axolotl · Issues· 239 open
Open on GitHubLocally synced open issues (discussions stay on GitHub)
- #3998
Export LoRA Adapters to GGUF via convert_lora_to_gguf.py
enhancementgood first issueUpdated Sep 11, 2026 - #2752
GRPO Single GPU / colocate with TRL==0.18.x
enhancementUpdated Sep 7, 2026 - #3981
Add support for new flash-attn (and kernel) interface in ring-flash-attention
enhancementUpdated Sep 4, 2026 - #3940
TokensPerSecondCallback captures resume_from_checkpoint before auto-resume detection resolves it, silently zeroing token metrics
Updated Aug 25, 2026 - #3203
OOM for causal lm evaluation and missing logging
bugwaiting for reporterUpdated Aug 20, 2026 - #3890
model support system follow-ups
enhancementUpdated Aug 10, 2026 - #3626
[Feature] Automatic LoRA rank recommendation based on dataset size
good first issueUpdated Aug 7, 2026 - #3908
MLflow/WandB/Comet/Trackio env vars not set on Ray Train worker (use_ray: true)
Updated Jul 30, 2026 - #3608
Ring Attention w/ document packing produces different results
bugUpdated Jul 21, 2026 - #3787
Expert-granularity CPU offload for quantized MoE: 30B-A3B QLoRA in ~7 GB (single resident expert layer, extends the layer_offloading idea)
waiting for reporterUpdated Jul 19, 2026 - #2396
EXTREMELY SLOW (unusable) towards end of tokenization of dataset with long multi turn conversations
bugUpdated Jul 18, 2026 - #3152
total_num_steps calculation is incorrect with sample_packing_eff_est
bugwaiting for reporterUpdated Jul 17, 2026 - #3846
[ Dep conflict ] FA4 uses cutlass 4.6.0.dev0 whereas sonicmoe EP 4.6.0
bugUpdated Jul 17, 2026 - #3845
[ Memory ] Compact then pad for sonicmoe EP sentinel
enhancementUpdated Jul 17, 2026 - #3824
DPO chat_template Strategy Drops tool_calls and Never Passes tools to the Chat Template
Updated Jul 10, 2026