web-llm icon indicating copy to clipboard operation
web-llm copied to clipboard

Where is the big vicuna-7b-v0 model file in local computer for chrome loading?

Open eatcosmos opened this issue 1 year ago • 3 comments

Have run well with GPU Intel(R) Iris(R) Xe Graphics 8G.

I want to see what the model like:

I search by file size or file keywords(pkl model wgsl) through the whole computer drivers, still cannot find the vicuna-7b-v0 model. I also search through the code by keywords like save、model,can not find where save downloaded file params_shard_1.bin...

It's strange, where is the model file and which code deal with the model file and save them to computer drivers for chrome?


image

eatcosmos avatar Apr 19 '23 18:04 eatcosmos

C:\Users\{YOUR_NAME}\AppData\Local\Google\Chrome SxS\User Data\Default\Service Worker\CacheStorage

yunghoy avatar Apr 19 '23 22:04 yunghoy

image

eatcosmos avatar Apr 20 '23 13:04 eatcosmos

@eatcosmos

I search by file size or file keywords(pkl model wgsl) through the whole computer drivers,

In similar situations, I search by date and see the most recently created files. I use Windwos 10 Home edition and Agent Ransack (it's free) which allows me to search what was created in the current day only.

If you want to see clean filenames (instead of random ones), here's what worked for me (plus, you can download them to a location you specify): https://github.com/mlc-ai/web-llm/issues/19#issuecomment-1511754031

It's strange, where is the model file

The model is downloaded and used as shards, without being glued into one big file: https://github.com/mlc-ai/web-llm/issues/39

zeritonius avatar Apr 22 '23 12:04 zeritonius