# tacotron-2-support
- u
{K EY1} (Kei)
08/22/2022, 11:32 PMI don't have training graphs in my notebook - u
{K EY1} (Kei)
08/22/2022, 11:32 PMWait Maybe the issue was the batch size was set to 6 Could that cause problems in such a huge dataset? - c
Cris140
08/22/2022, 11:32 PMIt looks like the model didn't learn - c
Cris140
08/22/2022, 11:33 PMI think so, I've used 18 to train a huge Portuguese dataset - u
{K EY1} (Kei)
08/22/2022, 11:33 PMSo 6 is too low and it didn't work - u
{K EY1} (Kei)
08/23/2022, 12:29 AMHow should i go about training this again then? Should i still use my notebook - c
Cris140
08/23/2022, 1:19 AMDo you want to use the model privately? - c
Cris140
08/23/2022, 1:32 AMIf yes, you can do a fork using pipeline or default tt2 - c
Cris140
08/23/2022, 1:33 AMThis way you can do it with any language that espeak has available - u
{K EY1} (Kei)
08/23/2022, 1:46 AMI want to release the model to the public - a
Alexis
08/23/2022, 3:28 PMHello ! I downloaded Cuda and everything by following this tutorial https://uberduck-guide.gitbook.io/uberduck-written-tutorial/training-time/setting-up-for-local-training-wip but tacotron use my CPU instead of my GPU. Is there something to do to activate it ? - c
Cris140
08/23/2022, 7:43 PMinstall pytorch using CUDA - c
Cris140
08/23/2022, 7:43 PMconda install pytorch torchvision torchaudio cudatoolkit=11.3 -c pytorch if you're using anaconda, you can use this command - a
Alexis
08/24/2022, 1:50 PMHello, yes I did that but it is still running on the cpu - c
Cris140
08/24/2022, 1:53 PMCan you type nvidia-smi on the prompt and send what it shows? - a
Alexis
08/24/2022, 3:56 PMYes, here it is - a
Alexis
08/24/2022, 3:57 PMAnd I have all the process using it on the bottom (not on the screenshot) - d
Dyno
08/24/2022, 3:57 PMRead #answers, @Alexis. - c
Cris140
08/24/2022, 4:36 PMThat's strange, are you able to synthesize using GPU? - a
Alexis
08/24/2022, 4:57 PMFor the moment I didn't synthesize anything - j
Justin
08/24/2022, 7:29 PM - t
the user
08/25/2022, 4:16 AMWhat is a good validation loss for a model with 8 WAVs and 47 seconds of data? (Training on Pipeline) - u
{K EY1} (Kei)
08/25/2022, 4:08 PM0.07 - u
(Dawn) Will Draw Fictional Women
08/25/2022, 4:59 PMDO LR DECAY - u
(Dawn) Will Draw Fictional Women
08/25/2022, 4:59 PMDAAAAAH - a
Alexis
08/26/2022, 12:15 AMStill can't running on the gpu. The only thing I have in difference with the tutorial is the version of the cuda toolkit (11.7 vs 11.6 on the tutorial). I don't know if it can be the reason why it is only launching on my cpu - n
Nesggy (leaving Disсord)
08/26/2022, 9:03 PMi still get this problem when synthezising - n
Nesggy (leaving Disсord)
08/26/2022, 9:03 PMI did the whole training again from scratch for the 3rd time - n
Nesggy (leaving Disсord)
08/26/2022, 9:04 PMall that and then at the very end of the process, I still get that error