Issue about model downloads #2
Description
This issue arises when using version 0.1.0 #1 on a Windows system.
When I started the FLM server for the first time, there were some issues with pulling the model file. Then I found the network connection caused it.
Model: llama3.2:1B
Name: Llama-3.2-1B-NPU2
Missing files:
config.json
- tokenizer.json
- attn.xclbin
- mm.xclbin
- dequant.xclbin
- layer.xclbin
- lm_head.xclbin
- model.q4nx
[PULL] Downloading 8 files...
Downloading 1/8: config.json?download=true
CURL error: Timeout was reached
Failed to download: https://huggingface.co/FastFlowLM/Llama-3.2-1B-NPU2/resolve/main/config.json?download=true
[ERROR] Failed to download model files.
[FLM] Loading model: C:\Users\wisemodel-aipc-04-b\flm/../../models/Llama-3.2-1B-NPU2
Failed to open file: C:\Users\wisemodel-aipc-04-b\flm/../../models/Llama-3.2-1B-NPU2
In some regions, model image pull may fail due to network issues.
To solve this issue. You can find mirrors of the project models on the Hugging Face mirror site in your area. For example, you can use the following link.
https://hf-mirror.com/FastFlowLM
After pulling the model, you should put the model file in the following path. For example
C:\Users\models\Llama-3.2-1B-NPU2
In this way, the model can be accessed by flm.
By the way. The path of the model file has been moved to the following path in v0.1.1#3.
https://github.com/FastFlowLM/FastFlowLM/releases/tag/v0.1.1
C:\Users\<USER>\Documents\flm\models\