Add sherpa-onnx int8 runtime files for offline desktop inference
Browse filesAdditive only: small encoder/decoder int8 ONNX + tokens. Existing weights untouched. Source: csukuangfj/sherpa-onnx-whisper-small. Tested locally ~2-3x realtime.
- small-decoder.int8.onnx +3 -0
- small-encoder.int8.onnx +3 -0
- small-tokens.txt +0 -0
small-decoder.int8.onnx
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:acad50b5c782696e91b55914cc5ab4f756f1532f76e22aa6fc615f39fb69a8ee
|
| 3 |
+
size 262226114
|
small-encoder.int8.onnx
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:4cbe7b22fa9026b843b60a68640c747de05bafb1a11b57edc0e66c232d9f33a9
|
| 3 |
+
size 112442483
|
small-tokens.txt
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|