See axolotl config

axolotl version: 0.4.1

adapter: lora
base_model: unsloth/Llama-3.2-3B-Instruct
bf16: true
chat_template: llama3
dataset_prepared_path: null
datasets:
- data_files:
  - 27f62aebddd38f97_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/27f62aebddd38f97_train_data.json
  type:
    field_instruction: instruction
    field_output: response
    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/6c2db466-5702-44b2-bc64-b973bd68bd8f
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: 1610
micro_batch_size: 4
mlflow_experiment_name: /tmp/27f62aebddd38f97_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: 71ab157c-b115-4e23-b527-3c35585c2108
wandb_project: Gradients-On-Demand
wandb_run: your_name
wandb_runid: 71ab157c-b115-4e23-b527-3c35585c2108
warmup_steps: 10
weight_decay: 0.0
xformers_attention: null

6c2db466-5702-44b2-bc64-b973bd68bd8f

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

• Loss: 0.5475

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: 1548

Training results

Training Loss	Epoch	Step	Validation Loss
1.614	0.0013	1	1.5537
0.6333	0.1292	100	0.6885
0.5737	0.2584	200	0.6389
0.6114	0.3876	300	0.6158
0.5759	0.5168	400	0.6013
0.5919	0.6460	500	0.5878
0.6375	0.7752	600	0.5805
0.539	0.9044	700	0.5745
0.5183	1.0336	800	0.5666
0.5742	1.1628	900	0.5643
0.4912	1.2920	1000	0.5590
0.5111	1.4212	1100	0.5551
0.5255	1.5504	1200	0.5519
0.5118	1.6796	1300	0.5494
0.4863	1.8088	1400	0.5480
0.5158	1.9380	1500	0.5475

Framework versions

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