Downloads · 30 days
103
22% of all-time downloads
JoaoZaokk/nemotron-3.5-asr-streaming-0.6b-gguf
nemotron-3.5-asr-streaming-0.6b-gguf is a automatic speech recognition model from JoaoZaokk. Use it when you need speech turned into text. It is set up for parakeet.cpp. The card lists the license as other.
GGUF of nvidia/nemotron-3.5-asr-streaming-0.6b (cache-aware streaming FastConformer-RNNT, 40+ locales, prompt-conditioned) for mudler/parakeet.cpp. License: NVIDIA OpenMDW-1.1 (see the source card).
Downloads · 30 days
103
22% of all-time downloads
All-time downloads
473
Public
Repo size
4.8 GB
Likes
0
Public
Click a slice to open those files.
.gguf4.8 GB · 100%
From the Hugging Face model README
GGUF of nvidia/nemotron-3.5-asr-streaming-0.6b (cache-aware streaming FastConformer-RNNT, 40+ locales, prompt-conditioned) for mudler/parakeet.cpp. License: NVIDIA OpenMDW-1.1 (see the source card).
Source checkpoint: nvidia/nemotron-3.5-asr-streaming-0.6b by NVIDIA (f16 GGUF by mudler) · License: other (unchanged; this repo only re-packages the weights)
Engine: Load with mudler/parakeet.cpp (parakeet-cli transcribe --model <file> --input audio.wav --lang pt-BR). Not compatible with whisper.cpp.
| File | Quantization | Size | Note |
|---|---|---|---|
nemotron-3.5-asr-streaming-0.6b-f16.gguf | f16 | 1484 MB | mudler's conversion, WER 0 vs NeMo |
nemotron-3.5-asr-streaming-0.6b-q8_0.gguf | q8_0 | 984 MB | quantized here from an F32 upcast of the f16 |
nemotron-3.5-asr-streaming-0.6b-q6_k.gguf | q6_k | 856 MB | quantized here from an F32 upcast of the f16 |
nemotron-3.5-asr-streaming-0.6b-q5_k.gguf | q5_k | 785 MB | quantized here from an F32 upcast of the f16 |
nemotron-3.5-asr-streaming-0.6b-q4_k.gguf | q4_k | 718 MB | quantized here from an F32 upcast of the f16 |
f16 is the lossless conversion; q8_0 is nearly identical in accuracy at ~55 % of the size; q5_0/q5_k are the phone-friendly choice; q4_* is smallest with a small accuracy cost.
Converted from the upstream checkpoint with the engine's own converter, then quantized with the engine's quantizer. Each variant was checked by transcribing short Portuguese and English samples before upload.
Weights are derivative works of the upstream model and keep its license. Please cite the original authors (NVIDIA (f16 GGUF by mudler)). The conversion and hosting here are maintained by JoaoZaokk so that the download links used by the Odysseus / Open WebUI native apps stay stable. No warranty.