base_model: mistralai/Mixtral-8x7B-v0.1
# optionally might have model_type or tokenizer_type
model_type: AutoModelForCausalLM
tokenizer_type: LlamaTokenizer
# Automatically upload checkpoint and final model to HF
# hub_model_id: username/custom_model_name

trust_remote_code: true

load_in_8bit: false
load_in_4bit: true

datasets:
  - path: tatsu-lab/alpaca
    type: alpaca
dataset_prepared_path: last_run_prepared
val_set_size: 0.02
output_dir: ./outputs/qlora-out

model_config:
  output_router_logits: true

adapter: qlora
lora_model_dir:

sequence_len: 1024
sample_packing: false
pad_to_sequence_len: false

lora_r: 32
lora_alpha: 16
lora_dropout: 0.05
lora_target_linear: true

wandb_project:
wandb_entity:
wandb_watch:
wandb_name:
wandb_log_model:

gradient_accumulation_steps: 4
micro_batch_size: 2
num_epochs: 1
optimizer: adamw_torch_fused
lr_scheduler: cosine
learning_rate: 0.0002

bf16: auto
tf32: true

gradient_checkpointing: true
resume_from_checkpoint:
logging_steps: 1
flash_attention: true

loss_watchdog_threshold: 5.0
loss_watchdog_patience: 3

warmup_steps: 10
evals_per_epoch: 4
saves_per_epoch: 1

weight_decay: 0.0
fsdp:
  - full_shard
  - auto_wrap
fsdp_config:
  fsdp_limit_all_gathers: true
  fsdp_sync_module_states: true
  fsdp_offload_params: true
  fsdp_use_orig_params: false
  fsdp_cpu_ram_efficient_loading: true
  fsdp_transformer_layer_cls_to_wrap: MixtralSparseMoeBlock
  fsdp_state_dict_type: FULL_STATE_DICT
  fsdp_auto_wrap_policy: TRANSFORMER_BASED_WRAP
  fsdp_sharding_strategy: FULL_SHARD
  fsdp_forward_prefetch: false
  fsdp_backward_prefetch: BACKWARD_PRE
special_tokens: