Commits

Branch selector

 main

User selector

All users

Datepicker

All time

Commit History

Commits on Jul 31, 2024

• fix roles to train defaults and make logging less verbose (axolotl-ai-cloud#1801)

  winglian

  authored
  78b42a3

  Copy full SHA for 78b42a3

Commits on Jul 30, 2024

• qlora-fsdp ram efficient loading with hf trainer (axolotl-ai-cloud#1791)

  Show description for 3ebf224
  winglian

  authored
  3ebf224

  Copy full SHA for 3ebf224

• publish axolotl images without extras in the tag name (axolotl-ai-cloud#1798)

  winglian

  authored
  dbf8fb5

  Copy full SHA for dbf8fb5

• update test and main/nightly builds (axolotl-ai-cloud#1797)

  Show description for 9a63884
  winglian

  authored
  9a63884

  Copy full SHA for 9a63884

• use 12.4.1 instead of 12.4 [skip-ci] (axolotl-ai-cloud#1796)

  winglian

  authored
  c5587b4

  Copy full SHA for c5587b4

• fix dockerfile and base builder (axolotl-ai-cloud#1795) [skip-ci]

  winglian

  authored
  d4f6a6b

  Copy full SHA for d4f6a6b

• move to supporting mostly 12.1 w 2.3.1 and add new 12.4 with 2.4.0 (axolotl-ai-cloud#1793)

  winglian

  authored
  d8d1788

  Copy full SHA for d8d1788

• Update README.md (axolotl-ai-cloud#1792)

  mhenrichsen

  authored
  3bc8e64

  Copy full SHA for 3bc8e64

Commits on Jul 29, 2024

• Add flexible configuration options for chat_template dataset training (axolotl-ai-cloud#1756)

  Show description for 55cc214
  

  Tostino

  and

  winglian

  authored
  55cc214

  Copy full SHA for 55cc214

Commits on Jul 28, 2024

• various batch of fixes (axolotl-ai-cloud#1785)

  Show description for 94ba932
  winglian

  authored
  94ba932

  Copy full SHA for 94ba932

Commits on Jul 27, 2024

• Bump deepspeed 20240727 (axolotl-ai-cloud#1790)

  Show description for 2268091
  winglian

  authored
  2268091

  Copy full SHA for 2268091

Commits on Jul 24, 2024

• add support for simpo via cpo trainer (axolotl-ai-cloud#1772)

  Show description for 6a9cfec
  winglian

  authored
  6a9cfec

  Copy full SHA for 6a9cfec

Commits on Jul 23, 2024

• fix fsdp loading of models, esp 70b (axolotl-ai-cloud#1780)

  winglian

  authored
  fe250ad

  Copy full SHA for fe250ad

• bump flash attention to 2.6.2 (axolotl-ai-cloud#1781) [skip ci]

  winglian

  authored
  e6b299d

  Copy full SHA for e6b299d

• bump transformers for updated llama 3.1 (axolotl-ai-cloud#1778)

  Show description for 608a2f3
  winglian

  authored
  608a2f3

  Copy full SHA for 608a2f3

• swaps to use newer sample packing for mistral (axolotl-ai-cloud#1773)

  Show description for 87455e7
  winglian

  authored
  87455e7

  Copy full SHA for 87455e7

Commits on Jul 21, 2024

• Add a chat_template prompt strategy for DPO (axolotl-ai-cloud#1725)

  Show description for 985819d
  

  Keith Stevens

  authored
  985819d

  Copy full SHA for 985819d

Commits on Jul 19, 2024

• Fix untrained tokens (axolotl-ai-cloud#1771)

  Show description for fa91b69
  winglian

  authored
  fa91b69

  Copy full SHA for fa91b69

• bump transformers and set roundup_power2_divisions for more VRAM improvements, low bit ao optimizers (axolotl-ai-cloud#1769)

  Show description for e4063d6
  winglian

  authored
  e4063d6

  Copy full SHA for e4063d6

Commits on Jul 18, 2024

• Unsloth rope (axolotl-ai-cloud#1767)

  Show description for 7830fe0
  winglian

  authored
  7830fe0

  Copy full SHA for 7830fe0

Commits on Jul 17, 2024

• set the number of dataset processes on the DPO Config rather than the trainer (axolotl-ai-cloud#1762)

  winglian

  authored
  c86c32a

  Copy full SHA for c86c32a

• re-enable PYTORCH_CUDA_ALLOC_CONF expandable_segments (axolotl-ai-cloud#1765) [skip ci]

  winglian

  authored
  8731b95

  Copy full SHA for 8731b95

• add torch_compile_mode options (axolotl-ai-cloud#1763) [skip ci]

  Show description for 8619b2d
  winglian

  authored
  8619b2d

  Copy full SHA for 8619b2d

• fixes to accelerator so that iterable pretraining datasets work (axolotl-ai-cloud#1759)

  Show description for 976f851
  winglian

  authored
  976f851

  Copy full SHA for 976f851

• fix num gpu check (axolotl-ai-cloud#1760)

  winglian

  authored
  152ab76

  Copy full SHA for 152ab76

Commits on Jul 16, 2024

• support for llama multipack using updated code/patches (axolotl-ai-cloud#1754)

  Show description for 5f58555
  winglian

  authored
  5f58555

  Copy full SHA for 5f58555

• torch compile and cuda alloc improvements (axolotl-ai-cloud#1755)

  Show description for cfc533a
  winglian

  authored
  cfc533a

  Copy full SHA for cfc533a

• update modal package and don't cache pip install (axolotl-ai-cloud#1757)

  Show description for e1725ae
  winglian

  authored
  e1725ae

  Copy full SHA for e1725ae

Commits on Jul 14, 2024

• add basic support for the optimi adamw optimizer (axolotl-ai-cloud#1727)

  Show description for 78e12f8
  winglian

  authored
  78e12f8

  Copy full SHA for 78e12f8

• bump flash attention 2.5.8 -> 2.6.1 (axolotl-ai-cloud#1738)

  Show description for 98af538
  winglian

  authored
  98af538

  Copy full SHA for 98af538

Commits on Jul 13, 2024

• Fix eval_sample_packing in llama-3 lora example (axolotl-ai-cloud#1716) [skip ci]

  Show description for 219cd0d
  

  RodriMora

  and

  winglian

  authored
  219cd0d

  Copy full SHA for 219cd0d

• Changed URL for dataset docs (axolotl-ai-cloud#1744)

  dameikle

  authored
  634f384

  Copy full SHA for 634f384

• bump xformers to 0.0.27 (axolotl-ai-cloud#1740)

  Show description for 4512738
  

  akshaylive

  and

  winglian

  authored
  4512738

  Copy full SHA for 4512738

• update to pytorch 2.3.1 (axolotl-ai-cloud#1746) [skip ci]

  winglian

  authored
  1e57b4c

  Copy full SHA for 1e57b4c

• fixes to prevent vram spike when train starts (axolotl-ai-cloud#1742)

  winglian

  authored
  a4a5bf0

  Copy full SHA for a4a5bf0

Pagination

Previous

Next