See axolotl config

axolotl version: 0.4.0

base_model: codellama/CodeLlama-7b-hf
model_type: LlamaForCausalLM
tokenizer_type: CodeLlamaTokenizer
is_llama_derived_model: true

hub_model_id: noeloco/modeltest1-dpo

load_in_8bit: false
load_in_4bit: true
strict: false

datasets:
  - path: noeloco/fizzbuzz-sft
    type: alpaca
    ds_type: json

hf_use_auth_token: true
push_dataset_to_hub: noeloco
val_set_size: 0.05
output_dir: ./lora-out
chat_template: chatml

rl: dpo
datasets:
    - path: noeloco/fizzbuzz-dpo
      split: train
      data_files: 
        - /tmp/fizzbuzz-ft/datasets/training-set-dpo.json  
        #type:
        # field_prompt: question
        # field_chosen: chosen
        # field_rejected: rejected
      ds_type: json
        #type: intel_apply_chatml
      type: chatml.intel

hf_use_auth_token: true
push_dataset_to_hub: noeloco
val_set_size: 0.05
output_dir: ./lora-out
chat_template: chatml


sequence_len: 2048
sample_packing: false
pad_to_sequence_len: true

adapter: lora
lora_model_dir:
lora_r: 16
lora_alpha: 8
lora_dropout: 0.05
lora_target_linear: true
lora_fan_in_fan_out:

wandb_project: runpod1
wandb_entity:
wandb_watch:
wandb_name: 
wandb_log_model:

gradient_accumulation_steps: 1
micro_batch_size: 2
num_epochs: 3
optimizer: paged_adamw_32bit
lr_scheduler: cosine
learning_rate: 0.0002

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

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

warmup_steps: 10
evals_per_epoch: 4
saves_per_epoch: 1
debug: true
deepspeed:
weight_decay: 0.0
fsdp:
fsdp_config:
special_tokens:
  bos_token: "<s>"
  eos_token: "</s>"
  unk_token: "<unk>"

modeltest1-dpo

This model is a fine-tuned version of codellama/CodeLlama-7b-hf 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: 0.0002
• train_batch_size: 2
• eval_batch_size: 8
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 10
• training_steps: 222

Training results

Framework versions

• PEFT 0.10.1.dev0
• Transformers 4.40.0.dev0
• Pytorch 2.1.2+cu118
• Datasets 2.15.0
• Tokenizers 0.15.0