See axolotl config

axolotl version: 0.4.1

adapter: lora
base_model: EleutherAI/gpt-neo-125m
bf16: true
chat_template: llama3
dataset_prepared_path: null
datasets:
- data_files:
  - aa838734bbefd9c8_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/aa838734bbefd9c8_train_data.json
  type:
    field_instruction: prompt
    field_output: GEITje-7B-ultra
    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: false
gradient_accumulation_steps: 8
gradient_checkpointing: true
group_by_length: false
hub_model_id: Alphatao/4ce2ceca-17a8-49c5-83c2-35618600ca7f
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: 32
lora_dropout: 0.05
lora_fan_in_fan_out: null
lora_model_dir: null
lora_r: 16
lora_target_linear: true
lora_target_modules: null
lr_scheduler: cosine
max_grad_norm: 1.0
max_steps: 4140
micro_batch_size: 4
mlflow_experiment_name: /tmp/aa838734bbefd9c8_train_data.json
model_type: AutoModelForCausalLM
num_epochs: 2
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: 1024
special_tokens:
  pad_token: <|endoftext|>
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: 0130bc26-f4a9-49f9-b382-a5b971eeaf04
wandb_project: Gradients-On-Demand
wandb_run: your_name
wandb_runid: 0130bc26-f4a9-49f9-b382-a5b971eeaf04
warmup_steps: 10
weight_decay: 0.0
xformers_attention: null

4ce2ceca-17a8-49c5-83c2-35618600ca7f

This model is a fine-tuned version of EleutherAI/gpt-neo-125m on the None dataset. It achieves the following results on the evaluation set:

• Loss: 2.2814

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: 3002

Training results

Training Loss	Epoch	Step	Validation Loss
22.3516	0.0007	1	2.7369
20.2093	0.0666	100	2.5747
19.7446	0.1333	200	2.5223
20.6209	0.1999	300	2.4892
20.299	0.2666	400	2.4635
19.2732	0.3332	500	2.4406
19.7515	0.3998	600	2.4214
18.9969	0.4665	700	2.4035
19.396	0.5331	800	2.3888
18.8607	0.5998	900	2.3739
18.908	0.6664	1000	2.3621
18.5906	0.7330	1100	2.3504
18.9578	0.7997	1200	2.3409
19.0391	0.8663	1300	2.3319
18.6685	0.9329	1400	2.3235
15.9064	0.9996	1500	2.3170
16.0333	1.0664	1600	2.3104
17.8283	1.1330	1700	2.3052
18.3232	1.1997	1800	2.3006
16.2907	1.2663	1900	2.2966
20.2129	1.3329	2000	2.2936
17.9803	1.3996	2100	2.2905
17.2894	1.4662	2200	2.2879
18.2833	1.5329	2300	2.2861
16.9763	1.5995	2400	2.2844
18.0936	1.6661	2500	2.2834
17.6034	1.7328	2600	2.2823
17.8223	1.7994	2700	2.2819
18.5859	1.8661	2800	2.2816
17.2502	1.9327	2900	2.2815
17.3237	1.9993	3000	2.2814

Framework versions

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