Spaces:
Running
Running
Commit
·
6c10fa6
1
Parent(s):
8b771ed
[TEMP] Offline
Browse files- app.py +1 -0
- src/leaderboard/read_evals.py +5 -1
- src/populate.py +1 -1
- src/submission/submit.py +28 -21
app.py
CHANGED
|
@@ -75,6 +75,7 @@ open_ended_leaderboard_df = open_ended_original_df.copy()
|
|
| 75 |
_, med_safety_original_df = get_leaderboard_df(EVAL_RESULTS_PATH, EVAL_REQUESTS_PATH, MED_SAFETY_COLS, MED_SAFETY_BENCHMARK_COLS, "score", "med_safety")
|
| 76 |
med_safety_leaderboard_df = med_safety_original_df.copy()
|
| 77 |
|
|
|
|
| 78 |
# breakpoint()
|
| 79 |
# # Token based results
|
| 80 |
# _, token_based_datasets_original_df = get_leaderboard_df(EVAL_RESULTS_PATH, EVAL_REQUESTS_PATH, DATASET_COLS, DATASET_BENCHMARK_COLS, "TokenBasedWithMacroAverage", "datasets")
|
|
|
|
| 75 |
_, med_safety_original_df = get_leaderboard_df(EVAL_RESULTS_PATH, EVAL_REQUESTS_PATH, MED_SAFETY_COLS, MED_SAFETY_BENCHMARK_COLS, "score", "med_safety")
|
| 76 |
med_safety_leaderboard_df = med_safety_original_df.copy()
|
| 77 |
|
| 78 |
+
# breakpoint()
|
| 79 |
# breakpoint()
|
| 80 |
# # Token based results
|
| 81 |
# _, token_based_datasets_original_df = get_leaderboard_df(EVAL_RESULTS_PATH, EVAL_REQUESTS_PATH, DATASET_COLS, DATASET_BENCHMARK_COLS, "TokenBasedWithMacroAverage", "datasets")
|
src/leaderboard/read_evals.py
CHANGED
|
@@ -45,7 +45,11 @@ class EvalResult:
|
|
| 45 |
def init_from_json_file(self, json_filepath, evaluation_metric):
|
| 46 |
"""Inits the result from the specific model result file"""
|
| 47 |
with open(json_filepath) as fp:
|
| 48 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
| 49 |
|
| 50 |
config = data.get("config")
|
| 51 |
|
|
|
|
| 45 |
def init_from_json_file(self, json_filepath, evaluation_metric):
|
| 46 |
"""Inits the result from the specific model result file"""
|
| 47 |
with open(json_filepath) as fp:
|
| 48 |
+
try:
|
| 49 |
+
data = json.load(fp)
|
| 50 |
+
except:
|
| 51 |
+
breakpoint()
|
| 52 |
+
|
| 53 |
|
| 54 |
config = data.get("config")
|
| 55 |
|
src/populate.py
CHANGED
|
@@ -68,7 +68,7 @@ def get_evaluation_queue_df(save_path: str, cols: list) -> list[pd.DataFrame]:
|
|
| 68 |
for run in all_evals:
|
| 69 |
# changes to be made here
|
| 70 |
status_list = [run["status"]["closed-ended"], run["status"]["open-ended"], run["status"]["med-safety"], run["status"]["cross-examination"]]
|
| 71 |
-
status_list = status_list[:
|
| 72 |
if "RUNNING" in status_list:
|
| 73 |
running_list.append(run)
|
| 74 |
elif "PENDING" in status_list or "RERUN" in status_list:
|
|
|
|
| 68 |
for run in all_evals:
|
| 69 |
# changes to be made here
|
| 70 |
status_list = [run["status"]["closed-ended"], run["status"]["open-ended"], run["status"]["med-safety"], run["status"]["cross-examination"]]
|
| 71 |
+
status_list = status_list[:3]
|
| 72 |
if "RUNNING" in status_list:
|
| 73 |
running_list.append(run)
|
| 74 |
elif "PENDING" in status_list or "RERUN" in status_list:
|
src/submission/submit.py
CHANGED
|
@@ -63,12 +63,17 @@ def add_new_eval(
|
|
| 63 |
global USERS_TO_SUBMISSION_DATES
|
| 64 |
if not REQUESTED_MODELS:
|
| 65 |
REQUESTED_MODELS, USERS_TO_SUBMISSION_DATES = already_submitted_models(EVAL_REQUESTS_PATH)
|
| 66 |
-
|
| 67 |
-
|
| 68 |
-
|
| 69 |
-
|
| 70 |
-
|
| 71 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 72 |
|
| 73 |
# precision = precision.split(" ")[0]
|
| 74 |
current_time = datetime.now(timezone.utc).strftime("%Y-%m-%dT%H:%M:%SZ")
|
|
@@ -96,22 +101,22 @@ def add_new_eval(
|
|
| 96 |
|
| 97 |
# Is the model info correctly filled?
|
| 98 |
try:
|
| 99 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 100 |
except Exception:
|
| 101 |
return styled_error("Could not get your model information. Please fill it up properly.")
