metadata

license: apache-2.0
datasets:
  - mlabonne/orpo-dpo-mix-40k
language:
  - en
library_name: transformers
base_model: h2oai/h2o-danube2-1.8b-base
tags:
  - llama-factory
  - unsloth

h2o-danube2 with ChatML template

This model was first fine-tuned with BAdam on mlabonne/orpo-dpo-mix-40k, but as SFT and not DPO, using LLama-Factory.

Template

<|im_start|>user
{{instruction}}<|im_end|>
<|im_start|>assistant
{{response}}<|im_end|>

BAdam config

### model
model_name_or_path: danube2-base-chatml

### method
stage: sft
do_train: true
finetuning_type: full
use_badam: true
badam_switch_mode: ascending
badam_switch_interval: 50
badam_verbose: 1
badam_start_block: 12
badam_mask_mode: scatter
seed: 314

### dataset
dataset: orpo_sft_mix_40k
template: hermes_chatml
cutoff_len: 8192
overwrite_cache: false
preprocessing_num_workers: 12

### output
output_dir: orpo-chatml-badam
logging_steps: 5
save_steps: 1
save_strategy: epoch
plot_loss: true
overwrite_output_dir: false

### train
per_device_train_batch_size: 2
gradient_accumulation_steps: 8
learning_rate: 0.00001
num_train_epochs: 2
lr_scheduler_type: cosine
warmup_ratio: 0.01
pure_bf16: true
flash_attn: fa2

### eval
val_size: 0.01
per_device_eval_batch_size: 1
eval_strategy: steps
eval_steps: 1000

BAdam training results

Training Loss	Epoch	Step	Validation Loss
0.7474	0.3653	1000	0.8887
0.9106	0.7306	2000	0.8681
0.8121	1.0958	3000	0.8635
0.8636	1.4611	4000	0.8562
0.8	1.8264	5000	0.8565