1. add a new streamlit ui to preview manifest
2. implement rpcy transcription client for files
parent
ae5586be72
commit
fa89775f86
|
|
@ -2,6 +2,10 @@ import os
|
|||
import logging
|
||||
import rpyc
|
||||
from functools import lru_cache
|
||||
import typer
|
||||
from pathlib import Path
|
||||
|
||||
app = typer.Typer()
|
||||
|
||||
logging.basicConfig(
|
||||
level=logging.INFO, format="%(asctime)s - %(name)s - %(levelname)s - %(message)s"
|
||||
|
|
@ -19,3 +23,28 @@ def transcribe_gen(asr_host=ASR_HOST, asr_port=ASR_PORT):
|
|||
asr = rpyc.connect(asr_host, asr_port).root
|
||||
logger.info(f"connected to asr server successfully")
|
||||
return asr.transcribe
|
||||
|
||||
|
||||
@app.command()
|
||||
def transcribe_file(audio_file: Path):
|
||||
from pydub import AudioSegment
|
||||
|
||||
transcriber = transcribe_gen()
|
||||
aud_seg = (
|
||||
AudioSegment.from_file_using_temporary_files(audio_file)
|
||||
.set_channels(1)
|
||||
.set_sample_width(2)
|
||||
.set_frame_rate(24000)
|
||||
)
|
||||
tscript_file_path = audio_file.with_suffix(".txt")
|
||||
transcription = transcriber(aud_seg.raw_data)
|
||||
with open(tscript_file_path, "w") as tf:
|
||||
tf.write(transcription)
|
||||
|
||||
|
||||
def main():
|
||||
app()
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
main()
|
||||
|
|
|
|||
|
|
@ -225,6 +225,12 @@ class ExtendedPath(type(Path())):
|
|||
with self.open("r") as jf:
|
||||
return json.load(jf)
|
||||
|
||||
def read_jsonl(self):
|
||||
print(f"reading jsonl from {self}")
|
||||
with self.open("r") as jf:
|
||||
for l in jf.readlines():
|
||||
yield json.loads(l)
|
||||
|
||||
def write_json(self, data):
|
||||
print(f"writing json to {self}")
|
||||
self.parent.mkdir(parents=True, exist_ok=True)
|
||||
|
|
|
|||
|
|
@ -0,0 +1,58 @@
|
|||
from pathlib import Path
|
||||
|
||||
import streamlit as st
|
||||
import typer
|
||||
from jasper.data.utils import ExtendedPath
|
||||
from jasper.data.validation.st_rerun import rerun
|
||||
|
||||
app = typer.Typer()
|
||||
|
||||
if not hasattr(st, "mongo_connected"):
|
||||
# st.task_id = str(uuid4())
|
||||
task_path = ExtendedPath("preview.lck")
|
||||
|
||||
def current_cursor_fn():
|
||||
return task_path.read_json()["current_cursor"]
|
||||
|
||||
def update_cursor_fn(val=0):
|
||||
task_path.write_json({"current_cursor": val})
|
||||
rerun()
|
||||
|
||||
st.get_current_cursor = current_cursor_fn
|
||||
st.update_cursor = update_cursor_fn
|
||||
st.mongo_connected = True
|
||||
# cursor_obj = mongo_conn.find_one({"type": "current_cursor", "task_id": st.task_id})
|
||||
# if not cursor_obj:
|
||||
update_cursor_fn(0)
|
||||
|
||||
|
||||
@st.cache()
|
||||
def load_ui_data(validation_ui_data_path: Path):
|
||||
typer.echo(f"Using validation ui data from {validation_ui_data_path}")
|
||||
return list(ExtendedPath(validation_ui_data_path).read_jsonl())
|
||||
|
||||
|
||||
@app.command()
|
||||
def main(manifest: Path):
|
||||
asr_data = load_ui_data(manifest)
|
||||
sample_no = st.get_current_cursor()
|
||||
if len(asr_data) - 1 < sample_no or sample_no < 0:
|
||||
print("Invalid samplno resetting to 0")
|
||||
st.update_cursor(0)
|
||||
sample = asr_data[sample_no]
|
||||
st.title(f"ASR Manifest Preview")
|
||||
st.markdown(f"{sample_no+1} of {len(asr_data)} : **{sample['text']}**")
|
||||
new_sample = st.number_input(
|
||||
"Go To Sample:", value=sample_no + 1, min_value=1, max_value=len(asr_data)
|
||||
)
|
||||
if new_sample != sample_no + 1:
|
||||
st.update_cursor(new_sample - 1)
|
||||
st.sidebar.markdown(f"Gold Text: **{sample['text']}**")
|
||||
st.audio(Path(sample["audio_filepath"]).open("rb"))
|
||||
|
||||
|
||||
if __name__ == "__main__":
|
||||
try:
|
||||
app()
|
||||
except SystemExit:
|
||||
pass
|
||||
Loading…
Reference in New Issue