base_model: timarni/qwen3_dpo_100k
# Automatically upload checkpoint and final model to HF
# hub_model_id: username/custom_model_name

plugins:
  - axolotl.integrations.cut_cross_entropy.CutCrossEntropyPlugin
strict: false

chat_template: qwen3
datasets:
  - path: timarni/MNLP_STEM_IT
    type: alpaca
    split: train

shuffle_merged_datasets: true

val_set_size: 0.1
output_dir: ./outputs/dpo_100k_STEM_IT
dataset_prepared_path: last_run_prepared

sequence_len: 4096 #2048
sample_packing: true # was true -> need to check if it actually learns on the samples or not (better understand te hyperparam and event. install axolotl to debug)
eval_sample_packing: true
pad_to_sequence_len: true
# train_on_inputs: true # NEW
# group_by_length: false NEW?

# To be sure that no LORA is done
adapter: null
lora: false
merge_lora: false

wandb_project: mnlp_project
wandb_entity: tim-arni
wandb_watch:
wandb_name: dpo_100k_STEM_IT
wandb_log_model:

gradient_accumulation_steps: 16 # 2
micro_batch_size: 2 # 1
num_epochs: 3
optimizer: adamw_torch
lr_scheduler: cosine
learning_rate: 0.00005 # 0.00005
# cosine_min_lr_ratio: 0.1

warmup_ratio: 0.05
weight_decay: 0.01

bf16: auto
tf32: true

gradient_checkpointing: offload
gradient_checkpointing_kwargs:
  use_reentrant: false
resume_from_checkpoint:
logging_steps: 1
gradient_clipping: 1.0 # or max_grad_norm?
flash_attention: true

evals_per_epoch: 4
saves_per_epoch: 2
save_total_limit: 20
special_tokens:

outputs/dpo_100k_STEM_IT

This model is a fine-tuned version of timarni/qwen3_dpo_100k on the timarni/MNLP_STEM_IT dataset. It achieves the following results on the evaluation set:

Loss: 0.1704

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

learning_rate: 5e-05
train_batch_size: 2
eval_batch_size: 2
seed: 42
gradient_accumulation_steps: 16
total_train_batch_size: 32
optimizer: Use adamw_torch with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
lr_scheduler_type: cosine
lr_scheduler_warmup_steps: 12
num_epochs: 3.0

Training results

Training Loss	Epoch	Step	Validation Loss
1.0297	0.0124	1	1.0920
0.1602	0.2479	20	0.1863
0.1495	0.4957	40	0.1758
0.1409	0.7436	60	0.1709
0.1497	0.9915	80	0.1653
0.1118	1.2479	100	0.1638
0.1119	1.4957	120	0.1595
0.1068	1.7436	140	0.1590
0.1085	1.9915	160	0.1571
0.0833	2.2479	180	0.1672
0.0759	2.4957	200	0.1706
0.0875	2.7436	220	0.1705
0.0756	2.9915	240	0.1704

Framework versions

Transformers 4.51.3
Pytorch 2.5.1+cu121
Datasets 3.5.1
Tokenizers 0.21.1

timarni/dpo_100k_STEM_IT_160

outputs/dpo_100k_STEM_IT

Model description

Intended uses & limitations

Training and evaluation data

Training procedure

Training hyperparameters

Training results

Framework versions

timarni/
dpo_100k_STEM_IT_160