Commit Graph

14 Commits

Author SHA1 Message Date
Wing Lian
4fde300e5f update outputs path so that we can mount workspace to /workspace/data (#1623)
* update outputs path so that we can mount workspace to /workspace/data

* fix ln order
2024-05-15 12:44:13 -04:00
Wing Lian
54d2ac155b Mixtral fixes 20240124 (#1192) [skip ci]
* mixtral nccl fixes

* make sure to patch for z3
2024-01-24 14:59:57 -05:00
Wing Lian
814aee6603 Phi2 multipack (#1173)
* phi2 multipack

* update validation and examples for phi

* more updates to phi examples

* make sure to use the correct collator for phi multipack

* phi needs attention mask now for multipack

* if the special token already exists in the tokenizer, don't require in lora modules to save

* fix qlora yml for phi, fix phi test validation

* test qlora too

* make sure flash attention is enabled for the test

* don't use remote code for phi anymore

* reduce sequence len for sample packing phi
2024-01-23 12:54:36 -05:00
Wing Lian
782b6a4216 set fp16 to false if bf16, update bf16: auto in example YAMLs (#1122) [skip ci]
* set fp16 to false if bf16, update bf16: auto in example YAMLs

* unset fp16 so that it fallsback properly if bf16 isn't available

* Update README.md [skip-ci]

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

* test that bf16 disables fp16

---------

Co-authored-by: NanoCode012 <kevinvong@rocketmail.com>
2024-01-22 18:44:01 -05:00
Wing Lian
c1b741d9fb pin model_revision for phi2 (#1123) 2024-01-14 17:31:51 -05:00
Wing Lian
732851f105 Phi2 rewrite (#1058)
* restore to current phi modeling code from phi-2

* enable gradient checkpointing

* don't cast everything to float32 all the time

* gradient checkpointing for phi2 ParallelBlock module too

* fix enabling flash attn for phi2

* add comment about import

* fix phi2 example

* fix model type check for tokenizer

* revert float32 -> bf16 casting changes

* support fused dense flash attn

* fix the repo for flash-attn

* add package name for subdir pkg

* fix the data collator when not using sample packing

* install packaging for pytests in ci

* also fix setup to not install flash attn fused dense subdir if not extras

* split out the fused-dense-lib in extra requires

* don't train w group_by_length for phi

* update integration test to use phi2

* set max steps and save steps for phi e2e tests

* try to workaround ssave issue in ci

* skip phi2 e2e test for now
2024-01-08 14:04:22 -05:00
Wing Lian
5f79b8242f new evals_per_epoch and saves_per_epoch to make things cleaner (#944)
* new evals_per_epoch and saves_per_epoch to make things cleaner

* update per PR feedback
2023-12-12 15:35:23 -05:00
NanoCode012
a1da39cd48 Feat(wandb): Refactor to be more flexible (#767)
* Feat: Update to handle wandb env better

* chore: rename wandb_run_id to wandb_name

* feat: add new recommendation and update config

* fix: indent and pop disabled env if project passed

* feat: test env set for wandb and recommendation

* feat: update to use wandb_name and allow id

* chore: add info to readme
2023-12-04 22:17:25 +09:00
Wing Lian
9bf854e59c Phi update 202311 (#876)
* add phi modeling from hf

* update for packing and use new modeling class for phi

* update e2e tests for phi to use new model name

* update example phi to also use new phi model name

* use AutoModelForCausalLM for phi lora since sample packing isn't supported
2023-11-17 12:47:17 -05:00
Wing Lian
2d8def68dc simplify by removing duplicate base_model_config (#772) 2023-10-23 01:42:38 -04:00
Wing Lian
e50a64e85e prepared dataset caching, other misc fixes (#665)
* prepared dataset caching, other misc fixes

* also don't load from disk cache unless explicit
2023-10-02 21:07:24 -04:00
Wing Lian
62eaee7649 make phi training work with Loras (#588)
* valdiation for phi loras

* fix model config class check

* update readme for phi traiing
2023-09-15 20:51:55 -04:00
Wing Lian
12a2dbbc2c Support Sample packing for phi arch (#586)
* phi sequence packing

* sample packing fixes

* fix linting

* fix inference and phi e2e tests

* update phi example now that sample packing works

* wandb import keeps getting moved around
2023-09-15 15:46:54 -04:00
Wing Lian
228420972e Phi examples (#569)
* add phi full ft example

* Add readme to point out that deepspeed should be used

* zero1 is better than zero2 for phi
2023-09-14 11:17:47 -04:00