huggingworld commited on
Commit
5e07315
·
verified ·
1 Parent(s): 5ac4ba5

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +6 -28
README.md CHANGED
@@ -46,9 +46,9 @@ import { VectorSearch } from 'https://cdn.jsdelivr.net/gh/jasonmayes/VectorSearc
46
 
47
  // Embedding Model Configuration.
48
  const MODEL_RUNTIME = 'litertjs'; // OR 'transformersjs'
49
- const MODEL_URL = 'model/embeddinggemma-300M_seq1024_mixed-precision.tflite'; // OR 'Xenova/all-MiniLM-L6-v2' if transformersjs runtime.
50
  const SEQ_LENGTH = 1024;
51
- const TOKENIZER = 'onnx-community/embeddinggemma-300m-ONNX';
52
  const EMBEDDING_MODEL_CONFIG = {
53
  runtime: MODEL_RUNTIME,
54
  url: MODEL_URL,
@@ -155,7 +155,7 @@ This model is available to download from HuggingFace which you must do yourself
155
  const MODEL_RUNTIME = 'litertjs';
156
  const MODEL_URL = 'model/embeddinggemma-300M_seq1024_mixed-precision.tflite';
157
  const SEQ_LENGTH = 1024;
158
- const TOKENIZER = 'onnx-community/embeddinggemma-300m-ONNX';
159
  const EMBEDDING_MODEL_CONFIG = {
160
  runtime: MODEL_RUNTIME,
161
  url: MODEL_URL,
@@ -178,9 +178,9 @@ If you wish to use the all-MiniLM-L6-v2 embedding model instead for speed you ca
178
  ```javascript
179
  // Embedding Model Configuration.
180
  const MODEL_RUNTIME = 'transformersjs';
181
- const MODEL_URL = 'Xenova/all-MiniLM-L6-v2';
182
  const SEQ_LENGTH = 128;
183
- const TOKENIZER = 'onnx-community/embeddinggemma-300m-ONNX';
184
  const EMBEDDING_MODEL_CONFIG = {
185
  runtime: MODEL_RUNTIME,
186
  url: MODEL_URL,
@@ -199,29 +199,7 @@ However please note this model is faster for a few reasons:
199
 
200
  ### LiteRT.js Wasm files (optional self host)
201
 
202
- See the demo folder in this repo that contains a "wasm" sub folder with all the Web Assembly files needed for the LiteRT.js runtime. You can choose to serve these files yourself and update the config object if you do so but remember to enable CORS headers on your server so the files can be used if you do that. If you are curious to learn more about these files see the [official LiteRT.js documentation](https://ai.google.dev/edge/litert/web).
203
-
204
- By default the library pulls in these Wasm files from JSDeliver CDN.
205
-
206
- If your hosted version is not in the same location update the config object to specify the new Wasm folder location on your webserver as follows:
207
-
208
- ```javascript
209
- // Embedding Model Configuration.
210
- const MODEL_RUNTIME = 'litertjs';
211
- const MODEL_URL = 'model/embeddinggemma-300M_seq1024_mixed-precision.tflite';
212
- const SEQ_LENGTH = 1024;
213
- const TOKENIZER = 'onnx-community/embeddinggemma-300m-ONNX';
214
- const EMBEDDING_MODEL_CONFIG = {
215
- runtime: MODEL_RUNTIME,
216
- litertjsWasmUrl: '/wasm', // Specify your path to your custom hosted Wasm files here!
217
- url: MODEL_URL,
218
- sequenceLength: SEQ_LENGTH,
219
- tokenizer: TOKENIZER
220
- };
221
-
222
- // Instantiate VectorSearch Master Class.
223
- const VECTOR_SEARCH = new VectorSearch(EMBEDDING_MODEL_CONFIG);
224
- ```
225
 
226
  Note when you call load you can also optionally specify a HTML element to render loading status updates to like this:
227
 
 
46
 
47
  // Embedding Model Configuration.
48
  const MODEL_RUNTIME = 'litertjs'; // OR 'transformersjs'
49
+ const MODEL_URL = 'model/embeddinggemma-300M_seq1024_mixed-precision.tflite'; // OR 'huggingworld/all-MiniLM-L6-v2' if transformersjs runtime.
50
  const SEQ_LENGTH = 1024;
51
+ const TOKENIZER = 'huggingworld/embeddinggemma-300m-ONNX';
52
  const EMBEDDING_MODEL_CONFIG = {
53
  runtime: MODEL_RUNTIME,
54
  url: MODEL_URL,
 
155
  const MODEL_RUNTIME = 'litertjs';
156
  const MODEL_URL = 'model/embeddinggemma-300M_seq1024_mixed-precision.tflite';
157
  const SEQ_LENGTH = 1024;
158
+ const TOKENIZER = 'huggingworld/embeddinggemma-300m-ONNX';
159
  const EMBEDDING_MODEL_CONFIG = {
160
  runtime: MODEL_RUNTIME,
161
  url: MODEL_URL,
 
178
  ```javascript
179
  // Embedding Model Configuration.
180
  const MODEL_RUNTIME = 'transformersjs';
181
+ const MODEL_URL = 'huggingworld/all-MiniLM-L6-v2';
182
  const SEQ_LENGTH = 128;
183
+ const TOKENIZER = 'huggingworld/embeddinggemma-300m-ONNX';
184
  const EMBEDDING_MODEL_CONFIG = {
185
  runtime: MODEL_RUNTIME,
186
  url: MODEL_URL,
 
199
 
200
  ### LiteRT.js Wasm files (optional self host)
201
 
202
+ Clone the repo of the space, git clone --depth 1 https://huggingface.co/spaces/huggingworld/vectorsearch-turbo-webgpu, all assets have been included.
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
203
 
204
  Note when you call load you can also optionally specify a HTML element to render loading status updates to like this:
205