See axolotl config

axolotl version: 0.3.0

base_model: HuggingFaceH4/mistral-7b-sft-beta
model_type: MistralForCausalLM
tokenizer_type: LlamaTokenizer
is_mistral_derived_model: true

load_in_8bit: false
load_in_4bit: false
strict: false

rl: kto_pair
datasets:
  - path: winglian/deita-nectar
    split: train_dpo
    type: zephyr.nectar
dataset_prepared_path: last_run_prepared
val_set_size: 0.0
output_dir: ./zephyr-deita-kto
save_total_limit: 3
hub_model_id: openaccess-ai-collective/kto-zephyr-deita-nectar

adapter: lora
lora_model_dir:

sequence_len: 2048
sample_packing: false
pad_to_sequence_len: false

lora_r: 128
lora_alpha: 64
lora_dropout: 0.05
lora_target_linear: true
lora_modules_to_save:
lora_fan_in_fan_out:
lora_target_modules:
  - gate_proj
  - down_proj
  - up_proj
  - q_proj
  - v_proj
  - k_proj
  - o_proj

wandb_project: dpo-zephyr-deita-nectar
wandb_entity: oaaic
wandb_watch:
wandb_run_id: kto
wandb_log_model:

gradient_accumulation_steps: 4
micro_batch_size: 4
num_epochs: 1
optimizer: paged_adamw_8bit
adam_beta2: 0.95
adam_epsilion: 0.00001
lr_scheduler: cosine
learning_rate: 1.0e-5

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

gradient_checkpointing: true
early_stopping_patience:
resume_from_checkpoint:
local_rank:
logging_steps: 1
xformers_attention:
flash_attention: true

warmup_steps: 100
eval_steps:
eval_table_size:
eval_table_max_new_tokens: 128
save_steps: 45
debug:
deepspeed:
weight_decay: 0.1
fsdp:
fsdp_config:
special_tokens:
save_safetensors: true

dataloader_num_workers: 16
dataloader_pin_memory: true

zephyr-deita-kto

This model is a fine-tuned version of HuggingFaceH4/mistral-7b-sft-beta 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: 1e-05
• train_batch_size: 4
• eval_batch_size: 8
• seed: 42
• distributed_type: multi-GPU
• num_devices: 4
• gradient_accumulation_steps: 4
• total_train_batch_size: 64
• total_eval_batch_size: 32
• optimizer: Adam with betas=(0.9,0.95) and epsilon=1e-08
• lr_scheduler_type: linear
• lr_scheduler_warmup_steps: 100
• training_steps: 135

Training results

Framework versions

• PEFT 0.7.0
• Transformers 4.37.0.dev0
• Pytorch 2.0.1+cu118
• Datasets 2.16.1
• Tokenizers 0.15.0