Search results

Learning rate estimation questions in 2.0.0rc2

Status: Open.

#31 In LoganBooker/prodigy-plus-schedule-free;

· qwe795138426 opened on Jun 19, 2025

Paper: Why Gradients Rapidly Increase Near the End of Training

Status: Open.

#30 In LoganBooker/prodigy-plus-schedule-free;

· rockerBOO opened on Jun 10, 2025

Log clarification

Status: Open.

#29 In LoganBooker/prodigy-plus-schedule-free;

· deGENERATIVE-SQUAD opened on Jun 2, 2025

AltLoRA: Towards Better Gradient Approximation in Low-Rank Adaptation with Alternating Projections

Status: Open.

#27 In LoganBooker/prodigy-plus-schedule-free;

· rockerBOO opened on May 22, 2025

OrthoGrad proposed in another paper

Status: Open.

#25 In LoganBooker/prodigy-plus-schedule-free;

· rockerBOO opened on May 5, 2025

d*LR report scales throughout training

Status: Open.

#23 In LoganBooker/prodigy-plus-schedule-free;

· rockerBOO opened on Apr 18, 2025

Flux finetuning consumes all VRAM(24GB) and shared gpu memeory(32GB) compared to adafactor which uses 20+10GB

Status: Open.

#22 In LoganBooker/prodigy-plus-schedule-free;

· mcmanus82 opened on Apr 18, 2025

Low d with Flux model, general feedback

Status: Open.

#14 In LoganBooker/prodigy-plus-schedule-free;

· rockerBOO opened on Jan 31, 2025