-
In #125 @martinholmer said:
> It seems to me the more important Phase 3 issues highlighted by the [examination results](https://github.com/PSLmodels/tax-microdata-benchmarking/blob/master/tax_micro…
-
**Is your feature request related to a problem? Please describe.**
I think it would be better if the colleges are sorted in desc order of their packages
I am a GSSOC'24 Contributor please assign me …
-
**Describe the bug**
CTC finetuning is not converging I tried to change hyperparameters, but still there is no luck. During training the model started to output empty strings.
**Expected beh…
-
hi, I am creating an ASR model on Korean dataset.
My E2E ASR model without language model works really well, wer around 1 on dev and evaluation set.
But When I apply language model fusion (deep an…
-
Hi espnet team,
Thank you for your amazing work on OWSM, this greatly helps the open-source community. Truly grateful for your efforts.
I assume this repo is a proper place to discuss OWSM relat…
wgb14 updated
2 weeks ago
-
In the paper entitled: "Improving Massively Multilingual ASR with Auxiliary CTC Objectives" you proposed a method where early encoder layers where conditioned to identify the language of speech, while…
-
Distributed training, single node, 4 GPUs.
```
...
ep 33 train, step 10, ctc_4 2.259, ctc_8 1.880, ctc 1.840, num_seqs 10, max_size:time 237360, max_size:out-spatial 52, mem_usage:cuda:1 6.6GB, 0…
-
**FEDERAL**
EITC
CTC
**OTHER**
CCDF
TANF
SSI
-
Sometimes we want to "brute force" or preview a code when we aren't sure if we remembered it correctly. There is CTCSS/DCS search, but that does not set the TX code, plus there is currently a bug wher…
-
作者您好,我在您代码中发现了上述两个函数 "ctc_ent_loss_log" 和 “ctc_ent_loss”。从代码实现上,“ctc_ent_loss”更符合直观的认识。我在读"ctc_ent_loss_log"函数时有点吃力,以下是几个问题想请教您:
(1)您为什么选择"ctc_ent_loss_log"函数,是因为网络的output存在负项吗?
(2)能解释下"ctc_ent_loss…