Can we have it in GGUF F16/32?

#9
by qdrddr - opened

Hi, can you please generate GGUF F16/32 for this model?

See https://ollama.com/Losspost/stella_en_1.5b_v5 , but very little information about it (e.g. dimensions). If anyone can figure out how to customize the dimensionality (along the lines of the different versions offered on the dunzhang files and versions page. I would like an Ollama (or gguf) version of the recommended 1024d version.

The config.json file for the 1024d version reads as follows, and the model card suggests that it can be customized to several different dimensionalities just by replacing the value of "out_features":
{
"in_features": 1536,
"out_features": 1024,
"bias": true,
"activation_function": "torch.nn.modules.linear.Identity"
}

Sign up or log in to comment