tocmo0nlord/axolotl
1
0
Fork 0
Code Issues Pull Requests Actions 3 Packages Projects Releases Wiki Activity

Commit Graph

  • b53a41372f feat: update transformers version to 4.49.0 (#2340) NanoCode012 2025-02-21 09:12:06 +07:00
  • 9413d27add Built site for gh-pages Quarto GHA Workflow Runner 2025-02-20 19:31:01 +00:00
  • 02f45e94be calculate sample length fixes and SFT splitting fixes (#2351) Wing Lian 2025-02-20 14:29:58 -05:00
  • 90dfcd8c03 Revert "Fix sample packing producing longer sequences than specified by `sequ…" revert-2332-fix_sample_packing Sung Ching Liu 2025-02-19 21:13:25 -05:00
  • 1a09d5e844 some refactoring grpo_liger Salman Mohammadi 2025-02-19 17:35:35 +00:00
  • cf61b4aba7 Merge branch 'main' into grpo_liger Salman Mohammadi 2025-02-19 16:17:42 +00:00
  • 14d274efe6 WIP liger support Salman Mohammadi 2025-02-19 15:34:32 +00:00
  • 109cae8122 Built site for gh-pages Quarto GHA Workflow Runner 2025-02-19 14:24:39 +00:00
  • 954e192f38 quick formatting fix for LoRA optims doc (#2349) Dan Saunders 2025-02-19 09:23:31 -05:00
  • 02efd7e83d quick formatting fix for LoRA optims doc lora-kernels-doc-fix Dan Saunders 2025-02-19 14:17:20 +00:00
  • d8295344ec Built site for gh-pages Quarto GHA Workflow Runner 2025-02-19 05:03:28 +00:00
  • 8dfadc2b3c Fix sample packing producing longer sequences than specified by sequence_len (#2332) Tobias 2025-02-19 06:02:35 +01:00
  • beec9a3d21 Built site for gh-pages Quarto GHA Workflow Runner 2025-02-18 21:09:41 +00:00
  • 23a9fcb0a7 make sure chatml dpo dataset loading works (#2333) Wing Lian 2025-02-18 16:08:40 -05:00
  • 822a8a6931 pylint patch_lora_post_model_load Dan Saunders 2025-02-18 19:59:17 +00:00
  • 1a51180637 removing unused function Dan Saunders 2025-02-18 19:36:03 +00:00
  • 7562aadf89 fix Dan Saunders 2025-02-18 19:13:09 +00:00
  • 479f5e18dd Small updates Dan Saunders 2025-02-18 19:08:27 +00:00
  • 945dcc5020 move patching to post-model load to improve applicability Dan Saunders 2025-02-18 19:00:12 +00:00
  • c6be6e42bd Built site for gh-pages Quarto GHA Workflow Runner 2025-02-18 15:07:29 +00:00
  • c3d4f6e295 Doc fix: TORCH_ROCM_AOTRITON_ENABLE_EXPERIMENTAL not necessary to use Triton kernel patches (#2343) Dan Saunders 2025-02-18 10:06:31 -05:00
  • 46dcae42f1 Built site for gh-pages Quarto GHA Workflow Runner 2025-02-18 09:31:59 +00:00
  • 7fa690fac8 bump dev version (#2342) Wing Lian 2025-02-18 04:30:59 -05:00
  • abcbb7138e Built site for gh-pages Quarto GHA Workflow Runner 2025-02-18 09:27:25 +00:00
  • 3c743c4bfb v0.7.0 for release (#2341) v0.7.0 Wing Lian 2025-02-18 04:26:21 -05:00
  • e847da6a85 Built site for gh-pages Quarto GHA Workflow Runner 2025-02-18 08:40:31 +00:00
  • 91bb95685a chore: cleanup deprecated config elements (#2309) NJordan72 2025-02-18 03:39:24 -05:00
  • fbf3ca86c9 feat: add support for qwen25 vl for multimodal pixtral_integration NanoCode012 2025-02-18 12:42:29 +07:00
  • fcea4633b6 Built site for gh-pages Quarto GHA Workflow Runner 2025-02-18 03:00:28 +00:00
  • b194e17c28 feat: add config for optional parameters in a chat message (#2260) NJordan72 2025-02-17 21:59:27 -05:00
  • 8d9f715cea Built site for gh-pages Quarto GHA Workflow Runner 2025-02-17 20:47:01 +00:00
  • 3aac3b1da9 Move sweeps code to another module (#2338) Dan Saunders 2025-02-17 15:46:04 -05:00
