msluszniak's picture
fix(v0.9.0): rename vision→image; restore text encoder from v0.8.0
68bad8b verified
{
"$schema": "https://huggingface.co/software-mansion/react-native-executorch-spec/resolve/main/config.schema.json",
"model": "clip_vit_base_patch32",
"family": "clip",
"capabilities": [
"text-embedding",
"image-embedding"
],
"backend": "xnnpack",
"license": "mit",
"tokenizer": "../tokenizer.json",
"variants": [
{
"file": null,
"components": {
"image": "clip_vit_base_patch32_image_xnnpack_fp32.pte",
"text": "clip_vit_base_patch32_text_xnnpack_fp32.pte"
},
"precision": "fp32",
"quantized": false,
"default": true,
"methods": {
"image": {
"inputs": [
{
"shape": [
1,
3,
224,
224
],
"dtype": "float32"
}
],
"outputs": [
{
"shape": [
1,
512
],
"dtype": "float32"
}
]
},
"text": {
"inputs": [
{
"shape": [
1,
77
],
"dtype": "int64"
}
],
"outputs": [
{
"shape": [
1,
512
],
"dtype": "float32"
}
]
}
}
},
{
"file": null,
"components": {
"image": "clip_vit_base_patch32_image_xnnpack_int8.pte",
"text": "clip_vit_base_patch32_text_xnnpack_fp32.pte"
},
"precision": "int8",
"quantized": true,
"default": true,
"methods": {
"image": {
"inputs": [
{
"shape": [
1,
3,
224,
224
],
"dtype": "float32"
}
],
"outputs": [
{
"shape": [
1,
512
],
"dtype": "float32"
}
]
},
"text": {
"inputs": [
{
"shape": [
1,
77
],
"dtype": "int64"
}
],
"outputs": [
{
"shape": [
1,
512
],
"dtype": "float32"
}
]
}
}
}
]
}