whistle-bf16 (bf16)

Full-precision (bf16) PyTorch port of Cactus-Compute/whistle, for vLLM / TGI / Xinference / transformers users who don't want the GGUF quant.

• Format: pytorch_model.bin (bf16). safetensors conversion in progress.
• Architecture: identical to the original — drop-in for AutoModelForCausalLM.from_pretrained("rty813/whistle-bf16").
• Also loads in vLLM (vllm serve rty813/whistle-bf16) and Xinference (search "whistle").

Original model card:

---

library_name: cactus-needle pipeline_tag: automatic-speech-recognition license: apache-2.0 language: - en - de - fr - es - it - nl - pl tags: - speech-recognition - speech-to-text - on-device - edge - quantization - webassembly

A speech-to-text model for mobiles, wearables, robots, smart home, automotive and microcontrollers. The whole model is a single 16.9 MB file, and it runs on the same CPU engine as Needle, from the same container and the same quantisation, with no dependencies and no GPU.

Whistle does three jobs, all of them on the device:

• Transcription: 16 kHz mono audio of any length, one pass up to 30 seconds and windowed beyond it, in English, German, French, Spanish, Italian, Dutch and Polish. The language is detected unless you name it, and silence returns an empty transcript rather than an invented sentence. Live audio streams for as long as you like, committing the words two passes agree on.
• Word timestamps: every word with its start, end and probability, aligned from the decoder's own attention, so an app can highlight, seek or cut on a word.
• Speech embedding: the encoder output, one row per 80 ms frame, for matching and retrieval without decoding a transcript at all.

Keyword biasing takes the names, places and product words your users actually say and favours them during the search, so the ones that matter survi