Hugging Face's logo

OpenVoiceOS
/
parakeet-tdt-0.6b-dutch-coreml-4bit

Automatic Speech Recognition
Core ML
NeMo
Dutch
coremltools
asr
speech
parakeet
yuriyvnv
4bit-palettize-kmeans
Model card Files
xet
 Community
Jarbas commited on
Commit
4e38e4a
·
verified ·
1 Parent(s): b284915

Upload folder using huggingface_hub

Browse files
Files changed (13) hide show
  1. README.md +59 -0
  2. metadata.json +121 -0
  3. parakeet_decoder.mlpackage/Data/com.apple.CoreML/model.mlmodel +3 -0
  4. parakeet_decoder.mlpackage/Data/com.apple.CoreML/weights/weight.bin +3 -0
  5. parakeet_decoder.mlpackage/Manifest.json +18 -0
  6. parakeet_joint_decision_single_step.mlpackage/Data/com.apple.CoreML/model.mlmodel +3 -0
  7. parakeet_joint_decision_single_step.mlpackage/Data/com.apple.CoreML/weights/weight.bin +3 -0
  8. parakeet_joint_decision_single_step.mlpackage/Manifest.json +18 -0
  9. parakeet_mel_encoder.mlpackage/Data/com.apple.CoreML/model.mlmodel +3 -0
  10. parakeet_mel_encoder.mlpackage/Data/com.apple.CoreML/weights/weight.bin +3 -0
  11. parakeet_mel_encoder.mlpackage/Manifest.json +18 -0
  12. quantization_summary.json +17 -0
  13. vocab.json +0 -0
README.md ADDED
@@ -0,0 +1,59 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ language:
4
+ - nl
5
+ tags:
6
+ - asr
7
+ - speech
8
+ - coreml
9
+ - nemo
10
+ - parakeet
11
+ - yuriyvnv
12
+ - 4bit-palettize-kmeans
13
+ library_name: coremltools
14
+ pipeline_tag: automatic-speech-recognition
15
+ base_model: yuriyvnv/parakeet-tdt-0.6b-dutch
16
+ ---
17
+
18
+ # parakeet-tdt-0.6b-dutch-coreml-4bit
19
+
20
+ CoreML conversion of [yuriyvnv/parakeet-tdt-0.6b-dutch](https://huggingface.co/yuriyvnv/parakeet-tdt-0.6b-dutch) — 4BIT PALETTIZE KMEANS quantized.
21
+
22
+ | | |
23
+ |---|---|
24
+ | **Architecture** | TDT (Token-and-Duration Transducer) |
25
+ | **Language** | Dutch |
26
+ | **Sample rate** | 16000 Hz |
27
+ | **Max audio** | 15.0s |
28
+ | **Vocab size** | 8192 |
29
+ | **Framework** | NVIDIA NeMo → CoreML (coremltools) |
30
+
31
+ ## Components
32
+
33
+ | File | Component | Best compute |
34
+ |------|-----------|--------------|
35
+ | `parakeet_mel_encoder.mlpackage` | mel_encoder | ANE / GPU |
36
+ | `parakeet_decoder.mlpackage` | decoder | CPU only |
37
+ | `parakeet_joint_decision_single_step.mlpackage` | joint_decision_single_step | ANE / GPU |
38
+
39
+ ## Usage
40
+
41
+ ```bash
42
+ pip install ovos-stt-plugin-coreml
43
+ ```
44
+
45
+ ```python
46
+ from ovos_stt_plugin_coreml import CoremlSTT
47
+ from ovos_plugin_manager.utils.audio import AudioFile
48
+
49
+ stt = CoremlSTT(config={"metadata": "metadata.json"})
50
+
51
+ with AudioFile("speech.wav") as f:
52
+ audio = f.read()
53
+ print(stt.execute(audio))
54
+ ```
55
+
56
+ ## Source model
57
+
58
+ [yuriyvnv/parakeet-tdt-0.6b-dutch](https://huggingface.co/yuriyvnv/parakeet-tdt-0.6b-dutch)
