base_model: mistralai/Mixtral-8x7B-v0.1
model_type: AutoModelForCausalLM
tokenizer_type: AutoTokenizer

model_config:
  output_router_logits: true
  router_aux_loss_coef: 0.02
  router_z_loss_coef: 0.001

load_in_8bit: false
load_in_4bit: true
strict: false

datasets:
  - path: /workspace/data/rpguild-grammar-filtered.jsonl
    type: rp_forum
    shards: 4
  - path: /workspace/limaerp-8192.jsonl
    type: rp_forum
    repeats: 2
  - path: /workspace/data/no-robots-rp.jsonl
    type: pippa
  - path: Open-Orca/SlimOrca
    type: sharegpt
    shards: 15
  - path: chargoddard/summarize_from_feedback_alpaca
    type: alpaca
    shards: 14
  - path: chargoddard/coedit-reworded
    type: alpaca
    shards: 14
prompt_format: economic_rp
dataset_prepared_path: last_run_prepared
val_set_size: 0.04
output_dir: ./mixtral-lora
save_safetensors: true

adapter: qlora
lora_model_dir:

sequence_len: 8192
sample_packing: true
pad_to_sequence_len: true

lora_r: 128
lora_alpha: 256
lora_dropout: 0.05
lora_target_linear: true
lora_fan_in_fan_out:
# lora_target_modules:

wandb_project: mixtral-lora
wandb_entity:
wandb_watch:
wandb_run_id:
wandb_log_model:

gradient_accumulation_steps: 16
micro_batch_size: 2
eval_batch_size: 2
num_epochs: 1
optimizer: adamw_bnb_8bit
lr_scheduler: cosine
learning_rate: 0.00025

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

gradient_checkpointing: true
early_stopping_patience:
resume_from_checkpoint:
auto_resume_from_checkpoints: true
local_rank:
logging_steps: 1
xformers_attention:
flash_attention: true
save_total_limit: 2

warmup_steps: 10
eval_steps: 0.05
save_steps: 0.05
eval_table_size:
eval_table_max_new_tokens: 128
weight_decay: 0.0
special_tokens:
  bos_token: "<s>"
  eos_token: "</s>"
  unk_token: "<unk>"