Spaces:
Sleeping
Sleeping
submit-system-update (#838)
Browse files- add login at submit and user info (0b31d4e1ffcd10bcdf9603d588d3b1d28c77fdc2)
- use gradio-leaderboard 0.0.11 (c8c479205c76ee4c961253c5909ffaa1eb6b03ed)
- update and revised submission info (8013f916ae1e9df83a51c72a1b736f51f10c8bc5)
- app.py +1 -0
- pyproject.toml +1 -1
- requirements.txt +1 -1
- src/display/about.py +26 -14
- src/submission/submit.py +22 -7
app.py
CHANGED
|
@@ -235,6 +235,7 @@ with main_block:
|
|
| 235 |
|
| 236 |
with gr.Row():
|
| 237 |
gr.Markdown("# ✉️✨ Submit your model here!", elem_classes="markdown-text")
|
|
|
|
| 238 |
|
| 239 |
with gr.Row():
|
| 240 |
with gr.Column():
|
|
|
|
| 235 |
|
| 236 |
with gr.Row():
|
| 237 |
gr.Markdown("# ✉️✨ Submit your model here!", elem_classes="markdown-text")
|
| 238 |
+
login_button = gr.LoginButton(elem_id="oauth-button")
|
| 239 |
|
| 240 |
with gr.Row():
|
| 241 |
with gr.Column():
|
pyproject.toml
CHANGED
|
@@ -45,7 +45,7 @@ tokenizers = ">=0.15.0"
|
|
| 45 |
gradio-space-ci = {git = "https://huggingface.co/spaces/Wauplin/gradio-space-ci", rev = "0.2.3"}
|
| 46 |
isort = "^5.13.2"
|
| 47 |
ruff = "^0.3.5"
|
| 48 |
-
gradio-leaderboard = "0.0.
|
| 49 |
gradio = {extras = ["oauth"], version = "^4.36.1"}
|
| 50 |
requests = "^2.31.0"
|
| 51 |
requests-oauthlib = "^1.3.1"
|
|
|
|
| 45 |
gradio-space-ci = {git = "https://huggingface.co/spaces/Wauplin/gradio-space-ci", rev = "0.2.3"}
|
| 46 |
isort = "^5.13.2"
|
| 47 |
ruff = "^0.3.5"
|
| 48 |
+
gradio-leaderboard = "0.0.11"
|
| 49 |
gradio = {extras = ["oauth"], version = "^4.36.1"}
|
| 50 |
requests = "^2.31.0"
|
| 51 |
requests-oauthlib = "^1.3.1"
|
requirements.txt
CHANGED
|
@@ -17,7 +17,7 @@ isort
|
|
| 17 |
ruff
|
| 18 |
gradio==4.31.0
|
| 19 |
gradio[oauth]
|
| 20 |
-
gradio_leaderboard==0.0.
|
| 21 |
requests==2.31.0
|
| 22 |
requests-oauthlib== 1.3.1
|
| 23 |
schedule == 1.2.2
|
|
|
|
| 17 |
ruff
|
| 18 |
gradio==4.31.0
|
| 19 |
gradio[oauth]
|
| 20 |
+
gradio_leaderboard==0.0.11
|
| 21 |
requests==2.31.0
|
| 22 |
requests-oauthlib== 1.3.1
|
| 23 |
schedule == 1.2.2
|
src/display/about.py
CHANGED
|
@@ -179,35 +179,47 @@ EVALUATION_QUEUE_TEXT = f"""
|
|
| 179 |
|
| 180 |
Models added here will be automatically evaluated on the 🤗 cluster.
|
| 181 |
|
| 182 |
-
|
| 183 |
|
| 184 |
-
##
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 185 |
|
| 186 |
-
### 1) Make sure you can load your model and tokenizer using AutoClasses:
|
| 187 |
```python
|
| 188 |
from transformers import AutoConfig, AutoModel, AutoTokenizer
|
| 189 |
config = AutoConfig.from_pretrained("your model name", revision=revision)
|
| 190 |
model = AutoModel.from_pretrained("your model name", revision=revision)
|
| 191 |
tokenizer = AutoTokenizer.from_pretrained("your model name", revision=revision)
|
| 192 |
```
|
| 193 |
-
|
|
|
|
|
|
|
|
|
|
| 194 |
|
| 195 |
-
|
| 196 |
-
|
| 197 |
|
| 198 |
-
###
|
| 199 |
-
It's a new format for storing weights which is safer and faster to load and use. It will also allow us to add the number of parameters of your model to the `Extended Viewer`!
|
| 200 |
-
|
| 201 |
-
### 3) Make sure your model has an open license!
|
| 202 |
This is a leaderboard for Open LLMs, and we'd love for as many people as possible to know they can use your model 🤗
|
| 203 |
|
| 204 |
-
### 4
|
| 205 |
When we add extra information about models to the leaderboard, it will be automatically taken from the model card
|
| 206 |
|
| 207 |
-
### 5
|
| 208 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
| 209 |
|
| 210 |
-
<b>Note:</b> Please be advised that when submitting, git <b>branches</b> and <b>tags</b> will be strictly tied to the <b>specific commit</b> present at the time of submission. This ensures revision consistency.
