split files?
This doesn't seem to load with any loader in oobabooga. Do we have to unsplit the model after downloading?
Yes you have to join the files together. Using the Q2 as an example, here's how you do it. If there are more split files, then you'll need all of them and you'll have to edit the windows command accordingly.
Linux and macOS:
cat falcon-180b.Q2_K.gguf-split-* > falcon-180b.Q2_K.gguf && rm falcon-180b.Q2_K.gguf-split-*
Windows command line:
COPY /B falcon-180b.Q2_K.gguf-split-a + falcon-180b.Q2_K.gguf-split-b falcon-180b.Q2_K.gguf
del falcon-180b.Q2_K.gguf-split-a falcon-180b.Q2_K.gguf-split-b
Yes you have to join the files together. Using the Q2 as an example, here's how you do it. If there are more split files, then you'll need all of them and you'll have to edit the windows command accordingly.
Linux and macOS:
cat falcon-180b.Q2_K.gguf-split-* > falcon-180b.Q2_K.gguf && rm falcon-180b.Q2_K.gguf-split-*
Windows command line:
COPY /B falcon-180b.Q2_K.gguf-split-a + falcon-180b.Q2_K.gguf-split-b falcon-180b.Q2_K.gguf
del falcon-180b.Q2_K.gguf-split-a falcon-180b.Q2_K.gguf-split-b
Getting an error in LM Studio 0.3.5 after merging (copy /b) from 14 fragments:
Failed to load the model
llama.cpp error: 'invalid split file: D:\LLModels\imported-models\Qwen\qwen2.5-72b-instruct\qwen2.5-72b-instruct-q5_k_m.gguf'
What can be the reason?
P.S. Saw comments that now there is no need to merge - just use? But I see 14 rows in Select a model to load - should i choose first file? Anyway there is a mess of identical 14 row in case of Qwen 14 fragments
P.P.S. Checked, no need to merge, works fine with all segments in one folder, LM Studio catches all of them as single model.