refactor: Update model handling and utility functions
Browse files- Removed BrainDAO from model type dropdown in app.py
- Commented out citation accordion in app.py
- Adjusted job scheduler interval from 1800 to 3600 seconds in app.py
- Updated utility functions in src/utils.py for model name extraction from file paths
- Refactored model name retrieval in src/leaderboard/read_evals.py, src/populate.py, and src/submission/submit.py
- Removed model likes from submission data in src/submission/submit.py
- Adjusted model type storage in submission data to exclude emoji
- app.py +11 -11
- src/leaderboard/read_evals.py +3 -3
- src/populate.py +4 -4
- src/submission/submit.py +18 -21
- src/utils.py +3 -3
app.py
CHANGED
@@ -170,7 +170,7 @@ with demo:
|
|
170 |
model_name_textbox = gr.Textbox(label="Model name")
|
171 |
revision_name_textbox = gr.Textbox(label="Revision commit", placeholder="main")
|
172 |
model_type = gr.Dropdown(
|
173 |
-
choices=[t.to_str(" ") for t in ModelType if t
|
174 |
label="Model type",
|
175 |
multiselect=False,
|
176 |
value=None,
|
@@ -207,18 +207,18 @@ with demo:
|
|
207 |
submission_result,
|
208 |
)
|
209 |
|
210 |
-
with gr.Row():
|
211 |
-
|
212 |
-
|
213 |
-
|
214 |
-
|
215 |
-
|
216 |
-
|
217 |
-
|
218 |
-
|
219 |
|
220 |
scheduler = BackgroundScheduler()
|
221 |
-
scheduler.add_job(restart_space, "interval", seconds=
|
222 |
scheduler.start()
|
223 |
demo.queue(default_concurrency_limit=40).launch(
|
224 |
server_name="0.0.0.0",
|
|
|
170 |
model_name_textbox = gr.Textbox(label="Model name")
|
171 |
revision_name_textbox = gr.Textbox(label="Revision commit", placeholder="main")
|
172 |
model_type = gr.Dropdown(
|
173 |
+
choices=[t.to_str(" ") for t in ModelType if t not in [ModelType.Unknown, ModelType.BrainDAO]],
|
174 |
label="Model type",
|
175 |
multiselect=False,
|
176 |
value=None,
|
|
|
207 |
submission_result,
|
208 |
)
|
209 |
|
210 |
+
# with gr.Row():
|
211 |
+
# with gr.Accordion("π Citation", open=False):
|
212 |
+
# citation_button = gr.Textbox(
|
213 |
+
# value=CITATION_BUTTON_TEXT,
|
214 |
+
# label=CITATION_BUTTON_LABEL,
|
215 |
+
# lines=20,
|
216 |
+
# elem_id="citation-button",
|
217 |
+
# show_copy_button=True,
|
218 |
+
# )
|
219 |
|
220 |
scheduler = BackgroundScheduler()
|
221 |
+
scheduler.add_job(restart_space, "interval", seconds=3600)
|
222 |
scheduler.start()
|
223 |
demo.queue(default_concurrency_limit=40).launch(
|
224 |
server_name="0.0.0.0",
|
src/leaderboard/read_evals.py
CHANGED
@@ -13,7 +13,7 @@ import numpy as np
|
|
13 |
from src.display.formatting import make_clickable_model
|
14 |
from src.display.utils import AutoEvalColumn, ModelType, Precision, Tasks, WeightType
|
15 |
from src.submission.check_validity import is_model_on_hub
|
16 |
-
from src.utils import
|
17 |
|
18 |
|
19 |
@dataclass
|
@@ -42,14 +42,14 @@ class EvalResult:
|
|
42 |
with open(json_filepath) as fp:
|
43 |
data = json.load(fp)
|
44 |
|
45 |
-
org, model =
|
46 |
config = data.get("config")
|
47 |
|
48 |
# Precision
|
49 |
precision = Precision.from_str(config.get("model_dtype"))
|
50 |
|
51 |
result_key = f"{org}_{model}_{precision.value.name}"
|
52 |
-
model_name =
|
53 |
|
54 |
still_on_hub, _, model_config = is_model_on_hub(
|
55 |
model_name,
|
|
|
13 |
from src.display.formatting import make_clickable_model
|
14 |
from src.display.utils import AutoEvalColumn, ModelType, Precision, Tasks, WeightType
|
15 |
from src.submission.check_validity import is_model_on_hub
|
16 |
+
from src.utils import get_model_name_from_filepath, get_org_and_model_names_from_filepath, get_request_hash
