Spaces:
Sleeping
Sleeping
Adding progress bar
Browse files
app.py
CHANGED
@@ -26,7 +26,7 @@ from sentence_transformers import SentenceTransformer
|
|
26 |
from dotenv import load_dotenv
|
27 |
import os
|
28 |
|
29 |
-
import spaces
|
30 |
import gradio as gr
|
31 |
|
32 |
|
@@ -132,13 +132,13 @@ def get_docs_from_parquet(parquet_urls, column, offset, limit):
|
|
132 |
return df[column].tolist()
|
133 |
|
134 |
|
135 |
-
@spaces.GPU
|
136 |
# TODO: Modify batch size to reduce memory consumption during embedding calculation, which value is better?
|
137 |
def calculate_embeddings(docs):
|
138 |
return sentence_model.encode(docs, show_progress_bar=True, batch_size=32)
|
139 |
|
140 |
|
141 |
-
@spaces.GPU
|
142 |
def fit_model(docs, embeddings):
|
143 |
global global_topic_model
|
144 |
|
@@ -177,6 +177,11 @@ def generate_topics(dataset, config, split, column, nested_column):
|
|
177 |
all_docs = []
|
178 |
reduced_embeddings_list = []
|
179 |
topics_info, topic_plot = None, None
|
|
|
|
|
|
|
|
|
|
|
180 |
while offset < limit:
|
181 |
docs = get_docs_from_parquet(parquet_urls, column, offset, chunk_size)
|
182 |
if not docs:
|
@@ -220,14 +225,23 @@ def generate_topics(dataset, config, split, column, nested_column):
|
|
220 |
)
|
221 |
|
222 |
logging.info(f"Topics: {repr_model_topics}")
|
|
|
223 |
|
224 |
-
yield
|
|
|
|
|
|
|
|
|
225 |
|
226 |
offset += chunk_size
|
227 |
|
228 |
logging.info("Finished processing all data")
|
229 |
cuda.empty_cache() # Clear cache at the end of each chunk
|
230 |
-
return
|
|
|
|
|
|
|
|
|
231 |
|
232 |
|
233 |
with gr.Blocks() as demo:
|
@@ -267,6 +281,7 @@ with gr.Blocks() as demo:
|
|
267 |
generate_button = gr.Button("Generate Topics", variant="primary")
|
268 |
|
269 |
gr.Markdown("## Datamap")
|
|
|
270 |
topics_plot = gr.Plot()
|
271 |
with gr.Accordion("Topics Info", open=False):
|
272 |
topics_df = gr.DataFrame(interactive=False, visible=True)
|
@@ -279,7 +294,7 @@ with gr.Blocks() as demo:
|
|
279 |
text_column_dropdown,
|
280 |
nested_text_column_dropdown,
|
281 |
],
|
282 |
-
outputs=[topics_df, topics_plot],
|
283 |
)
|
284 |
|
285 |
def _resolve_dataset_selection(
|
|
|
26 |
from dotenv import load_dotenv
|
27 |
import os
|
28 |
|
29 |
+
# import spaces
|
30 |
import gradio as gr
|
31 |
|
32 |
|
|
|
132 |
return df[column].tolist()
|
133 |
|
134 |
|
135 |
+
# @spaces.GPU
|
136 |
# TODO: Modify batch size to reduce memory consumption during embedding calculation, which value is better?
|
137 |
def calculate_embeddings(docs):
|
138 |
return sentence_model.encode(docs, show_progress_bar=True, batch_size=32)
|
139 |
|
140 |
|
141 |
+
# @spaces.GPU
|
142 |
def fit_model(docs, embeddings):
|
143 |
global global_topic_model
|
144 |
|
|
|
177 |
all_docs = []
|
178 |
reduced_embeddings_list = []
|
179 |
topics_info, topic_plot = None, None
|
180 |
+
yield (
|
181 |
+
gr.DataFrame(interactive=False, visible=True),
|
182 |
+
gr.Plot(visible=True),
|
183 |
+
gr.Label({f"⚙️ Generating topics {dataset}": 0.0}, visible=True),
|
184 |
+
)
|
185 |
while offset < limit:
|
186 |
docs = get_docs_from_parquet(parquet_urls, column, offset, chunk_size)
|
187 |
if not docs:
|
|
|
225 |
)
|
226 |
|
227 |
logging.info(f"Topics: {repr_model_topics}")
|
228 |
+
progress = min(offset / limit, 1.0)
|
229 |
|
230 |
+
yield (
|
231 |
+
topics_info,
|
232 |
+
topic_plot,
|
233 |
+
gr.Label({f"⚙️ Generating topics {dataset}": progress}, visible=True),
|
234 |
+
)
|
235 |
|
236 |
offset += chunk_size
|
237 |
|
238 |
logging.info("Finished processing all data")
|
239 |
cuda.empty_cache() # Clear cache at the end of each chunk
|
240 |
+
return (
|
241 |
+
topics_info,
|
242 |
+
topic_plot,
|
243 |
+
gr.Label({f"⚙️ Generating topics {dataset}": 1.0}, visible=True),
|
244 |
+
)
|
245 |
|
246 |
|
247 |
with gr.Blocks() as demo:
|
|
|
281 |
generate_button = gr.Button("Generate Topics", variant="primary")
|
282 |
|
283 |
gr.Markdown("## Datamap")
|
284 |
+
full_topics_generation_label = gr.Label(visible=False, show_label=False)
|
285 |
topics_plot = gr.Plot()
|
286 |
with gr.Accordion("Topics Info", open=False):
|
287 |
topics_df = gr.DataFrame(interactive=False, visible=True)
|
|
|
294 |
text_column_dropdown,
|
295 |
nested_text_column_dropdown,
|
296 |
],
|
297 |
+
outputs=[topics_df, topics_plot, full_topics_generation_label],
|
298 |
)
|
299 |
|
300 |
def _resolve_dataset_selection(
|