See axolotl config

axolotl version: 0.4.1

adapter: lora
base_model: JackFram/llama-68m
bf16: true
chat_template: llama3
dataset_prepared_path: null
datasets:
- data_files:
  - 623d9787d7fd7d0e_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/623d9787d7fd7d0e_train_data.json
  type:
    field_instruction: text
    field_output: title
    format: '{instruction}'
    no_input_format: '{instruction}'
    system_format: '{system}'
    system_prompt: ''
debug: null
device_map:
  ? ''
  : 0,1,2,3,4,5,6,7
early_stopping_patience: 2
eval_max_new_tokens: 128
eval_steps: 100
eval_table_size: null
flash_attention: true
gradient_accumulation_steps: 8
gradient_checkpointing: true
group_by_length: false
hub_model_id: Alphatao/ffcfbd55-9faa-4516-b0c4-a4012948b3ab
hub_repo: null
hub_strategy: null
hub_token: null
learning_rate: 0.0002
load_best_model_at_end: true
load_in_4bit: false
load_in_8bit: false
local_rank: null
logging_steps: 1
lora_alpha: 128
lora_dropout: 0.1
lora_fan_in_fan_out: null
lora_model_dir: null
lora_r: 64
lora_target_linear: true
lora_target_modules:
- q_proj
- k_proj
- v_proj
- o_proj
lr_scheduler: cosine
max_grad_norm: 1.0
max_steps: 34737
micro_batch_size: 4
mlflow_experiment_name: /tmp/623d9787d7fd7d0e_train_data.json
model_type: AutoModelForCausalLM
num_epochs: 10
optimizer: adamw_bnb_8bit
output_dir: miner_id_24
pad_to_sequence_len: true
resume_from_checkpoint: null
s2_attention: null
sample_packing: false
save_steps: 100
sequence_len: 2048
special_tokens:
  pad_token: </s>
strict: false
tf32: true
tokenizer_type: AutoTokenizer
train_on_inputs: false
trust_remote_code: true
val_set_size: 0.04
wandb_entity: null
wandb_mode: online
wandb_name: af6de128-8865-42e8-800b-ff2d2b1acccd
wandb_project: Gradients-On-Demand
wandb_run: your_name
wandb_runid: af6de128-8865-42e8-800b-ff2d2b1acccd
warmup_steps: 10
weight_decay: 0.0
xformers_attention: null

ffcfbd55-9faa-4516-b0c4-a4012948b3ab

This model is a fine-tuned version of JackFram/llama-68m on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.0142

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: 4
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 8
• total_train_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: 10
• training_steps: 9750

Training results

Training Loss	Epoch	Step	Validation Loss
2.344	0.0010	1	2.4328
0.0466	0.1026	100	0.0487
0.019	0.2051	200	0.0343
0.0141	0.3077	300	0.0274
0.0057	0.4103	400	0.0277
0.009	0.5128	500	0.0213
0.0014	0.6154	600	0.0186
0.0021	0.7179	700	0.0183
0.0019	0.8205	800	0.0179
0.0059	0.9231	900	0.0169
0.0029	1.0256	1000	0.0160
0.017	1.1282	1100	0.0167
0.0233	1.2308	1200	0.0135
0.0014	1.3333	1300	0.0154
0.185	1.4359	1400	0.0142

Framework versions

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