Bartosz Hanc commited on
Commit ·
d71eaff
1
Parent(s): ec14be9
Publish schema-compliant artifact method signatures
Browse files- e2b/mlx/config.json +33 -101
- e2b/vulkan/config.json +33 -101
- e2b/xnnpack/config.json +33 -99
e2b/mlx/config.json
CHANGED
|
@@ -13,66 +13,36 @@
|
|
| 13 |
{
|
| 14 |
"file": "gemma_4_e2b_mlx_int4.pte",
|
| 15 |
"precision": "int4",
|
| 16 |
-
"quantized": true,
|
| 17 |
-
"default": true,
|
| 18 |
"methods": {
|
| 19 |
-
"
|
| 20 |
-
"get_bos_id": 2,
|
| 21 |
-
"get_eos_ids": [
|
| 22 |
-
1,
|
| 23 |
-
106,
|
| 24 |
-
50
|
| 25 |
-
],
|
| 26 |
-
"get_max_context_len": 2048,
|
| 27 |
-
"get_max_seq_len": 511,
|
| 28 |
-
"get_vocab_size": 262144,
|
| 29 |
-
"use_kv_cache": true,
|
| 30 |
-
"has_ple": true,
|
| 31 |
-
"image_placeholder_id": 258880,
|
| 32 |
-
"audio_placeholder_id": 258881,
|
| 33 |
-
"audio_sample_rate": 16000,
|
| 34 |
-
"audio_block_samples": 7680,
|
| 35 |
-
"audio_block_k_min": 1,
|
| 36 |
-
"audio_block_k_max": 62,
|
| 37 |
-
"audio_soft_tokens_max": 744,
|
| 38 |
-
"audio_sample_count_min": 7680,
|
| 39 |
-
"audio_sample_count_max": 476160,
|
| 40 |
-
"token_embedding": {
|
| 41 |
"inputs": [
|
| 42 |
{
|
| 43 |
-
"name": "input_ids",
|
| 44 |
"shape": [
|
| 45 |
1,
|
| 46 |
-
|
| 47 |
-
|
| 48 |
-
|
| 49 |
-
|
|
|
|
|
|
|
|
|
|
| 50 |
],
|
| 51 |
"dtype": "int64"
|
| 52 |
}
|
| 53 |
],
|
| 54 |
"outputs": [
|
| 55 |
{
|
| 56 |
-
"name": "embeds",
|
| 57 |
"shape": [
|
| 58 |
1,
|
| 59 |
-
|
| 60 |
-
|
| 61 |
-
"max": 511
|
| 62 |
-
}
|
| 63 |
],
|
| 64 |
"dtype": "float32"
|
| 65 |
},
|
| 66 |
{
|
| 67 |
-
"name": "ple_tokens",
|
| 68 |
"shape": [
|
| 69 |
1,
|
| 70 |
-
|
| 71 |
-
"min": 1,
|
| 72 |
-
"max": 511
|
| 73 |
-
},
|
| 74 |
-
35,
|
| 75 |
-
256
|
| 76 |
],
|
| 77 |
"dtype": "float32"
|
| 78 |
}
|
|
@@ -81,44 +51,31 @@
|
|
| 81 |
"text_decoder": {
|
| 82 |
"inputs": [
|
| 83 |
{
|
| 84 |
-
"name": "embeds",
|
| 85 |
"shape": [
|
| 86 |
1,
|
| 87 |
-
|
| 88 |
-
"min": 1,
|
| 89 |
-
"max": 511
|
| 90 |
-
},
|
| 91 |
1536
|
| 92 |
],
|
| 93 |
"dtype": "float32"
|
| 94 |
},
|
| 95 |
{
|
| 96 |
-
"name": "ple_tokens",
|
| 97 |
"shape": [
|
| 98 |
1,
|
| 99 |
-
|
| 100 |
-
"min": 1,
|
| 101 |
-
"max": 511
|
| 102 |
-
},
|
| 103 |
35,
|
| 104 |
256
|
| 105 |
],
|
| 106 |
"dtype": "float32"
|
| 107 |
},
|
| 108 |
{
|
| 109 |
-
"name": "input_pos",
|
| 110 |
"shape": [
|
| 111 |
-
|
| 112 |
-
"min": 1,
|
| 113 |
-
"max": 511
|
| 114 |
-
}
|
| 115 |
],
|
| 116 |
"dtype": "int64"
|
| 117 |
}
|
| 118 |
],
|
| 119 |
