean Prompts",
|
54 |
+
"Exclude Ties": "Exclude Ties and Bothbad",
|
55 |
+
"Exclude Short Query (< 5 tokens)": "Exclude Short User Query (< 5 tokens)",
|
56 |
+
"Exclude Refusal": 'Exclude model responses with refusal (e.g., "I cannot answer")',
|
57 |
+
"overall_limit_5_user_vote": "overall_limit_5_user_vote",
|
58 |
+
"Overall (Deprecated)": "Overall without De-duplicating Top Redundant Queries (top 0.1%). See details in [blog post](https://lmsys.org/blog/2024-05-17-category-hard/#note-enhancing-quality-through-de-duplication).",
|
59 |
+
}
|
60 |
+
|
61 |
+
PROPRIETARY_LICENSES = ["Proprietary", "Proprietory"]
|
62 |
+
|
63 |
+
|
64 |
+
def download_latest_data_from_space(
|
65 |
+
repo_id: str, file_type: Literal["pkl", "csv"]
|
66 |
+
) -> str:
|
67 |
+
"""
|
68 |
+
Downloads the latest data file of the specified file type from the given repository space.
|
69 |
+
|
70 |
+
Args:
|
71 |
+
repo_id (str): The ID of the repository space.
|
72 |
+
file_type (Literal["pkl", "csv"]): The type of the data file to download. Must be either "pkl" or "csv".
|
73 |
+
|
74 |
+
Returns:
|
75 |
+
str: The local file path of the downloaded data file.
|
76 |
+
"""
|
77 |
+
|
78 |
+
def extract_date(filename):
|
79 |
+
return filename.split("/")[-1].split(".")[0].split("_")[-1]
|
80 |
+
|
81 |
+
fs = HfFileSystem()
|
82 |
+
data_file_path = f"spaces/{repo_id}/*.{file_type}"
|
83 |
+
files = fs.glob(data_file_path)
|
84 |
+
files = [
|
85 |
+
file for file in files if "leaderboard_table" in file or "elo_results" in file
|
86 |
+
]
|
87 |
+
latest_file = sorted(files, key=extract_date, reverse=True)[0]
|
88 |
+
|
89 |
+
latest_filepath_local = hf_hub_download(
|
90 |
+
repo_id=repo_id,
|
91 |
+
filename=latest_file.split("/")[-1],
|
92 |
+
repo_type="space",
|
93 |
+
)
|
94 |
+
print(latest_file.split("/")[-1])
|
95 |
+
return latest_filepath_local
|
96 |
+
|
97 |
+
|
98 |
+
def get_constants(dfs):
|
99 |
+
"""
|
100 |
+
Calculate and return the minimum and maximum Elo scores, as well as the maximum number of models per month.
|
101 |
+
|
102 |
+
Parameters:
|
103 |
+
- dfs (dict): A dictionary containing DataFrames for different categories.
|
104 |
+
|
105 |
+
Returns:
|
106 |
+
- min_elo_score (float): The minimum Elo score across all DataFrames.
|
107 |
+
- max_elo_score (float): The maximum Elo score across all DataFrames.
|
108 |
+
- upper_models_per_month (int): The maximum number of models per month per license across all DataFrames.
|
109 |
+
"""
|
110 |
+
filter_ranges = {}
|
111 |
+
for k, df in dfs.items():
|
112 |
+
filter_ranges[k] = {
|
113 |
+
"min_elo_score": df["rating"].min().round(),
|
114 |
+
"max_elo_score": df["rating"].max().round(),
|
115 |
+
"upper_models_per_month": int(
|
116 |
+
df.groupby(["Month-Year", "License"])["rating"]
|
117 |
+
.apply(lambda x: x.count())
|
118 |
+
.max()
|
119 |
+
),
|
120 |
+
}
|
121 |
+
|
122 |
+
min_elo_score = float("inf")
|
123 |
+
max_elo_score = float("-inf")
|
124 |
+
upper_models_per_month = 0
|
125 |
+
|
126 |
+
for _, value in filter_ranges.items():
|
127 |
+
min_elo_score = min(min_elo_score, value["min_elo_score"])
|
128 |
+
max_elo_score = max(max_elo_score, value["max_elo_score"])
|
129 |
+
upper_models_per_month = max(
|
130 |
+
upper_models_per_month, value["upper_models_per_month"]
|
131 |
+
)
|
132 |
+
return min_elo_score, max_elo_score, upper_models_per_month
|
133 |
+
|
134 |
+
|
135 |
+
def update_release_date_mapping(
|
136 |
+
new_model_keys_to_add: List[str],
|
137 |
+
leaderboard_df: pd.DataFrame,
|
138 |
+
release_date_mapping: pd.DataFrame,
|
139 |
+
) -> pd.DataFrame:
|
140 |
+
"""
|
141 |
+
Update the release date mapping with new model keys.
