v1.1.1 - bring on the potato models

[bghira]

Trained with NF4 via PagedLion8Bit.

• New custom timestep distribution for Flux via --flux_use_beta_schedule, --flux_beta_schedule_alpha, --flux_beta_schedule_beta (Add the ability to use a Beta schedule to select Flux timesteps #1023)
• The trendy AdEMAMix, its 8bit and paged counterparts are all now available as bnb-ademamix, bnb-ademamix-8bit, and bnb-ademamix8bit-paged`
• All low-bit optimisers from Bits n Bytes are now included for NVIDIA and ROCm systems
• NF4 training on NVIDIA systems down to 9090M total using Lion8Bit and 512px training at 1.5 sec/iter on a 4090

What's Changed

• int8-quanto followup fixes (batch size > 1) by @bghira in int8-quanto followup fixes (batch size > 1) #1016
• merge by @bghira in merge #1018
• update doc by @bghira in update doc #1019
• update docs by @bghira in update docs #1025
• Add the ability to use a Beta schedule to select Flux timesteps by @AmericanPresidentJimmyCarter in Add the ability to use a Beta schedule to select Flux timesteps #1023
• AdEMAMix, 8bit Adam/AdamW/Lion/Adagrad, Paged optimisers by @bghira in AdEMAMix, 8bit Adam/AdamW/Lion/Adagrad, Paged optimisers #1026
• Bits n Bytes NF4 training by @bghira in Bits n Bytes NF4 training #1028
• merge by @bghira in merge #1029

Full Changelog: v1.1...v1.1.1

---

This discussion was created from the release v1.1.1 - bring on the potato models.