Hugging Face's logo

OpenVoiceOS
/
parakeet-rnnt-110m-da-coreml-4bit

Automatic Speech Recognition
Core ML
NeMo
Danish
coremltools
asr
speech
parakeet
nvidia
4bit-palettize-kmeans
Model card Files
xet
 Community
Jarbas commited on
Commit
d3175e2
·
verified ·
1 Parent(s): b655571

Upload folder using huggingface_hub

Browse files
Files changed (13) hide show
  1. README.md +59 -0
  2. metadata.json +113 -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 +46 -0
README.md ADDED
@@ -0,0 +1,59 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: cc-by-4.0
3
+ language:
4
+ - da
5
+ tags:
6
+ - asr
7
+ - speech
8
+ - coreml
9
+ - nemo
10
+ - parakeet
11
+ - nvidia
12
+ - 4bit-palettize-kmeans
13
+ library_name: coremltools
14
+ pipeline_tag: automatic-speech-recognition
15
+ base_model: nvidia/parakeet-rnnt-110m-da-dk
16
+ ---
17
+
18
+ # parakeet-rnnt-110m-da-coreml-4bit
19
+
20
+ CoreML conversion of [nvidia/parakeet-rnnt-110m-da-dk](https://huggingface.co/nvidia/parakeet-rnnt-110m-da-dk) — 4BIT PALETTIZE KMEANS quantized.
21
+
22
+ | | |
23
+ |---|---|
24
+ | **Architecture** | RNNT |
25
+ | **Language** | Danish |
26
+ | **Sample rate** | 16000 Hz |
27
+ | **Max audio** | 15.0s |
28
+ | **Vocab size** | 44 |
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
+ [nvidia/parakeet-rnnt-110m-da-dk](https://huggingface.co/nvidia/parakeet-rnnt-110m-da-dk)
59
+
metadata.json ADDED
@@ -0,0 +1,113 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "model_id": "nvidia/parakeet-rnnt-110m-da-dk",
3
+ "model_type": "parakeet_rnnt",
4
+ "language": "da",
5
+ "sample_rate": 16000,
6
+ "max_audio_seconds": 15.0,
7
+ "max_audio_samples": 240000,
8
+ "vocab_size": 44,
9
+ "blank_id": 44,
10
+ "checkpoint": {
11
+ "type": "pretrained",
12
+ "model_id": "nvidia/parakeet-rnnt-110m-da-dk"
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
+ 512,
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
+ 1,
53
+ 1,
54
+ 640
55
+ ],
56
+ "c_in": [
57
+ 1,
58
+ 1,
59
+ 640
60
+ ]
61
+ },
62
+ "outputs": {
63
+ "decoder": [
64
+ 1,
65
+ 640,
66
+ 1
67
+ ],
68
+ "h_out": [
69
+ 1,
70
+ 1,
71
+ 640
72
+ ],
73
+ "c_out": [
74
+ 1,
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
+ 512,
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
+ }
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 sha256:31834ecf58229fc19ddb94493660287acd987f3f35dbba1878cf13b86377e5c6
3
+ size 9040
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:e13c8d3af7e5469ec8f7213c270931b0a052a7f0038b93317af959e1790b1801
3
+ size 1658624
parakeet_decoder.mlpackage/Manifest.json ADDED
@@ -0,0 +1,18 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "fileFormatVersion": "1.0.0",
3
+ "itemInfoEntries": {
4
+ "896B10BC-C907-4BAF-AE81-946A36B2913B": {
5
+ "author": "com.apple.CoreML",
6
+ "description": "CoreML Model Specification",
7
+ "name": "model.mlmodel",
8
+ "path": "com.apple.CoreML/model.mlmodel"
9
+ },
10
+ "F5A413E3-3B01-444B-B320-1A27D55BB7F1": {
