tocmo0nlord/axolotl
1
0
Fork 0
Code Issues Pull Requests Actions 2 Packages Projects Releases Wiki Activity

improve how we setup eval/save strategies and steps (#547)

Browse Source 
* setup save end eval strategies to be consistent with trainer logic

* add comments

* better eval handling
This commit is contained in:
Wing Lian
2023-09-13 11:37:23 -04:00
committed by GitHub
parent e7aa7b1a1e
commit 36e53c7442
1 changed files with 18 additions and 6 deletions
Download Patch File Download Diff File Expand all files Collapse all files

24
src/axolotl/utils/trainer.py
View File

@@ -567,21 +567,33 @@ def setup_trainer(cfg, train_dataset, eval_dataset, model, tokenizer, total_num_
"sample_packing_efficiency" "sample_packing_efficiency"
] = cfg.sample_packing_eff_est ] = cfg.sample_packing_eff_est
if cfg.val_set_size == 0: if cfg.eval_steps and cfg.evaluation_strategy:
# assume if the user set both, they know what they're doing
training_arguments_kwargs["evaluation_strategy"] = cfg.evaluation_strategy
training_arguments_kwargs["eval_steps"] = cfg.eval_steps
elif cfg.val_set_size == 0:
# no eval set, so don't eval
training_arguments_kwargs["evaluation_strategy"] = "no" training_arguments_kwargs["evaluation_strategy"] = "no"
elif cfg.evaluation_strategy and cfg.evaluation_strategy in ["epoch", "no"]:
# if explicitly set for epoch, just set, and eval steps don't matter
training_arguments_kwargs["evaluation_strategy"] = cfg.evaluation_strategy
elif cfg.eval_steps: elif cfg.eval_steps:
# steps isn't used w/ epochs
training_arguments_kwargs["evaluation_strategy"] = "steps" training_arguments_kwargs["evaluation_strategy"] = "steps"
training_arguments_kwargs["eval_steps"] = cfg.eval_steps training_arguments_kwargs["eval_steps"] = cfg.eval_steps
else: else:
# we have an eval set, but no steps defined, use epoch # we have an eval set, but no steps defined, default to use epoch
training_arguments_kwargs["evaluation_strategy"] = "epoch" training_arguments_kwargs["evaluation_strategy"] = "epoch"
if cfg.save_strategy: if cfg.save_steps:
# save_steps implies save_strategy of steps
training_arguments_kwargs["save_strategy"] = "steps"
training_arguments_kwargs["save_steps"] = cfg.save_steps
elif cfg.save_strategy:
training_arguments_kwargs["save_strategy"] = cfg.save_strategy training_arguments_kwargs["save_strategy"] = cfg.save_strategy
else: else:
training_arguments_kwargs["save_strategy"] = ( # default to saving each epoch if not defined
"steps" if cfg.save_steps else "epoch" training_arguments_kwargs["save_strategy"] = "epoch"
)
if cfg.do_bench_eval: if cfg.do_bench_eval:
training_arguments_kwargs["do_bench_eval"] = cfg.do_bench_eval training_arguments_kwargs["do_bench_eval"] = cfg.do_bench_eval