base_model: microsoft/phi-2
model_type: AutoModelForCausalLM
tokenizer_type: AutoTokenizer
is_llama_derived_model: false
trust_remote_code: true

load_in_8bit: false
load_in_4bit: true
strict: false

datasets:
  - path: s3nh/alpaca-dolly-instruction-only-polish
    type: alpaca

dataset_prepared_path:
val_set_size: 0.05
output_dir: ./phi-2-sft-out

sequence_len: 1024
sample_packing: false  # not CURRENTLY compatible with LoRAs
pad_to_sequence_len:

adapter: qlora
lora_model_dir:
lora_r: 64
lora_alpha: 32
lora_dropout: 0.05
lora_target_linear: true
lora_fan_in_fan_out:

wandb_project:
wandb_entity:
wandb_watch:
wandb_name:
wandb_log_model:

gradient_accumulation_steps: 1
micro_batch_size: 1
num_epochs: 4
optimizer: adamw_torch
adam_beta2: 0.95
adam_epsilon: 0.00001
max_grad_norm: 1.0
lr_scheduler: cosine
learning_rate: 0.000003

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

gradient_checkpointing:
early_stopping_patience:
resume_from_checkpoint: false
local_rank:
logging_steps: 100
xformers_attention:
flash_attention: true

warmup_steps: 10
evals_per_epoch: 4
saves_per_epoch:
save_strategy: steps
save_steps: 5000
debug:
deepspeed:
weight_decay: 0.1
fsdp:
fsdp_config:
resize_token_embeddings_to_32x: true
special_tokens:
  bos_token: "<|endoftext|>"
  eos_token: "<|endoftext|>"
  unk_token: "<|endoftext|>"
  pad_token: "<|endoftext|>"