|
17 |
|
18 |
|
19 |
@dataclass
|
|
|
42 |
with open(json_filepath) as fp:
|
43 |
data = json.load(fp)
|
44 |
|
45 |
+
org, model = get_org_and_model_names_from_filepath(json_filepath)
|
46 |
config = data.get("config")
|
47 |
|
48 |
# Precision
|
49 |
precision = Precision.from_str(config.get("model_dtype"))
|
50 |
|
51 |
result_key = f"{org}_{model}_{precision.value.name}"
|
52 |
+
model_name = get_model_name_from_filepath(json_filepath)
|
53 |
|
54 |
still_on_hub, _, model_config = is_model_on_hub(
|
55 |
model_name,
|
src/populate.py
CHANGED
@@ -10,7 +10,7 @@ import pandas as pd
|
|
10 |
from src.display.formatting import has_no_nan_values, make_clickable_model
|
11 |
from src.display.utils import AutoEvalColumn, EvalQueueColumn
|
12 |
from src.leaderboard.read_evals import get_raw_eval_results
|
13 |
-
from src.utils import
|
14 |
|
15 |
|
16 |
def get_leaderboard_df(results_path: str, requests_path: str, cols: list, benchmark_cols: list) -> pd.DataFrame:
|
@@ -43,7 +43,7 @@ def get_leaderboard_df(results_path: str, requests_path: str, cols: list, benchm
|
|
43 |
# continue
|
44 |
|
45 |
# # data[EvalQueueColumn.model.name] = make_clickable_model(data["model"])
|
46 |
-
# model_name =
|
47 |
# data[EvalQueueColumn.model.name] = make_clickable_model(model_name)
|
48 |
|
49 |
# data[EvalQueueColumn.revision.name] = data.get("revision", "main")
|
@@ -66,7 +66,7 @@ def get_leaderboard_df(results_path: str, requests_path: str, cols: list, benchm
|
|
66 |
# continue
|
67 |
|
68 |
# # data[EvalQueueColumn.model.name] = make_clickable_model(data["model"])
|
69 |
-
# model_name =
|
70 |
# data[EvalQueueColumn.model.name] = make_clickable_model(model_name)
|
71 |
|
72 |
# data[EvalQueueColumn.revision.name] = data.get("revision", "main")
|
@@ -92,7 +92,7 @@ def get_evaluation_requests_df(save_path: str, cols: list) -> list[pd.DataFrame]
|
|
92 |
print(f"Error reading or decoding {file_path}: {e}")
|
93 |
return None
|
94 |
|
95 |
-
model_name =
|
96 |
# data[EvalQueueColumn.model.name] = make_clickable_model(data["model"])
|
97 |
data[EvalQueueColumn.model.name] = make_clickable_model(model_name)
|
98 |
data[EvalQueueColumn.revision.name] = data.get("revision", "main")
|
|
|
10 |
from src.display.formatting import has_no_nan_values, make_clickable_model
|
11 |
from src.display.utils import AutoEvalColumn, EvalQueueColumn
|
12 |
from src.leaderboard.read_evals import get_raw_eval_results
|
13 |
+
from src.utils import get_model_name_from_filepath
|
14 |
|
15 |
|
16 |
def get_leaderboard_df(results_path: str, requests_path: str, cols: list, benchmark_cols: list) -> pd.DataFrame:
|
|
|
43 |
# continue
|
44 |
|
45 |
# # data[EvalQueueColumn.model.name] = make_clickable_model(data["model"])
|
46 |
+
# model_name = get_model_name_from_filepath(file_path)
|
47 |
# data[EvalQueueColumn.model.name] = make_clickable_model(model_name)
|
48 |
|
49 |
# data[EvalQueueColumn.revision.name] = data.get("revision", "main")
|
|
|
66 |
# continue
|
67 |
|
68 |
# # data[EvalQueueColumn.model.name] = make_clickable_model(data["model"])
|
69 |
+
# model_name = get_model_name_from_filepath(file_path)
|
70 |
# data[EvalQueueColumn.model.name] = make_clickable_model(model_name)
|
71 |
|
72 |
# data[EvalQueueColumn.revision.name] = data.get("revision", "main")
|
|
|
92 |
print(f"Error reading or decoding {file_path}: {e}")
|
93 |
return None
|
94 |
|
95 |
+
model_name = get_model_name_from_filepath(file_path)
|
96 |
# data[EvalQueueColumn.model.name] = make_clickable_model(data["model"])
|
97 |