"outputs": [
|
| 120 |
{
|
| 121 |
-
"name": "logits",
|
| 122 |
"shape": [
|
| 123 |
1,
|
| 124 |
1,
|
|
@@ -128,86 +85,61 @@
|
|
| 128 |
}
|
| 129 |
]
|
| 130 |
},
|
| 131 |
-
"
|
| 132 |
"inputs": [
|
| 133 |
{
|
| 134 |
-
"name": "pixel_values",
|
| 135 |
"shape": [
|
| 136 |
1,
|
| 137 |
-
|
| 138 |
-
448,
|
| 139 |
-
640
|
| 140 |
],
|
| 141 |
-
"dtype": "
|
| 142 |
}
|
| 143 |
],
|
| 144 |
"outputs": [
|
| 145 |
{
|
| 146 |
-
"name": "embeds",
|
| 147 |
"shape": [
|
| 148 |
1,
|
| 149 |
-
|
| 150 |
1536
|
| 151 |
],
|
| 152 |
"dtype": "float32"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 153 |
}
|
| 154 |
]
|
| 155 |
},
|
| 156 |
-
"
|
| 157 |
"inputs": [
|
| 158 |
{
|
| 159 |
-
"name": "waveform",
|
| 160 |
"shape": [
|
| 161 |
1,
|
| 162 |
-
|
| 163 |
-
|
| 164 |
-
|
| 165 |
-
}
|
| 166 |
],
|
| 167 |
"dtype": "float32"
|
| 168 |
-
},
|
| 169 |
-
{
|
| 170 |
-
"name": "num_valid_samples",
|
| 171 |
-
"shape": [
|
| 172 |
-
1
|
| 173 |
-
],
|
| 174 |
-
"dtype": "int64"
|
| 175 |
}
|
| 176 |
],
|
| 177 |
"outputs": [
|
| 178 |
{
|
| 179 |
-
"name": "audio_embeds",
|
| 180 |
"shape": [
|
| 181 |
1,
|
| 182 |
-
|
| 183 |
-
"min": 12,
|
| 184 |
-
"max": 744
|
| 185 |
-
},
|
| 186 |
1536
|
| 187 |
],
|
| 188 |
"dtype": "float32"
|
| 189 |
-
},
|
| 190 |
-
{
|
| 191 |
-
"name": "audio_mask",
|
| 192 |
-
"shape": [
|
| 193 |
-
1,
|
| 194 |
-
{
|
| 195 |
-
"min": 12,
|
| 196 |
-
"max": 744
|
| 197 |
-
}
|
| 198 |
-
],
|
| 199 |
-
"dtype": "float32"
|
| 200 |
}
|
| 201 |
]
|
| 202 |
}
|
| 203 |
}
|
| 204 |
}
|
| 205 |
],
|
| 206 |
-
"size": "e2b"
|
| 207 |
-
"_methods_needing_shapes": [
|
| 208 |
-
"audio_encoder",
|
| 209 |
-
"text_decoder",
|
| 210 |
-
"token_embedding",
|
| 211 |
-
"vision_encoder"
|
| 212 |
-
]
|
| 213 |
}
|
|
|
|
| 13 |
{
|
| 14 |
"file": "gemma_4_e2b_mlx_int4.pte",
|
| 15 |
"precision": "int4",
|
|
|
|
|
|
|
| 16 |
"methods": {
|
| 17 |
+
"audio_encoder": {
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 18 |
"inputs": [
|
| 19 |
{
|
|
|
|
| 20 |
"shape": [
|
| 21 |
1,
|
| 22 |
+
476160
|
| 23 |
+
],
|
| 24 |
+
"dtype": "float32"
|
| 25 |
+
},
|
| 26 |
+
{
|
| 27 |
+
"shape": [
|
| 28 |
+
1
|
| 29 |
],
|
| 30 |
"dtype": "int64"
|
| 31 |
}
|
| 32 |
],
|
| 33 |
"outputs": [
|
| 34 |
{
|
|
|
|
| 35 |
"shape": [
|
| 36 |
1,
|
| 37 |
+
744,
|
| 38 |
+
1536
|
|
|
|
|
|
|
| 39 |
],
|
| 40 |
"dtype": "float32"
|
| 41 |
},
|
| 42 |
{
|
|
|
|
| 43 |
"shape": [
|
| 44 |
1,
|
| 45 |
+
744
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 46 |
],