59
+
metadata.json ADDED
@@ -0,0 +1,121 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model_id": "yuriyvnv/parakeet-tdt-0.6b-dutch",
3
+ "model_type": "parakeet_tdt_rnnt",
4
+ "language": "nl",
5
+ "sample_rate": 16000,
6
+ "max_audio_seconds": 15.0,
7
+ "max_audio_samples": 240000,
8
+ "vocab_size": 8192,
9
+ "blank_id": 8192,
10
+ "checkpoint": {
11
+ "type": "pretrained",
12
+ "model_id": "yuriyvnv/parakeet-tdt-0.6b-dutch"
13
+ },
14
+ "coreml": {
15
+ "compute_precision": "FLOAT32",
16
+ "quantization": "4bit_palettize_kmeans"
17
+ },
18
+ "components": {
19
+ "mel_encoder": {
20
+ "path": "parakeet_mel_encoder.mlpackage",
21
+ "inputs": {
22
+ "audio_signal": [
23
+ 1,
24
+ 240000
25
+ ],
26
+ "audio_length": [
27
+ 1
28
+ ]
29
+ },
30
+ "outputs": {
31
+ "encoder": [
32
+ 1,
33
+ 1024,
34
+ 188
35
+ ],
36
+ "encoder_length": [
37
+ 1
38
+ ]
39
+ }
40
+ },
41
+ "decoder": {
42
+ "path": "parakeet_decoder.mlpackage",
43
+ "inputs": {
44
+ "targets": [
45
+ 1,
46
+ 1
47
+ ],
48
+ "target_length": [
49
+ 1
50
+ ],
51
+ "h_in": [
52
+ 2,
53
+ 1,
54
+ 640
55
+ ],
56
+ "c_in": [
57
+ 2,
58
+ 1,
59
+ 640
60
+ ]
61
+ },
62
+ "outputs": {
63
+ "decoder": [
64
+ 1,
65
+ 640,
66
+ 1
67
+ ],
68
+ "h_out": [
69
+ 2,
70
+ 1,
71
+ 640
72
+ ],
73
+ "c_out": [
74
+ 2,
75
+ 1,
76
+ 640
77
+ ]
78
+ }
79
+ },
80
+ "joint_decision_single_step": {
81
+ "path": "parakeet_joint_decision_single_step.mlpackage",
82
+ "inputs": {
83
+ "encoder_step": [
84
+ 1,
85
+ 1024,
86
+ 1
87
+ ],
88
+ "decoder_step": [
89
+ 1,
90
+ 640,
91
+ 1
92
+ ]
93
+ },
94
+ "outputs": {
95
+ "token_id": [
96
+ 1,
97
+ 1,
98
+ 1
99
+ ],
100
+ "token_prob": [
101
+ 1,
102
+ 1,
103
+ 1
104
+ ],
105
+ "duration": [
106
+ 1,
107
+ 1,
108
+ 1
109
+ ]
110
+ }
111
+ }
112
+ },
113
+ "joint_extra_outputs": 5,
114
+ "duration_bins": [
115
+ 0,
116
+ 1,
117
+ 2,
118
+ 3,
119
+ 4
120
+ ]
121
+ }
parakeet_decoder.mlpackage/Data/com.apple.CoreML/model.mlmodel ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9f9cb0ce2251e781044b1c50054f6ce17148d5a2822f5ac42d5ff8f66821a191
3
+ size 13061
parakeet_decoder.mlpackage/Data/com.apple.CoreML/weights/weight.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:560e02c955e3dc254196c8bcd5a6b80db040b24b7d593ff2be09e64061c3bef0
3
+ size 5909952
parakeet_decoder.mlpackage/Manifest.json ADDED
@@ -0,0 +1,18 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "fileFormatVersion": "1.0.0",
3
+ "itemInfoEntries": {
4
+ "3D927FED-5249-4512-91E5-AA60DBB4A00B": {
5
+ "author": "com.apple.CoreML",
6
+ "description": "CoreML Model Weights",
7
+ "name": "weights",
8
+ "path": "com.apple.CoreML/weights"
9
+ },
10
+ "CC903519-4F15-40D6-95E8-053620A30E86": {
11
+ "author": "com.apple.CoreML",
12
+ "description": "CoreML Model Specification",
13
+ "name": "model.mlmodel",
