config.yml

llama2.yml

base_model: conceptofmind/LLongMA-2-13b-16k
base_model_config: conceptofmind/LLongMA-2-13b-16k
model_type: LlamaForCausalLM
tokenizer_type: LlamaTokenizer

load_in_8bit: false
load_in_4bit: true
strict: false

datasets:
  - path: ehartford/wizard_vicuna_70k_unfiltered
    data_files: wizard_vicuna_dataset_unfiltered.json
    type: sharegpt:chat

dataset_prepared_path: last_run_prepared
val_set_size: 0.001
output_dir: ./qlora-out

adapter: qlora
lora_model_dir:

sequence_len: 16000
max_packed_sequence_len: 16000
lora_r: 32
lora_alpha: 16
lora_dropout: 0.1
lora_target_modules:
lora_target_linear: true
lora_fan_in_fan_out:

wandb_project:
wandb_watch:
wandb_run_id:
wandb_log_model:

gradient_accumulation_steps: 4
micro_batch_size: 1
num_epochs: 3
learning_rate: 0.0002
train_on_inputs: false
group_by_length: false
bf16: false
fp16: true
tf32: false

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

warmup_steps: 10
eval_steps: 20
save_steps:
debug:
deepspeed:
weight_decay: 0.0
fsdp:
fsdp_config:
special_tokens:
  bos_token: "<s>"
  eos_token: "</s>"
  unk_token: "<unk>"
  pad_token: "<pad>"