The download is too slow. Is there any way to improve the download speed

[LoveSlience]

I spent several hours downloading model， Is there any way to improve the download speed， or download it locally and then load it

[felladrin]

Yes, you can. You can clone any mlc-chat repository from HuggingFace (that has been built for WebGPU) and load them from localhost using a config like this:

import * as webLLM from "@mlc-ai/web-llm";

const modelId = "Some-Model-Name-q4f32_1";

const appConfig = {
  model_list: [
    {
      model_url:
        "http://localhost:8080/params/",
      local_id: modelId,
    },
  ],
  model_lib_map: {
    [modelId]:
      "http://localhost:8080/Some-Model-Name-q4f32_1-webgpu.wasm",
  },
};

const chatConfig = {
  temperature: 0.7,
  repetition_penalty: 1.2
};

const chat = new webLLM.ChatModule();

await chat.reload(modelId, chatConfig, appConfig);

Note
Some models uploaded to HuggingFace, won't contain the .wasm file. Their repository will contain only the params at the root of the repository. In those cases, usually, it's because the .wasm should be loaded from https://github.com/mlc-ai/binary-mlc-llm-libs.
Examples:
Repository without .wasm file (params are in the root)
Repository containing .wasm file (params are inside params folder)

[LoveSlience]

Thank you very much, I download this modal mlc-ai/mlc-chat-vicuna-v1-7b-q4f32_0 but I can't find the wasm file. I can only use an online address or download it locally separately. Can you add wasm files to the warehouse

const modelId = "vicuna-v1-7b-q4f32_0";

const appConfig = {
    model_list: [
      {
        model_url:
          "http://localhost:3000",
        local_id: modelId,
      },
    ],
    model_lib_map: {
      [modelId]:
        "https://raw.githubusercontent.com/mlc-ai/binary-mlc-llm-libs/main/vicuna-v1-7b-q4f32_0-webgpu-v1.wasm",
    },
  };

const chatConfig = {
    temperature: 0.7,
    repetition_penalty: 1.2
  };

[felladrin]

Great you got it working, @LoveSlience!
That's good that you pointed that out. I've updated my comment above to clarify that for future readers.
About mlc-ai/mlc-chat-vicuna-v1-7b-q4f32_0, unfortunately, I can't change that repo, as I'm not part of the mlc-ai team.