index_timestep_zero possible solution
#5
by patientxtr - opened
The full safetensors have "model.diffusion_model.index_timestep_zero" tensor with shape of "0". The illama_quantize is not allowing to create the ggufs with the shape of "0".
So, I removed that tensor and quantized.
There are no errors in comfyui for me when using this test ggufs.
I am now trying to add back this tensor. Will update the ggufs if successful.
Added the index_timestep_zero tensor after quantizing. Uploading the updated ggufs from now
Arunk25 changed discussion status to closed
