* Implement mixed precision operations with a registry design and metadate for quant spec in checkpoint.
* Updated design using Tensor Subclasses
* Fix FP8 MM
* An actually functional POC
* Remove CK reference and ensure correct compute dtype
* Update unit tests
* ruff lint
* Implement mixed precision operations with a registry design and metadate for quant spec in checkpoint.
* Updated design using Tensor Subclasses
* Fix FP8 MM
* An actually functional POC
* Remove CK reference and ensure correct compute dtype
* Update unit tests
* ruff lint
* Fix missing keys
* Rename quant dtype parameter
* Rename quant dtype parameter
* Fix unittests for CPU build
Turns out torch.compile has some gaps in context manager decorator
syntax support. I've sent patches to fix that in PyTorch, but it won't
be available for all the folks running older versions of PyTorch, hence
this trivial patch.
This should speed up the lowvram mode a bit. It currently is only enabled when --async-offload is used but it will be enabled by default in the future if there are no problems.
The idea is that you can indicate how much quality vs speed you want.
At the moment:
--fast 2 enables fp16 accumulation if your pytorch supports it.
--fast 5 enables fp8 matrix mult on fp8 models and the optimization above.
--fast without a number enables all optimizations.
When generating images with fp8_e4_m3 Flux and batch size >1, using --fast, ComfyUI throws a "view size is not compatible with input tensor's size and stride" error pointing at the first of these two calls to view.
As reshape is semantically equivalent to view except for working on a broader set of inputs, there should be no downside to changing this. The only difference is that it clones the underlying data in cases where .view would error out. I have confirmed that the output still looks as expected, but cannot confirm that no mutable use is made of the tensors anywhere.
Note that --fast is only marginally faster than the default.
Optimizations that might break things/lower quality will be put behind
this flag first and might be enabled by default in the future.
Currently the only optimization is float8_e4m3fn matrix multiplication on
4000/ADA series Nvidia cards or later. If you have one of these cards you
will see a speed boost when using fp8_e4m3fn flux for example.
