tocmo0nlord/axolotl
1
0
Fork 0
Code Issues Pull Requests Actions 3 Packages Projects Releases Wiki Activity
1,044 Commits 319 Branches 32 Tags
cdc71f73c84efffcdcc2c4039b5fa1fe684226fa
Commit Graph

207 Commits

Author SHA1 Message Date
Wing Lian
cdc71f73c8 update table for rwkv4 support, fix process count for dataset (#822) 2023-11-04 23:45:44 -04:00
Wing Lian
8b79ff0e94 fix eval_steps to be a sane default (#797) 
* fix eval_steps to be a sane default

* update docs for fractional eval_steps
 2023-10-27 22:36:30 -04:00
Aleksa Gordić
2e71ff03a6 Add docker advanced instruction to README (#792) 2023-10-27 09:24:04 -04:00
Casper
e50ab072e2 Create preprocess CLI (#785) 
* Create preprocess CLI

* Print prompt template if debugging

* Add print for unsupported prompters

* Formatting

* Formatting

* Refactor variables

* Formatting

* Formatting

* Formatting

* Formatting
 2023-10-26 09:35:42 -04:00
NanoCode012
20aa4b57d2 chore(readme): Improve documentation on conversation field (#782) 
* chore(readme): Improve documentation on conversation field

* fix: clarify where the option is
 2023-10-24 12:52:32 +09:00
NanoCode012
afedc470bd Fix: Cannot tokenize with bf16 and on cpu (#766) 2023-10-23 01:32:26 +09:00
Casper
15d3a654bf Implement fused modules (#747) 
* MLP: Memory saving

* Remove RMSNorm restrictions

* Map packed weights to original

* FusedAttention module

* Simplify code

* Move fused modules

* Fix critical typo

* Split inplace

* Add FFT config

* Add validation of fused arguments

* Add fused arguments to config

* Update docs

* Fix validation logic

* Add fused modules to flash attn

* Only fuse during training

* Remove timing

* Formatting

* Formatting

* Formatting

* chore: lint

* chore: lint

* add e2e tests for fused llama

* no lora for tests

---------

Co-authored-by: Wing Lian <wing.lian@gmail.com>
 2023-10-21 16:08:25 -04:00
Wing Lian
a21935f07a add to docs (#703) 2023-10-19 21:32:30 -04:00
Casper
e1b214c62b Clarify custom format example (#729) 
* Clarify custom prompt format

* Simplify format
 2023-10-14 09:28:12 -04:00
Maxime
3bd9528390 add noisy embedding (#721) 
* add noisy embedding

* fix format

* Update README.md

* Update README.md

* linter issues

* caseus fixes

---------

Co-authored-by: Maxime <maxime@nope.no>
 2023-10-13 10:00:42 -04:00
NanoCode012
5855dded3d fix(doc): update default doc according to arg (#714) 2023-10-10 21:51:56 +09:00
NanoCode012
11c48c5e03 fix(doc): Add note on inference w sample packing (#712) 2023-10-10 21:08:17 +09:00
seungduk.kim.2304
77c84e02fd Update README with some explanations (#700) 
* Update README with some explanations

* revert commit-hook change

* add more explanation about batch size and gradient accum

* not use latex foromat

* decorate

* git hook again

* Attach a link that explains about LoRA hyperparameters

* update table of content

* Explanation about lora_modules_to_save
 2023-10-08 13:37:54 -04:00
Wing Lian
2642caedf2 refactor to set eval_batch_size earlier if unset, so we can warn if mismatched (#662) 2023-10-02 21:08:07 -04:00
Kyle Corbitt
9ec20777ba Make dataset_processes configurable (#651) 
I'm using the Axolotl script to train models on https://modal.com serverless GPUs. Unfortunately, their environment seems to have some kind of bug where if I try to run `datasets.filter` with too high a `num_proc`, it throws an error and dies.

This PR adds a new configuration option `dataset_processes`, which lets you explicitly set the number of processes used to map/filter the dataset. If not included, this defaults to the current behavior of setting that to `os.cpu_count()`.
 2023-09-29 00:22:22 -04:00
ich
590d6032fd Fix bug when using pretokenized datasets (#652) 
* fix pretokenized datasets readme

* check if dataset type is not set to handle pretokenized datasets
 2023-09-28 22:54:10 -04:00
Wing Lian
409ca0f21c add support for defined train split (#654) 2023-09-28 20:14:14 -04:00
NanoCode012
eb41f76f92 Feat: Add example for Mistral (#644) 
* Feat: Add example for Mistral

* chore: turn off flash

* chore: add is_mistral_derived_model

* chore: update following PR
 2023-09-28 20:15:00 +09:00
Napuh
85b0be2ba7 Warn users to login to HuggingFace (#645) 
* added warning if user is not logged in HF