|
| 47 |
"dtype": "float32"
|
| 48 |
}
|
|
|
|
| 51 |
"text_decoder": {
|
| 52 |
"inputs": [
|
| 53 |
{
|
|
|
|
| 54 |
"shape": [
|
| 55 |
1,
|
| 56 |
+
511,
|
|
|
|
|
|
|
|
|
|
| 57 |
1536
|
| 58 |
],
|
| 59 |
"dtype": "float32"
|
| 60 |
},
|
| 61 |
{
|
|
|
|
| 62 |
"shape": [
|
| 63 |
1,
|
| 64 |
+
511,
|
|
|
|
|
|
|
|
|
|
| 65 |
35,
|
| 66 |
256
|
| 67 |
],
|
| 68 |
"dtype": "float32"
|
| 69 |
},
|
| 70 |
{
|
|
|
|
| 71 |
"shape": [
|
| 72 |
+
511
|
|
|
|
|
|
|
|
|
|
| 73 |
],
|
| 74 |
"dtype": "int64"
|
| 75 |
}
|
| 76 |
],
|
| 77 |
"outputs": [
|
| 78 |
{
|
|
|
|
| 79 |
"shape": [
|
| 80 |
1,
|
| 81 |
1,
|
|
|
|
| 85 |
}
|
| 86 |
]
|
| 87 |
},
|
| 88 |
+
"token_embedding": {
|
| 89 |
"inputs": [
|
| 90 |
{
|
|
|
|
| 91 |
"shape": [
|
| 92 |
1,
|
| 93 |
+
511
|
|
|
|
|
|
|
| 94 |
],
|
| 95 |
+
"dtype": "int64"
|
| 96 |
}
|
| 97 |
],
|
| 98 |
"outputs": [
|
| 99 |
{
|
|
|
|
| 100 |
"shape": [
|
| 101 |
1,
|
| 102 |
+
511,
|
| 103 |
1536
|
| 104 |
],
|
| 105 |
"dtype": "float32"
|
| 106 |
+
},
|
| 107 |
+
{
|
| 108 |
+
"shape": [
|
| 109 |
+
1,
|
| 110 |
+
511,
|
| 111 |
+
35,
|
| 112 |
+
256
|
| 113 |
+
],
|
| 114 |
+
"dtype": "float32"
|
| 115 |
}
|
| 116 |
]
|
| 117 |
},
|
| 118 |
+
"vision_encoder": {
|
| 119 |
"inputs": [
|
| 120 |
{
|
|
|
|
| 121 |
"shape": [
|
| 122 |
1,
|
| 123 |
+
3,
|
| 124 |
+
448,
|
| 125 |
+
640
|
|
|
|
| 126 |
],
|
| 127 |
"dtype": "float32"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 128 |
}
|
| 129 |
],
|
| 130 |
"outputs": [
|
| 131 |
{
|
|
|
|
| 132 |
"shape": [
|
| 133 |
1,
|
| 134 |
+
280,
|
|
|
|
|
|
|
|
|
|
| 135 |
1536
|
| 136 |
],
|
| 137 |
"dtype": "float32"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 138 |
}
|
| 139 |
]
|
| 140 |
}
|
| 141 |
}
|
| 142 |
}
|
| 143 |
],
|
| 144 |
+
"size": "e2b"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 145 |
}
|
e2b/vulkan/config.json
CHANGED
|
@@ -13,66 +13,36 @@
|
|
| 13 |
{
|
| 14 |
"file": "gemma_4_e2b_vulkan_8da4w.pte",
|
| 15 |
"precision": "8da4w",
|
| 16 |
-
"quantized": true,
|
| 17 |
-
"default": true,
|
| 18 |
"methods": {
|
| 19 |
-
"
|
| 20 |
-
"get_bos_id": 2,
|
| 21 |
-
"get_eos_ids": [
|
| 22 |
-
1,
|
| 23 |
-
106,
|
| 24 |
-
50
|
| 25 |
-
],
|
| 26 |
-
"get_max_context_len": 2048,
|
| 27 |
-
"get_max_seq_len": 128,
|
| 28 |
-
"get_vocab_size": 262144,
|
| 29 |
-
"use_kv_cache": true,
|
| 30 |
-
"has_ple": true,
|
| 31 |
-
"image_placeholder_id": 258880,
|
| 32 |
-
"audio_placeholder_id": 258881,
|
| 33 |
-
"audio_sample_rate": 16000,
|
| 34 |
-
