base_model: TinyLlama/TinyLlama-1.1B-intermediate-step-1195k-token-2.5T
model_type: LlamaForCausalLM
tokenizer_type: LlamaTokenizer
is_llama_derived_model: true

load_in_8bit: true
load_in_4bit: false
strict: false

datasets:
  - path: jondurbin/airoboros-3.1
    type: sharegpt
dataset_prepared_path:
val_set_size: 0.01
output_dir: ./airoboros-lora-out

sequence_len: 2048
sample_packing: true

adapter: lora
lora_model_dir:
lora_r: 128
lora_alpha: 64
lora_dropout: 0.1
lora_target_linear: true
lora_fan_in_fan_out:

wandb_project: airoboros-tinyllama
wandb_entity:
wandb_watch:
wandb_name:
wandb_log_model:

gradient_accumulation_steps: 1
micro_batch_size: 8
num_epochs: 1
optimizer: adamw_bnb_8bit
lr_scheduler: cosine
learning_rate: 0.0002
adam_beta1: 0.999
adam_beta2: 0.95
adam_eps: 0.00001

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

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

warmup_steps: 100
evals_per_epoch: 2
eval_table_size:
saves_per_epoch: 2
debug:
deepspeed:
weight_decay: 0.0
fsdp:
fsdp_config:
special_tokens: