AI Models - Computer Vision, Conversational AI, and More

NGC | Catalog

Many AI applications have common needs: classification, object detection, language translation, text-to-speech, recommender engines, sentiment analysis, and more. When developing applications with these capabilities, it is much faster to start with a model that is pre-trained and then tune it for a specific use case. The NGC catalog offers pre-trained models for a variety of common AI tasks that are optimized for NVIDIA Tensor Core GPUs, and can be easily re-trained by updating just a few layers, saving valuable time.

Sort: Last Modified

RIVA EnglishUS Hifigan

Model

Riva multisepaker with IPA for G2P

RIVA EnglishUS Fastpitch

Model

Riva multisepaker with IPA for G2P

DLRM checkpoint (TensorFlow2, TF32, BS64k, Base, FL15)

Model

DLRM TensorFlow2 checkpoint trained on Criteo Dataset with FreqLimit=15 on A100 with TF32

BERT PaddlePaddle checkpoint (Large, Pretraining, AMP, LAMB)

Model

BERT Large PaddlePaddle checkpoint pretrained with LAMB optimizer using AMP

BERT PaddlePaddle checkpoint (Large, QA, SQuAD1.1, AMP)

Model

BERT-Large PaddlePaddle checkpoint finetuned for QA on SQuAD v1.1

Riva ASR Mandarin LM

Model

Base Mandarin 4-gram LM

Riva ASR Arabic Inverse Normalization Grammar

Model

Base Arabic grammar

Riva ASR Mandarin Inverse Normalization Grammar

Model

Base Mandarin grammar

Riva TTS English US Auxiliary Files

Model

Contains files used in rmir creation

TTS Es Multispeaker FastPitch HiFiGAN

Model

This collection contains two models. 1) Multi-speaker 44100Hz FastPitch trained on approximately 20 hours of Latin American Spanish speech from 174 speakers. 2) HiFiGAN trained on mel spectrograms produced by the Multi-speaker FastPitch in (1).

RIVA Punctuation and Capitalization for Spanish

Model

For each word in the input text, the model: 1) predicts a punctuation mark that should follow the word (if any), the model supports commas, periods and question marks) and 2) predicts if the word should be capitalized or not.

Riva ASR Korean LM

Model

Base Korean 4-gram LM

Riva ASR Brazilian Portuguese LM

Model

Base Brazilian Portuguese 4-gram LM

RIVA Conformer ASR Korean

Model

Korean (ko-KR) Conformer ASR model trained on ASR set 1.0

RIVA Citrinet-1024 ASR Korean

Model

Korean (ko-KR) Citrinet-1024 ASR model trained on ASR set 1.0

RIVA Conformer ASR Brazilian Portuguese

Model

Brazilian Portuguese (pt-BR) Conformer ASR model trained on ASR set 1.0

RIVA Citrinet-1024 ASR Brazilian Portuguese

Model

Brazilian Portuguese (pt-BR) Citrinet-1024 ASR model trained on ASR set 1.0

STT It Conformer-Transducer Large

Model

Conformer-Transducer-Large model for Italian Automatic Speech Recognition, Trained on NeMo ASRSET 2.0

STT It Conformer CTC Large

Model

Conformer-CTC-Large model for Italian Automatic Speech Recognition, Trained on NeMo ASRSET 2.0

STT Ru Conformer-CTC Large

Model

Conformer-CTC-Large model for Russian Automatic Speech Recognition, trained on Mozilla Common Voice 10.0 (Russian), Golos (Russian), Russian LibriSpeech (RuLS) and SOVA (RuAudiobooksDevices, RuDevices) datasets.

TTS En RAD-TTS Aligner

Model

RAD-TTS Aligner model trained on female English speech.

STT Ru Conformer-Transducer Large

Model

Conformer-Transducer-Large model for Russian Automatic Speech Recognition, trained on Mozilla Common Voice 10.0 (Russian), Golos (Russian), Russian LibriSpeech (RuLS) and SOVA (RuAudiobooksDevices, RuDevices) datasets.

Riva ASR French Inverse Normalization Grammar

Model

Base French grammar

TTS En FastPitch

Model

FastPitch Speech Synthesis model trained on female English speech.

Bi3D Proximity Segmentation

Model

Bi3D is a binary depth classification network that is used to classify the depth of objects at a given distance.