data[EvalQueueColumn.model.name] = make_clickable_model(model_name)
|
98 |
data[EvalQueueColumn.revision.name] = data.get("revision", "main")
|
src/submission/submit.py
CHANGED
@@ -16,7 +16,7 @@ USERS_TO_SUBMISSION_DATES = None
|
|
16 |
|
17 |
|
18 |
def add_new_eval(
|
19 |
-
|
20 |
# base_model: str,
|
21 |
revision: str,
|
22 |
# precision: str,
|
@@ -32,12 +32,6 @@ def add_new_eval(
|
|
32 |
if not REQUESTED_MODELS:
|
33 |
REQUESTED_MODELS, USERS_TO_SUBMISSION_DATES = already_submitted_models(EVAL_REQUESTS_PATH)
|
34 |
|
35 |
-
user_name = ""
|
36 |
-
model_path = model
|
37 |
-
if "/" in model:
|
38 |
-
user_name = model.split("/")[0]
|
39 |
-
model_path = model.split("/")[1]
|
40 |
-
|
41 |
precision = precision.split(" ")[0]
|
42 |
current_time = datetime.now(timezone.utc).strftime("%Y-%m-%dT%H:%M:%SZ")
|
43 |
|
@@ -55,13 +49,13 @@ def add_new_eval(
|
|
55 |
# return styled_error(f'Base model "{base_model}" {error}')
|
56 |
|
57 |
if weight_type != "Adapter":
|
58 |
-
model_on_hub, error, _ = is_model_on_hub(model_name=
|
59 |
if not model_on_hub:
|
60 |
-
return styled_error(f'Model "{
|
61 |
|
62 |
# Is the model info correctly filled?
|
63 |
try:
|
64 |
-
model_info = API.model_info(repo_id=
|
65 |
except Exception:
|
66 |
return styled_error("Could not get your model information. Please fill it up properly.")
|
67 |
|
@@ -76,7 +70,7 @@ def add_new_eval(
|
|
76 |
except Exception:
|
77 |
return styled_error("Please select a license for your model")
|
78 |
|
79 |
-
is_model_card_ok, error_msg = check_model_card(
|
80 |
if not is_model_card_ok:
|
81 |
return styled_error(error_msg)
|
82 |
|
@@ -91,23 +85,26 @@ def add_new_eval(
|
|
91 |
"weight_type": weight_type,
|
92 |
"status": "PENDING",
|
93 |
"submitted_time": current_time,
|
94 |
-
"model_type": model_type,
|
95 |
-
"likes": model_info.likes,
|
96 |
"params": model_size,
|
97 |
"license": license_title,
|
98 |
-
"private": False,
|
99 |
}
|
100 |
|
101 |
# Check for duplicate submission
|
102 |
-
request_id = get_request_id(
|
103 |
if request_id in REQUESTED_MODELS:
|
104 |
return styled_warning("This model has been already submitted.")
|
105 |
-
request_hash = get_request_hash(
|
106 |
|
107 |
print("Creating eval file")
|
108 |
-
OUT_DIR = f"{EVAL_REQUESTS_PATH}/{
|
109 |
os.makedirs(OUT_DIR, exist_ok=True)
|
110 |
-
|
|
|
|
|
|
|
111 |
|
112 |
with open(out_path, "w") as f:
|
113 |
f.write(json.dumps(eval_entry))
|
@@ -115,15 +112,15 @@ def add_new_eval(
|
|
115 |
print("Uploading eval file")
|
116 |
API.upload_file(
|
117 |
path_or_fileobj=out_path,
|
118 |
-
path_in_repo=
|
119 |
repo_id=REQUESTS_REPO,
|
120 |
repo_type="dataset",
|
121 |
-
commit_message=f"Add {
|
122 |
)
|
123 |
|
124 |
# Remove the local file
|
125 |
os.remove(out_path)
|
126 |
|
127 |
return styled_message(
|
128 |
-
"Your
|
129 |
)
|
|
|
16 |
|
17 |
|
18 |
def add_new_eval(
|
19 |
+
model_name: str,
|
20 |
# base_model: str,
|
21 |
revision: str,
|
22 |
# precision: str,
|
|
|
32 |
if not REQUESTED_MODELS:
|
33 |
REQUESTED_MODELS, USERS_TO_SUBMISSION_DATES = already_submitted_models(EVAL_REQUESTS_PATH)
|
34 |
|
|
|
|
|
|
|
|
|
|
|
|
|
35 |
precision = precision.split(" ")[0]
|
36 |
current_time = datetime.now(timezone.utc).strftime("%Y-%m-%dT%H:%M:%SZ")
|
37 |
|
|
|
49 |
# return styled_error(f'Base model "{base_model}" {error}')
|
50 |
|
51 |
if weight_type != "Adapter":
|
52 |
+
model_on_hub, error, _ = is_model_on_hub(model_name=model_name, revision=revision, token=TOKEN, test_tokenizer=True)
|
53 |
if not model_on_hub:
|
54 |
+
return styled_error(f'Model "{model_name}" {error}')
