tocmo0nlord
  • Joined on 2026-02-11
tocmo0nlord created branch sp-rl in tocmo0nlord/axolotl 2026-04-26 15:08:00 +00:00
tocmo0nlord pushed to sp-rl at tocmo0nlord/axolotl 2026-04-26 15:08:00 +00:00
9f30d3d33a reworking SP logic into composed handler
ce07081d6c doc updates; config fix
3ce43b6db9 simplifying trainer mixins and adding to rl trainers
Compare 3 commits »
tocmo0nlord created branch sp-rl-v3 in tocmo0nlord/axolotl 2026-04-26 15:08:00 +00:00
tocmo0nlord created branch soap-optim in tocmo0nlord/axolotl 2026-04-26 15:07:59 +00:00
tocmo0nlord pushed to soap-optim at tocmo0nlord/axolotl 2026-04-26 15:07:59 +00:00
efa1209a92 add smoke test training
67b9e31bbc make sure to set alternate optimizer and set lr and eps from adam
ad60916323 add soap optimizer support
Compare 3 commits »
tocmo0nlord created branch sp-fix-masking in tocmo0nlord/axolotl 2026-04-26 15:07:59 +00:00
tocmo0nlord pushed to sp-fix-masking at tocmo0nlord/axolotl 2026-04-26 15:07:59 +00:00
954b989e88 log warning re: logged losses / gradient scaling per rank
c64c881460 using existing packed seqlens util
cefd57cecb adding smoke test
2f3c52ea2f pre-commit fix
741015b3cf refactor and fix multipack seqlens
Compare 6 commits »
tocmo0nlord created branch sp-restore-buffers in tocmo0nlord/axolotl 2026-04-26 15:07:59 +00:00
tocmo0nlord pushed to sp-restore-buffers at tocmo0nlord/axolotl 2026-04-26 15:07:59 +00:00
979632f59c SP restore buffers
tocmo0nlord created branch sharegpt-batched in tocmo0nlord/axolotl 2026-04-26 15:07:58 +00:00
tocmo0nlord pushed to sharegpt-batched at tocmo0nlord/axolotl 2026-04-26 15:07:58 +00:00
b4d84d56d5 support for batched sharegpt tokenization to skip bad data
tocmo0nlord created branch sharegpt-field-conversations in tocmo0nlord/axolotl 2026-04-26 15:07:58 +00:00
tocmo0nlord pushed to sharegpt-field-conversations at tocmo0nlord/axolotl 2026-04-26 15:07:58 +00:00
b7fe46579d make the conversations/messages field configurable for sharegpt
tocmo0nlord created branch smaller-rand-model in tocmo0nlord/axolotl 2026-04-26 15:07:58 +00:00
tocmo0nlord pushed to smaller-rand-model at tocmo0nlord/axolotl 2026-04-26 15:07:58 +00:00
a0670abc94 add output for train loss in assertian err
08f287b57f swap llama tests for 7m param model
b4c7d9c29d fix perplexity scores
d2637fb01d first pass at modifying tests to use llama-7m
Compare 4 commits »
tocmo0nlord created branch smol-ci in tocmo0nlord/axolotl 2026-04-26 15:07:58 +00:00
tocmo0nlord pushed to smol-ci at tocmo0nlord/axolotl 2026-04-26 15:07:58 +00:00
993db05b3a fix losses
1b9520cc8b more train steps
f77408a3d0 fix tests
5db4272f69 more steps for loss check
431888c1de use smaller pretrained models for ci
Compare 5 commits »
tocmo0nlord pushed to shampoo-low_bit at tocmo0nlord/axolotl 2026-04-26 15:07:57 +00:00
f1b4030cdd WIP shampoo low bit optimizers
tocmo0nlord created branch shared-prepared-ci in tocmo0nlord/axolotl 2026-04-26 15:07:57 +00:00
tocmo0nlord pushed to shared-prepared-ci at tocmo0nlord/axolotl 2026-04-26 15:07:57 +00:00
b79996bdc4 tweak loss
68368de7ed add seed for stable reproducibility
a94c4a014b tweak acceptable loss from changed hyperparams
0102ca5943 fix cfg merge
97e8c01a70 tweak losses
Compare 9 commits »