* updated doc to suggest logging in to HF
 2023-09-27 17:43:35 -04:00
Wing Lian
895f0a0723 skip some flash attn patches unless explicitly enabled (#643) 
* skip some flash attn patches if explicitly disabled

* make the other patches optional
 2023-09-27 12:11:07 -04:00
Wing Lian
e7d3e2dbb6 use fastchat conversations template (#578) 
* use fastchat conversations template

* require fastchat (fschat) pip install

* handle roles dynamically from conversation

* tweak fastchat conversation with a monkeypatch to get individual turns

* fix up so it works with multiple conversation styles, and don't strip the turns

* fix sharegpt fixture now that we're using a more correct tokenization

* use a new prompter and support fastchat conversation type

* use sharegpt from prompt strategies now

* update docs, add chatml template

* add a newline after im_end token

* ensure we correctly set system message

* update per PR feedback to handle deprecated sharegpt types

* don't add duplicate wandb req

* make sharegpt fields configurable from yml

* llama2 fixes

* don't fail fatally when turns are improper
 2023-09-27 12:10:45 -04:00
NanoCode012
19a600a8b8 Feat: Add support for upstream FA2 (#626) 
* Feat: Add support for upstream FA2

* chore: add is_falcon_derived_model: true to examples

* chore: add config to readme for documentation

* feat: add extra model types

* fix: remove old falcon flash patch

* chore: pin transformers and accelerate
 2023-09-26 09:53:28 -04:00
Fernando Tarin Morales
5e5296a77c Added quotes to the pip install -e command to fix an incompatibility with shells that do glob expansion like zsh (#632) 2023-09-25 11:50:14 -04:00
NanoCode012
67b9888630 Feat(doc): Add eval_sample_packing to doc (#625) 2023-09-23 13:11:27 +09:00
Wing Lian
c25ba7939b update README w deepspeed info (#605) 2023-09-22 00:15:52 -04:00
NanoCode012
00dce35fb2 Feat(data): Allow loading local csv and text (#594) 
* Feat(data): Allow loading local csv and text

* chore: update readme for loading data
 2023-09-17 11:32:27 -04:00
NanoCode012
3a2edc85c3 Feat(doc): Add features to doc (#583) 2023-09-16 01:14:15 +09:00
Wing Lian
f7a22632d7 support custom field for completion from yml (#580) 
* support custom field for completion from yml

* remove legacy completion check and add doc

* update README docs
 2023-09-15 07:48:21 -04:00
Wing Lian
a5a625f47e update support matrix with btlm and phi (#579) 2023-09-15 02:46:15 -04:00
Wing Lian
861cecac2a refactor scripts/finetune.py into new cli modules (#550) 
* refactor scripts/finetune.py into new cli modules

* continue to support scripts/finetune.py

* update readme with updated cli commands

* Update scripts/finetune.py

Co-authored-by: NanoCode012 <kevinvong@rocketmail.com>

---------

Co-authored-by: NanoCode012 <kevinvong@rocketmail.com>
 2023-09-15 01:43:52 -04:00
Wing Lian
a4e1bb6606 let hf trainer handle torch compile (#516) 
* let hf trainer handle torch compile

* remove torch compile checks, include option for backend

* suppress torch errors to get further

* require min torch version of 2.1.0 for torch compile to work

---------

Co-authored-by: Aman Karmani <aman@tmm1.net>
 2023-09-13 11:42:12 -04:00
Glavin Wiechert
5b67ea98a6 Add training callback to send predictions to WandB table (#521) 
* WIP Add training callback to send predictions to WandB table

* WIP improve wandb table reporting callback

* WIP improve wandb table reporting callback (cont)

* Add VSCode launching for debugging

* Add tiny llama example

* WIP attempt to improve post-eval prediction generation for table

* WIP attempt to improve post-eval prediction generation for table - part 2

* WIP batch generation

* WIP attempt to handle sample_packing using position_ids for wandb prediction table

* WIP add code for debugging

* Fix sample_packing support for wandb prediction table

* Clean up code for PR review

* Add eval_table_size, eval_table_max_new_tokens configs & clean up code

* Clean up PR, delete VSCode config, add tiny-llama example

* Add eval_table_size, eval_table_max_new_tokens documentation. Fix linting/formatting
 2023-09-13 09:51:08 -04:00
Wing Lian
9845c5e12d document that packaging needs to be installed before flash-attn (#559) 2023-09-12 12:18:30 -04:00
The Objective Dad
6d57f2f0f0 ergonomic update to optimizer config doc (#548) 2023-09-11 12:35:45 -04:00
Wing Lian
34c0a86a11 update readme to point to direct link to runpod template, cleanup install instrucitons (#532) 
* update readme to point to direct link to runpod template, cleanup install instrucitons

