See axolotl config

axolotl version: 0.4.1

adapter: lora
base_model: unsloth/Llama-3.2-3B
bf16: true
chat_template: llama3
dataset_prepared_path: null
datasets:
- data_files:
  - 9b4cab2992cdb07f_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/9b4cab2992cdb07f_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: 400
eval_table_size: null
flash_attention: true
gradient_accumulation_steps: 4
gradient_checkpointing: true
group_by_length: false
hub_model_id: Alphatao/a3fbf6bf-297b-4b51-af46-de552ea133f3
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:
- q_proj
- k_proj
- v_proj
- o_proj
- down_proj
- up_proj
lr_scheduler: cosine
max_grad_norm: 1.0
max_steps: 7351
micro_batch_size: 2
mlflow_experiment_name: /tmp/9b4cab2992cdb07f_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: 400
sequence_len: 2048
strict: false
tf32: true
tokenizer_type: AutoTokenizer
train_on_inputs: false
trust_remote_code: true
val_set_size: 0.033916240452578315
wandb_entity: null
wandb_mode: online
wandb_name: 30612698-6e4a-41b9-a416-94b6b03904c8
wandb_project: Gradients-On-Demand
wandb_run: your_name
wandb_runid: 30612698-6e4a-41b9-a416-94b6b03904c8
warmup_steps: 10
weight_decay: 0.0
xformers_attention: null

a3fbf6bf-297b-4b51-af46-de552ea133f3

This model is a fine-tuned version of unsloth/Llama-3.2-3B on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.7379

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: 2
• eval_batch_size: 2
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 8
• 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: 7351

Training results

Training Loss	Epoch	Step	Validation Loss
1.3563	0.0001	1	2.0488
0.7624	0.0225	400	0.8732
0.975	0.0449	800	0.8350
0.4809	0.0674	1200	0.8289
0.7347	0.0899	1600	0.8194
0.9339	0.1123	2000	0.8096
1.1196	0.1348	2400	0.8037
1.5178	0.1573	2800	0.7933
1.0605	0.1797	3200	0.7887
0.8303	0.2022	3600	0.7788
0.6516	0.2247	4000	0.7725
0.8837	0.2472	4400	0.7635
1.2204	0.2696	4800	0.7572
1.1298	0.2921	5200	0.7494
0.9194	0.3146	5600	0.7461
0.9396	0.3370	6000	0.7426
1.0739	0.3595	6400	0.7393
0.7616	0.3820	6800	0.7378
1.2642	0.4044	7200	0.7379

Framework versions

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