|
| 211 |
## Model types
|
| 212 |
{icons}
|
| 213 |
"""
|
|
|
|
| 179 |
|
| 180 |
Models added here will be automatically evaluated on the 🤗 cluster.
|
| 181 |
|
| 182 |
+
> **Important:** Don't forget to read the [FAQ](https://huggingface.co/docs/leaderboards/open_llm_leaderboard/faq) and [documentation](https://huggingface.co/docs/leaderboards/open_llm_leaderboard/about) for more information! 📄
|
| 183 |
|
| 184 |
+
## Submission Disclaimer
|
| 185 |
+
**By submitting a model, you acknowledge that:**
|
| 186 |
+
- We store information about who submitted each model in [Requests dataset](https://huggingface.co/datasets/open-llm-leaderboard/requests).
|
| 187 |
+
- This practice helps maintain the integrity of our leaderboard, prevent spam, and ensure responsible submissions.
|
| 188 |
+
- Your submission will be visible to the community and you may be contacted regarding your model.
|
| 189 |
+
- Please submit carefully and responsibly 💛
|
| 190 |
+
|
| 191 |
+
## First Steps Before Submitting a Model
|
| 192 |
+
|
| 193 |
+
### 1. Ensure Your Model Loads with AutoClasses
|
| 194 |
+
Verify that you can load your model and tokenizer using AutoClasses:
|
| 195 |
|
|
|
|
| 196 |
```python
|
| 197 |
from transformers import AutoConfig, AutoModel, AutoTokenizer
|
| 198 |
config = AutoConfig.from_pretrained("your model name", revision=revision)
|
| 199 |
model = AutoModel.from_pretrained("your model name", revision=revision)
|
| 200 |
tokenizer = AutoTokenizer.from_pretrained("your model name", revision=revision)
|
| 201 |
```
|
| 202 |
+
Note:
|
| 203 |
+
- If this step fails, debug your model before submitting.
|
| 204 |
+
- Ensure your model is public.
|
| 205 |
+
- We are working on adding support for models requiring `use_remote_code=True`.
|
| 206 |
|
| 207 |
+
### 2. Convert Weights to Safetensors
|
| 208 |
+
[Safetensors](https://huggingface.co/docs/safetensors/index) is a new format for storing weights which is safer and faster to load and use. It will also allow us to add the number of parameters of your model to the `Extended Viewer`!
|
| 209 |
|
| 210 |
+
### 3. Verify Your Model Open License
|
|
|
|
|
|
|
|
|
|
| 211 |
This is a leaderboard for Open LLMs, and we'd love for as many people as possible to know they can use your model 🤗
|
| 212 |
|
| 213 |
+
### 4. Complete Your Model Card
|
| 214 |
When we add extra information about models to the leaderboard, it will be automatically taken from the model card
|
| 215 |
|
| 216 |
+
### 5. Select Correct Precision
|
| 217 |
+
Choose the right precision to avoid evaluation errors:
|
| 218 |
+
- Not all models convert properly from float16 to bfloat16.
|
| 219 |
+
- Incorrect precision can cause issues (e.g., loading a bf16 model in fp16 may generate NaNs).
|
| 220 |
+
|
| 221 |
+
> **Important:** When submitting, git branches and tags will be strictly tied to the specific commit present at the time of submission to ensure revision consistency.
|
| 222 |
|
|
|
|
| 223 |
## Model types
|
| 224 |
{icons}
|
| 225 |
"""
|
src/submission/submit.py
CHANGED
|
@@ -1,5 +1,6 @@
|
|
| 1 |
import json
|
| 2 |
import os
|
|
|
|
| 3 |
from datetime import datetime, timezone
|
| 4 |
|
| 5 |
from dataclasses import dataclass
|
|
@@ -59,16 +60,25 @@ def add_new_eval(
|
|
| 59 |
weight_type: str,
|
| 60 |
model_type: str,
|
| 61 |
use_chat_template: bool,
|
| 62 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 63 |
global REQUESTED_MODELS
|
| 64 |
global USERS_TO_SUBMISSION_DATES
|
| 65 |
if not REQUESTED_MODELS:
|
| 66 |
REQUESTED_MODELS, USERS_TO_SUBMISSION_DATES = already_submitted_models(EVAL_REQUESTS_PATH)
|
| 67 |
|
| 68 |
-
|
|
|
|
| 69 |
model_path = model
|
| 70 |
if "/" in model:
|
| 71 |
-
|
| 72 |
model_path = model.split("/")[1]
|
| 73 |
|
| 74 |
precision = precision.split(" ")[0]
|
|
@@ -77,10 +87,14 @@ def add_new_eval(
|
|
| 77 |
if model_type is None or model_type == "":
|
| 78 |
return styled_error("Please select a model type.")