* default install flash-attn and auto-gptq now too

* update readme w flash-attn extra

* fix version in setup
 2023-09-08 11:58:54 -04:00
The Objective Dad
5e2d8a42d9 Adding NCCL Timeout Guide (#536) 
* fixes NCCL_P2P_LEVEL=NVL #429

* adding more insights into verious values of NCCL_P2P_LEVEL
 2023-09-08 11:57:47 -04:00
NanoCode012
f51c9c56c6 Fix(doc): Inform Windows users to use WSL/docker (#518) 2023-09-01 00:08:21 -07:00
Jan Philipp Harries
396a7a74fc Added advanced DDP args (#515) 
* add ddp_config

* add advanced ddp config

* add ddp_config

* add advanced ddp config

---------

Co-authored-by: Jan Philipp Harries <jphme@users.noreply.github.com>
 2023-08-31 10:37:47 -07:00
Wing Lian
5ac3392075 support for datasets with multiple names (#480) 
* support for datasets with multiple names

* update docs
 2023-08-29 06:18:17 -07:00
NanoCode012
48c56470d0 Fix(doc): Clarify no amp to full yaml docs (#496) 2023-08-29 06:17:37 -07:00
Birch-san
8e197f6fb4 pad_to_worst_case_seq_len boolean, for testing memory limits (#498) 
* pad_to_worst_case_seq_len boolean, for testing memory limits

* remove collator_pad_to_longest option since it does nothing

see docs: https://huggingface.co/docs/transformers/main_classes/data_collator#transformers.DataCollatorWithPadding.padding

True and "longest" mean the same thing

* rename to `pad_to_sequence_len, and ensure 64 alignment

---------

Co-authored-by: Aman Karmani <aman@tmm1.net>
 2023-08-28 18:47:16 -04:00
NanoCode012
ad8be435ad Feat(doc): Update eval_steps doc (#487) 2023-08-27 10:09:09 +09:00
Charles O. Goddard
bde3c5a478 ReLoRA implementation (with quantization) (#322) 
* Experimental ReLoRA (+qlora) implementation

* Add CPU offload

* Remove local config

* Fix saving logic

* Remove redundant assert

* Fix logic errors

* Move ReLoRA into its own trainer class with a method override to create the proper scheduler

* Formatting & typing fixes

* Use safe_serialization

* Don't allow fsdp/deepspeed with ReLoRA

* Fix cpu-offload logic, enable multi gpu

* Document parameters and add comment

* Fix merge issue

* Smooth over some sharp edges

* Implement resume from checkpoint for relora

* Address review comments

* Fix saving logic

* Add necessary metadata to safetensors

---------

Co-authored-by: Wing Lian <wing.lian@gmail.com>
 2023-08-23 23:07:18 -04:00
NanoCode012
55c23c7bcb Fix(doc): Clarify config (#466) 2023-08-23 11:56:01 -04:00
TearGosling
f4746507f6 feat: add Metharme prompt strategy (#446) 
* Add Metharme tokenizing strategy

This strategy accounts for how the Metharme JSONLs are formatted as well as adds duplicated EOS tokens which can help trim model output length.
I haven't gotten the chance to test this yet, and probably won't have the chance for quite a bit, so I'm committing this now.

* Redo Metharme tokenizing strategy

lol

* fix: oops

* Rearrange a conditional

* chore: reformat code in accordance with linter

* chore: Make lint not freak out

* chore: fix lint

---------

Co-authored-by: NanoCode012 <kevinvong@rocketmail.com>
 2023-08-22 11:21:45 +09:00
NanoCode012
04a42b6db1 feat(docs): improve user customized prompts (#443) 
* feat(docs): improve user customized prompts

* feat(doc): add custom pretokenized instructions

* chore: clean old data folder

* chore: add new line
 2023-08-20 23:59:43 -04:00
NanoCode012
919f4cac90 feat(doc): add pillow to lambda instructions (#445) 2023-08-20 23:59:23 -04:00
Wing Lian
d2e7f27240 support user defined prompters, pretokenized datasets in config, local parquet, local arrow files (#348) 
* support user defined prompters, pretokenized datasets in config, local parquet, local arrow files

* fix user defined dataset types

* fix for system prompts

* fix tests

* fix checks for parquet and arrow

* aha moment that d.data_files isn't used

* add documentation for ds_type to add support for parquet and arrow
 2023-08-20 09:17:49 -04:00
Philpax
d21318dfb9 docs(readme): add cd axolotl (#440) 2023-08-19 19:14:05 -04:00
Wing Lian
b3f5e00ff5 use save_strategy from config if available (#434) 
* use save_strategy from config if available

* update docs for save_strategy
 2023-08-18 20:28:23 -04:00