Adaptor size increases significantly when PT with LORA done with modules to save has #lora_modules_to_save: # - embed_tokens # - lm_head

Please check that this issue hasn't been reported before.

[X] I searched previous Bug Reports didn't find any similar reports.

Expected Behavior

Lora Adaptor size shouldnt be in GBs on completeing Lora Pretraining. This happens due to following config: lora_modules_to_save:

embed_tokens
lm_head

Without this it works fine. There was Peft Library issue for this but it still seems to be there. https://github.com/huggingface/trl/issues/1287

Current behaviour

On PT Lora Training Gemma 2B it creates an adaptor of 4GB almost same the size of the base model.

This happens due to the mentioned modules to save config. If they are removed it works fine. Given below is the cfg file i used.

Steps to reproduce

Pretrain Gemma with Lora 2B with given config. Check the adaptor size.

Config yaml

base_model: google/gemma-2b
model_type: AutoModelForCausalLM
tokenizer_type: AutoTokenizer

load_in_8bit: false
load_in_4bit: true
strict: false

datasets:
  - path: /content/test_txt_data-10exmpl.json
    type: completion
    field: text
dataset_prepared_path:
val_set_size: 0
output_dir: ./qlora-out

adapter: qlora
lora_model_dir:

sequence_len: 600
sample_packing: true
pad_to_sequence_len: true

lora_r: 64
lora_alpha: 128
lora_dropout: 0.05
lora_target_modules: 
  - q_proj
  - v_proj
  - k_proj
  - o_proj
  - gate_proj
  - down_proj
  - up_proj
lora_modules_to_save:
  - embed_tokens
  - lm_head
lora_target_linear: true
lora_fan_in_fan_out:

gradient_accumulation_steps: 1
micro_batch_size: 1
num_epochs: 1
optimizer: adamw_bnb_8bit
lr_scheduler: cosine
learning_rate: 0.0002

train_on_inputs: false
group_by_length: false
bf16: false
fp16: false
tf32: false

gradient_checkpointing: true
early_stopping_patience:
resume_from_checkpoint:
local_rank:
logging_steps: 1
xformers_attention:
flash_attention: False

warmup_ratio: 0.1
evals_per_epoch: 1
eval_table_size:
eval_max_new_tokens: 128
eval_sample_packing: False
saves_per_epoch: 1
debug:
deepspeed:
weight_decay: 0.0
fsdp:
fsdp_config:
special_tokens:

save_safetensors: True

Possible solution

No response

Which Operating Systems are you using?

[X] Linux
[ ] macOS
[ ] Windows

Python Version

3.10

axolotl branch-commit

latest

Acknowledgements

[X] My issue title is concise, descriptive, and in title casing.
[X] I have searched the existing issues to make sure this bug has not been reported yet.
[X] I am using the latest version of axolotl.
[X] I have provided enough information for the maintainers to reproduce and diagnose the issue.

axolotl-ai-cloud / axolotl

Adaptor size increases significantly when PT with LORA done with modules to save has #lora_modules_to_save: # - embed_tokens # - lm_head #1511