k2-fsa icefall Q A · Discussions · GitHub

Slow training of pruned transducer is normal?
Q&A RuABraun

Sort by: Latest activity

Categories

Q&A Discussions

Ask the community for help

You must be logged in to vote

Knowledge distillation from a large zipformer2 to a smaller one

joazoa asked Aug 19, 2023 in Q&A · Unanswered

4
You must be logged in to vote

Can transcript_words.txt file contain only unique and sorted lines ?

ChrystianKacki asked May 4, 2024 in Q&A · Closed · Answered

12
You must be logged in to vote

Repository template for model publishing

daniel-dona asked Apr 26, 2024 in Q&A · Answered

1
You must be logged in to vote

Halving BPE vocab size does not reduce mem usage ???

iggygeek asked Apr 17, 2024 in Q&A · Closed · Answered

2
You must be logged in to vote

Data requirements and recommendations for training large-scaled Zipformer2

bharathraj-v asked Apr 3, 2024 in Q&A · Unanswered

0
You must be logged in to vote

Have you tried to initialize the encoder part from a pretrained model ?

iggygeek asked Mar 26, 2024 in Q&A · Answered

5
You must be logged in to vote

am-scale and lm-scale for "Simple RNNT" loss smoothing

YuriiMytiai asked Feb 7, 2024 in Q&A · Answered

4
You must be logged in to vote

Training strategy for Zipformer using fp16 ??

ZQuang2202 asked Jan 14, 2024 in Q&A · Unanswered

2
You must be logged in to vote

Is the zipformer2 supported by ncnn ?

joazoa asked Nov 7, 2023 in Q&A · Answered

2
You must be logged in to vote

Training diverges

RuABraun asked Oct 23, 2023 in Q&A · Unanswered

7
You must be logged in to vote

python Segmentation fault

DeityBoom asked Oct 27, 2023 in Q&A · Unanswered

0
You must be logged in to vote

Shallow fusion or rescoring with word-level n-grams for modified_beam_search

mrsrikanth asked Sep 14, 2023 in Q&A · Unanswered

7
You must be logged in to vote

Bypass Temporal Classification: Weakly Supervised Automatic Speech Recognition with Imperfect Transcripts

joazoa asked Sep 12, 2023 in Q&A · Closed · Unanswered

1
You must be logged in to vote

Finetuning a model to "forget" all but some sentences

joazoa asked Aug 19, 2023 in Q&A · Unanswered

0
You must be logged in to vote

Is it possible to bring back old Kaldi's chain-model style training back to Icefall

davidlin409 asked Apr 12, 2023 in Q&A · Answered

2
You must be logged in to vote

Suggestion neural-network architecture with small parameter size

davidlin409 asked Jan 5, 2023 in Q&A · Closed · Unanswered

2
You must be logged in to vote

Multi-host Training

usoylu2 asked May 23, 2023 in Q&A · Answered

1
You must be logged in to vote

Expected training time

usoylu2 asked May 23, 2023 in Q&A · Answered

1
You must be logged in to vote

Models architecture and params

Cirediallo asked May 23, 2023 in Q&A · Answered

1
You must be logged in to vote

How can we do fine Tuning / Transfer Learning for a new language

joazoa asked Feb 22, 2023 in Q&A · Answered

2
You must be logged in to vote

TLG for k2

homink asked Apr 28, 2023 in Q&A · Answered

16
You must be logged in to vote

Result Gap between CTC and RNN-T Multi-Task Training

teowenshen asked Mar 29, 2023 in Q&A · Closed · Answered

2
You must be logged in to vote

can we training emformer transducer stateless with phoneme base?

trangtv57 asked Sep 26, 2022 in Q&A · Answered

1
You must be logged in to vote

use of external LM to improve transcriptions

iggygeek asked Mar 9, 2023 in Q&A · Answered

6
You must be logged in to vote

ValueError: not enough values to unpack (expected 2, got 0)

efrathason asked Dec 1, 2022 in Q&A · Unanswered

18