-
paddlepaddle-engine
你好,你提供的asr_sdk使用的是paddlepaddle-engine运行
```
ai.djl.paddlepaddle
paddlepaddle-engine
${djl.version}
```
…
-
## 重现步骤
1. 启动服务端
```
paddlespeech_server start --config_file config.yaml
```
- 配置文件:
```
host: 0.0.0.0
port: 5000
protocol: 'http'
engine_list: ['asr_python', 't…
-
python==3.10
paddlepaddle==2.5.0
paddlespeech=1.4.1
错误信息1:
```
KeyError Traceback (most recent call last)
Cell In[4], line 11
4 # 读取wav
5 # 格式要求…
-
Hi all,
The reason I've been slow on convnet-benchmarks these days is because i've been working on the side on DeepMark.
I initially wrote _convnet-benchmarks_ to increase competition among framewor…
-
when i run your code
it happened RAM OOM in eval part
i don't know why happened this problem?
my desktop ram size is 128GB and using 4-ways gpu
and it was increase memory every eval batch
also 4…
-
----------- Configuration Arguments -----------
check_task: image_classification
debug: False
fluid_model: /home/work/luqiankun/tools_for_renfeng/qt_batchlist_31w_bak/params/
fluid_model_name:
f…
-
- Train an **Engish** language model (Kneser-Ney smoothed 5-gram, with pruning), with [KenLM toolkit](https://kheafield.com/papers/edinburgh/estimate_paper.pdf), on cleaned text from the [Common Crawl…
-
Hello @saghul @nikvaessen!
Recently came across Jitsi while researching GSOC projects. Love the platform! Wanted to contribute to the real time speech to text feature. The options considered:
- …
-
When I train to epoch 3, the following errors have been shown. Could you help? thanks.
> ===========master initModelParams=================
> fail to restore model params from /data1/asr/savemode…
-
conda虚拟环境python=3.8
在examples/aishell/asr0下 首先第一步处理数据完成
然后执行第二步骤训练时,命令:bash run.sh --gpus 0 --stage 1 --stop_stage 1
报错信息如下:
checkpoint name deepspeech2
using 1 gpus...
LAUNCH INFO 2023-08…