Skip to content

Commit e00ce98

Browse files
mudlermudler
authored
fix: try to add whisperx and faster-whisper for more variants (#9278)
Signed-off-by: Ettore Di Giacinto <mudler@localai.io>
1 parent 285f7d4 commit e00ce98

12 files changed

Lines changed: 206 additions & 3 deletions

File tree

‎.github/workflows/backend.yml‎

Lines changed: 65 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -105,6 +105,19 @@ jobs:
105105
dockerfile: "./backend/Dockerfile.python"
106106
context: "./"
107107
ubuntu-version: '2404'
108+
- build-type: ''
109+
cuda-major-version: ""
110+
cuda-minor-version: ""
111+
platforms: 'linux/amd64'
112+
tag-latest: 'auto'
113+
tag-suffix: '-cpu-faster-whisper'
114+
runs-on: 'ubuntu-latest'
115+
base-image: "ubuntu:24.04"
116+
skip-drivers: 'true'
117+
backend: "faster-whisper"
118+
dockerfile: "./backend/Dockerfile.python"
119+
context: "./"
120+
ubuntu-version: '2404'
108121
- build-type: ''
109122
cuda-major-version: ""
110123
cuda-minor-version: ""
@@ -965,6 +978,32 @@ jobs:
965978
backend: "mlx-distributed"
966979
dockerfile: "./backend/Dockerfile.python"
967980
context: "./"
981+
- build-type: 'l4t'
982+
cuda-major-version: "13"
983+
cuda-minor-version: "0"
984+
platforms: 'linux/arm64'
985+
tag-latest: 'auto'
986+
tag-suffix: '-nvidia-l4t-cuda-13-arm64-whisperx'
987+
runs-on: 'ubuntu-24.04-arm'
988+
base-image: "ubuntu:24.04"
989+
skip-drivers: 'false'
990+
ubuntu-version: '2404'
991+
backend: "whisperx"
992+
dockerfile: "./backend/Dockerfile.python"
993+
context: "./"
994+
- build-type: 'l4t'
995+
cuda-major-version: "13"
996+
cuda-minor-version: "0"
997+
platforms: 'linux/arm64'
998+
tag-latest: 'auto'
999+
tag-suffix: '-nvidia-l4t-cuda-13-arm64-faster-whisper'
1000+
runs-on: 'ubuntu-24.04-arm'
1001+
base-image: "ubuntu:24.04"
1002+
skip-drivers: 'false'
1003+
ubuntu-version: '2404'
1004+
backend: "faster-whisper"
1005+
dockerfile: "./backend/Dockerfile.python"
1006+
context: "./"
9681007
- build-type: 'cublas'
9691008
cuda-major-version: "13"
9701009
cuda-minor-version: "0"
@@ -1644,6 +1683,32 @@ jobs:
16441683
dockerfile: "./backend/Dockerfile.python"
16451684
context: "./"
16461685
ubuntu-version: '2204'
1686+
- build-type: 'l4t'
1687+
cuda-major-version: "12"
1688+
cuda-minor-version: "0"
1689+
platforms: 'linux/arm64'
1690+
tag-latest: 'auto'
1691+
tag-suffix: '-nvidia-l4t-whisperx'
1692+
runs-on: 'ubuntu-24.04-arm'
1693+
base-image: "nvcr.io/nvidia/l4t-jetpack:r36.4.0"
1694+
skip-drivers: 'true'
1695+
backend: "whisperx"
1696+
dockerfile: "./backend/Dockerfile.python"
1697+
context: "./"
1698+
ubuntu-version: '2204'
1699+
- build-type: 'l4t'
1700+
cuda-major-version: "12"
1701+
cuda-minor-version: "0"
1702+
platforms: 'linux/arm64'
1703+
tag-latest: 'auto'
1704+
tag-suffix: '-nvidia-l4t-faster-whisper'
1705+
runs-on: 'ubuntu-24.04-arm'
1706+
base-image: "nvcr.io/nvidia/l4t-jetpack:r36.4.0"
1707+
skip-drivers: 'true'
1708+
backend: "faster-whisper"
1709+
dockerfile: "./backend/Dockerfile.python"
1710+
context: "./"
1711+
ubuntu-version: '2204'
16471712
# SYCL additional backends
16481713
- build-type: 'intel'
16491714
cuda-major-version: ""

‎Makefile‎

Lines changed: 2 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -428,6 +428,7 @@ prepare-test-extra: protogen-python
428428
$(MAKE) -C backend/python/qwen-asr
429429
$(MAKE) -C backend/python/nemo
430430
$(MAKE) -C backend/python/voxcpm
431+
$(MAKE) -C backend/python/faster-whisper
431432
$(MAKE) -C backend/python/whisperx
432433
$(MAKE) -C backend/python/ace-step
433434
$(MAKE) -C backend/python/trl
@@ -449,6 +450,7 @@ test-extra: prepare-test-extra
449450
$(MAKE) -C backend/python/qwen-asr test
450451
$(MAKE) -C backend/python/nemo test
451452
$(MAKE) -C backend/python/voxcpm test
453+
$(MAKE) -C backend/python/faster-whisper test
452454
$(MAKE) -C backend/python/whisperx test
453455
$(MAKE) -C backend/python/ace-step test
454456
$(MAKE) -C backend/python/trl test

