base_model: nvidia/Llama-3.1-Minitron-4B-Width-Base
model_type: LlamaForCausalLM
tokenizer_type: AutoTokenizer

load_in_8bit: false
load_in_4bit: false
strict: false

datasets:
  - path: mlabonne/FineTome-100k
    type: chat_template
    split: train
    train_on_eos: turn
dataset_prepared_path: last_run_prepared
val_set_size: 0.0
output_dir: ./outputs/out

sequence_len: 2048
sample_packing: true
pad_to_sequence_len: true

wandb_project: device_mesh-test
wandb_entity: axolotl-ai
wandb_watch:
wandb_name:
wandb_log_model:

gradient_accumulation_steps: 1
micro_batch_size: 4
num_epochs: 1
optimizer: adamw_torch
lr_scheduler: cosine
learning_rate: 2e-5

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

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

warmup_steps: 100
evals_per_epoch: 1
saves_per_epoch: 1
weight_decay: 0.0
fsdp:
  - auto_wrap
fsdp_config:
  fsdp_use_orig_params: true
  fsdp_auto_wrap_policy: TRANSFORMER_BASED_WRAP
  fsdp_transformer_layer_cls_to_wrap: LlamaDecoderLayer
special_tokens:
  pad_token: <|end_of_text|>