* Implement GGUF metadata extraction function
Added function to extract metadata from GGUF files.
* Updated the GGUF model loading and patching classes to include metadata handling.
* Clean up return logic for extra metadata
This should be more future proof in case we need to return other attributes in the future. Possible breaking change for anyone using `gguf_sd_loader` directly either way, though.
---------
Co-authored-by: City <125218114+city96@users.noreply.github.com>
* add default named_modules_to_munmap variable
Default this to handle to case of cloning a GGUFModelPatcher which then
expects it to exist on the clone.
I could never reproduce this condition, however 4 users have the error.
This error path is very sensitive to VRAM condtions and workflow and I
never got the details so its a bit of a guess.
But from a code point of view, its the right thing anyway
* nodes: carry the mmap_released state through a clone()
There are workflows where the same underlying model will be loaded
twice by two different ModelPatchers.
With the current code as-is, the first patcher will load fine and
intercept the pins while removing already done pins from the free
modules list.
It then will load again with the second MP however, this one will
not get the callbacks for the already pinned stuff and then rip
through and .to().to() them all. This destroys the pinning setup
while comfy core still keeps the modules registered as unpinned.
This causes a crash on einval when core goes to unpin the not pinned
tensors.
Comfy core now has a guard against unpinning the not-pinned, however
what I think is happening, is the RAM of the pin in freed in the
.to().to() and then malloc re-allocates it to new tensors with cuda
keeping its records of previous pinning. einval can happen when
you try and unpin a tensor in the middle of the block which is
definately possible on this use after free pattern.
Comfy core recently introduced a feature where weights may be pinned
when loading, particularly for the case of offloading.
Intercept this, and immediately detached each weight before the pinning.
This avoids a crash that at least some users are experiencing.
Use a little dict on the modules to keep track of whats already done,
and when the catch-all detacher loop comes through, use this dict
(which has already done modules removed) as the iterator basis.
Attempts to override the unet_gguf key with our current one if it exists + print a warning.
Should only be triggered if another custom node edits the expected key in some way.
Tokenizer shape may be mismatched depending which model was used as source (128320 VS 128256) - warning instead of error for now.
k and q attn keys need to be reshaped back to HF format (inverse of convert_hf_to_gguf.py logic)
Original logic was triggering SDXL specific code to reshape img_in weight, causing forge not to work with it.
Tagging @blepping since this might change how the SDXL code is evaluated, but based on a quick check it should work.
#83
needs updated ComfyUI as well to fix it, issue was with the direct load logic in model management putting the model on the GPU before the weights were loaded, at which point the loading would overwrite the weights with the CPU ones
#98
Instead of adding/removing patches, sync with `self.patches[key]` - old patch should be GC'd. Most likely no legitimate reasons for double patching. Can simplify logic in ops.py if no issues.
Potential fix, lowvram weights are called via `patch_weight_to_device` on each run and it was possible to apply the same patch multiple times on subsequent generations.
City