Issue with this repository having 2+ "Q8_0" models.
One issue that I stumbled upon, when getting the model I was using the site on mobile and in the "Files and versions" tab I could see the first few letters of the filenames (couldn't see that there were 2 sets of models). When I tapped on the white button after the "Safe" button I used the popup to select "Use this model" and Ollama on the last file. Copied and pasted "ollama run hf.co/DavidAU/Gemma-The-Writer-N-Restless-Quill-V2-Enhanced32-10B-Uncensored-GGUF:Q8_0", replaced "run" with "pull".
When filing the previous issue yesterday I checked the model's hash and it didn't match until I found the model that it actually downloaded.
Instead of downloading "Gemma-The-Writer-N-Restless-Quill-V2-10B-max-cpu-D_AU-Q8_0.gguf"
it downloaded (had the hash of) "Gemma-The-Writer-N-Restless-Quill-V2-10B-D_AU-Q8_0.gguf"
which also generates the same command to download/run the model.
"ollama run hf.co/DavidAU/Gemma-The-Writer-N-Restless-Quill-V2-Enhanced32-10B-Uncensored-GGUF:Q8_0"
I'm guessing that huggingface has a bug or issue generating "Use this model" info with duplicate quant values "Q8_0" in my case, and ollama grabs the first "Q8:0" file.
I don't think I've seen other repositories have variations like that. Probably best to limit each repository to 1 set of matching models, cause I'm not sure ollama can pull the 2nd "Q8_0" if someone needed that.
hmm ; it should see the "unique" file name, and use that - might be a bug at HF and/or OLLAMA.
It all depends on how the systems are parsing the directory of files.