|
|
|
@ -4,10 +4,11 @@ High-performance inference of [OpenAI's Whisper](https://github.com/openai/whisp
|
|
|
|
|
|
|
|
|
|
|
|
- Plain C/C++ implementation without dependencies
|
|
|
|
- Plain C/C++ implementation without dependencies
|
|
|
|
- ARM_NEON and AVX intrinsics support
|
|
|
|
- ARM_NEON and AVX intrinsics support
|
|
|
|
- Mixed F16 / F32 support
|
|
|
|
- Mixed F16 / F32 precision
|
|
|
|
- Low memory usage (Flash Attention + Flash Forward)
|
|
|
|
- Low memory usage (Flash Attention + Flash Forward)
|
|
|
|
- Zero memory allocations at runtime
|
|
|
|
- Zero memory allocations at runtime
|
|
|
|
- Runs on the CPU (Mac and Linux support)
|
|
|
|
- Runs on the CPU (Mac and Linux)
|
|
|
|
|
|
|
|
- Runs on a Raspberry Pi [#7](https://github.com/ggerganov/whisper.cpp/issues/7)
|
|
|
|
|
|
|
|
|
|
|
|
## Usage
|
|
|
|
## Usage
|
|
|
|
|
|
|
|
|
|
|
|
|
Georgi Gerganov
2 years ago
committed by
GitHub