‎backend/index.yaml‎

Lines changed: 48 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -400,12 +400,15 @@
400400
license: MIT
401401
name: "faster-whisper"
402402
capabilities:
403+
default: "cpu-faster-whisper"
403404
nvidia: "cuda12-faster-whisper"
404405
intel: "intel-faster-whisper"
405406
amd: "rocm-faster-whisper"
406407
metal: "metal-faster-whisper"
407408
nvidia-cuda-13: "cuda13-faster-whisper"
408409
nvidia-cuda-12: "cuda12-faster-whisper"
410+
nvidia-l4t: "nvidia-l4t-arm64-faster-whisper"
411+
nvidia-l4t-cuda-12: "nvidia-l4t-arm64-faster-whisper"
409412
- &moonshine
410413
description: |
411414
Moonshine is a fast, accurate, and efficient speech-to-text transcription model using ONNX Runtime.
@@ -446,6 +449,8 @@
446449
default: "cpu-whisperx"
447450
nvidia-cuda-13: "cuda13-whisperx"
448451
nvidia-cuda-12: "cuda12-whisperx"
452+
nvidia-l4t: "nvidia-l4t-arm64-whisperx"
453+
nvidia-l4t-cuda-12: "nvidia-l4t-arm64-whisperx"
449454
- &kokoro
450455
icon: https://avatars.githubusercontent.com/u/166769057?v=4
451456
description: |
@@ -2082,11 +2087,13 @@
20822087
- !!merge <<: *faster-whisper
20832088
name: "faster-whisper-development"
20842089
capabilities:
2090+
default: "cpu-faster-whisper-development"
20852091
nvidia: "cuda12-faster-whisper-development"
20862092
intel: "intel-faster-whisper-development"
20872093
amd: "rocm-faster-whisper-development"
20882094
metal: "metal-faster-whisper-development"
20892095
nvidia-cuda-13: "cuda13-faster-whisper-development"
2096+
nvidia-l4t: "nvidia-l4t-arm64-faster-whisper-development"
20902097
- !!merge <<: *faster-whisper
20912098
name: "cuda12-faster-whisper-development"
20922099
uri: "quay.io/go-skynet/local-ai-backends:master-gpu-nvidia-cuda-12-faster-whisper"
@@ -2127,6 +2134,36 @@
21272134
uri: "quay.io/go-skynet/local-ai-backends:master-metal-darwin-arm64-faster-whisper"
21282135
mirrors:
21292136
- localai/localai-backends:master-metal-darwin-arm64-faster-whisper
2137+
- !!merge <<: *faster-whisper
2138+
name: "cuda12-faster-whisper"
2139+
uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-nvidia-cuda-12-faster-whisper"
2140+
mirrors:
2141+
- localai/localai-backends:latest-gpu-nvidia-cuda-12-faster-whisper
2142+
- !!merge <<: *faster-whisper
2143+
name: "rocm-faster-whisper"
2144+
uri: "quay.io/go-skynet/local-ai-backends:latest-gpu-rocm-hipblas-faster-whisper"
2145+
mirrors:
2146+
- localai/localai-backends:latest-gpu-rocm-hipblas-faster-whisper
2147+
- !!merge <<: *faster-whisper
2148+
name: "cpu-faster-whisper"
2149+
uri: "quay.io/go-skynet/local-ai-backends:latest-cpu-faster-whisper"
2150+
mirrors:
2151+
- localai/localai-backends:latest-cpu-faster-whisper
2152+
- !!merge <<: *faster-whisper
2153+
name: "cpu-faster-whisper-development"
2154+
uri: "quay.io/go-skynet/local-ai-backends:master-cpu-faster-whisper"
2155+
mirrors:
2156+
- localai/localai-backends:master-cpu-faster-whisper
2157+
- !!merge <<: *faster-whisper
2158+
name: "nvidia-l4t-arm64-faster-whisper"
2159+
uri: "quay.io/go-skynet/local-ai-backends:latest-nvidia-l4t-faster-whisper"
2160+
mirrors:
2161+
- localai/localai-backends:latest-nvidia-l4t-faster-whisper
2162+
- !!merge <<: *faster-whisper
2163+
name: "nvidia-l4t-arm64-faster-whisper-development"
2164+
uri: "quay.io/go-skynet/local-ai-backends:master-nvidia-l4t-faster-whisper"
2165+
mirrors:
2166+
- localai/localai-backends:master-nvidia-l4t-faster-whisper
21302167
## moonshine
21312168
- !!merge <<: *moonshine
21322169
name: "moonshine-development"
@@ -2185,6 +2222,7 @@
21852222
default: "cpu-whisperx-development"
21862223
nvidia-cuda-13: "cuda13-whisperx-development"
21872224
nvidia-cuda-12: "cuda12-whisperx-development"
2225+
nvidia-l4t: "nvidia-l4t-arm64-whisperx-development"
21882226
- !!merge <<: *whisperx
21892227
name: "cpu-whisperx"
21902228
uri: "quay.io/go-skynet/local-ai-backends:latest-cpu-whisperx"
@@ -2235,6 +2273,16 @@
22352273
uri: "quay.io/go-skynet/local-ai-backends:master-metal-darwin-arm64-whisperx"
22362274
mirrors:
22372275
- localai/localai-backends:master-metal-darwin-arm64-whisperx
2276+
- !!merge <<: *whisperx
2277+
name: "nvidia-l4t-arm64-whisperx"
2278+
uri: "quay.io/go-skynet/local-ai-backends:latest-nvidia-l4t-whisperx"
2279+
mirrors:
2280+
- localai/localai-backends:latest-nvidia-l4t-whisperx
2281+
- !!merge <<: *whisperx
2282+
name: "nvidia-l4t-arm64-whisperx-development"
2283+
uri: "quay.io/go-skynet/local-ai-backends:master-nvidia-l4t-whisperx"
2284+
mirrors:
2285+
- localai/localai-backends:master-nvidia-l4t-whisperx
22382286
## coqui
22392287