"audio_block_samples": 7680,
|
| 35 |
-
"audio_block_k_min": 1,
|
| 36 |
-
"audio_block_k_max": 62,
|
| 37 |
-
"audio_soft_tokens_max": 744,
|
| 38 |
-
"audio_sample_count_min": 7680,
|
| 39 |
-
"audio_sample_count_max": 476160,
|
| 40 |
-
"token_embedding": {
|
| 41 |
"inputs": [
|
| 42 |
{
|
| 43 |
-
"name": "input_ids",
|
| 44 |
"shape": [
|
| 45 |
1,
|
| 46 |
-
|
| 47 |
-
|
| 48 |
-
|
| 49 |
-
|
|
|
|
|
|
|
|
|
|
| 50 |
],
|
| 51 |
"dtype": "int64"
|
| 52 |
}
|
| 53 |
],
|
| 54 |
"outputs": [
|
| 55 |
{
|
| 56 |
-
"name": "embeds",
|
| 57 |
"shape": [
|
| 58 |
1,
|
| 59 |
-
|
| 60 |
-
|
| 61 |
-
"max": 128
|
| 62 |
-
}
|
| 63 |
],
|
| 64 |
"dtype": "float32"
|
| 65 |
},
|
| 66 |
{
|
| 67 |
-
"name": "ple_tokens",
|
| 68 |
"shape": [
|
| 69 |
1,
|
| 70 |
-
|
| 71 |
-
"min": 1,
|
| 72 |
-
"max": 128
|
| 73 |
-
},
|
| 74 |
-
35,
|
| 75 |
-
256
|
| 76 |
],
|
| 77 |
"dtype": "float32"
|
| 78 |
}
|
|
@@ -81,44 +51,31 @@
|
|
| 81 |
"text_decoder": {
|
| 82 |
"inputs": [
|
| 83 |
{
|
| 84 |
-
"name": "embeds",
|
| 85 |
"shape": [
|
| 86 |
1,
|
| 87 |
-
|
| 88 |
-
"min": 1,
|
| 89 |
-
"max": 128
|
| 90 |
-
},
|
| 91 |
1536
|
| 92 |
],
|
| 93 |
"dtype": "float32"
|
| 94 |
},
|
| 95 |
{
|
| 96 |
-
"name": "ple_tokens",
|
| 97 |
"shape": [
|
| 98 |
1,
|
| 99 |
-
|
| 100 |
-
"min": 1,
|
| 101 |
-
"max": 128
|
| 102 |
-
},
|
| 103 |
35,
|
| 104 |
256
|
| 105 |
],
|
| 106 |
"dtype": "float32"
|
| 107 |
},
|
| 108 |
{
|
| 109 |
-
"name": "input_pos",
|
| 110 |
"shape": [
|
| 111 |
-
|
| 112 |
-
"min": 1,
|
| 113 |
-
"max": 128
|
| 114 |
-
}
|
| 115 |
],
|
| 116 |
"dtype": "int64"
|
| 117 |
}
|
| 118 |
],
|
| 119 |
"outputs": [
|
| 120 |
{
|
| 121 |
-
"name": "logits",
|
| 122 |
"shape": [
|
| 123 |
1,
|
| 124 |
1,
|
|
@@ -128,86 +85,61 @@
|
|
| 128 |
}
|
| 129 |
]
|
| 130 |
},
|
| 131 |
-
"
|
| 132 |
"inputs": [
|
| 133 |
{
|
| 134 |
-
"name": "pixel_values",
|
| 135 |
"shape": [
|
| 136 |
1,
|
| 137 |
-
|
| 138 |
-
448,
|
| 139 |
-
640
|
| 140 |
],
|
| 141 |
-
"dtype": "
|
| 142 |
}
|
| 143 |
],
|
| 144 |
"outputs": [
|
| 145 |
{
|
| 146 |
-
"name": "embeds",
|
| 147 |
"shape": [
|
| 148 |
1,
|
| 149 |
-
|
| 150 |
1536
|
| 151 |
],
|
| 152 |
"dtype": "float32"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 153 |
}
|
| 154 |
]
|
| 155 |
},
|
| 156 |
-
"
|
| 157 |
"inputs": [
|
| 158 |
{
|
| 159 |
-
"name": "waveform",
|
| 160 |
"shape": [
|
| 161 |
1,
|
| 162 |
-
|
| 163 |
-
|
| 164 |
-
|
| 165 |
-
}
|
| 166 |
],
|
| 167 |
"dtype": "float32"
|
| 168 |
-
},
|
| 169 |
-
{
|
| 170 |
