See axolotl config

axolotl version: 0.4.1

adapter: lora
auto_find_batch_size: true
base_model: princeton-nlp/gemma-2-9b-it-SimPO
bf16: auto
chat_template: llama3
dataset_prepared_path: null
datasets:
- data_files:
  - 9d8623bfae84f70c_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/9d8623bfae84f70c_train_data.json
  type:
    field_instruction: am
    field_output: en
    format: '{instruction}'
    no_input_format: '{instruction}'
    system_format: '{system}'
    system_prompt: ''
debug: null
deepspeed: null
early_stopping_patience: 3
eval_max_new_tokens: 128
eval_steps: 50
eval_table_size: null
flash_attention: true
fp16: null
fsdp: null
fsdp_config: null
gradient_accumulation_steps: 8
gradient_checkpointing: false
group_by_length: true
hub_model_id: tuantmdev/73844773-e275-4b01-ab0e-ffb7451f7acf
hub_repo: null
hub_strategy: checkpoint
hub_token: null
learning_rate: 1e-4
load_in_4bit: false
load_in_8bit: false
local_rank: null
logging_steps: 40
lora_alpha: 64
lora_dropout: 0.05
lora_fan_in_fan_out: null
lora_model_dir: null
lora_r: 32
lora_target_linear: true
lr_scheduler: cosine
max_grad_norm: 1.0
max_steps: 400
micro_batch_size: 2
mlflow_experiment_name: /tmp/9d8623bfae84f70c_train_data.json
model_type: AutoModelForCausalLM
num_epochs: 1
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: 50
save_strategy: steps
sequence_len: 512
strict: false
tf32: false
tokenizer_type: AutoTokenizer
train_on_inputs: false
trust_remote_code: true
val_set_size: 0.05
wandb_entity: null
wandb_mode: online
wandb_name: 4bf73c1d-dfe2-48f5-b525-0bfa7c993024
wandb_project: Gradients-On-Demand
wandb_run: unknown
wandb_runid: 4bf73c1d-dfe2-48f5-b525-0bfa7c993024
warmup_steps: 80
weight_decay: 0.0
xformers_attention: null

73844773-e275-4b01-ab0e-ffb7451f7acf

This model is a fine-tuned version of princeton-nlp/gemma-2-9b-it-SimPO on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.9382

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: 0.0001
• train_batch_size: 2
• eval_batch_size: 2
• seed: 42
• gradient_accumulation_steps: 8
• total_train_batch_size: 16
• 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: 80
• training_steps: 400

Training results

Training Loss	Epoch	Step	Validation Loss
No log	0.0018	1	4.0493
2.083	0.0886	50	1.4012
1.262	0.1773	100	1.2690
1.1674	0.2659	150	1.1629
1.0414	0.3545	200	1.0874
0.9941	0.4432	250	1.0058
0.9774	0.5318	300	0.9604
0.9046	0.6204	350	0.9389
0.9285	0.7091	400	0.9382

Framework versions

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