-
**Problem with decode result on SPGISpeech dataset**
Hi, I downloaded the pretrained model from [https://zenodo.org/record/4585546](https://zenodo.org/record/4585546) and inference with differe…
-
你好,请教一个问题:
比如我在推理的时候设置batch_size=8,指的是在处理一条语音时,同时解码的token数,还是同时进行8条语音推理?
-
### System Info
Framework: Vuejs 3
Transformer.js version: 2.17.1
### Environment/Platform
- [X] Website/web-app
- [ ] Browser extension
- [ ] Server-side (e.g., Node.js, Deno, Bun)
- [ ] Deskt…
-
非常感谢你们的开源工作,给了我很大的帮助。关于新发布的评估部分,我有几个问题请教下,期待你们的回复:
1.为什么复现结果有差距?librispeech test clean这个集合我离线测试wer是1.73左右,官方结果是1.6
2.为什么qwen-audio和qwen2-audio在同样的wer评估准则下效果差不多?我把qwen-audio的tn部分换成qwen2-audio的代码,两者we…
fzhml updated
1 month ago
-
I've been using the ESPHome variant for a few days and today set up an automation to press M1 / M2 every 30 minutes. After some time it failed and the display showed "ASR".
Couldn't control anythin…
-
Hello All,
we are training a zip former model for about 3400 hours of Tamil data.
We were facing this issue:
RuntimeError:
grad_scale is too small, exiting: 1.4901161193847656e-08
…
-
I try to reconstruct ancestral sequence of protein. I use the parameters under (i cut protein for you to see better) :
begin data;
dimensions ntax=4 nchar=1057;
format datatype=protein missing…
-
Currently with have `ptr_loads` all over the LLVM backend (345 times!) that determines if an LLVM register has to be loaded first to access the value (and many other cases where we use ASR types direc…
-
Hello,
When installed on a local Ubuntu WSL under windows 11,
It currently works fine with --cpu_mode but throws "RuntimeError: CUDA failed with error out of memory" in standard mode. Unable to us…
-
Most likely this has to be a new `expr` node, something like `RealNearest`. Let's first figure out an LLVM (and WASM) instructions to compute it, as well as x86/arm assembly instructions, so that we k…