Infra Flashcards
What is horovod? How does it works?
What is cycle cloud?
Personal cluster orchestrate a Linux environment with different cpus scheduling jobs
How can you look for data and results of the cloud works?
With the azure data storage
What is an nsf blob storage?
NFS Network File System
BLOB STORAGE:
Use on A Zune To Store BIG
DATA AND’ SECURE THEM
BLOB (BINARY LARGE OBJECT)
VERY GOOD FOR MEDIA AND
CHANGING LOGS
BLOBS AN STORED IN CONTAINERS
(SIMILAR TO FOLDERS)
What is panini?
Where and when does rescoring happen in our ASR decoding?
Rescoring is a panini model that happens on the output of the decoder, like the one you find on the .json output of the unidec_r
Where can you find the rescoring output in pdecode?
What is the difference between the TiP and PPE environments?
TiP (Test in Production) is an unstable test environment, while PPE (Pre-Production Environment) is a stable test environment with two regions.
What does the backend team manage? Tolking about Ivo’s team
The backend team manages the APIs.
What should be done before whitelisting a subscription for TiP deployment?
Before whitelisting a subscription, create a project in Speech Studio to ensure it is added to the TiP database during the initial request.
Before whitelisting a subscription, create a project in Speech Studio to ensure it is added to the TiP database during the initial request.
Before whitelisting a subscription, create a project in Speech Studio to ensure it is added to the TiP database during the initial request.
Before whitelisting a subscription, create a project in Speech Studio to ensure it is added to the TiP database during the initial request.
f a subscription is not created in WUS2, the subscription ID won’t be added to the database, causing issues with locating the subscription.
Where are the output of the segmentation service located?
in spsvctestusw2 under custom speech artifacts, acoustic data there you have a folder with the returned id.
Once you have the ID of the segmentation service datasets, how can you see that i the database?
In the context of changing the behaviour of the ASR model what are reformulation rules? where is the file located?
They are simple rules that map one answer to the predicted corrected one. The file is in the display-post-processing (not in the ITN model in LMTools)