|
| 79 |
|
|
|
|
|
|
|
|
|
|
|
|
|
| 80 |
# Is the user rate limited?
|
| 81 |
-
if
|
| 82 |
user_can_submit, error_msg = user_submission_permission(
|
| 83 |
-
|
| 84 |
)
|
| 85 |
if not user_can_submit:
|
| 86 |
return styled_error(error_msg)
|
|
@@ -144,7 +158,6 @@ def add_new_eval(
|
|
| 144 |
|
| 145 |
# Seems good, creating the eval
|
| 146 |
print("Adding new eval")
|
| 147 |
-
|
| 148 |
eval_entry = {
|
| 149 |
"model": model,
|
| 150 |
"base_model": base_model,
|
|
@@ -159,10 +172,11 @@ def add_new_eval(
|
|
| 159 |
"job_id": -1,
|
| 160 |
"job_start_time": None,
|
| 161 |
"use_chat_template": use_chat_template,
|
|
|
|
| 162 |
}
|
| 163 |
|
| 164 |
print("Creating eval file")
|
| 165 |
-
OUT_DIR = f"{EVAL_REQUESTS_PATH}/{
|
| 166 |
os.makedirs(OUT_DIR, exist_ok=True)
|
| 167 |
out_path = f"{OUT_DIR}/{model_path}_eval_request_False_{precision}_{weight_type}.json"
|
| 168 |
|
|
@@ -170,6 +184,7 @@ def add_new_eval(
|
|
| 170 |
f.write(json.dumps(eval_entry))
|
| 171 |
|
| 172 |
print("Uploading eval file")
|
|
|
|
| 173 |
API.upload_file(
|
| 174 |
path_or_fileobj=out_path,
|
| 175 |
path_in_repo=out_path.split("eval-queue/")[1],
|
|
|
|
| 1 |
import json
|
| 2 |
import os
|
| 3 |
+
import gradio as gr
|
| 4 |
from datetime import datetime, timezone
|
| 5 |
|
| 6 |
from dataclasses import dataclass
|
|
|
|
| 60 |
weight_type: str,
|
| 61 |
model_type: str,
|
| 62 |
use_chat_template: bool,
|
| 63 |
+
profile: gr.OAuthProfile | None
|
| 64 |
+
):
|
| 65 |
+
# Login require
|
| 66 |
+
if profile is None:
|
| 67 |
+
return styled_error("Hub Login Required")
|
| 68 |
+
|
| 69 |
+
# Name of the actual user who sent the request
|
| 70 |
+
username = profile.username
|
| 71 |
+
|
| 72 |
global REQUESTED_MODELS
|
| 73 |
global USERS_TO_SUBMISSION_DATES
|
| 74 |
if not REQUESTED_MODELS:
|
| 75 |
REQUESTED_MODELS, USERS_TO_SUBMISSION_DATES = already_submitted_models(EVAL_REQUESTS_PATH)
|
| 76 |
|
| 77 |
+
|
| 78 |
+
org_or_user = ""
|
| 79 |
model_path = model
|
| 80 |
if "/" in model:
|
| 81 |
+
org_or_user = model.split("/")[0]
|
| 82 |
model_path = model.split("/")[1]
|
| 83 |
|
| 84 |
precision = precision.split(" ")[0]
|
|
|
|
| 87 |
if model_type is None or model_type == "":
|
| 88 |
return styled_error("Please select a model type.")
|
| 89 |
|
| 90 |
+
# Is user submitting own model?
|
| 91 |
+
# Check that username in the org.
|
| 92 |
+
# if org_or_user != profile.username:
|
| 93 |
+
|
| 94 |
# Is the user rate limited?
|
| 95 |
+
if org_or_user != "":
|
| 96 |
user_can_submit, error_msg = user_submission_permission(
|
| 97 |
+
org_or_user, USERS_TO_SUBMISSION_DATES, RATE_LIMIT_PERIOD, RATE_LIMIT_QUOTA
|
| 98 |
)
|
| 99 |
if not user_can_submit:
|
| 100 |
return styled_error(error_msg)
|
|
|
|
| 158 |
|
| 159 |
# Seems good, creating the eval
|
| 160 |
print("Adding new eval")
|
|
|
|
| 161 |
eval_entry = {
|
| 162 |
"model": model,
|
| 163 |
"base_model": base_model,
|
|
|
|
| 172 |
"job_id": -1,
|
| 173 |
"job_start_time": None,
|
| 174 |
"use_chat_template": use_chat_template,
|
| 175 |
+
"sender": username
|
| 176 |
}
|
| 177 |
|
| 178 |
print("Creating eval file")
|
| 179 |
+
OUT_DIR = f"{EVAL_REQUESTS_PATH}/{org_or_user}"
|
| 180 |
os.makedirs(OUT_DIR, exist_ok=True)
|
| 181 |
out_path = f"{OUT_DIR}/{model_path}_eval_request_False_{precision}_{weight_type}.json"
|
| 182 |
|
|
|
|
| 184 |
f.write(json.dumps(eval_entry))
|
| 185 |
|
| 186 |
print("Uploading eval file")
|
| 187 |
+
print(eval_entry)
|
| 188 |
API.upload_file(
|
| 189 |
path_or_fileobj=out_path,
|
| 190 |
path_in_repo=out_path.split("eval-queue/")[1],
|