axolotl version: 0.4.0
base_model: four-two-labs/tinyllama-moe-nord-completion-6B
model_type: AutoModelForCausalLM
tokenizer_type: LlamaTokenizer
load_in_8bit: false
load_in_4bit: false
strict: false
rl: orpo
orpo_alpha: 0.1
remove_unused_columns: false
chat_template: chatml
datasets:
- path: four-two-labs/nord-dpo-mix-181k-axolotl
type: chat_template.argilla
split: train
output_dir: ./runs/model/tinyllama-moe-orpo
dataset_prepared_path: ./runs/data/tinyllama-dpo-data
val_set_size: 0.01
sequence_len: 2048
sample_packing: false
pad_to_sequence_len: false
wandb_project: axolotl
wandb_entity:
wandb_watch:
wandb_name:
wandb_log_model:
gradient_accumulation_steps: 1
micro_batch_size: 2
num_epochs: 3
optimizer: paged_adamw_8bit
lr_scheduler: cosine
learning_rate: 3e-5
train_on_inputs: false
group_by_length: false
bf16: true
fp16:
tf32: true
gradient_checkpointing: true
early_stopping_patience:
resume_from_checkpoint:
local_rank:
logging_steps: 1
xformers_attention:
flash_attention: true
loss_watchdog_threshold: 5.0
loss_watchdog_patience: 3
warmup_steps: 10
evals_per_epoch: 4
eval_table_size:
eval_max_new_tokens: 128
saves_per_epoch: 1
debug:
deepspeed:
weight_decay: 0.0
fsdp:
fsdp_config:
special_tokens:
runs/model/tinyllama-moe-orpo
This model is a fine-tuned version of four-two-labs/tinyllama-moe-nord-completion-6B on the None dataset.
Model description
More information needed
Intended uses & limitations
More information needed
Training and evaluation data
More information needed
Training procedure
Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 3e-05
- train_batch_size: 2
- eval_batch_size: 8
- seed: 42
- distributed_type: multi-GPU
- num_devices: 10
- total_train_batch_size: 20
- total_eval_batch_size: 80
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: cosine
- lr_scheduler_warmup_steps: 10
- training_steps: 27511
Training results
Framework versions
- Transformers 4.40.0.dev0
- Pytorch 2.1.2+cu118
- Datasets 2.15.0
- Tokenizers 0.15.0