See axolotl config

axolotl version: 0.4.1

adapter: lora
base_model: unsloth/Llama-3.2-3B-Instruct
bf16: auto
chat_template: llama3
dataset_prepared_path: null
datasets:
- data_files:
  - alpaca-cleaned_train_data.json
  ds_type: json
  path: /workspace/input_data/alpaca-cleaned_train_data.json
  type:
    field_input: input
    field_instruction: instruction
    field_output: output
    system_format: '{system}'
    system_prompt: ''
debug: null
deepspeed: null
early_stopping_patience: null
eval_max_new_tokens: 128
eval_table_size: null
evals_per_epoch: 2
flash_attention: true
fp16: null
fsdp: null
fsdp_config: null
gradient_accumulation_steps: 8
gradient_checkpointing: true
group_by_length: false
hub_model_id: masatochi/tuning-cf8d7e96-9f53-41c1-bbef-58adbf0ac75f
hub_strategy: checkpoint
hub_token: null
learning_rate: 0.0002
load_in_4bit: false
load_in_8bit: true
local_rank: null
logging_steps: 1
lora_alpha: 16
lora_dropout: 0.06
lora_fan_in_fan_out: null
lora_model_dir: null
lora_r: 8
lora_target_linear: true
lr_scheduler: cosine
max_steps: 200
micro_batch_size: 3
mlflow_experiment_name: /tmp/alpaca-cleaned_train_data.json
model_type: LlamaForCausalLM
num_epochs: 3
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: 5
save_strategy: steps
sequence_len: 4096
strict: false
tf32: false
tokenizer_type: AutoTokenizer
train_on_inputs: false
val_set_size: 0.05
wandb_entity: lkotbimehdi
wandb_mode: online
wandb_project: lko
wandb_run: miner_id_24
wandb_runid: cf8d7e96-9f53-41c1-bbef-58adbf0ac75f
warmup_steps: 30
weight_decay: 0.0
xformers_attention: null

tuning-cf8d7e96-9f53-41c1-bbef-58adbf0ac75f

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: 1.0825

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: 3
• eval_batch_size: 3
• seed: 42
• gradient_accumulation_steps: 8
• total_train_batch_size: 24
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 30
• training_steps: 200

Training results

Training Loss	Epoch	Step	Validation Loss
1.3759	0.0005	1	1.4346
1.2609	0.0166	34	1.1323
1.1674	0.0333	68	1.0982
1.0521	0.0499	102	1.0889
1.1228	0.0665	136	1.0842
0.9368	0.0831	170	1.0825

Framework versions

• PEFT 0.13.2
• Transformers 4.45.2
• Pytorch 2.4.1+cu124
• Datasets 3.0.1
• Tokenizers 0.20.1