See axolotl config

axolotl version: 0.4.1

adapter: lora
base_model: berkeley-nest/Starling-LM-7B-alpha
bf16: auto
chat_template: llama3
dataloader_num_workers: 12
dataset_prepared_path: null
datasets:
- data_files:
  - 2614f613adf7b61b_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/2614f613adf7b61b_train_data.json
  type:
    field_instruction: source
    field_output: target
    format: '{instruction}'
    no_input_format: '{instruction}'
    system_format: '{system}'
    system_prompt: ''
debug: null
deepspeed: null
device_map: auto
do_eval: true
early_stopping_patience: 2
eval_batch_size: 8
eval_max_new_tokens: 128
eval_steps: 300
eval_table_size: null
evals_per_epoch: null
flash_attention: true
fp16: null
fsdp: null
fsdp_config: null
gradient_accumulation_steps: 8
gradient_checkpointing: true
group_by_length: true
hub_model_id: nttx/6a03cbfb-f508-48da-82ce-1fec35e9804b
hub_private_repo: false
hub_repo: null
hub_strategy: checkpoint
hub_token: null
learning_rate: 0.0004
load_in_4bit: false
load_in_8bit: false
local_rank: null
logging_steps: 300
lora_alpha: 16
lora_dropout: 0.05
lora_fan_in_fan_out: null
lora_model_dir: null
lora_r: 8
lora_target_linear: true
lr_scheduler: cosine
max_grad_norm: 1.0
max_memory:
  0: 75GB
max_steps: 3000
micro_batch_size: 8
mlflow_experiment_name: /tmp/2614f613adf7b61b_train_data.json
model_type: AutoModelForCausalLM
num_epochs: 100
optim_args:
  adam_beta1: 0.9
  adam_beta2: 0.95
  adam_epsilon: 1e-5
optimizer: adamw_bnb_8bit
output_dir: miner_id_24
pad_to_sequence_len: true
push_to_hub: true
resume_from_checkpoint: null
s2_attention: null
sample_packing: false
save_steps: 300
saves_per_epoch: null
sequence_len: 512
strict: false
tf32: false
tokenizer_type: AutoTokenizer
train_on_inputs: false
trust_remote_code: true
val_set_size: 0.05
wandb_entity: null
wandb_mode: online
wandb_name: 387b315f-1ade-4d76-8f5b-08bf2466aaeb
wandb_project: Gradients-On-Demand
wandb_run: your_name
wandb_runid: 387b315f-1ade-4d76-8f5b-08bf2466aaeb
warmup_steps: 100
weight_decay: 0.0
xformers_attention: null

6a03cbfb-f508-48da-82ce-1fec35e9804b

This model is a fine-tuned version of berkeley-nest/Starling-LM-7B-alpha on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.7999

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: 0.0004
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• gradient_accumulation_steps: 8
• total_train_batch_size: 64
• optimizer: Use OptimizerNames.ADAMW_BNB with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=adam_beta1=0.9,adam_beta2=0.95,adam_epsilon=1e-5
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 100
• training_steps: 3000

Training results

Training Loss	Epoch	Step	Validation Loss
No log	0.0001	1	1.9461
7.1434	0.0397	300	1.1098
6.9099	0.0793	600	0.9761
7.0094	0.1190	900	1.0105
6.9951	0.1586	1200	0.9529
6.9451	0.1983	1500	0.9290
6.8013	0.2380	1800	0.8606
6.7064	0.2776	2100	0.8218
6.5338	0.3173	2400	0.8056
6.5503	0.3570	2700	0.8002
6.4917	0.3966	3000	0.7999

Framework versions

• PEFT 0.13.2
• Transformers 4.46.0
• Pytorch 2.5.0+cu124
• Datasets 3.0.1
• Tokenizers 0.20.1