Code

Feat/sharegpt multirole#1

Code

Open

teknium1 wants to merge 60 commits intoNousResearch:mainNousResearch/axolotl-func-calling:mainfrom axolotl-ai-cloud:feat/sharegpt_multiroleaxolotl-ai-cloud/axolotl:feat/sharegpt_multiroleCopy head branch name to clipboard

+4,195-642

Commits

Commits on Feb 22, 2024

• feat(prompt): support multiple roles for sharegpt

  NanoCode012

  committed
  1623a50

• fix: add handling of empty role back

  NanoCode012

  committed
  85ddde2

• feat: rebased and allowed more dynamic roles via config

  NanoCode012

  committed
  33bcf57

• fix: variable

  NanoCode012

  committed
  b54869c

• chore: update message

  NanoCode012

  committed
  8751ffb

Commits on Feb 23, 2024

• feat: add vicuna format

  NanoCode012

  committed
  24591e8

• fix: JSON serializable error

  NanoCode012

  committed
  9e171a9

Commits on Feb 24, 2024

• fix(readme): Clarify doc for tokenizer_config (#1323) [skip ci]

  NanoCode012

  authored
  2ed52bd

• Use yaml codeblock for config.yaml field (#1303) [skip ci]

  kallewoof

  authored
  5cf226e

Commits on Feb 26, 2024

• make mlflow optional (#1317)

  Show description for 5894f0e
  winglian

  authored
  5894f0e

• Pydantic 2.x cfg (#1239)

  Show description for cc3cebf
  winglian

  authored
  cc3cebf

• chore: update readme to be more clear (#1326) [skip ci]

  NanoCode012

  authored
  c6b01e0

• ADD: push checkpoints to mlflow artifact registry (#1295) [skip ci]

  Show description for d756534
  

  3 peopleauthored
  d756534

• hotfix for capabilities loading (#1331)

  winglian

  authored
  7de912e

• hotfix for lora rank (#1332)

  winglian

  authored
  cf00231

• hotfix for missing outputs params (#1333)

  winglian

  authored
  e7eed20

• hotfix to exclude_unset from pydantic config when converting back to a dict (#1334)

  winglian

  authored
  269c543

• Add StableLM 2 Example Scripts (#1327) [skip ci]

  Show description for f30d062
  ncoop57

  authored
  f30d062

• add lion-pytorch optimizer (#1299) [skip ci]

  Show description for 1648279
  

  maximegmd

  and

  winglian

  authored
  1648279

• Support user-defined prompt processing strategies for dpo (#1248)

  Show description for 1e3d530
  

  nopperl

  and

  winglian

  authored
  1e3d530

Commits on Feb 27, 2024

• more pydantic fixes (#1338)

  winglian

  authored
  3f69571

• Mps mistral lora (#1292) [skip ci]

  Show description for 0f6af36
  

  3 peopleauthored
  0f6af36

• fix: checkpoint saving with deepspeed (#1321)

  NanoCode012

  authored
  5be8b55

• Update debugging.md (#1339) [skip ci]

  hamelsmu

  authored
  5265cd6

• fix steps check for anneal on first cycle (#1316)

  winglian

  authored
  2c9c88b

• Update fastchat_conversation_turns.py (#1294) [skip ci]

  Show description for 2b9687f
  eltociear

  authored
  2b9687f

• add gemma instruct chat template (#1341)

  Show description for c1a7b3d
  winglian

  authored
  c1a7b3d

Commits on Feb 28, 2024

• more fixes 20240228 (#1342) [skip ci]

  Show description for 0f985e1
  winglian

  authored
  0f985e1

• deprecate py 3.9 support, set min pytorch version (#1343) [skip ci]

  winglian

  authored
  6d4bbb8

• Fix use_mlflow to be bool instead of str (#1344)

  chiragjn

  authored
  3a5a2d2

• fix for protected model_ namespace w pydantic (#1345)

  winglian

  authored
  6b3b271

Commits on Feb 29, 2024

• run tests again on Modal (#1289) [skip ci]

  Show description for 0001862
  winglian

  authored
  0001862

Commits on Mar 2, 2024

• chore: enable sample_packing for Gemma (#1351)

  NanoCode012

  authored
  170d4d7

Commits on Mar 4, 2024

• Fix validation for early stopping (#1358)

  chiragjn

  authored
  b5b4492

• plain input/output prompt strategy w/o chat templates (#1346)

  Show description for 4d09b42
  winglian

  authored
  4d09b42

Commits on Mar 5, 2024

• lora+ support (#1352)

  Show description for decb66e
  winglian

  authored
  decb66e

• allow the sharegpt handler to also better handle datasets destined for openai finetuning (#1361)

  Show description for 2598c9f
  winglian

  authored
  2598c9f

• Update tinyllama lora.yml to fix eval packing issue (#1362)

  rasbt

  authored
  8984bf1

Commits on Mar 6, 2024

• add starcoder2 (#1349)

  Show description for e0f1895
  

  3 peopleauthored
  e0f1895

• Remove unsupported python version 3.9 from README (#1364) [skip ci]

  nirogu

  authored
  3765747

• support for DoRA w/ PEFT (#1363)

  winglian

  authored
  0cfdb2c

• add docs for input_output format (#1367) [skip ci]

  Show description for ed70a08
  hamelsmu

  authored
  ed70a08

• update flash attention for gemma support: (#1368)

  winglian

  authored
  58b0d4b

Commits on Mar 7, 2024

• JarvisLabs (#1372)

  Show description for 638c2da
  winglian

  authored
  638c2da

Commits on Mar 8, 2024

• FDSP + QLoRA (#1378)

  Show description for 9b6ee83
  winglian

  authored
  9b6ee83

Commits on Mar 11, 2024

• validation for fsdp and deepspeed (#1388) [skip ci]

  Show description for 3fd8093
  winglian

  authored
  3fd8093

• support for rslora (#1387) [skip ci]

  winglian

  authored
  7659c00

• Fix pydantic configuration for the max_memory input (#1385) [skip ci]

  Show description for 0bc114d
  

  dandm1

  and

  winglian

  authored
  0bc114d

• Set gradient_clipping to auto in DeepSpeed configs (#1382) [skip ci]

  seungduk-yanolja

  authored
  b0ee9ec

• Add Glaive conversation format support (#1365)

  Show description for b7d8a7d
  

  brianfitzgerald

  and

  winglian

  authored
  b7d8a7d

• chore: lint (#1389)

  winglian

  authored
  4326520

Commits on Mar 12, 2024

• Merge branch 'main' into feat/sharegpt_multirole

  NanoCode012

  committed
  58fa1ee

• fix: typing

  NanoCode012

  committed
  2b9d66c

• fix: don't remap for unknown keys

  NanoCode012

  committed
  ad70d34

• fix: add roles to pydantic

  NanoCode012

  committed
  3031632

• feat: add test

  NanoCode012

  committed
  c2738b3

• chore: remove leftover print

  NanoCode012

  committed
  320312f

• chore: remove leftover comment

  NanoCode012

  committed
  4b86dd8

• chore: remove print

  NanoCode012

  committed
  f00f63b

Commits on Mar 13, 2024

• fix: update test to use chatml

  NanoCode012

  committed
  cc545db