11
+ "author": "com.apple.CoreML",
12
+ "description": "CoreML Model Weights",
13
+ "name": "weights",
14
+ "path": "com.apple.CoreML/weights"
15
+ }
16
+ },
17
+ "rootModelIdentifier": "896B10BC-C907-4BAF-AE81-946A36B2913B"
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 sha256:8740a71ff984d12f2cd850e1fc561e1718df16feadbfa49491b6408981ac48e3
3
+ size 7252
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 sha256:2502972ce373c220a88d92ab9c410f538233f88f3276261c37f79e822d2fc2dd
3
+ size 386522
parakeet_joint_decision_single_step.mlpackage/Manifest.json ADDED
@@ -0,0 +1,18 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "fileFormatVersion": "1.0.0",
3
+ "itemInfoEntries": {
4
+ "0A5822CA-A8B2-474F-8FB2-9858FEBF3D35": {
5
+ "author": "com.apple.CoreML",
6
+ "description": "CoreML Model Weights",
7
+ "name": "weights",
8
+ "path": "com.apple.CoreML/weights"
9
+ },
10
+ "8EE4709B-C1EC-40F5-AF62-4998E8295FB7": {
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": "8EE4709B-C1EC-40F5-AF62-4998E8295FB7"
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 sha256:1fbc0582a336d6c77f76b67d5d95092afbd186c8ddb7bfc2ee79a84fe68a3608
3
+ size 585058
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 sha256:542c7c3129dc7eee4c3f39490c0b965948d70949122e577b91116e7377cd64ea
3
+ size 54476032
parakeet_mel_encoder.mlpackage/Manifest.json ADDED
@@ -0,0 +1,18 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "fileFormatVersion": "1.0.0",
3
+ "itemInfoEntries": {
4
+ "2F8B52BC-69B1-4E96-AC94-11F8879880E8": {
5
+ "author": "com.apple.CoreML",
6
+ "description": "CoreML Model Specification",
7
+ "name": "model.mlmodel",
8
+ "path": "com.apple.CoreML/model.mlmodel"
9
+ },
10
+ "52301B89-7DE3-46C2-B538-0076FC158D67": {
11
+ "author": "com.apple.CoreML",
12
+ "description": "CoreML Model Weights",
13
+ "name": "weights",
14
+ "path": "com.apple.CoreML/weights"
15
+ }
16
+ },
17
+ "rootModelIdentifier": "2F8B52BC-69B1-4E96-AC94-11F8879880E8"
18
+ }
quantization_summary.json ADDED
@@ -0,0 +1,17 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "parakeet_decoder": {
3
+ "original_mb": 6.3,
4
+ "4bit_mb": 1.6,
5
+ "compression": 3.97
6
+ },
7
+ "parakeet_joint_decision_single_step": {
8
+ "original_mb": 1.5,
9
+ "4bit_mb": 0.4,
10
+ "compression": 3.91
11
+ },
12
+ "parakeet_mel_encoder": {
13
+ "original_mb": 206.5,
14
+ "4bit_mb": 52.5,
15
+ "compression": 3.93
16
+ }
17
+ }
vocab.json ADDED
@@ -0,0 +1,46 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ [
2
+ "<unk>",
3
+ "er",
4
+ "▁",
5
+ "e",
6
+ "r",
7
+ "n",
8
+ "t",
9
+ "i",
10
+ "a",
11
+ "s",
12
+ "d",
13
+ "l",
14
+ "o",
15
+ "g",
16
+ "m",
17
+ "k",
18
+ "f",
19
+ "v",
20
+ "b",
21
+ "u",
22
+ "h",
23
+ "p",
24
+ "å",
25
+ "ø",
26
+ "æ",
27
+ "y",
28
+ "j",
29
+ "c",
30
+ "1",
31
+ "0",
32
+ "9",
33
+ "2",
34
+ "8",
35
+ "w",
36
+ "7",
37
+ "6",
38
+ "5",
39
+ "4",
40
+ "3",
41
+ "z",
42
+ "x",
43
+ "é",
44
+ "q",
45
+ "ü"
46
+ ]