Skip to content

Releases

v3.1.4

Choose a tag to compare

@github-actions github-actions released this 06 Nov 20:05

Full Changelog: v3.1.3...v3.1.4

v3.1.3

Choose a tag to compare

@github-actions github-actions released this 02 Nov 08:43

Full Changelog: v3.1.2...v3.1.3

v3.1.2

Choose a tag to compare

@github-actions github-actions released this 01 Nov 20:26

Full Changelog: v3.1.1...v3.1.2

v3.1.1

Choose a tag to compare

@github-actions github-actions released this 01 Nov 09:54

Full Changelog: v3.1.0...v3.1.1

v3.1.0

Choose a tag to compare

@github-actions github-actions released this 30 Oct 15:08

Full Changelog: v3.0.2...v3.1.0

v3.0.2

Choose a tag to compare

@github-actions github-actions released this 29 Oct 16:09

Full Changelog: v3.0.1...v3.0.2

v3.0.1

Choose a tag to compare

@github-actions github-actions released this 29 Oct 01:18

Full Changelog: v3.0.0...v3.0.1

v3.0.0

Choose a tag to compare

@github-actions github-actions released this 28 Oct 21:52

Full Changelog: v2.5.8...v3.0.0

v2.5.8

Choose a tag to compare

@github-actions github-actions released this 27 Sep 13:01

ComfyUI-ThinkingLLM v2.5.8

SageAttention 1.x, including 1.0.6, can now be selected for Hugging Face Qwen nodes on ROCm through its Triton backend. Set attention_mode=sage; auto tries Sage before Flash Attention 2 and PyTorch SDPA.

  • NVIDIA CUDA kernel selection remains available.
  • Unsupported masks, head sizes, and model layers use the original SDPA path.
  • The SageAttention 1 path copies keys before the kernel to protect the KV cache from in-place smoothing.
  • The GGUF flash_attn setting belongs to llama.cpp and does not select SageAttention.

Validation: 7 focused tests, 153 feature and compatibility tests, and catalog validation passed on the local Windows CPU test path. End-to-end ROCm inference has not been verified on AMD hardware.

v2.5.7

Choose a tag to compare

@github-actions github-actions released this 26 Aug 10:43

Full Changelog: v2.5.6...v2.5.7