  • 31095fb186 Built site for gh-pages Quarto GHA Workflow Runner 2025-02-17 19:24:11 +00:00
  • 3d8425fa91 Activation function Triton kernels, LoRA custom autograd functions (#2324) Dan Saunders 2025-02-17 14:23:15 -05:00
  • 1405e0f108 Built site for gh-pages Quarto GHA Workflow Runner 2025-02-17 05:08:26 +00:00
  • 97a2fa2781 Select input_ids explicitly after panda conversion (#2335) Seungduk Kim 2025-02-17 14:07:27 +09:00
  • 125e30da31 Built site for gh-pages Quarto GHA Workflow Runner 2025-02-13 22:40:20 +00:00
  • a98526ef78 add support for include_tokens_per_second in training args (#2269) Wing Lian 2025-02-13 17:39:19 -05:00
  • 365f71e3ff Built site for gh-pages Quarto GHA Workflow Runner 2025-02-13 22:29:18 +00:00
  • 7a13442b94 Built site for gh-pages Quarto GHA Workflow Runner 2025-02-13 22:28:50 +00:00
  • 2e57391bf8 fix: add missing shards_idx, preprocess_shards to docs and validator (#2331) NanoCode012 2025-02-14 05:28:21 +07:00
  • aa45fed451 Add bos_token and add_generation_prompt to the alpaca chat template (#2322) minpeter 2025-02-14 07:27:55 +09:00
  • 49e480e240 Built site for gh-pages Quarto GHA Workflow Runner 2025-02-13 21:03:12 +00:00
  • 77764de0dd Built site for gh-pages Quarto GHA Workflow Runner 2025-02-13 21:02:41 +00:00
  • a09a5cfd1c feat(doc): add tensorboard config to docs (#2329) NanoCode012 2025-02-14 04:02:16 +07:00
  • e16307c27a Built site for gh-pages Quarto GHA Workflow Runner 2025-02-13 21:02:01 +00:00
  • 40362d60e0 feat(doc): Improve guide to dataset types with better examples (#2286) NanoCode012 2025-02-14 04:01:41 +07:00
  • ffae8d6a95 GRPO (#2307) Wing Lian 2025-02-13 16:01:01 -05:00
  • da69624dee Built site for gh-pages Quarto GHA Workflow Runner 2025-02-13 13:32:00 +00:00
  • fdbb1a207c [Fixing #2149] load_from_disk for RL-type training (#2193) Lee Park 2025-02-13 08:31:07 -05:00
  • 82d04ea060 test v2batch w/ flex attn bursteratom 2025-02-13 00:11:45 -05:00
  • 96836d1b5b Built site for gh-pages Quarto GHA Workflow Runner 2025-02-12 23:30:03 +00:00
  • 30046315d9 disable ray tests for latest torch release (#2328) Wing Lian 2025-02-12 18:29:02 -05:00
  • d63283f386 Built site for gh-pages Quarto GHA Workflow Runner 2025-02-12 15:05:30 +00:00
  • e37a4a536a lint docs (#2327) Wing Lian 2025-02-12 10:04:26 -05:00
  • f24efd77a1 lint docs docs-lint-20250212 Wing Lian 2025-02-12 10:04:01 -05:00
  • 0ef1f011fe Merge branch 'main' into flx_attn_support Sung Ching Liu 2025-02-11 23:31:56 -05:00
  • 3755ac54f9 Built site for gh-pages Quarto GHA Workflow Runner 2025-02-11 18:19:28 +00:00
  • 44f64ab627 Update faq.qmd (#2319) Sung Ching Liu 2025-02-11 13:18:31 -05:00
  • a44f398b14 Update faq.qmd bursteratom-doc-faq-update Sung Ching Liu 2025-02-11 09:17:49 -05:00
  • 356d9b6458 Update faq.qmd Sung Ching Liu 2025-02-11 09:16:40 -05:00
  • 4393d8e649 Merge branch 'main' into bursteratom-doc-faq-update Sung Ching Liu 2025-02-11 08:59:51 -05:00
  • 7ef6b7ee2d Built site for gh-pages Quarto GHA Workflow Runner 2025-02-08 11:02:53 +00:00
  • 826f1b1494 feat(doc): Add multi-node torchrun info (#2304) NanoCode012 2025-02-08 18:02:02 +07:00
  • 526e5ee8b8 fix(config): missing config not being documented and fix model_ override (#2317) NanoCode012 2025-02-08 18:01:48 +07:00