|
| 102 |
|
| 103 |
-
|
| 104 |
-
|
| 105 |
-
# Were the model card and license filled?
|
| 106 |
-
try:
|
| 107 |
-
license = model_info.cardData["license"]
|
| 108 |
-
except Exception:
|
| 109 |
-
return styled_error("Please select a license for your model")
|
| 110 |
-
|
| 111 |
-
modelcard_OK, error_msg = check_model_card(model)
|
| 112 |
-
if not modelcard_OK:
|
| 113 |
-
return styled_error(error_msg)
|
| 114 |
-
|
| 115 |
# Verify the inference config now
|
| 116 |
# try:
|
| 117 |
# label_normalization_map = ast.literal_eval(label_normalization_map)
|
|
@@ -143,10 +148,10 @@ def add_new_eval(
|
|
| 143 |
},
|
| 144 |
"submitted_time": current_time,
|
| 145 |
"model_type": model_type,
|
| 146 |
-
"likes":
|
| 147 |
"num_params": model_size,
|
| 148 |
"license": license,
|
| 149 |
-
"private":
|
| 150 |
"slurm_id": None
|
| 151 |
}
|
| 152 |
|
|
@@ -158,6 +163,8 @@ def add_new_eval(
|
|
| 158 |
print("Creating eval file")
|
| 159 |
OUT_DIR = f"{EVAL_REQUESTS_PATH}/{user_name}"
|
| 160 |
os.makedirs(OUT_DIR, exist_ok=True)
|
|
|
|
|
|
|
| 161 |
out_path = f"{OUT_DIR}/{model_path}_{revision}_{precision}_{weight_type}_eval_request.json"
|
| 162 |
|
| 163 |
with open(out_path, "w") as f:
|
|
|
|
| 63 |
global USERS_TO_SUBMISSION_DATES
|
| 64 |
if not REQUESTED_MODELS:
|
| 65 |
REQUESTED_MODELS, USERS_TO_SUBMISSION_DATES = already_submitted_models(EVAL_REQUESTS_PATH)
|
| 66 |
+
if model.startswith("/"):
|
| 67 |
+
user_name = ""
|
| 68 |
+
model_path = model
|
| 69 |
+
private = True
|
| 70 |
+
else:
|
| 71 |
+
user_name = ""
|
| 72 |
+
model_path = model
|
| 73 |
+
if "/" in model:
|
| 74 |
+
user_name = model.split("/")[0]
|
| 75 |
+
model_path = model.split("/")[1]
|
| 76 |
+
private = False
|
| 77 |
|
| 78 |
# precision = precision.split(" ")[0]
|
| 79 |
current_time = datetime.now(timezone.utc).strftime("%Y-%m-%dT%H:%M:%SZ")
|
|
|
|
| 101 |
|
| 102 |
# Is the model info correctly filled?
|
| 103 |
try:
|
| 104 |
+
if model.startswith("/"):
|
| 105 |
+
model_info = API.model_info(repo_id=model, revision=revision)
|
| 106 |
+
model_size = get_model_size(model_info=model_info)
|
| 107 |
+
license = model_info.cardData["license"]
|
| 108 |
+
modelcard_OK, error_msg = check_model_card(model)
|
| 109 |
+
if not modelcard_OK:
|
| 110 |
+
return styled_error(error_msg)
|
| 111 |
+
likes = model_info.likes
|
| 112 |
+
else:
|
| 113 |
+
model_size = None
|
| 114 |
+
license = None
|
| 115 |
+
likes = 0
|
| 116 |
except Exception:
|
| 117 |
return styled_error("Could not get your model information. Please fill it up properly.")
|
| 118 |
|
| 119 |
+
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 120 |
# Verify the inference config now
|
| 121 |
# try:
|
| 122 |
# label_normalization_map = ast.literal_eval(label_normalization_map)
|
|
|
|
| 148 |
},
|
| 149 |
"submitted_time": current_time,
|
| 150 |
"model_type": model_type,
|
| 151 |
+
"likes": likes,
|
| 152 |
"num_params": model_size,
|
| 153 |
"license": license,
|
| 154 |
+
"private": private,
|
| 155 |
"slurm_id": None
|
| 156 |
}
|
| 157 |
|
|
|
|
| 163 |
print("Creating eval file")
|
| 164 |
OUT_DIR = f"{EVAL_REQUESTS_PATH}/{user_name}"
|
| 165 |
os.makedirs(OUT_DIR, exist_ok=True)
|
| 166 |
+
if model_path.startswith("/"):
|
| 167 |
+
os.makedirs(f"{OUT_DIR}/{model_path}", exist_ok=True)
|
| 168 |
out_path = f"{OUT_DIR}/{model_path}_{revision}_{precision}_{weight_type}_eval_request.json"
|
| 169 |
|
| 170 |
with open(out_path, "w") as f:
|