-
I did experiments on wsj0_mix dataset but got worse results with the current recipe. I tried to change the default parameters (different initialization seeds and different scheduled sampling probs) bu…
-
I run the HKUST and Librispeech recipes, but achieved the performances far behind the published ones. I didn't use rnnlm since it costs much time. I wander whether it is normal to get there results s…
-
Hi @rmiglior,
I see that there are now a couple of E-Models appearing on the Single Cell category
As you can see here:
![Screenshot 2019-03-28 at 10 54 38](https://user-images.githubusercontent.…
-
I used the default parameters of the script without training a language model because I don't have wsj1 transcriptions. I set the language model weight to be 0 and use spa. I used 1 gpu throughout the…
-
**What would you like to be added**:
supports other modules to subscribe the message from edge
**Why is this needed**:
I want to extend a module to handle the message from the edge. But I
f…
-
There's something wrong with the json IO format in the decoding step. I checked the generated json in the tmp directory. The generated result_r*h*.wrd.json files are normal, but the format of all_resu…
-
When I user docker to run tedlium asr1 recipe
```
./run.sh --docker_gpu 0,1,2,3 --docker_egs tedlium/asr1 --ngpu 4
```
I got some bugs in stage 5
```
stage 5: Decoding
2019-04-22 17:47:02,929 (…
fanlu updated
5 years ago
-
I have been testing the docker containers with the AN4/ASR1 recipe before uploading them, and found that the backend chainer is not working, neither in cpu or gpu mode. So I tested it without docker.…
-
I am analyzing a python project and srccheck is not detecting abstractness in my project.
Understand detects the class abstract, from ABC, and methods are annotated as abstractmethods,
![evid…
-
## Details description of the task
We moved all the optimized models (including all the releases) from the github repository to the following container
https://object.cscs.ch/v1/AUTH_c0a333ecf7c04…