|
142 |
+
|
143 |
+
Args:
|
144 |
+
new_model_keys_to_add (List[str]): A list of new model keys to add to the release date mapping.
|
145 |
+
leaderboard_df (pd.DataFrame): The leaderboard DataFrame containing the model information.
|
146 |
+
release_date_mapping (pd.DataFrame): The current release date mapping DataFrame.
|
147 |
+
|
148 |
+
Returns:
|
149 |
+
pd.DataFrame: The updated release date mapping DataFrame.
|
150 |
+
"""
|
151 |
+
# if any, add those to the release date mapping
|
152 |
+
if new_model_keys_to_add:
|
153 |
+
for key in new_model_keys_to_add:
|
154 |
+
new_entry = {
|
155 |
+
"key": key,
|
156 |
+
"Model": leaderboard_df[leaderboard_df["key"] == key]["Model"].values[
|
157 |
+
0
|
158 |
+
],
|
159 |
+
"Release Date": datetime.today().strftime("%Y-%m-%d"),
|
160 |
+
}
|
161 |
+
|
162 |
+
with open("release_date_mapping.json", "r") as file:
|
163 |
+
data = json.load(file)
|
164 |
+
|
165 |
+
data.append(new_entry)
|
166 |
+
|
167 |
+
with open("release_date_mapping.json", "w") as file:
|
168 |
+
json.dump(data, file, indent=4)
|
169 |
+
|
170 |
+
print(f"Added {key} to release_date_mapping.json")
|
171 |
+
|
172 |
+
# reload the release date mapping
|
173 |
+
release_date_mapping = pd.read_json(
|
174 |
+
"release_date_mapping.json", orient="records"
|
175 |
+
)
|
176 |
+
return release_date_mapping
|
177 |
+
|
178 |
+
|
179 |
+
def format_data(df):
|
180 |
+
"""
|
181 |
+
Formats the given DataFrame by performing the following operations:
|
182 |
+
- Converts the 'License' column values to 'Proprietary LLM' if they are in PROPRIETARY_LICENSES, otherwise 'Open LLM'.
|
183 |
+
- Converts the 'Release Date' column to datetime format.
|
184 |
+
- Adds a new 'Month-Year' column by extracting the month and year from the 'Release Date' column.
|
185 |
+
- Rounds the 'rating' column to the nearest integer.
|
186 |
+
- Resets the index of the DataFrame.
|
187 |
+
|
188 |
+
Args:
|
189 |
+
df (pandas.DataFrame): The DataFrame to be formatted.
|
190 |
+
|
191 |
+
Returns:
|
192 |
+
pandas.DataFrame: The formatted DataFrame.
|
193 |
+
"""
|
194 |
+
df["License"] = df["License"].apply(
|
195 |
+
lambda x: "Proprietary LLM" if x in PROPRIETARY_LICENSES else "Open LLM"
|
196 |
+
)
|
197 |
+
df["Release Date"] = pd.to_datetime(df["Release Date"])
|
198 |
+
df["Month-Year"] = df["Release Date"].dt.to_period("M")
|
199 |
+
df["rating"] = df["rating"].round()
|
200 |
+
return df.reset_index(drop=True)
|
201 |
+
|
202 |
+
|
203 |
+
def get_trendlines(fig):
|
204 |
+
|
205 |
+
trend_lines = px.get_trendline_results(fig)
|
206 |
+
|
207 |
+
return [
|
208 |
+
trend_lines.iloc[i]["px_fit_results"].params.tolist()
|
209 |
+
for i in range(len(trend_lines))
|
210 |
+
]
|
211 |
+
|
212 |
+
|
213 |
+
def find_crossover_point(b1, m1, b2, m2):
|
214 |
+
"""
|
215 |
+
Determine the X value at which two trendlines will cross over.
|
216 |
+
|
217 |
+
Parameters:
|
218 |
+
m1 (float): Slope of the first trendline.
|
219 |
+
b1 (float): Intercept of the first trendline.
|
220 |
+
m2 (float): Slope of the second trendline.
|
221 |
+
b2 (float): Intercept of the second trendline.
|
222 |
+
|
223 |
+
Returns:
|
224 |
+
float: The X value where the two trendlines cross.
|
225 |
+
"""
|
226 |
+
if m1 == m2:
|
227 |
+
raise ValueError("The trendlines are parallel and do not cross.")
|
228 |
+
|
229 |
+
x_crossover = (b2 - b1) / (m1 - m2)
|
230 |
+
return x_crossover
|
231 |
+
|
232 |
+
# Function to create sigmoid transition
|
233 |
+
def sigmoid_transition(x, x0, k=0.1):
|
234 |
+
return expit(k * (x - x0))
|