Update index.js
Browse files
index.js
CHANGED
@@ -19,7 +19,7 @@ const fileUpload = document.getElementById('upload');
|
|
19 |
const imageContainer = document.getElementById('container');
|
20 |
const example = document.getElementById('example');
|
21 |
|
22 |
-
let ortSessionA, ortSessionB, ortSessionC;
|
23 |
let config;
|
24 |
|
25 |
async function initializeSessions() {
|
@@ -29,19 +29,16 @@ async function initializeSessions() {
|
|
29 |
await getModelFile(ONNX_MODEL, `onnx/QwenVL_A_${QUANT}.onnx`),
|
30 |
{ executionProviders: ["webgpu"] }
|
31 |
);
|
32 |
-
console.log({ortSessionA});
|
33 |
|
34 |
ortSessionB = await ort.InferenceSession.create(
|
35 |
await getModelFile(ONNX_MODEL, `onnx/QwenVL_B_${QUANT}.onnx`),
|
36 |
{ executionProviders: ["webgpu"] }
|
37 |
);
|
38 |
-
console.log({ortSessionB});
|
39 |
|
40 |
ortSessionC = await ort.InferenceSession.create(
|
41 |
await getModelFile(ONNX_MODEL, `onnx/QwenVL_C_${QUANT}.onnx`),
|
42 |
{ executionProviders: ["webgpu"] }
|
43 |
);
|
44 |
-
console.log({ortSessionC});
|
45 |
|
46 |
config = (await getModelJSON(BASE_MODEL, "config.json"));
|
47 |
|
@@ -77,7 +74,6 @@ export async function imageTextToText(
|
|
77 |
query,
|
78 |
vision = true
|
79 |
) {
|
80 |
-
let ortSessionA, ortSessionB, ortSessionC, ortSessionD, ortSessionE;
|
81 |
|
82 |
const prompt_head_len = new Tensor("int64", new BigInt64Array([5n]), [1]);
|
83 |
|
@@ -136,12 +132,6 @@ export async function imageTextToText(
|
|
136 |
|
137 |
const dummy = new ort.Tensor("int32", new Int32Array([0]), []);
|
138 |
|
139 |
-
if (!ortSessionB) {
|
140 |
-
await ort.InferenceSession.create(
|
141 |
-
await getModelFile(ONNX_MODEL, `onnx/QwenVL_B_${QUANT}.onnx`),
|
142 |
-
{ executionProviders: ["webgpu"] }
|
143 |
-
);
|
144 |
-
}
|
145 |
let { hidden_states } = await ortSessionB.run({
|
146 |
input_ids: input_ids,
|
147 |
ids_len: ids_len,
|
|
|
19 |
const imageContainer = document.getElementById('container');
|
20 |
const example = document.getElementById('example');
|
21 |
|
22 |
+
let ortSessionA, ortSessionB, ortSessionC, ortSessionD, ortSessionE;
|
23 |
let config;
|
24 |
|
25 |
async function initializeSessions() {
|
|
|
29 |
await getModelFile(ONNX_MODEL, `onnx/QwenVL_A_${QUANT}.onnx`),
|
30 |
{ executionProviders: ["webgpu"] }
|
31 |
);
|
|
|
32 |
|
33 |
ortSessionB = await ort.InferenceSession.create(
|
34 |
await getModelFile(ONNX_MODEL, `onnx/QwenVL_B_${QUANT}.onnx`),
|
35 |
{ executionProviders: ["webgpu"] }
|
36 |
);
|
|
|
37 |
|
38 |
ortSessionC = await ort.InferenceSession.create(
|
39 |
await getModelFile(ONNX_MODEL, `onnx/QwenVL_C_${QUANT}.onnx`),
|
40 |
{ executionProviders: ["webgpu"] }
|
41 |
);
|
|
|
42 |
|
43 |
config = (await getModelJSON(BASE_MODEL, "config.json"));
|
44 |
|
|
|
74 |
query,
|
75 |
vision = true
|
76 |
) {
|
|
|
77 |
|
78 |
const prompt_head_len = new Tensor("int64", new BigInt64Array([5n]), [1]);
|
79 |
|
|
|
132 |
|
133 |
const dummy = new ort.Tensor("int32", new Int32Array([0]), []);
|
134 |
|
|
|
|
|
|
|
|
|
|
|
|
|
135 |
let { hidden_states } = await ortSessionB.run({
|
136 |
input_ids: input_ids,
|
137 |
ids_len: ids_len,
|