-
"name": "num_valid_samples",
|
| 171 |
-
"shape": [
|
| 172 |
-
1
|
| 173 |
-
],
|
| 174 |
-
"dtype": "int64"
|
| 175 |
}
|
| 176 |
],
|
| 177 |
"outputs": [
|
| 178 |
{
|
| 179 |
-
"name": "audio_embeds",
|
| 180 |
"shape": [
|
| 181 |
1,
|
| 182 |
-
|
| 183 |
-
"min": 12,
|
| 184 |
-
"max": 744
|
| 185 |
-
},
|
| 186 |
1536
|
| 187 |
],
|
| 188 |
"dtype": "float32"
|
| 189 |
-
},
|
| 190 |
-
{
|
| 191 |
-
"name": "audio_mask",
|
| 192 |
-
"shape": [
|
| 193 |
-
1,
|
| 194 |
-
{
|
| 195 |
-
"min": 12,
|
| 196 |
-
"max": 744
|
| 197 |
-
}
|
| 198 |
-
],
|
| 199 |
-
"dtype": "float32"
|
| 200 |
}
|
| 201 |
]
|
| 202 |
}
|
| 203 |
}
|
| 204 |
}
|
| 205 |
],
|
| 206 |
-
"size": "e2b"
|
| 207 |
-
"_methods_needing_shapes": [
|
| 208 |
-
"audio_encoder",
|
| 209 |
-
"text_decoder",
|
| 210 |
-
"token_embedding",
|
| 211 |
-
"vision_encoder"
|
| 212 |
-
]
|
| 213 |
}
|
|
|
|
| 13 |
{
|
| 14 |
"file": "gemma_4_e2b_vulkan_8da4w.pte",
|
| 15 |
"precision": "8da4w",
|
|
|
|
|
|
|
| 16 |
"methods": {
|
| 17 |
+
"audio_encoder": {
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 18 |
"inputs": [
|
| 19 |
{
|
|
|
|
| 20 |
"shape": [
|
| 21 |
1,
|
| 22 |
+
476160
|
| 23 |
+
],
|
| 24 |
+
"dtype": "float32"
|
| 25 |
+
},
|
| 26 |
+
{
|
| 27 |
+
"shape": [
|
| 28 |
+
1
|
| 29 |
],
|
| 30 |
"dtype": "int64"
|
| 31 |
}
|
| 32 |
],
|
| 33 |
"outputs": [
|
| 34 |
{
|
|
|
|
| 35 |
"shape": [
|
| 36 |
1,
|
| 37 |
+
744,
|
| 38 |
+
1536
|
|
|
|
|
|
|
| 39 |
],
|
| 40 |
"dtype": "float32"
|
| 41 |
},
|
| 42 |
{
|
|
|
|
| 43 |
"shape": [
|
| 44 |
1,
|
| 45 |
+
744
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 46 |
],
|
| 47 |
"dtype": "float32"
|
| 48 |
}
|
|
|
|
| 51 |
"text_decoder": {
|
| 52 |
"inputs": [
|
| 53 |
{
|
|
|
|
| 54 |
"shape": [
|
| 55 |
1,
|
| 56 |
+
128,
|
|
|
|
|
|
|
|
|
|
| 57 |
1536
|
| 58 |
],
|
| 59 |
"dtype": "float32"
|
| 60 |
},
|
| 61 |
{
|
|
|
|
| 62 |
"shape": [
|
| 63 |
1,
|
| 64 |
+
128,
|
|
|
|
|
|
|
|
|
|
| 65 |
35,
|
| 66 |
256
|
| 67 |
],
|
| 68 |
"dtype": "float32"
|
| 69 |
},
|
| 70 |
{
|
|
|
|
| 71 |
"shape": [
|
| 72 |
+
128
|
|
|
|
|
|
|
|
|
|
| 73 |
],
|
| 74 |
"dtype": "int64"
|
| 75 |
}
|
| 76 |
],
|
| 77 |
"outputs": [
|
| 78 |
{
|
|
|
|
| 79 |
"shape": [
|
| 80 |
1,
|
| 81 |
1,
|
|
|
|
| 85 |
}
|
| 86 |
]
|
| 87 |
},
|
| 88 |
+
"token_embedding": {
|
| 89 |
"inputs": [
|
| 90 |
{
|
|
|
|
| 91 |
"shape": [
|
| 92 |
1,
|
| 93 |
+
128
|
|
|
|
|
|
|
| 94 |
],
|
| 95 |
+
"dtype": "int64"
|
| 96 |
}
|
| 97 |
],
|
| 98 |
"outputs": [
|
| 99 |
