Hugging Face's logo

software-mansion
/
react-native-executorch-lfm-3

ExecuTorch
react-native-executorch
lfm3
Model card Files
xet
 Community
msluszniak commited on
Commit
a9f3e17
·
verified ·
1 Parent(s): ad2cbc1

Update LFM3-Audio-2B exports

Browse files
Files changed (3) hide show
  1. .gitattributes +1 -0
  2. audio_2b/mlx/config.json +192 -0
  3. audio_2b/mlx/lfm_3_audio_2b_mlx_int4.pte +3 -0
.gitattributes CHANGED
@@ -43,3 +43,4 @@ vl_2b/tokenizer.json filter=lfs diff=lfs merge=lfs -text
43
  vl_2b/xnnpack/lfm_3_vl_2b_xnnpack_8da4w.pte filter=lfs diff=lfs merge=lfs -text
44
  audio_2b/tokenizer.json filter=lfs diff=lfs merge=lfs -text
45
  audio_2b/xnnpack/lfm_3_audio_2b_xnnpack_8da4w.pte filter=lfs diff=lfs merge=lfs -text
 
 
43
  vl_2b/xnnpack/lfm_3_vl_2b_xnnpack_8da4w.pte filter=lfs diff=lfs merge=lfs -text
44
  audio_2b/tokenizer.json filter=lfs diff=lfs merge=lfs -text
45
  audio_2b/xnnpack/lfm_3_audio_2b_xnnpack_8da4w.pte filter=lfs diff=lfs merge=lfs -text
46
+ audio_2b/mlx/lfm_3_audio_2b_mlx_int4.pte filter=lfs diff=lfs merge=lfs -text
audio_2b/mlx/config.json ADDED
@@ -0,0 +1,192 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "$schema": "https://huggingface.co/software-mansion/react-native-executorch-spec/resolve/main/config.schema.json",
3
+ "model": "lfm3_audio_2b",
4
+ "family": "lfm3",
5
+ "capabilities": [
6
+ "text-generation",
7
+ "audio-text-to-text",
8
+ "speech-to-text"
9
+ ],
10
+ "backend": "mlx",
11
+ "license": "lfm-1.0",
12
+ "variants": [
13
+ {
14
+ "file": "lfm_3_audio_2b_mlx_int4.pte",
15
+ "precision": "int4",
16
+ "quantized": true,
17
+ "default": true,
18
+ "methods": {
19
+ "audio_encoder": {
20
+ "inputs": [
21
+ {
22
+ "shape": [
23
+ 1,
24
+ 479920
25
+ ],
26
+ "dtype": "float32"
27
+ }
28
+ ],
29
+ "outputs": [
30
+ {
31
+ "shape": [
32
+ 1,
33
+ 375,
34
+ 2048
35
+ ],
36
+ "dtype": "float32"
37
+ }
38
+ ]
39
+ },
40
+ "token_embedding": {
41
+ "inputs": [
42
+ {
43
+ "shape": [
44
+ 1,
45
+ 256
46
+ ],
47
+ "dtype": "int64"
48
+ }
49
+ ],
50
+ "outputs": [
51
+ {
52
+ "shape": [
53
+ 1,
54
+ 256,
55
+ 2048
56
+ ],
57
+ "dtype": "float32"
58
+ }
59
+ ]
60
+ },
61
+ "text_decoder": {
62
+ "inputs": [
63
+ {
64
+ "shape": [
65
+ 1,
66
+ 256,
67
+ 2048
68
+ ],
69
+ "dtype": "float32"
70
+ },
71
+ {
72
+ "shape": [
73
+ 256
74
+ ],
75
+ "dtype": "int64"
76
+ }
77
+ ],
78
+ "outputs": [
79
+ {
80
+ "shape": [
81
+ 1,
82
+ 128000
83
+ ],
84
+ "dtype": "float32"
85
+ }
86
+ ]
87
+ },
88
+ "enable_dynamic_shape": true,
89
+ "get_bos_id": 124894,
90
+ "get_eos_ids": [
91
+ 124900
92
+ ],
93
+ "get_max_context_len": 8192,
94
+ "get_max_seq_len": 256,
95
+ "get_n_layers": 24,
96
+ "get_vocab_size": 128000,
97
+ "use_kv_cache": true,
98
+ "use_sdpa_with_kv_cache": true,
99
+ "get_audio_frames_per_token": 8
100
+ }
101
+ },
102
+ {
103
+ "file": "lfm_3_audio_2b_mlx_int8.pte",
104
+ "precision": "int8",
105
+ "quantized": true,
106
+ "default": false,
107
+ "methods": {
108
+ "audio_encoder": {
109
+ "inputs": [
110
+ {
111
+ "shape": [
112
+ 1,
113
+ 479920
114
+ ],
115
+ "dtype": "float32"
116
+ }
117
+ ],
118
+ "outputs": [
119
+ {
120
+ "shape": [
121
+ 1,
122
+ 375,
123
+ 2048
124
+ ],
125
+ "dtype": "float32"
126
+ }
127
+ ]
128
+ },
129
+ "token_embedding": {
130
+ "inputs": [
131
+ {
132
+ "shape": [
133
+ 1,
134
+ 256
135
+ ],
136
+ "dtype": "int64"
137
+ }
138
+ ],
139
+ "outputs": [
140
+ {
141
+ "shape": [
142
+ 1,
143
+ 256,
144
+ 2048
145
+ ],
146
+ "dtype": "float32"
147
+ }
148
+ ]
149
+ },
150
+ "text_decoder": {
151
+ "inputs": [
152
+ {
153
+ "shape": [
154
+ 1,
155
+ 256,
156
+ 2048
157
+ ],
158
+ "dtype": "float32"
159
+ },
160
+ {
161
+ "shape": [
162
+ 256
163
+ ],
164
+ "dtype": "int64"
165
+ }
166
+ ],
167
+ "outputs": [
168
+ {
169
+ "shape": [
170
+ 1,
171
+ 128000
172
+ ],
173
+ "dtype": "float32"
174
+ }
175
+ ]
176
+ },
177
+ "enable_dynamic_shape": true,
178
+ "get_bos_id": 124894,
179
+ "get_eos_ids": [
180
+ 124900
181
+ ],
182
+ "get_max_context_len": 8192,
183
+ "get_max_seq_len": 256,
184
+ "get_n_layers": 24,
185
+ "get_vocab_size": 128000,
186
+ "use_kv_cache": true,
187
+ "use_sdpa_with_kv_cache": true,
188
+ "get_audio_frames_per_token": 8
189
+ }
190
+ }
191
+ ]
192
+ }
audio_2b/mlx/lfm_3_audio_2b_mlx_int4.pte ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8a3af1ad4ea483a6933a97f66882e55ac9b598e3030c32c4a90aa8890e18a87d
3
+ size 1159583360