Remove unused debug print. Add addition check for request files
Browse files- app.py +0 -10
- src/leaderboard/read_evals.py +8 -1
app.py
CHANGED
@@ -134,16 +134,12 @@ def filter_queries(query: str, filtered_df: pd.DataFrame) -> pd.DataFrame:
|
|
134 |
def filter_models(
|
135 |
df: pd.DataFrame, type_query: list, size_query: list, precision_query: list, show_deleted: bool
|
136 |
) -> pd.DataFrame:
|
137 |
-
print("Initial number of models:", len(df))
|
138 |
-
|
139 |
# Show all models
|
140 |
if show_deleted:
|
141 |
filtered_df = df
|
142 |
else:
|
143 |
filtered_df = df[df[AutoEvalColumn.still_on_hub.name] == True]
|
144 |
|
145 |
-
print("After hub filter:", len(filtered_df))
|
146 |
-
|
147 |
if "All" not in type_query:
|
148 |
if "?" in type_query:
|
149 |
filtered_df = filtered_df.loc[~df[AutoEvalColumn.model_type_symbol.name].isin([t for t in ModelType if t != "?"])]
|
@@ -151,16 +147,12 @@ def filter_models(
|
|
151 |
type_emoji = [t[0] for t in type_query]
|
152 |
filtered_df = filtered_df.loc[df[AutoEvalColumn.model_type_symbol.name].isin(type_emoji)]
|
153 |
|
154 |
-
print("After type filter:", len(filtered_df))
|
155 |
-
|
156 |
if "All" not in precision_query:
|
157 |
if "?" in precision_query:
|
158 |
filtered_df = filtered_df.loc[df[AutoEvalColumn.precision.name].isna()]
|
159 |
else:
|
160 |
filtered_df = filtered_df.loc[df[AutoEvalColumn.precision.name].isin(precision_query + ["None"])]
|
161 |
|
162 |
-
print("After precision filter:", len(filtered_df))
|
163 |
-
|
164 |
if "All" not in size_query:
|
165 |
if "?" in size_query:
|
166 |
filtered_df = filtered_df.loc[df[AutoEvalColumn.params.name].isna()]
|
@@ -170,8 +162,6 @@ def filter_models(
|
|
170 |
mask = params_column.apply(lambda x: any(numeric_interval.contains(x)))
|
171 |
filtered_df = filtered_df.loc[mask]
|
172 |
|
173 |
-
print("After size filter:", len(filtered_df))
|
174 |
-
|
175 |
return filtered_df
|
176 |
|
177 |
|
|
|
134 |
def filter_models(
|
135 |
df: pd.DataFrame, type_query: list, size_query: list, precision_query: list, show_deleted: bool
|
136 |
) -> pd.DataFrame:
|
|
|
|
|
137 |
# Show all models
|
138 |
if show_deleted:
|
139 |
filtered_df = df
|
140 |
else:
|
141 |
filtered_df = df[df[AutoEvalColumn.still_on_hub.name] == True]
|
142 |
|
|
|
|
|
143 |
if "All" not in type_query:
|
144 |
if "?" in type_query:
|
145 |
filtered_df = filtered_df.loc[~df[AutoEvalColumn.model_type_symbol.name].isin([t for t in ModelType if t != "?"])]
|
|
|
147 |
type_emoji = [t[0] for t in type_query]
|
148 |
filtered_df = filtered_df.loc[df[AutoEvalColumn.model_type_symbol.name].isin(type_emoji)]
|
149 |
|
|
|
|
|
150 |
if "All" not in precision_query:
|
151 |
if "?" in precision_query:
|
152 |
filtered_df = filtered_df.loc[df[AutoEvalColumn.precision.name].isna()]
|
153 |
else:
|
154 |
filtered_df = filtered_df.loc[df[AutoEvalColumn.precision.name].isin(precision_query + ["None"])]
|
155 |
|
|
|
|
|
156 |
if "All" not in size_query:
|
157 |
if "?" in size_query:
|
158 |
filtered_df = filtered_df.loc[df[AutoEvalColumn.params.name].isna()]
|
|
|
162 |
mask = params_column.apply(lambda x: any(numeric_interval.contains(x)))
|
163 |
filtered_df = filtered_df.loc[mask]
|
164 |
|
|
|
|
|
165 |
return filtered_df
|
166 |
|
167 |
|
src/leaderboard/read_evals.py
CHANGED
@@ -38,8 +38,9 @@ class EvalResult:
|
|
38 |
with open(json_filepath) as fp:
|
39 |
data = json.load(fp)
|
40 |
|
41 |
-
|
42 |
|
|
|
43 |
# Precision
|
44 |
precision = Precision.from_str(config.get("model_dtype"))
|
45 |
|
@@ -82,6 +83,8 @@ class EvalResult:
|
|
82 |
mean_acc = np.mean(accs) * 100.0
|
83 |
results[task.benchmark] = mean_acc
|
84 |
|
|
|
|
|
85 |
return self(
|
86 |
eval_name=result_key,
|
87 |
full_model=full_model,
|
@@ -95,6 +98,7 @@ class EvalResult:
|
|
95 |
model_type=model_type
|
96 |
)
|
97 |
|
|
|
98 |
def update_with_request_file(self, requests_path):
|
99 |
"""Finds the relevant request file for the current model and updates info with it"""
|
100 |
request_file = get_request_file_for_model(requests_path, self.full_model, self.precision.value.name)
|
@@ -176,6 +180,8 @@ def get_raw_eval_results(results_path: str, requests_path: str) -> list[EvalResu
|
|
176 |
for file in files:
|
177 |
model_result_filepaths.append(os.path.join(root, file))
|
178 |
|
|
|
|
|
179 |
eval_results = {}
|
180 |
for model_result_filepath in model_result_filepaths:
|
181 |
# Creation of result
|
@@ -197,4 +203,5 @@ def get_raw_eval_results(results_path: str, requests_path: str) -> list[EvalResu
|
|
197 |
except KeyError: # not all eval values present
|
198 |
continue
|
199 |
|
|
|
200 |
return results
|
|
|
38 |
with open(json_filepath) as fp:
|
39 |
data = json.load(fp)
|
40 |
|
41 |
+
print(f"Processing file: {json_filepath}")
|
42 |
|
43 |
+
config = data.get("config")
|
44 |
# Precision
|
45 |
precision = Precision.from_str(config.get("model_dtype"))
|
46 |
|
|
|
83 |
mean_acc = np.mean(accs) * 100.0
|
84 |
results[task.benchmark] = mean_acc
|
85 |
|
86 |
+
print(f"Model: {model}, Org: {org}, Results: {results.keys()}")
|
87 |
+
|
88 |
return self(
|
89 |
eval_name=result_key,
|
90 |
full_model=full_model,
|
|
|
98 |
model_type=model_type
|
99 |
)
|
100 |
|
101 |
+
|
102 |
def update_with_request_file(self, requests_path):
|
103 |
"""Finds the relevant request file for the current model and updates info with it"""
|
104 |
request_file = get_request_file_for_model(requests_path, self.full_model, self.precision.value.name)
|
|
|
180 |
for file in files:
|
181 |
model_result_filepaths.append(os.path.join(root, file))
|
182 |
|
183 |
+
print(f"Found {len(model_result_filepaths)} JSON files to process.")
|
184 |
+
|
185 |
eval_results = {}
|
186 |
for model_result_filepath in model_result_filepaths:
|
187 |
# Creation of result
|
|
|
203 |
except KeyError: # not all eval values present
|
204 |
continue
|
205 |
|
206 |
+
print(f"Successfully loaded {len(results)} models.")
|
207 |
return results
|