icefall
icefall copied to clipboard
k2-fsa
When initial an encoder like this ``` model = Conformer(80) device = torch.device('cuda') model.to(device) model.device = device ``` All parameters on GPU, But when convert to non scaled one, lke...
Hello, I see that in Tedlium3 recipe you use 500 word pieces. Did you try another numbers? (1000, 2000). Is there some influence on the decoding results? Thanks, AlexG.
The next generation Kaldi is developing rapidly. We have gotten some competitive results in some large and popular datasets based on k2, icefall, and Lhotse. Now, we want to apply...
Hello guys, I was able to run icefall docker on GPU machine according to https://github.com/k2-fsa/icefall/tree/master/docker But I have problems with running the docker on CPU machine. I tried following: docker...
This PR depends on https://github.com/k2-fsa/k2/pull/1057 in k2.
ha
foo) root@iZ2zeh94phzrw5upket4jsZ:~# export PYTHONPATH=/tmp/icefall:$PYTHONPATH (foo) root@iZ2zeh94phzrw5upket4jsZ:~# cd /tmp/icefall (foo) root@iZ2zeh94phzrw5upket4jsZ:/tmp/icefall# cd egs/yesno/ASR (foo) root@iZ2zeh94phzrw5upket4jsZ:/tmp/icefall/egs/yesno/ASR# ./prepare.sh 2022-08-28 15:42:25 (prepare.sh:24:main) dl_dir: /tmp/icefall/egs/yesno/ASR/download 2022-08-28 15:42:25 (prepare.sh:27:main) Stage 0: Download data Downloading waves_yesno.tar.gz: 4.49MB...
Hi it seems this variable is missing from the arguments of the optimizer initialization in train.py
Hello guys, I have a question regarding tedlium3 recipe. I see that during tedlium data preparation, fbank is extracted for tedlium and for musan. Therefore I assume that during the...
Training command is: ```shell ./pruned_transducer_stateless7/train.py \ --world-size 8 \ --num-epochs 90 \ --use-fp16 1 \ --max-duration 200 \ --exp-dir pruned_transducer_stateless7/exp \ --feedforward-dims "1024,1024,2048,2048,1024" \ --master-port 12535 ``` The tensorboard log...
