llama-3.1-orm.yaml

base_model: meta-llama/Llama-3.1-8B-Instruct
model_type: AutoModelForCausalLM
tokenizer_type: AutoTokenizer

load_in_8bit: false
load_in_4bit: false
strict: false

datasets:
  - path: RLHFlow/Mistral-ORM-Data
    conversation: llama3
    type: sharegpt
    split: "train"
    train_on_split: "train"

warmup_ratio: 0.05
val_set_size: 0.0
output_dir: ./orm
#wandb_project: preference-models
#wandb_entity: domain-generalization
wandb_watch:
wandb_name: "llama-31-8b-bs32_lr2e-6_orm"
wandb_log_model:

train_on_inputs: false

save_safetensors: true
#noisy_embedding_alpha: 10.0 # default for sharegpt type
dataset_prepared_path: ~/data/preference-models/last_run_prepared


dataset_processes: 48
#torch_compile: true
sequence_len: 8192
sample_packing: true
pad_to_sequence_len: true

trust_remote_code: True
adapter:
lora_model_dir:
#lora_r: 32
#lora_alpha: 16
#lora_dropout: 0.05
#lora_target_linear: true
#lora_fan_in_fan_out:




gradient_checkpointing: True

#warmup_ratio: 0.1
gradient_accumulation_steps: 4
micro_batch_size: 1
num_epochs: 1
#max_steps: 10
#optimizer: adamw_torch_fused
optimizer: paged_adamw_32bit
#lr_scheduler: constant_with_warmup
lr_scheduler: cosine
learning_rate: 2.0e-6

weight_decay: 0.0
max_grad_norm: 1.0


group_by_length: false
bf16: auto
fp16: false
tf32: true

early_stopping_patience:
local_rank:
logging_steps: 2
xformers_attention:
flash_attention: true


eval_steps:
eval_table_size:
eval_table_max_new_tokens:
#save_steps: 100
save_strategy: "epoch"
save_total_limit: 4
#save_safetensors: false
debug:


ddp: #true
deepspeed: #deepspeed/zero1.json # multi-gpu only

fsdp:
fsdp_config:
special_tokens:
        pad_token: <|end_of_text|>