Commit ·
3cdead0
1
Parent(s): 193632d
Use min/max dynamic shape bounds for forward inputs/outputs
Browse files- e2b/mlx/config.json +28 -6
- e2b/vulkan/config.json +28 -6
- e2b/xnnpack/config.json +28 -6
e2b/mlx/config.json
CHANGED
|
@@ -25,14 +25,36 @@
|
|
| 25 |
"get_max_seq_len": 2048,
|
| 26 |
"use_kv_cache": true,
|
| 27 |
"forward": {
|
| 28 |
-
"inputs": [
|
| 29 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 30 |
}
|
| 31 |
}
|
| 32 |
}
|
| 33 |
],
|
| 34 |
-
"size": "e2b"
|
| 35 |
-
"_methods_needing_shapes": [
|
| 36 |
-
"forward"
|
| 37 |
-
]
|
| 38 |
}
|
|
|
|
| 25 |
"get_max_seq_len": 2048,
|
| 26 |
"use_kv_cache": true,
|
| 27 |
"forward": {
|
| 28 |
+
"inputs": [
|
| 29 |
+
{
|
| 30 |
+
"name": "tokens",
|
| 31 |
+
"shape": [
|
| 32 |
+
1,
|
| 33 |
+
{
|
| 34 |
+
"min": 1,
|
| 35 |
+
"max": 64
|
| 36 |
+
}
|
| 37 |
+
],
|
| 38 |
+
"dtype": "int64"
|
| 39 |
+
}
|
| 40 |
+
],
|
| 41 |
+
"outputs": [
|
| 42 |
+
{
|
| 43 |
+
"name": "logits",
|
| 44 |
+
"shape": [
|
| 45 |
+
1,
|
| 46 |
+
{
|
| 47 |
+
"min": 1,
|
| 48 |
+
"max": 64
|
| 49 |
+
},
|
| 50 |
+
262144
|
| 51 |
+
],
|
| 52 |
+
"dtype": "float32"
|
| 53 |
+
}
|
| 54 |
+
]
|
| 55 |
}
|
| 56 |
}
|
| 57 |
}
|
| 58 |
],
|
| 59 |
+
"size": "e2b"
|
|
|
|
|
|
|
|
|
|
| 60 |
}
|
e2b/vulkan/config.json
CHANGED
|
@@ -26,14 +26,36 @@
|
|
| 26 |
"get_vocab_size": 262144,
|
| 27 |
"use_kv_cache": true,
|
| 28 |
"forward": {
|
| 29 |
-
"inputs": [
|
| 30 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 31 |
}
|
| 32 |
}
|
| 33 |
}
|
| 34 |
],
|
| 35 |
-
"size": "e2b"
|
| 36 |
-
"_methods_needing_shapes": [
|
| 37 |
-
"forward"
|
| 38 |
-
]
|
| 39 |
}
|
|
|
|
| 26 |
"get_vocab_size": 262144,
|
| 27 |
"use_kv_cache": true,
|
| 28 |
"forward": {
|
| 29 |
+
"inputs": [
|
| 30 |
+
{
|
| 31 |
+
"name": "tokens",
|
| 32 |
+
"shape": [
|
| 33 |
+
1,
|
| 34 |
+
{
|
| 35 |
+
"min": 1,
|
| 36 |
+
"max": 2048
|
| 37 |
+
}
|
| 38 |
+
],
|
| 39 |
+
"dtype": "int64"
|
| 40 |
+
}
|
| 41 |
+
],
|
| 42 |
+
"outputs": [
|
| 43 |
+
{
|
| 44 |
+
"name": "logits",
|
| 45 |
+
"shape": [
|
| 46 |
+
1,
|
| 47 |
+
{
|
| 48 |
+
"min": 1,
|
| 49 |
+
"max": 2048
|
| 50 |
+
},
|
| 51 |
+
262144
|
| 52 |
+
],
|
| 53 |
+
"dtype": "float32"
|
| 54 |
+
}
|
| 55 |
+
]
|
| 56 |
}
|
| 57 |
}
|
| 58 |
}
|
| 59 |
],
|
| 60 |
+
"size": "e2b"
|
|
|
|
|
|
|
|
|
|
| 61 |
}
|
e2b/xnnpack/config.json
CHANGED
|
@@ -26,14 +26,36 @@
|
|
| 26 |
"get_vocab_size": 262144,
|
| 27 |
"use_kv_cache": true,
|
| 28 |
"forward": {
|
| 29 |
-
"inputs": [
|
| 30 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 31 |
}
|
| 32 |
}
|
| 33 |
}
|
| 34 |
],
|
| 35 |
-
"size": "e2b"
|
| 36 |
-
"_methods_needing_shapes": [
|
| 37 |
-
"forward"
|
| 38 |
-
]
|
| 39 |
}
|
|
|
|
| 26 |
"get_vocab_size": 262144,
|
| 27 |
"use_kv_cache": true,
|
| 28 |
"forward": {
|
| 29 |
+
"inputs": [
|
| 30 |
+
{
|
| 31 |
+
"name": "tokens",
|
| 32 |
+
"shape": [
|
| 33 |
+
1,
|
| 34 |
+
{
|
| 35 |
+
"min": 1,
|
| 36 |
+
"max": 2048
|
| 37 |
+
}
|
| 38 |
+
],
|
| 39 |
+
"dtype": "int64"
|
| 40 |
+
}
|
| 41 |
+
],
|
| 42 |
+
"outputs": [
|
| 43 |
+
{
|
| 44 |
+
"name": "logits",
|
| 45 |
+
"shape": [
|
| 46 |
+
1,
|
| 47 |
+
{
|
| 48 |
+
"min": 1,
|
| 49 |
+
"max": 2048
|
| 50 |
+
},
|
| 51 |
+
262144
|
| 52 |
+
],
|
| 53 |
+
"dtype": "float32"
|
| 54 |
+
}
|
| 55 |
+
]
|
| 56 |
}
|
| 57 |
}
|
| 58 |
}
|
| 59 |
],
|
| 60 |
+
"size": "e2b"
|
|
|
|
|
|
|
|
|
|
| 61 |
}
|