base_model: /data/models/Qwen1.5-72B
model_type: AutoModelForCausalLM
tokenizer_type: AutoTokenizer

trust_remote_code: true

load_in_8bit: false
load_in_4bit: true
strict: false

datasets:
  - path: /data/eric/datasets/m-a-p_Code-Feedback-sharegpt.jsonl
    type: sharegpt
    conversation: chatml
  - path: /data/eric/datasets/m-a-p_CodeFeedback-Filtered-Instruction-sharegpt.jsonl
    type: sharegpt
    conversation: chatml
  - path: /data/eric/datasets/openhermes2_5-sharegpt.jsonl
    type: sharegpt
    conversation: chatml
  - path: /data/eric/datasets/gorilla_openfunctions_train_sharegpt.jsonl
    type: sharegpt
    conversation: chatml
  - path: /data/eric/datasets/SystemConversations.jsonl
    type: sharegpt
    conversation: chatml
  - path: /data/eric/datasets/identity_sharegpt.jsonl
    type: sharegpt
    conversation: chatml

chat_template: chatml

dataset_prepared_path: thingy
val_set_size: 0
output_dir: /data/eric/Liberated-Qwen-1.5-72b

sequence_len: 8192
sample_packing: true
pad_to_sequence_len: true

adapter: qlora
lora_model_dir:
lora_r: 32
lora_alpha: 16
lora_dropout: 0.05
lora_target_linear: true
lora_fan_in_fan_out:

wandb_project: AbacusLiberated
wandb_entity:
wandb_watch:
wandb_name:
wandb_log_model:

gradient_accumulation_steps: 8
micro_batch_size: 1
num_epochs: 3
optimizer: adamw_bnb_8bit
lr_scheduler: cosine
learning_rate: 0.0002

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

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

warmup_steps: 10
evals_per_epoch: 4
eval_table_size:
eval_max_new_tokens: 128
saves_per_epoch: 4
debug:
deepspeed: deepspeed_configs/zero2.json
weight_decay: 0.0
fsdp:
fsdp_config:

special_tokens:
  eos_token: "<|im_end|>"