{
|
|
|
|
| 100 |
"shape": [
|
| 101 |
1,
|
| 102 |
+
128,
|
| 103 |
1536
|
| 104 |
],
|
| 105 |
"dtype": "float32"
|
| 106 |
+
},
|
| 107 |
+
{
|
| 108 |
+
"shape": [
|
| 109 |
+
1,
|
| 110 |
+
128,
|
| 111 |
+
35,
|
| 112 |
+
256
|
| 113 |
+
],
|
| 114 |
+
"dtype": "float32"
|
| 115 |
}
|
| 116 |
]
|
| 117 |
},
|
| 118 |
+
"vision_encoder": {
|
| 119 |
"inputs": [
|
| 120 |
{
|
|
|
|
| 121 |
"shape": [
|
| 122 |
1,
|
| 123 |
+
3,
|
| 124 |
+
448,
|
| 125 |
+
640
|
|
|
|
| 126 |
],
|
| 127 |
"dtype": "float32"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 128 |
}
|
| 129 |
],
|
| 130 |
"outputs": [
|
| 131 |
{
|
|
|
|
| 132 |
"shape": [
|
| 133 |
1,
|
| 134 |
+
280,
|
|
|
|
|
|
|
|
|
|
| 135 |
1536
|
| 136 |
],
|
| 137 |
"dtype": "float32"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 138 |
}
|
| 139 |
]
|
| 140 |
}
|
| 141 |
}
|
| 142 |
}
|
| 143 |
],
|
| 144 |
+
"size": "e2b"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 145 |
}
|
e2b/xnnpack/config.json
CHANGED
|
@@ -14,63 +14,35 @@
|
|
| 14 |
"file": "gemma_4_e2b_xnnpack_8da4w.pte",
|
| 15 |
"precision": "8da4w",
|
| 16 |
"methods": {
|
| 17 |
-
"
|
| 18 |
-
"get_bos_id": 2,
|
| 19 |
-
"get_eos_ids": [
|
| 20 |
-
1,
|
| 21 |
-
106,
|
| 22 |
-
50
|
| 23 |
-
],
|
| 24 |
-
"get_max_context_len": 4096,
|
| 25 |
-
"get_max_seq_len": 512,
|
| 26 |
-
"get_vocab_size": 262144,
|
| 27 |
-
"use_kv_cache": true,
|
| 28 |
-
"has_ple": true,
|
| 29 |
-
"image_placeholder_id": 258880,
|
| 30 |
-
"audio_placeholder_id": 258881,
|
| 31 |
-
"audio_sample_rate": 16000,
|
| 32 |
-
"audio_block_samples": 7680,
|
| 33 |
-
"audio_block_k_min": 1,
|
| 34 |
-
"audio_block_k_max": 62,
|
| 35 |
-
"audio_soft_tokens_max": 744,
|
| 36 |
-
"audio_sample_count_min": 7680,
|
| 37 |
-
"audio_sample_count_max": 476160,
|
| 38 |
-
"token_embedding": {
|
| 39 |
"inputs": [
|
| 40 |
{
|
| 41 |
-
"name": "input_ids",
|
| 42 |
"shape": [
|
| 43 |
1,
|
| 44 |
-
|
| 45 |
-
|
| 46 |
-
|
| 47 |
-
|
|
|
|
|
|
|
|
|
|
| 48 |
],
|
| 49 |
"dtype": "int64"
|
| 50 |
}
|
| 51 |
],
|
| 52 |
"outputs": [
|
| 53 |
{
|
| 54 |
-
"name": "embeds",
|
| 55 |
"shape": [
|
| 56 |
1,
|
| 57 |
-
|
| 58 |
-
|
| 59 |
-
"max": 1024
|
| 60 |
-
}
|
| 61 |
],
|
| 62 |
"dtype": "float32"
|
| 63 |
},
|
| 64 |
{
|
| 65 |
-
"name": "ple_tokens",
|
| 66 |
"shape": [
|
| 67 |
1,
|
| 68 |
-
|
| 69 |
-
"min": 1,
|
| 70 |
-
"max": 1024
|
| 71 |
-
},
|
| 72 |
-
35,
|
| 73 |
-
256
|
| 74 |
],
|
| 75 |
"dtype": "float32"
|
| 76 |
}
|
|
@@ -79,44 +51,31 @@
|
|
| 79 |
