See axolotl config

axolotl version: 0.6.0

adapter: lora
base_model: oopsung/llama2-7b-koNqa-test-v1
bf16: auto
data_collator:
  max_length: 8192
  padding: true
  type: dynamic_padding
dataset_prepared_path: null
datasets:
- data_files:
  - 7897b36af6847987_train_data.json
  ds_type: json
  format: custom
  path: 7897b36af6847987_train_data.json
  preprocessing:
  - shuffle: true
  type:
    field: null
    field_input: null
    field_instruction: mood
    field_output: lyrics
    field_system: null
    format: null
    no_input_format: null
    system_format: '{system}'
    system_prompt: ''
debug: null
deepspeed: null
early_stopping_patience: null
eval_max_new_tokens: 128
eval_table_size: null
evals_per_epoch: 4
flash_attention: true
fp16: null
fsdp: null
fsdp_config: null
gradient_accumulation_steps: 4
gradient_checkpointing: true
group_by_length: false
hub_model_id: taopanda/test-llama2-7b-koNqa-test-v1
learning_rate: 0.0001980900647573094
load_in_4bit: false
load_in_8bit: false
local_rank: null
logging_steps: 1
lora_alpha: 32
lora_dropout: 0.05
lora_fan_in_fan_out: null
lora_model_dir: null
lora_r: 16
lora_target_linear: true
lr_scheduler: cosine
max_grad_norm: 1.0
max_steps: 600
micro_batch_size: 8
model_type: LlamaForCausalLM
num_epochs: 2
optimizer: adamw_bnb_8bit
output_dir: ./outputs/lora-out/taopanda_test-llama2-7b-koNqa-test-v1
resume_from_checkpoint: null
s2_attention: null
save_safetensors: true
save_steps: 0.15
save_total_limit: 1
seed: 26232
special_tokens: null
strict: false
tf32: true
tokenizer_type: AutoTokenizer
train_on_inputs: false
trust_remote_code: true
val_set_size: 0.1
wandb_entity: fatcat87-taopanda
wandb_log_model: null
wandb_mode: online
wandb_name: taopanda_test-llama2-7b-koNqa-test-v1
wandb_project: subnet56-test
wandb_runid: taopanda_test-llama2-7b-koNqa-test-v1
wandb_watch: null
warmup_ratio: 0.06
weight_decay: 0.0
xformers_attention: null

test-llama2-7b-koNqa-test-v1

This model is a fine-tuned version of oopsung/llama2-7b-koNqa-test-v1 on the 7897b36af6847987_train_data.json dataset. It achieves the following results on the evaluation set:

• Loss: 1.5600

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: 0.0001980900647573094
• train_batch_size: 8
• eval_batch_size: 8
• seed: 26232
• distributed_type: multi-GPU
• num_devices: 4
• gradient_accumulation_steps: 4
• total_train_batch_size: 128
• total_eval_batch_size: 32
• optimizer: Use OptimizerNames.ADAMW_BNB with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 12
• training_steps: 205

Training results

Training Loss	Epoch	Step	Validation Loss
2.0028	0.0098	1	2.0533
1.7152	0.2543	26	1.6692
1.5552	0.5086	52	1.6252
1.648	0.7628	78	1.6017
1.5565	1.0098	104	1.5852
1.5165	1.2641	130	1.5732
1.5192	1.5183	156	1.5643
1.5389	1.7726	182	1.5600

Framework versions

• PEFT 0.14.0
• Transformers 4.48.1
• Pytorch 2.5.1+cu124
• Datasets 3.2.0
• Tokenizers 0.21.0