8de452c18b
* whisper : prepare infra for new decoding strategies * whisper : apply logit filters and compute logprobs * whisper : add whisper_get_logits() * whisper : separate self and cross attention memory Initial step needed for supporting parallel decoders * whisper : move probs_id buffer to whisper_context * whisper : refactor kv cache into separate struct * whisper : move self-attention kv cache to whisper_decoder * whisper : wip decoding parameters + strategies * whisper : wip decoding parameters + strategies (part 2) * whisper : wip decoding parameters + strategies (part 3) * whisper : wip decoding parameters + strategies (part 4) * whisper : fix prompt_past update to not include prompt_init * whisper : temperature + best_of support * whisper : support for compression_ration_threshold We actually use entropy, but it is similar * command : fix example to use logits instead of obsolete probs * whisper : handle empty sequence ranking * whisper : add WHISPER_DEBUG + diagnostic prints + new main args * whisper : minor fixes * whisper : add beam-search support * whisper : bug fix when there no previous context * whisper : add comments * stream : disable temperature fallback For real-time processing, we always want a single decoder running at T=0 * whisper.swiftui : update example - fix paths + add empty folders |
||
|---|---|---|
| .. | ||
| CMakeLists.txt | ||
| README.md | ||
| command.cpp | ||
| commands.txt | ||
README.md
command
This is a basic Voice Assistant example that accepts voice commands from the microphone. More info is available in issue #171.
# Run with default arguments and small model
./command -m ./models/ggml-small.en.bin -t 8
# On Raspberry Pi, use tiny or base models + "-ac 768" for better performance
./command -m ./models/ggml-tiny.en.bin -ac 768 -t 3 -c 0
https://user-images.githubusercontent.com/1991296/204038393-2f846eae-c255-4099-a76d-5735c25c49da.mp4
Web version: examples/command.wasm
Guided mode
"Guided mode" allows you to specify a list of commands (i.e. strings) and the transcription will be guided to classify your command into one from the list. This can be useful in situations where a device is listening only for a small subset of commands.
Initial tests show that this approach might be extremely efficient in terms of performance, since it integrates very well with the "partial Encoder" idea from #137.
# Run in guided mode, the list of allowed commands is in commands.txt
./command -m ./models/ggml-base.en.bin -cmd ./examples/command/commands.txt
# On Raspberry Pi, in guided mode you can use "-ac 128" for extra performance
./command -m ./models/ggml-tiny.en.bin -cmd ./examples/command/commands.txt -ac 128 -t 3 -c 0
https://user-images.githubusercontent.com/1991296/207435352-8fc4ed3f-bde5-4555-9b8b-aeeb76bee969.mp4
Building
The command tool depends on SDL2 library to capture audio from the microphone. You can build it like this:
# Install SDL2 on Linux
sudo apt-get install libsdl2-dev
# Install SDL2 on Mac OS
brew install sdl2
make command