"text_decoder": {
|
| 80 |
"inputs": [
|
| 81 |
{
|
| 82 |
-
"name": "embeds",
|
| 83 |
"shape": [
|
| 84 |
1,
|
| 85 |
-
|
| 86 |
-
"min": 1,
|
| 87 |
-
"max": 512
|
| 88 |
-
},
|
| 89 |
1536
|
| 90 |
],
|
| 91 |
"dtype": "float32"
|
| 92 |
},
|
| 93 |
{
|
| 94 |
-
"name": "ple_tokens",
|
| 95 |
"shape": [
|
| 96 |
1,
|
| 97 |
-
|
| 98 |
-
"min": 1,
|
| 99 |
-
"max": 512
|
| 100 |
-
},
|
| 101 |
35,
|
| 102 |
256
|
| 103 |
],
|
| 104 |
"dtype": "float32"
|
| 105 |
},
|
| 106 |
{
|
| 107 |
-
"name": "input_pos",
|
| 108 |
"shape": [
|
| 109 |
-
|
| 110 |
-
"min": 1,
|
| 111 |
-
"max": 512
|
| 112 |
-
}
|
| 113 |
],
|
| 114 |
"dtype": "int64"
|
| 115 |
}
|
| 116 |
],
|
| 117 |
"outputs": [
|
| 118 |
{
|
| 119 |
-
"name": "logits",
|
| 120 |
"shape": [
|
| 121 |
1,
|
| 122 |
1,
|
|
@@ -126,86 +85,61 @@
|
|
| 126 |
}
|
| 127 |
]
|
| 128 |
},
|
| 129 |
-
"
|
| 130 |
"inputs": [
|
| 131 |
{
|
| 132 |
-
"name": "pixel_values",
|
| 133 |
"shape": [
|
| 134 |
1,
|
| 135 |
-
|
| 136 |
-
448,
|
| 137 |
-
640
|
| 138 |
],
|
| 139 |
-
"dtype": "
|
| 140 |
}
|
| 141 |
],
|
| 142 |
"outputs": [
|
| 143 |
{
|
| 144 |
-
"name": "embeds",
|
| 145 |
"shape": [
|
| 146 |
1,
|
| 147 |
-
|
| 148 |
1536
|
| 149 |
],
|
| 150 |
"dtype": "float32"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 151 |
}
|
| 152 |
]
|
| 153 |
},
|
| 154 |
-
"
|
| 155 |
"inputs": [
|
| 156 |
{
|
| 157 |
-
"name": "waveform",
|
| 158 |
"shape": [
|
| 159 |
1,
|
| 160 |
-
|
| 161 |
-
|
| 162 |
-
|
| 163 |
-
}
|
| 164 |
],
|
| 165 |
"dtype": "float32"
|
| 166 |
-
},
|
| 167 |
-
{
|
| 168 |
-
"name": "num_valid_samples",
|
| 169 |
-
"shape": [
|
| 170 |
-
1
|
| 171 |
-
],
|
| 172 |
-
"dtype": "int64"
|
| 173 |
}
|
| 174 |
],
|
| 175 |
"outputs": [
|
| 176 |
{
|
| 177 |
-
"name": "audio_embeds",
|
| 178 |
"shape": [
|
| 179 |
1,
|
| 180 |
-
|
| 181 |
-
"min": 12,
|
| 182 |
-
"max": 744
|
| 183 |
-
},
|
| 184 |
1536
|
| 185 |
],
|
| 186 |
"dtype": "float32"
|
| 187 |
-
},
|
| 188 |
-
{
|
| 189 |
-
"name": "audio_mask",
|
| 190 |
-
"shape": [
|
| 191 |
-
1,
|
| 192 |
-
{
|
| 193 |
-
"min": 12,
|
| 194 |
-
"max": 744
|
| 195 |
-
}
|
| 196 |
-
],
|
| 197 |
-
"dtype": "float32"
|
| 198 |
}
|
| 199 |
]
|
| 200 |
}
|
| 201 |
}
|
| 202 |
}
|
| 203 |
],
|
| 204 |
-
"size": "e2b"
|
| 205 |
-
"_methods_needing_shapes": [
|
| 206 |
-
"audio_encoder",
|
| 207 |
-
"text_decoder",
|
| 208 |
-
"token_embedding",
|
| 209 |
-
"vision_encoder"
|
| 210 |
-
]
|
| 211 |
}
|
|
|
|
| 14 |
"file": "gemma_4_e2b_xnnpack_8da4w.pte",
|
| 15 |
