Five deprecated partner V1 nodes shared an exact display_name with their
live V2 replacements, so the node picker (and any object_info consumer)
showed two identically-named nodes and users/agents routinely landed on
the deprecated one. The headline case: OpenAIGPTImage1 (deprecated) and
OpenAIGPTImageNodeV2 (live) were both "OpenAI GPT Image 2", and the
deprecated node listed gpt-image-1 first in its model combo, nudging
picks onto an older, costlier-for-worse-output model even though it
already defaults to gpt-image-2.
For each of the five deprecated twins:
- append " (legacy)" to display_name so it no longer collides with the
live node
- add a one-sentence description note: the class name is historical,
kept only for existing-workflow compatibility, supports the current
models, and names the live V2 class to prefer for new workflows
- reorder the model combo so the current model leads where it was
stale-first (OpenAIGPTImage1: gpt-image-2 now leads; GeminiNode: the
gemini-3.1 generation now leads). Grok, Nano Banana 2, and ByteDance
Seedream already led with the current model, so their combos are
unchanged.
Class names (node_id / class_type) are untouched, so existing workflows
referencing these nodes still load and run. Only display_name,
description, and combo order change; no object_info schema change.
The six ControlNet display_name collisions noted in the same analysis
live in core nodes, not comfy_api_nodes, and are left for a follow-up.
When a prompt is submitted without client_id and its output nodes are
served from cache, _send_cached_ui returned early before recording the
cached UI outputs, so /api/jobs/{job_id} (and /history) reported success
with empty outputs. Record the outputs before the client_id check.
#14862 auto-enables the comfy-kitchen Triton backend whenever torch.version.hip
is set and Triton >= 3.7. The INT8 matmul kernels compile tl.dot to matrix-core
instructions (WMMA on RDNA3+/gfx11xx-gfx12xx, MFMA on CDNA/gfx9xx); RDNA1/RDNA2
(gfx10xx) have neither, so the auto-enabled INT8 path hangs the GPU there
(reported on RDNA2 + triton-windows 3.7.1: native and custom-node INT8 freeze
until reset).
Gate the automatic ROCm default on GPU architecture as well as Triton version so
RDNA1/RDNA2 stay on the working eager fallback. Add --disable-triton-backend as
an explicit override; --enable-triton-backend still force-enables on any arch.
On AMD/ROCm the CUDA backend is unavailable, so Triton is the only accelerated
comfy-kitchen backend. It was disabled by default (opt-in --enable-triton-backend),
leaving AMD on the slow eager path. Enable it by default when torch.version.hip is
set AND Triton is >= 3.7 -- older Triton lacks libdevice.rint on the HIP backend and
hard-crashes the INT8 path, so on Triton < 3.7 it stays disabled with a log line.
NVIDIA behavior is unchanged; the explicit --enable-triton-backend flag still works
as an override.
Fixes#14861
mixed_precision_ops.Linear.forward only quantized activations that were 2D, or
3D (reshaped to 2D). Inputs with rank >= 4 (e.g. Anima's MLP activations, which
are not reshaped to 3D the way the attention path is) fell through the
`input_reshaped.ndim == 2` guard and reached scaled_mm as bf16, silently
dispatching a bf16 kernel instead of FP8. Since MLP is roughly half the compute,
the FP8 speedup was far below expectation.
Generalize the existing 3D->2D reshape to any rank >= 3 (flatten the leading
dims, keep the contraction dim) and reshape the output back to the original
leading dims. 2D and 3D inputs are handled exactly as before; only rank >= 4
inputs change (now quantized instead of skipped). This matches the rank-agnostic
handling already used by the training path (flatten(0, -2) / unflatten).
Fixes#14595.
linear_dtype in comfy_quant metadata can be used to set if the int4 op does
the matrix multiplication in int8 or int4, the default is int4 on GPUs that
support it with fallback to int8 for GPUs that don't.
Each folder in the listing now carries its registered extension
allowlist verbatim; an empty array means the folder accepts any
extension (match-all), mirroring filter_files_extensions semantics.
Gives consumers the filtering rule itself rather than just its output:
/models/{folder} lists files by the per-folder rule but the rule is not
exposed anywhere, and /experiment/models/{folder} filters everything by
the global supported_pt_extensions regardless of registration.
Presentation-level filtering of match-all folders (e.g. hiding
README/config noise that repository-downloading custom nodes leave in
model directories) is deliberately left to the consumer.
Co-authored-by: guill <jacob.e.segal@gmail.com>
Matt Miller
Alexander Piskun
Daxiong (Lin)
Comfy Org PR Bot
comfyanonymous
彼彼
Alexis Rolland
comfyanonymous
Terry Jia
Gustavo Schneiter
Yousef R. Gamaleldin
liminfei-amd
John Pollock
Simon Pinfold