svale-600M-hf

3dio-ai/svale-600M converted from its NeMo archive to transformers format (ParakeetForTDT). Weights are unchanged — this is a format conversion only, produced with the conversion script from transformers:

python src/transformers/models/parakeet/convert_nemo_to_hf.py \
  --hf_repo_id 3dio-ai/svale-600M --model_type tdt --output_dir .

Why

The original ships only as svale-600M.nemo, so using it requires nemo_toolkit. In a Hugging Face ZeroGPU Space that is a problem beyond dependency weight: NeMo's import in the main process breaks the fork() ZeroGPU uses to start GPU workers, and every request fails with "GPU task aborted". Converting removes NeMo from the runtime.

Use

from transformers import pipeline

asr = pipeline("automatic-speech-recognition", model="kasper-cbx/svale-600M-hf", device="cuda")
asr({"raw": audio, "sampling_rate": 16000})  # 16 kHz mono float32

Output is lowercase and unpunctuated, like the original. Danish only — it takes no language or task argument.

Decode one clip at a time. Batched calls (batch_size > 1) over-generate on the padded tail: on a 45-turn fixture, batching reproduced the NeMo reference on 23/45 turns, while batch_size=1 reproduced it on 42/45 (the remaining three differ only as uden for/udenfor and one nonsense utterance).

Licence

Inherits the NVIDIA Open Model License from the base model, and the CoRal OpenRAIL-D use restrictions that come with its training data: no speech synthesis, no biometric identification.