tocmo0nlord
  • Joined on 2026-02-11
tocmo0nlord created branch fa3-hopper in tocmo0nlord/axolotl 2026-04-26 15:07:02 +00:00
tocmo0nlord pushed to fa3-hopper at tocmo0nlord/axolotl 2026-04-26 15:07:02 +00:00
9bdf4b1c23 improve handling and error if fa3 requested but not installeD
d6f64a3684 handle args to drop dropout
0735454782 move fa3 tests to multigpu since we only run those on hopper
bb6464c4c6 use get_device_capability since CI setting in cfg is unreliable
323a9cb153 handle return sig change for fa3
Compare 10 commits »
tocmo0nlord created branch feat/beautiful-readme in tocmo0nlord/axolotl 2026-04-26 15:07:02 +00:00
tocmo0nlord created branch fa-261 in tocmo0nlord/axolotl 2026-04-26 15:07:01 +00:00
tocmo0nlord pushed to fa-261 at tocmo0nlord/axolotl 2026-04-26 15:07:01 +00:00
tocmo0nlord created branch fa-check in tocmo0nlord/axolotl 2026-04-26 15:07:01 +00:00
tocmo0nlord pushed to fa-check at tocmo0nlord/axolotl 2026-04-26 15:07:01 +00:00
9c0fa60220 fsdp2 w evals fixed upstream
8efdc59796 just assume that fa supports window
172b08b209 integration check for transformers#40002
Compare 3 commits »
tocmo0nlord created branch eos-hell in tocmo0nlord/axolotl 2026-04-26 15:07:00 +00:00
tocmo0nlord pushed to eos-hell at tocmo0nlord/axolotl 2026-04-26 15:07:00 +00:00
6c49083d8b improve check for base case
94c226edb3 fixes last eos token not in labels on basic use case
Compare 2 commits »
tocmo0nlord created branch exp-expand-len in tocmo0nlord/axolotl 2026-04-26 15:07:00 +00:00
tocmo0nlord pushed to exp-expand-len at tocmo0nlord/axolotl 2026-04-26 15:07:00 +00:00
6fcb73faaa more gpt-neox long ctx fixes
a32cc1d021 fix bettertransformers save, force it to skip after saving correctly in callback
86bd9fcff4 more tweaks to do pre-training with bettertransformers
ed7531abb8 experimental expansion of ctx len
bdb547b830 add validation/warning for bettertransformers and torch version
Compare 8 commits »
tocmo0nlord created branch enable_tp in tocmo0nlord/axolotl 2026-04-26 15:06:59 +00:00
tocmo0nlord pushed to enable_tp at tocmo0nlord/axolotl 2026-04-26 15:06:59 +00:00
60c98a4353 stuff
c760d2b815 test accelerator
2014f58181 set os environ RANK
b5f9dd44f2 set os environ RANK
b17b1aada7 initialise process group for tp
Compare 10 commits »
tocmo0nlord pushed to dump-config at tocmo0nlord/axolotl 2026-04-26 15:06:58 +00:00
b594f18f6e just redact api keys
700791deb9 Merge branch 'main' into dump-config
d6d2cc673b remove none-valued config before dumping
Compare 3 commits »
tocmo0nlord created branch dynamic-sft in tocmo0nlord/axolotl 2026-04-26 15:06:58 +00:00
tocmo0nlord pushed to dynamic-sft at tocmo0nlord/axolotl 2026-04-26 15:06:58 +00:00
208f8b253f add validation for DFT
75ad1a9932 use dynamic finetuning with chunked cross entropy
Compare 2 commits »
tocmo0nlord created branch e2e-fsdp-trainer in tocmo0nlord/axolotl 2026-04-26 15:06:58 +00:00
tocmo0nlord pushed to e2e-fsdp-trainer at tocmo0nlord/axolotl 2026-04-26 15:06:58 +00:00
39ab9626f1 add transformers module to cleanup
26bd81cec0 re-enable tests w change in patching
Compare 2 commits »
tocmo0nlord created branch embeddings-resize in tocmo0nlord/axolotl 2026-04-26 15:06:58 +00:00
tocmo0nlord pushed to embeddings-resize at tocmo0nlord/axolotl 2026-04-26 15:06:58 +00:00
31079cd5fd smart resize embeddings