master-473-9565c7f

add support for flux2 klein (#1193)

* add support for flux2 klein 4b

* add support for flux2 klein 8b

* use attention_mask in Flux.2 klein LLMEmbedder

* update docs

master-472-fbce16e

fix: avoid undefined behavior on image mask allocation failure (#1198)

master-471-7010bb4

feat: support for SDXS-512 model (#1180)

* feat: add U-Net specials of SDXS

* docs: update distilled_sd.md for SDXS-512

* feat: for SDXS use AutoencoderTiny as the primary VAE

* docs: update distilled_sd.md for SDXS-512

* fix: SDXS code cleaning after review by stduhpf

* format code

* fix sdxs with --taesd-preview-only

---------

Co-authored-by: leejet <[email protected]>

master-470-48d3161

feat: add sd-server API support for steps, sampler and scheduler (#1173)

master-468-885e62e

refactor: replace ggml_ext_attention with ggml_ext_attention_ext (#1185)

master-467-0e52afc

feat: enable vae tiling for vid gen (#1152)

* enable vae tiling for vid gen

* format code

* eliminate compilation warning

---------

Co-authored-by: leejet <[email protected]>

master-466-27b5f17

ci: only push Docker images on master or release

master-463-e7e83ed

fix(server): use has_file for mask multipart detection (#1178)

master-462-c5602a6

feat: prioritize gguf and safetensors formats for embeddings and LoRA…

master-457-b90b1ee

chore: eliminate compilation warnings under MSVC (#1170)