search

NVIDIA / NeMo

A scalable generative AI framework built for researchers and developers working on Large Language Models, Multimodal, and Speech AI (Automatic Speech Recognition and Text-to-Speech)
https://docs.nvidia.com/nemo-framework/user-guide/latest/overview.html
Apache License 2.0
11.5k stars 2.41k forks source link

Combining Nemo with Pre-existing Classification Models #249

Closed eanmikale closed 3 years ago

eanmikale commented 4 years ago

Hello! Great work here. We'd like to know if you could guide us toward the proper resources concerning how to combine NeMo with pre-existing Image Classification or even Object Detection models. Thank you for your time.

okuchaiev commented 4 years ago

currently we only have "conversational AI" collections: nemo_asr, nemo_nlp and nemo_tts. But you can start your own with image classification modules

eanmikale commented 4 years ago

Hi! I'm sorry I am still confused. We already have image classification models. We would like to integrate these models with conversational A.I. using nemo. Is there a way to do so using the current modules? Or would we need to utilize other tools outside of the software suite? Thank you.

Warm Regards,

Ean Mikale, J.D.

Chief A.I. Architect

Infinite 8 Institute, L3C

1111 N 13th Street #417

Omaha, NE 68102

1-(202)-492-6119

mailto:ean@infinite8institute.com

http://www.infinite8institute.com

Accelerated technology for tomorrow, today.

---- On Wed, 01 Jan 2020 17:39:32 -0600 Oleksii Kuchaiev notifications@github.com wrote ----

currently we only have "conversational AI" collections: nemo_asr, nemo_nlp and nemo_tts. But you can start your own with image classification modules — You are receiving this because you authored the thread. Reply to this email directly, https://github.com/NVIDIA/NeMo/issues/249?email_source=notifications&email_token=AGCCBHWNHHY4RRVMOE7EYTTQ3USTJA5CNFSM4KBWIO3KYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEH5O4SQ#issuecomment-570093130, or https://github.com/notifications/unsubscribe-auth/AGCCBHTXRNBZTR4UYFUGL33Q3USTJANCNFSM4KBWIO3A.

okuchaiev commented 4 years ago

You would need to add your own Neural Modules (see https://nvidia.github.io/NeMo/tutorials/custommodules.html). You can put them into your own collection under your own license terms. If your existing image classification models are in PyTorch it should be relatively easy to convert then into neural modules

okuchaiev commented 3 years ago

closing as this is related to the old version