See axolotl config

axolotl version: 0.4.1

adapter: lora
base_model: unsloth/gemma-2b-it
bf16: true
chat_template: llama3
dataset_prepared_path: null
datasets:
- data_files:
  - 52b51d4a8d8dae82_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/52b51d4a8d8dae82_train_data.json
  type:
    field_instruction: instruction
    field_output: output
    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: 400
eval_table_size: null
flash_attention: false
gradient_accumulation_steps: 4
gradient_checkpointing: true
group_by_length: false
hub_model_id: Alphatao/9d65626e-53aa-48e7-9c43-1cd15808de5a
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
lr_scheduler: cosine
max_grad_norm: 1.0
max_steps: 11867
micro_batch_size: 2
mlflow_experiment_name: /tmp/52b51d4a8d8dae82_train_data.json
model_type: AutoModelForCausalLM
num_epochs: 10
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: 1024
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: ff680a59-ee3c-407a-8506-d9507d1c8bd0
wandb_project: Gradients-On-Demand
wandb_run: your_name
wandb_runid: ff680a59-ee3c-407a-8506-d9507d1c8bd0
warmup_steps: 10
weight_decay: 0.0
xformers_attention: null

9d65626e-53aa-48e7-9c43-1cd15808de5a

This model is a fine-tuned version of unsloth/gemma-2b-it on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.7440

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: 11867

Training results

Training Loss	Epoch	Step	Validation Loss
1.6257	0.0001	1	1.3826
0.823	0.0341	400	0.8594
0.8278	0.0682	800	0.8445
0.9407	0.1023	1200	0.8366
0.8239	0.1364	1600	0.8304
1.0677	0.1705	2000	0.8265
0.8125	0.2046	2400	0.8186
0.9515	0.2387	2800	0.8148
0.7069	0.2727	3200	0.8099
0.7106	0.3068	3600	0.8049
0.9589	0.3409	4000	0.8002
0.9544	0.3750	4400	0.7965
0.833	0.4091	4800	0.7919
0.8829	0.4432	5200	0.7867
1.0338	0.4773	5600	0.7828
0.7217	0.5114	6000	0.7784
0.7006	0.5455	6400	0.7739
0.8399	0.5796	6800	0.7699
0.5234	0.6137	7200	0.7656
0.6645	0.6478	7600	0.7614
0.8158	0.6819	8000	0.7582
0.5981	0.7160	8400	0.7548
0.6689	0.7501	8800	0.7522
0.854	0.7841	9200	0.7498
0.5398	0.8182	9600	0.7478
0.7944	0.8523	10000	0.7464
1.061	0.8864	10400	0.7452
0.7511	0.9205	10800	0.7445
0.6627	0.9546	11200	0.7441
0.8566	0.9887	11600	0.7440

Framework versions

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