22402288
- !!merge <<: *coqui

‎backend/python/faster-whisper/install.sh‎

Lines changed: 10 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -16,4 +16,14 @@ if [ "x${BUILD_PROFILE}" == "xintel" ]; then
1616
EXTRA_PIP_INSTALL_FLAGS+=" --upgrade --index-strategy=unsafe-first-match"
1717
fi
1818

19+
if [ "x${BUILD_PROFILE}" == "xl4t13" ]; then
20+
PYTHON_VERSION="3.12"
21+
PYTHON_PATCH="12"
22+
PY_STANDALONE_TAG="20251120"
23+
fi
24+
25+
if [ "x${BUILD_PROFILE}" == "xl4t12" ]; then
26+
USE_PIP=true
27+
fi
28+
1929
installRequirements

‎backend/python/faster-whisper/requirements-l4t12.txt‎

Lines changed: 3 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,3 @@
1+
--extra-index-url https://pypi.jetson-ai-lab.io/jp6/cu129/
2+
torch
3+
faster-whisper

‎backend/python/faster-whisper/requirements-l4t13.txt‎

Lines changed: 3 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,3 @@
1+
--extra-index-url https://download.pytorch.org/whl/cu130
2+
torch
3+
faster-whisper

‎backend/python/whisperx/install.sh‎

Lines changed: 15 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -8,8 +8,21 @@ else
88
source $backend_dir/../common/libbackend.sh
99
fi
1010

11-
if [ "x${BUILD_PROFILE}" != "xmetal" ] && [ "x${BUILD_PROFILE}" != "xmps" ]; then
12-
EXTRA_PIP_INSTALL_FLAGS+=" --index-strategy unsafe-best-match"
11+
if [ "x${BUILD_PROFILE}" == "xl4t13" ]; then
12+
PYTHON_VERSION="3.12"
13+
PYTHON_PATCH="12"
14+
PY_STANDALONE_TAG="20251120"
15+
fi
16+
17+
if [ "x${BUILD_PROFILE}" == "xl4t12" ]; then
18+
USE_PIP=true
19+
fi
20+
21+
# --index-strategy is a uv-only flag; skip it when using pip
22+
if [ "x${USE_PIP}" != "xtrue" ]; then
23+
if [ "x${BUILD_PROFILE}" != "xmetal" ] && [ "x${BUILD_PROFILE}" != "xmps" ]; then
24+
EXTRA_PIP_INSTALL_FLAGS+=" --index-strategy unsafe-best-match"
25+
fi
1326
fi
1427

1528
installRequirements

‎backend/python/whisperx/requirements-l4t12.txt‎

Lines changed: 3 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,3 @@
1+
--extra-index-url https://pypi.jetson-ai-lab.io/jp6/cu129/
2+
torch
3+
whisperx @ git+https://github.com/m-bain/whisperX.git

‎backend/python/whisperx/requirements-l4t13.txt‎

Lines changed: 3 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,3 @@
1+
--extra-index-url https://download.pytorch.org/whl/cu130
2+
torch
3+
whisperx @ git+https://github.com/m-bain/whisperX.git

‎core/gallery/backend_types.go‎

Lines changed: 4 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -71,6 +71,10 @@ func (m *GalleryBackend) IsCompatibleWith(systemState *system.SystemState) bool
7171
return true
7272
}
7373

74+
if systemState.CapabilityFilterDisabled() {
75+
return true
76+
}
77+
7478
// Meta backends are compatible if the system capability matches one of the keys
7579
if m.IsMeta() {
7680
capability := systemState.Capability(m.CapabilitiesMap)

0 commit comments

Comments
 (0)