See axolotl config

axolotl version: 0.4.1

adapter: lora
base_model: unsloth/Llama-3.2-1B
bf16: true
chat_template: llama3
dataset_prepared_path: null
datasets:
- data_files:
  - 1ba279f183337309_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/1ba279f183337309_train_data.json
  type:
    field_input: input
    field_instruction: instruction
    field_output: output
    format: '{instruction} {input}'
    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/c131672e-e94c-4ddc-a4e7-8b9339d6dd97
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: 2778
micro_batch_size: 4
mlflow_experiment_name: /tmp/1ba279f183337309_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: 2048
strict: false
tf32: true
tokenizer_type: AutoTokenizer
train_on_inputs: false
trust_remote_code: true
val_set_size: 0.05
wandb_entity: null
wandb_mode: online
wandb_name: 42ec8344-4e7f-44ae-abba-aec8960a0c35
wandb_project: Gradients-On-Demand
wandb_run: your_name
wandb_runid: 42ec8344-4e7f-44ae-abba-aec8960a0c35
warmup_steps: 10
weight_decay: 0.0
xformers_attention: null

c131672e-e94c-4ddc-a4e7-8b9339d6dd97

This model is a fine-tuned version of unsloth/Llama-3.2-1B on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.8596

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: 2778

Training results

Training Loss	Epoch	Step	Validation Loss
2.0421	0.0005	1	2.2055
1.3315	0.0501	100	1.0804
1.1595	0.1002	200	1.0350
1.1229	0.1503	300	1.0188
1.115	0.2004	400	0.9978
1.1094	0.2505	500	0.9956
0.9105	0.3006	600	0.9812
1.1083	0.3507	700	0.9692
0.9175	0.4009	800	0.9570
1.1945	0.4510	900	0.9546
1.0475	0.5011	1000	0.9424
0.9024	0.5512	1100	0.9365
1.1053	0.6013	1200	0.9273
1.0829	0.6514	1300	0.9192
1.0007	0.7015	1400	0.9078
1.0743	0.7516	1500	0.9008
1.0054	0.8017	1600	0.8941
0.9514	0.8518	1700	0.8860
1.0254	0.9019	1800	0.8799
0.9989	0.9520	1900	0.8739
0.8056	1.0022	2000	0.8687
0.7755	1.0523	2100	0.8702
0.7177	1.1024	2200	0.8662
1.0004	1.1525	2300	0.8649
0.9104	1.2026	2400	0.8632
0.7172	1.2527	2500	0.8611
0.7679	1.3028	2600	0.8598
0.6684	1.3529	2700	0.8596

Framework versions

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