See axolotl config

axolotl version: 0.12.2

base_model: Qwen/Qwen3-8B

strict: false

chat_template: tokenizer_default
datasets:
  - path: trillionlabs/android_control_ER_index_1000
    type: chat_template
    split: train
    field_messages: messages

adapter: lora
lora_model_dir:
lora_r: 32
lora_alpha: 64
lora_dropout: 0.05
lora_target_linear: true

dataset_prepared_path: datasets/android_control_ER_index_1000_prepared
val_set_size: 0.01
output_dir: ./outputs/sft_android_control_ER_index_1000
hub_model_id: trillionlabs/android_control_ER_index_1000

sequence_len: 6144
sample_packing: false
pad_to_sequence_len: false

wandb_project: axolotl
wandb_entity: suyeong_korea_univ-korea-university
wandb_name: android_control_ER_index_1000

gradient_accumulation_steps: 2
micro_batch_size: 2
num_epochs: 2
optimizer: adamw_torch_fused
lr_scheduler: cosine
learning_rate: 3e-6

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

gradient_checkpointing: true
gradient_checkpointing_kwargs:
  use_reentrant: false
early_stopping_patience:
resume_from_checkpoint:
logging_steps: 1
xformers_attention:
flash_attention: true
max_prompt_len: 6144


warmup_steps: 50
evals_per_epoch: 0
eval_table_size:
saves_per_epoch: 1
debug:
deepspeed:
weight_decay: 0.0
fsdp:
  - full_shard
  - auto_wrap
fsdp_config:
  fsdp_limit_all_gathers: true
  fsdp_sync_module_states: true
  fsdp_offload_params: true
  fsdp_use_orig_params: false
  fsdp_cpu_ram_efficient_loading: true
  fsdp_auto_wrap_policy: TRANSFORMER_BASED_WRAP
  fsdp_state_dict_type: FULL_STATE_DICT
  fsdp_sharding_strategy: FULL_SHARD
  fsdp_backward_prefetch: BACKWARD_PRE
special_tokens:
  pad_token: <|pad_token|>
  eos_token: <|im_end|>

seed: 11

android_control_ER_index_1000

This model is a fine-tuned version of Qwen/Qwen3-8B on the trillionlabs/android_control_ER_index_1000 dataset.

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: 3e-06
• train_batch_size: 2
• eval_batch_size: 2
• seed: 11
• distributed_type: multi-GPU
• num_devices: 4
• gradient_accumulation_steps: 2
• total_train_batch_size: 16
• total_eval_batch_size: 8
• optimizer: Use OptimizerNames.ADAMW_TORCH_FUSED 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: 50
• training_steps: 689

Training results

Framework versions

• PEFT 0.17.0
• Transformers 4.56.0
• Pytorch 2.7.1+cu126
• Datasets 4.0.0
• Tokenizers 0.22.0