ContainersNVIDIADeepSeek-R1NVIDIALlama-3.1-Nemotron-70B-InstructNVIDIAPyTorch
CollectionsNVIDIAOmniverse Kit (FB)NVIDIADeepStream SDKNVIDIAOmniverse Kit App Streaming
ModelsNVIDIAStyleGAN3 pretrained modelsNVIDIAPeopleNetNVIDIATrafficCamNet
ResourcesNVIDIARiva Skills Quick Start
Helm ChartsNVIDIAGPU OperatorNVIDIANVIDIA NIM Operator
NVIDIA
NVIDIA
SpeechLLM FastConformer Llama2-7B
Model
NVIDIA
NVIDIA
SpeechLLM FastConformer Llama2-7B

A 7B SpeechLLM model trained on speech-to-text recognition (ASR), speech-to-text translation (AST) and audio/speech question answering (SpeechQA, AudioQA) data.

1 Version
1.23.1Selected
04/23/2024 11:11 PM UTC25.81 GB
Accuracy
KeyValue
MCV-7.1 Test WER(%)8.53
LibriSpeech-test-other WER(%)4.65
FLEURS En->Es Test BLEU (%)16.97
FLEURS En->De Test BLEU (%)27.41
FLEURS En->Fr Test BLEU (%)25.79
Model
KeyValue
ArchitectureFastConformer-large-LLaMA2-7b
OutputsText response
InputsAudio context and text prompt
Number of Weights7B