* Initial HiDream01-image support
* Cleanup nodes
* Cleaner handling of empty placeholder models
* Remove snap_to_predefined, prefer tooltip for the trained resolutions
* Add model and block wrappers
* Fix shift tooltip
* Add node to work around the patch tile issue
Experimental, runs multiple passes with the patch grid offset and blends with various different methods.
* Qwen35 vision rotary_pos_emb cast fix
* Fix embedding layout type
* Some small optimizations
* Cleanup, don't need this fallback
* Prefix KV cache, cleanup
Bit of speed, reduce redundant code
* Get rid of redundant custom sampler, refactor noise scaling
Our existing lcm sampler is mathematically same, just added the missing options to it instead and a node to control them. Refactored the noise scaling and fix it for the stochastic samplers, add a generic node to control the initial noise scale.
* Update nodes_hidream_o1.py
* Fix some cache validation cases
* Keep existing sampling params
* Remove redundant video vision path
* Replace some numpy ops with torch
* Fx RoPE index for batch size > 1
* Prefer torch preprocessing
* Rename block_type to be compatible with existing patch nodes
* Fixes and tweaks
* initial WanDancer support
* nodes_wandancer: Add list form of chunker.
Create an alternate list form of the node so the chunk gens can be
trivially looped by the comfy executor.
* Closer match to original soxr resampling
* Remove librosa node
* Cleanup
---------
Co-authored-by: Rattus <rattus128@gmail.com>
If the same weight is used multiple times within the same prefetch
window, it should only apply compute state mutations once. Mark the
weight as fully resident on the first pass accordingly.
* initial gemma4 support
* parity with reference implementation
outputs can 100% match transformers with same sdpa flags, checkpoint this and then optimize
* Cleanup, video fixes
* cleanup, enable fused rms norm by default
* update comment
* Cleanup
* Update sd.py
* Various fixes
* Add fp8 scaled embedding support
* small fixes
* Translate think tokens
* Fix image encoder attention mask type
So it works with basic attention
* Handle thinking tokens different only for Gemma4
* Code cleanup
* Update nodes_textgen.py
* Use embed scale class instead of buffer
Slight difference to HF, but technically more accurate and simpler code
* Default to fused rms_norm
* Update gemma4.py
* mm: Use Aimdo raw allocator for cast buffers
pytorch manages allocation of growing buffers on streams poorly. Pyt
has no windows support for the expandable segments allocator (which is
the right tool for this job), while also segmenting the memory by
stream such that it can be generally re-used. So kick the problem to
aimdo which can just grow a virtual region thats freed per stream.
* plan
* ops: move cpu handler up to the caller
* ops: split up prefetch from weight prep block prefetching API
Split up the casting and weight formating/lora stuff in prep for
arbitrary prefetch support.
* ops: implement block prefetching API
allow a model to construct a prefetch list and operate it for increased
async offload.
* ltxv2: Implement block prefetching
* Implement lora async offload
Implement async offload of loras.
* pinned_memory: remove JIT RAM pressure release
This doesn't work, as freeing intermediates for pins needs to be
higher-priority than freeing pins-for-pins if and when you are going
to do that. So this is too late as pins-for-pins is model load time
and we dont have JIT pins-for-pins.
* cacheing: Add a filter to only free intermediates from inactive wfs
This is to get priorities in amongst pins straight.
* mm: free inactive-ram from RAM cache first
Stuff from inactive workflows should be freed before anything else.
* caching: purge old ModelPatchers first
Dont try and score them, just dump them at the first sign of trouble
if they arent part of the workflow.
Wires comfy-kitchen's TensorCoreAWQW4A16Layout (introduced on
feat/awq-w4a16-modulation) into ComfyUI's MixedPrecisionOps so checkpoints
that tag modulation linears with comfy_quant.format = "awq_w4a16" get
their (qweight, weight_scale, weight_zero) loaded into the kitchen layout
class instead of being dequantized to bf16 plain Linear at conversion time.
quant_ops.py:
- detect TensorCoreAWQW4A16Layout availability and stub it out for the
no-kitchen fallback (mirrors the SVDQuant W4A4 pattern)
- register the layout class + add "awq_w4a16" to QUANT_ALGOS
(storage_t = int8 packed uint4, parameters = {weight_scale, weight_zero},
default group_size = 64)
ops.py: add the awq_w4a16 branch in MixedPrecisionOps.Linear._load_from_state_dict
that constructs Params(scale, zeros, group_size, ...) and wraps qweight
into a QuantizedTensor — F.linear then dispatches to ck.gemv_awq_w4a16
via the layout's aten handlers.
Pairs with comfy-kitchen feat/awq-w4a16-modulation. Targets the ~10 GB
inflation in Qwen-Image-Edit kitchen-native checkpoints, where the
modulation linears (img_mod.1 / txt_mod.1) currently dominate disk + VRAM
because they're materialized as plain bf16 Linear during conversion.
quant_ops.py: register TensorCoreSVDQuantW4A4Layout when comfy-kitchen exposes
it; gate the kitchen CUDA backend on cuda >= 13 (the optimized kitchen CUDA
ops are validated against cu13+ runtimes; on older cu the backend falls back
to eager).
ops.py: handle svdquant_w4a4 quant_format by loading weight_scale / proj_down /
proj_up / smooth_factor into TensorCoreSVDQuantW4A4Layout.Params, with the
img_mlp.net.2 / txt_mlp.net.2 fallback for act_unsigned. Targets the row-major
kitchen-native kernels on feat/svdquant-w4a4-kitchen-native; the verbatim
zgemm path is a sibling branch.
the mixed_precision ops can have input_scale parameters that are used
in tensor math but arent a weight or bias so dont get proper VRAM
management. Treat these as force-castable parameters like the non comfy
weight, random params are buffers already are.
On Windows with aimdo enabled, disable_weight_init.Linear uses lazy
initialization that sets weight and bias to None to avoid unnecessary
memory allocation. This caused a crash when copy_() was called on the
None weight attribute in Stable_Zero123.__init__.
Replace copy_() with direct torch.nn.Parameter assignment, which works
correctly on both Windows (aimdo enabled) and other platforms.
