Issues / #444
#444 How do I make it use the models I already have?
closed · @mkultra333 · 3 コメント · GitHub で見る
本文
I already have an SSD drive with multiple versions of Qwen 3.8 Flash Next, including Unsloth's Qwen3.8-Flash-Next-UD-IQ3_XXS and Qwen3.8-Flash-Next-UD-Q2_K_XL. I intend to download coder and put that there too. The folder is D:\LLM_Small How do I make strata use those models instead of trying to download them? Also, if you're using the HF huggingface downloader, it's bugged. It doesn't resume partial downloads, instead if it gets cut off it restarts from the beginning of the latest incomplete file. So if you've got 29GB downloaded of a 30GB file and it stops, you have to restart that 30GB file from the very beginning. https://github.com/huggingface/huggingface_hub/issues/4196 So for people with slower connections like me, it's necessary to download models with an ordinary download manager that can properly resume partial downloads. I don't know if you use the HF downloader library, I mention it just in case because it's another reason people may need to manually set an existing model folder when setting up Strata.
関連リンク
インストール・モデル・リリースへのサイト内リンク。