File size: 3,235 Bytes
9346f1c 4ccccc6 4505a7e 4ccccc6 614ee1f 4ccccc6 1f60a20 4ccccc6 4b4926f 4ccccc6 818f024 4ccccc6 818f024 4ccccc6 818f024 4ccccc6 f3b890b 4ccccc6 8daa060 4ccccc6 8a21df1 4ccccc6 8daa060 10f9b3c 4ccccc6 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 |
import gradio as gr
import pandas as pd
from pathlib import Path
from src.asr.build_dataframe import per_dataset_tables, overall_table, list_corpora
RESULTS_DIR = Path(__file__).parent / "asr_data" / "results"
CSS = """
#df * { white-space: nowrap; } /* keep single-line model names */
.gradio-container { max-width: 1200px !important; }
th, td { font-size: 14px; }
"""
def get_overall_df() -> pd.DataFrame:
try:
df = overall_table(str(RESULTS_DIR))
return df.fillna("")
except Exception as e:
return pd.DataFrame({"Error": [f"{type(e).__name__}: {e}"]})
def get_per_corpus_df(corpus: str) -> pd.DataFrame:
try:
tables = per_dataset_tables(str(RESULTS_DIR))
df = tables.get(corpus, pd.DataFrame())
return df.fillna("")
except Exception as e:
return pd.DataFrame({"Error": [f"{type(e).__name__}: {e}"]})
with gr.Blocks(css=CSS) as demo:
gr.Markdown("# 🇱🇹 Lithuanian ASR Leaderboard (Local Results)")
with gr.Tab("Overall"):
df_overall = gr.Dataframe(
value=get_overall_df(),
interactive=False,
elem_id="df",
label="Overall (averages)",
wrap=False,
)
gr.Button("Reload overall").click(lambda: get_overall_df(), outputs=df_overall)
# One tab per dataset
corpora = list_corpora(str(RESULTS_DIR))
for corpus in corpora:
with gr.Tab(corpus.upper()):
df_c = gr.Dataframe(
value=get_per_corpus_df(corpus),
interactive=False,
elem_id="df",
label=f"{corpus.upper()} (WER, CER)",
wrap=False,
)
gr.Button(f"Reload {corpus}").click(lambda c=corpus: get_per_corpus_df(c), outputs=df_c)
gr.Markdown(
"""
### Evaluation
- GPU: 1 × NVIDIA GeForce RTX 4090
### 📚 Models
- [DrishtiSharma/whisper-large-v2-lithuanian](https://huggingface.co/DrishtiSharma/whisper-large-v2-lithuanian)
- [meldynamics/whisper-large-v2-lithuanian-ggml](https://huggingface.co/meldynamics/whisper-large-v2-lithuanian-ggml)
- [sam8000/whisper-large-v3-turbo-lithuanian-lithuania-0.06-cer-filtered](https://huggingface.co/sam8000/whisper-large-v3-turbo-lithuanian-lithuania-0.06-cer-filtered)
- [meldynamics/whisper-large-v3-turbo-lithuanian-lithuania-0.06-cer-filtered-ggml](https://huggingface.co/meldynamics/whisper-large-v3-turbo-lithuanian-lithuania-0.06-cer-filtered-ggml)
- [Aismantas/whisper-base-lithuanian](https://huggingface.co/Aismantas/whisper-base-lithuanian)
- [meldynamics/whisper-base-lithuanian-ggml](https://huggingface.co/meldynamics/whisper-base-lithuanian-ggml)
### 🗂 Datasets
- [Lithuania 0.06 CER Filtered](https://huggingface.co/datasets/sam8000/lithuania)
- [Liepa-2](https://huggingface.co/datasets/isLucid/liepa-2)
- [Common Voice 13.0 (Lithuanian subset)](https://huggingface.co/datasets/mozilla-foundation/common_voice_13_0)
- [Common Voice 17.0 (Lithuanian subset)](https://huggingface.co/datasets/mozilla-foundation/common_voice_17_0)
---
*Tip:* add new JSON files to `asr_data/results/` and click **Reload** to refresh the tables.
"""
)
if __name__ == "__main__":
demo.launch()
|