llama-3-orpo-ml

<!-- This model card has been generated automatically according to the information the Trainer had access to. You should probably proofread and complete it, then remove this comment. -->

WandB: https://wandb.ai/oaaic/orpo-llama-3/runs/gc2d3cxp

Benchmarks: TBD

<img src="https://raw.githubusercontent.com/OpenAccess-AI-Collective/axolotl/main/image/axolotl-badge-web.png" alt="Built with Axolotl" width="200" height="32"/>

<details><summary>See axolotl config</summary>

axolotl version: 0.4.0

base_model: winglian/meta-llama3-chatml
model_type: LlamaForCausalLM
tokenizer_type: AutoTokenizer

load_in_4bit: true

rl: orpo
orpo_alpha: 0.1
chat_template: chatml
datasets:
  - path: mlabonne/orpo-dpo-mix-40k
    type: chat_template.argilla
    chat_template: chatml

dataset_prepared_path: last_run_prepared
val_set_size: 0.01
output_dir: ./llama-3-orpo-qlora

sequence_len: 2048
sample_packing: false
pad_to_sequence_len: false

adapter: qlora
lora_r: 16
lora_alpha: 32
lora_dropout: 0.05
lora_target_modules:
  - q_proj
  - k_proj
  - v_proj
  - o_proj
  - gate_proj
  - up_proj
  - down_proj

wandb_project: orpo-llama-3
wandb_entity: oaaic
wandb_watch:
wandb_name:
wandb_log_model:

gradient_accumulation_steps: 4
micro_batch_size: 8
num_epochs: 1
optimizer: paged_adamw_8bit
lr_scheduler: cosine
learning_rate: 1.4e-5
max_grad_norm: 1.0

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

gradient_checkpointing: true
gradient_checkpointing_kwargs:
  use_reentrant: true
logging_steps: 1
flash_attention: true

warmup_steps: 10
evals_per_epoch: 5
saves_per_epoch: 1
weight_decay: 0.0
special_tokens:
  pad_token: <|end_of_text|>

</details><br>

llama-3-orpo-qlora

This model was trained from scratch 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: 1.4e-05
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 32
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 10
• training_steps: 1241

Training results

Framework versions

• PEFT 0.10.0
• Transformers 4.40.0.dev0
• Pytorch 2.1.2+cu118
• Datasets 2.15.0
• Tokenizers 0.15.0