unsloth/studio/backend/core
Daniel Han-Chen fb0a31b9a5 Fix/adjust diffusion: forward true_cfg_scale on Qwen/Flux for negative prompt (PR #5754)
QwenImagePipeline and FluxPipeline treat guidance_scale as the
distilled CFG factor and expose true_cfg_scale as the real
classifier-free guidance knob. Negative prompts only steer the
output when true_cfg_scale > 1, so forwarding only guidance_scale
left Qwen-Image on the default true_cfg_scale=4.0 and the user's
slider value silently ineffective for negative prompts.

When the loaded pipeline accepts both negative_prompt and
true_cfg_scale and the caller supplies a non-empty negative
prompt, forward guidance_scale through both kwargs so the
negative prompt actually steers generation. When no negative
prompt is supplied, true_cfg_scale is left at the model default
to avoid switching distilled CFG models into real-CFG mode (which
would double inference cost and degrade quality).

Adds two regression tests covering the forward-when-negative and
skip-when-no-negative paths.
2026-05-25 00:46:25 +00:00
..
data_recipe chore(deps): bump the npm-oxc-validator group across 1 directory with 2 updates (#5667) 2026-05-22 04:46:20 -07:00
export feat(studio): MLX training tab on Apple Silicon (LoRA / full FT, VLM, export) (#5265) 2026-05-05 23:54:58 -07:00
inference Fix/adjust diffusion: forward true_cfg_scale on Qwen/Flux for negative prompt (PR #5754) 2026-05-25 00:46:25 +00:00
training studio: install flash-linear-attention and tilelang for Qwen3.5 family (#5434) 2026-05-18 03:49:06 -07:00
__init__.py [Studio] Show non exported models in chat UI (#4892) 2026-04-14 15:03:58 +04:00
tool_healing.py studio: extract tool-call XML parser into a reusable helper module (#5583) 2026-05-19 05:06:17 -07:00