Azure AI Speech
An Azure service that integrates speech processing into apps and services.
1,516 questions
Custom neural voice data size is at 0 after training. Should I deploy the model?
![](data:image/svg+xml, %3Csvg xmlns='http://www.w3.org/2000/svg' height='64' class='font-weight-bold' style='font: 600 30.11764705882353px "SegoeUI", Arial' width='64'%3E%3Ccircle fill='hsl(259.20000000000005, 56.00000000000001%, 35%)' cx='32' cy='32' r='32' /%3E%3Ctext x='50%25' y='55%25' dominant-baseline='middle' text-anchor='middle' fill='%23FFF' %3EPP%3C/text%3E%3C/svg%3E)
PAVAGEAU Perrine
80
Reputation points
Hello,
We prepared 1749 utterances in order to create a Custom Neural Voice. In Step 3, which is "Train Model", it identified these 1749 utterances and suggested 25 hours of training time (see image attached).
The training has finished in over 14-15 hours, but the Data Size in the finished report shows "0" (see image attached). Is this something that would be updated after "deploying the model"? Kindly suggest.
{count} votes
-
romungi-MSFT 43,656 Reputation points Microsoft Employee2024-06-10T11:16:17.6833333+00:00 @PAVAGEAU Perrine What does the information placeholder say, next to Data size? I think this used to be the number of utterances used for training, I am not sure if anything changed with the display recently. If the model used 1749 utterances, I think it should display the corresponding data size or the no. of utterances used.
-
PAVAGEAU Perrine 80 Reputation points
2024-06-11T10:56:52.9033333+00:00 @romungi-MSFT thank you for responding!
The info placeholder indicates the data size is supposed to be "the total number of Utterances used to train the model"
Additionally, the "testing" phase was successful with 100 utterances and you can also see the utterance data get populated under "training set", all while the total utterances used for the full model training remains 0 (see image attachments for both).
This is exactly similar to how we trained other datasets, so I'm not sure why the Dataset this time has recorded 0 utterances.
Could it have something to do with the "engine version"?
Sign in to comment