  • ca4cd4192e Built site for gh-pages Quarto GHA Workflow Runner 2025-02-08 02:35:08 +00:00
  • 8c11f6aa3b Built site for gh-pages Quarto GHA Workflow Runner 2025-02-08 02:34:39 +00:00
  • fd8cb32547 chore: remove redundant py310 from tests (#2316) NanoCode012 2025-02-08 09:34:16 +07:00
  • e48e2df4dd feat: update FA to 2.7.4.post1 which includes torch2.6 binary (#2315) NanoCode012 2025-02-08 09:34:01 +07:00
  • b7616022ab bump transformers to 4.48.3 (#2318) Wing Lian 2025-02-07 21:33:44 -05:00
  • 1faf1a5c5a batch add of spectrum snr results (#2320) Wing Lian 2025-02-07 21:33:14 -05:00
  • c0a1d205c7 packed doc mask starts at 1, 0 means masked out Sunny Liu 2025-02-07 14:44:52 -05:00
  • f7bf883470 Update faq.qmd Sung Ching Liu 2025-02-07 11:57:22 -05:00
  • 7591331d0c Update faq.qmd Sung Ching Liu 2025-02-07 11:42:23 -05:00
  • f89ddf7e23 Built site for gh-pages Quarto GHA Workflow Runner 2025-02-07 12:29:51 +00:00
  • 5bbad5ef93 feat: add torch2.6 to ci (#2311) NanoCode012 2025-02-07 19:28:54 +07:00
  • 6905711e45 set max steps to -1 when empty grpo-path-v2 Wing Lian 2025-02-06 17:27:52 -05:00
  • bb5a6135eb don't set total num steps for grpo Wing Lian 2025-02-06 17:23:13 -05:00
  • e637f9b1a4 cleanup pythonpath if axo in it Wing Lian 2025-02-06 17:03:21 -05:00
  • 1a3bfd6e0f test not deleting pythonpath for custom code bundling Wing Lian 2025-02-06 13:40:30 -05:00
  • 3df4df868c make sure to pass kwargs when using accelerate Wing Lian 2025-02-06 14:00:15 -05:00
  • c82cbdc6d9 make sure to handle num-processes with cloud Wing Lian 2025-02-06 13:50:39 -05:00
  • c9d842ef2e test not deleting pythonpath for custom code bundling grpo-path Wing Lian 2025-02-06 13:40:30 -05:00
  • ecea44c902 fix num_processes in passing to accelerate Wing Lian 2025-02-06 13:39:46 -05:00
  • 4f9c57e95d check for src axolotl in PYTHONPATH before removing it Wing Lian 2025-02-06 13:26:23 -05:00
  • 3d38bc82b8 include vllm in build Wing Lian 2025-02-06 11:09:42 -05:00
  • 13d458d0ae feat: update readme with inference instructions feat/linearize NanoCode012 2025-02-06 21:29:36 +07:00
  • ebd406af1d fix: lin_attn_mask in wrong dtype NanoCode012 2025-02-06 15:25:33 +07:00
  • 756a8332d6 set default on trl config Wing Lian 2025-02-05 22:17:10 -05:00
  • aded9c500d refactor cfg.grpo_* to use cfg.trl.* Wing Lian 2025-02-05 20:41:14 -05:00
  • 3659d812f7 use cfg.max_completion_length, not sequence_len Wing Lian 2025-02-05 13:20:17 -05:00
  • bdb0f97082 adding 'reward_processing_classes' Salman Mohammadi 2025-02-05 18:18:42 +00:00
  • 65b6519447 adding 'reward_processing_classes' Salman Mohammadi 2025-02-05 18:13:05 +00:00
  • a1958b09de seperately include max_completion_len Wing Lian 2025-02-05 13:01:52 -05:00
  • caa49a9d7d fix: use existing model config NanoCode012 2025-02-06 00:12:14 +07:00
  • c15ea6b956 fix: load vocab_size NanoCode012 2025-02-05 23:46:59 +07:00
  • dc05fff332 Built site for gh-pages Quarto GHA Workflow Runner 2025-02-05 14:25:04 +00:00
  • a971eb4ce6 Torch 2.6 support for base docker image (#2312) Wing Lian 2025-02-05 09:24:02 -05:00
  • b8f258817e adding reward fn verification Salman Mohammadi 2025-02-05 13:30:02 +00:00
  • 578fa764c8 chore: moved feature map into linear attention NanoCode012 2025-02-05 19:40:11 +07:00