"precision": "8da4w",
|
| 16 |
"methods": {
|
| 17 |
+
"audio_encoder": {
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 18 |
"inputs": [
|
| 19 |
{
|
|
|
|
| 20 |
"shape": [
|
| 21 |
1,
|
| 22 |
+
476160
|
| 23 |
+
],
|
| 24 |
+
"dtype": "float32"
|
| 25 |
+
},
|
| 26 |
+
{
|
| 27 |
+
"shape": [
|
| 28 |
+
1
|
| 29 |
],
|
| 30 |
"dtype": "int64"
|
| 31 |
}
|
| 32 |
],
|
| 33 |
"outputs": [
|
| 34 |
{
|
|
|
|
| 35 |
"shape": [
|
| 36 |
1,
|
| 37 |
+
744,
|
| 38 |
+
1536
|
|
|
|
|
|
|
| 39 |
],
|
| 40 |
"dtype": "float32"
|
| 41 |
},
|
| 42 |
{
|
|
|
|
| 43 |
"shape": [
|
| 44 |
1,
|
| 45 |
+
744
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 46 |
],
|
| 47 |
"dtype": "float32"
|
| 48 |
}
|
|
|
|
| 51 |
"text_decoder": {
|
| 52 |
"inputs": [
|
| 53 |
{
|
|
|
|
| 54 |
"shape": [
|
| 55 |
1,
|
| 56 |
+
512,
|
|
|
|
|
|
|
|
|
|
| 57 |
1536
|
| 58 |
],
|
| 59 |
"dtype": "float32"
|
| 60 |
},
|
| 61 |
{
|
|
|
|
| 62 |
"shape": [
|
| 63 |
1,
|
| 64 |
+
512,
|
|
|
|
|
|
|
|
|
|
| 65 |
35,
|
| 66 |
256
|
| 67 |
],
|
| 68 |
"dtype": "float32"
|
| 69 |
},
|
| 70 |
{
|
|
|
|
| 71 |
"shape": [
|
| 72 |
+
512
|
|
|
|
|
|
|
|
|
|
| 73 |
],
|
| 74 |
"dtype": "int64"
|
| 75 |
}
|
| 76 |
],
|
| 77 |
"outputs": [
|
| 78 |
{
|
|
|
|
| 79 |
"shape": [
|
| 80 |
1,
|
| 81 |
1,
|
|
|
|
| 85 |
}
|
| 86 |
]
|
| 87 |
},
|
| 88 |
+
"token_embedding": {
|
| 89 |
"inputs": [
|
| 90 |
{
|
|
|
|
| 91 |
"shape": [
|
| 92 |
1,
|
| 93 |
+
512
|
|
|
|
|
|
|
| 94 |
],
|
| 95 |
+
"dtype": "int64"
|
| 96 |
}
|
| 97 |
],
|
| 98 |
"outputs": [
|
| 99 |
{
|
|
|
|
| 100 |
"shape": [
|
| 101 |
1,
|
| 102 |
+
512,
|
| 103 |
1536
|
| 104 |
],
|
| 105 |
"dtype": "float32"
|
| 106 |
+
},
|
| 107 |
+
{
|
| 108 |
+
"shape": [
|
| 109 |
+
1,
|
| 110 |
+
512,
|
| 111 |
+
35,
|
| 112 |
+
256
|
| 113 |
+
],
|
| 114 |
+
"dtype": "float32"
|
| 115 |
}
|
| 116 |
]
|
| 117 |
},
|
| 118 |
+
"vision_encoder": {
|
| 119 |
"inputs": [
|
| 120 |
{
|
|
|
|
| 121 |
"shape": [
|
| 122 |
1,
|
| 123 |
+
3,
|
| 124 |
+
448,
|
| 125 |
+
640
|
|
|
|
| 126 |
],
|
| 127 |
"dtype": "float32"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 128 |
}
|
| 129 |
],
|
| 130 |
"outputs": [
|
| 131 |
{
|
|
|
|
| 132 |
"shape": [
|
| 133 |
1,
|
| 134 |
+
280,
|
|
|
|
|
|
|
|
|
|
| 135 |
1536
|
| 136 |
],
|
| 137 |
"dtype": "float32"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 138 |
}
|
| 139 |
]
|
| 140 |
}
|
| 141 |
}
|
| 142 |
}
|
| 143 |
],
|
| 144 |
+
"size": "e2b"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 145 |
}
|