14
+ "path": "com.apple.CoreML/model.mlmodel"
15
+ }
16
+ },
17
+ "rootModelIdentifier": "CC903519-4F15-40D6-95E8-053620A30E86"
18
+ }
parakeet_joint_decision_single_step.mlpackage/Data/com.apple.CoreML/model.mlmodel ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d450bff77133f085758e75cb31f647833e6d3badd6ebdc6fc4457319dfea2c61
3
+ size 8906
parakeet_joint_decision_single_step.mlpackage/Data/com.apple.CoreML/weights/weight.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4ef3a3360b1f73436ed0c15d0918c80e27ad37e4315168caae0b04797b34fd69
3
+ size 3175628
parakeet_joint_decision_single_step.mlpackage/Manifest.json ADDED
@@ -0,0 +1,18 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "fileFormatVersion": "1.0.0",
3
+ "itemInfoEntries": {
4
+ "E4F159F9-0F23-4089-A6CF-61015B7B7576": {
5
+ "author": "com.apple.CoreML",
6
+ "description": "CoreML Model Specification",
7
+ "name": "model.mlmodel",
8
+ "path": "com.apple.CoreML/model.mlmodel"
9
+ },
10
+ "F994A227-F072-4E5A-8E07-2AF24816294D": {
11
+ "author": "com.apple.CoreML",
12
+ "description": "CoreML Model Weights",
13
+ "name": "weights",
14
+ "path": "com.apple.CoreML/weights"
15
+ }
16
+ },
17
+ "rootModelIdentifier": "E4F159F9-0F23-4089-A6CF-61015B7B7576"
18
+ }
parakeet_mel_encoder.mlpackage/Data/com.apple.CoreML/model.mlmodel ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f47427b2927165c27e47fee4ebf7c8a30cc730b3d56f43e9191e24e4b4f80b42
3
+ size 694501
parakeet_mel_encoder.mlpackage/Data/com.apple.CoreML/weights/weight.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ed1f384cf1f539d6250f40a2f6caa1602a37e06f4bbdfa79e81764ec4fae34ec
3
+ size 297267392
parakeet_mel_encoder.mlpackage/Manifest.json ADDED
@@ -0,0 +1,18 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "fileFormatVersion": "1.0.0",
3
+ "itemInfoEntries": {
4
+ "4E21D5FD-6688-43A1-A8B7-E659B7B08732": {
5
+ "author": "com.apple.CoreML",
6
+ "description": "CoreML Model Specification",
7
+ "name": "model.mlmodel",
8
+ "path": "com.apple.CoreML/model.mlmodel"
9
+ },
10
+ "8B1CB281-BF50-40D9-85E1-FDA3ECD00D23": {
11
+ "author": "com.apple.CoreML",
12
+ "description": "CoreML Model Weights",
13
+ "name": "weights",
14
+ "path": "com.apple.CoreML/weights"
15
+ }
16
+ },
17
+ "rootModelIdentifier": "4E21D5FD-6688-43A1-A8B7-E659B7B08732"
18
+ }
quantization_summary.json ADDED
@@ -0,0 +1,17 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "parakeet_decoder": {
3
+ "original_mb": 22.5,
4
+ "4bit_mb": 5.6,
5
+ "compression": 3.99
6
+ },
7
+ "parakeet_joint_decision_single_step": {
8
+ "original_mb": 12.1,
9
+ "4bit_mb": 3.0,
10
+ "compression": 3.97
11
+ },
12
+ "parakeet_mel_encoder": {
13
+ "original_mb": 1132.5,
14
+ "4bit_mb": 284.2,
15
+ "compression": 3.99
16
+ }
17
+ }
vocab.json ADDED
The diff for this file is too large to render. See raw diff