|
55 |
|
56 |
# Is the model info correctly filled?
|
57 |
try:
|
58 |
+
model_info = API.model_info(repo_id=model_name, revision=revision)
|
59 |
except Exception:
|
60 |
return styled_error("Could not get your model information. Please fill it up properly.")
|
61 |
|
|
|
70 |
except Exception:
|
71 |
return styled_error("Please select a license for your model")
|
72 |
|
73 |
+
is_model_card_ok, error_msg = check_model_card(model_name)
|
74 |
if not is_model_card_ok:
|
75 |
return styled_error(error_msg)
|
76 |
|
|
|
85 |
"weight_type": weight_type,
|
86 |
"status": "PENDING",
|
87 |
"submitted_time": current_time,
|
88 |
+
"model_type": model_type.split()[1], # remove the emoji
|
89 |
+
# "likes": model_info.likes,
|
90 |
"params": model_size,
|
91 |
"license": license_title,
|
92 |
+
# "private": False,
|
93 |
}
|
94 |
|
95 |
# Check for duplicate submission
|
96 |
+
request_id = get_request_id(model_name, revision, precision)
|
97 |
if request_id in REQUESTED_MODELS:
|
98 |
return styled_warning("This model has been already submitted.")
|
99 |
+
request_hash = get_request_hash(model_name, revision, precision)
|
100 |
|
101 |
print("Creating eval file")
|
102 |
+
OUT_DIR = f"{EVAL_REQUESTS_PATH}/{model_name}"
|
103 |
os.makedirs(OUT_DIR, exist_ok=True)
|
104 |
+
|
105 |
+
out_path = f"{OUT_DIR}/{request_hash}.json"
|
106 |
+
if os.path.exists(out_path):
|
107 |
+
os.remove(out_path)
|
108 |
|
109 |
with open(out_path, "w") as f:
|
110 |
f.write(json.dumps(eval_entry))
|
|
|
112 |
print("Uploading eval file")
|
113 |
API.upload_file(
|
114 |
path_or_fileobj=out_path,
|
115 |
+
path_in_repo='{}/{}.json'.format(model_name, request_hash),
|
116 |
repo_id=REQUESTS_REPO,
|
117 |
repo_type="dataset",
|
118 |
+
commit_message=f"Add {model_name} to eval requests",
|
119 |
)
|
120 |
|
121 |
# Remove the local file
|
122 |
os.remove(out_path)
|
123 |
|
124 |
return styled_message(
|
125 |
+
"Your model has been submitted."
|
126 |
)
|
src/utils.py
CHANGED
@@ -4,15 +4,15 @@
|
|
4 |
import hashlib
|
5 |
|
6 |
|
7 |
-
def
|
8 |
org, model = file_path.split("/")[-3:-1]
|
9 |
model = model.removesuffix(".json")
|
10 |
model = model.split('_request_')[0]
|
11 |
return org, model
|
12 |
|
13 |
|
14 |
-
def
|
15 |
-
org, model =
|
16 |
return f"{org}/{model}"
|
17 |
|
18 |
|
|
|
4 |
import hashlib
|
5 |
|
6 |
|
7 |
+
def get_org_and_model_names_from_filepath(file_path: str) -> str:
|
8 |
org, model = file_path.split("/")[-3:-1]
|
9 |
model = model.removesuffix(".json")
|
10 |
model = model.split('_request_')[0]
|
11 |
return org, model
|
12 |
|
13 |
|
14 |
+
def get_model_name_from_filepath(file_path: str) -> str:
|
15 |
+
org, model = get_org_and_model_names_from_filepath(file_path)
|
16 |
return f"{org}/{model}"
|
17 |
|
18 |
|