See axolotl config

axolotl version: 0.4.1

adapter: lora
base_model: unsloth/gemma-2-2b
bf16: true
chat_template: llama3
data_processes: 56
dataset_prepared_path: null
datasets:
- data_files:
  - ad15d34493388077_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/ad15d34493388077_train_data.json
  type:
    field_input: document_description
    field_instruction: document_type
    field_output: generated_text
    format: '{instruction} {input}'
    no_input_format: '{instruction}'
    system_format: '{system}'
    system_prompt: ''
debug: null
deepspeed: null
device_map: auto
distributed_training:
  multi_gpu: true
  num_gpus: 2
do_eval: true
early_stopping_patience: 4
eval_batch_size: 8
eval_max_new_tokens: 128
eval_steps: 150
eval_table_size: null
evals_per_epoch: null
flash_attention: true
fp16: false
fsdp:
- full_shard
- cpu_offload
fsdp_config:
  cpu_offload: true
  mixed_precision: true
  sharding_strategy: FULL_SHARD
gradient_accumulation_steps: 4
gradient_checkpointing: false
group_by_length: true
hub_model_id: cimol/ec60756e-f241-4bf6-b4fc-27db961e8a33
hub_repo: null
hub_strategy: checkpoint
hub_token: null
learning_rate: 8.0e-05
load_in_4bit: false
load_in_8bit: false
local_rank: null
logging_steps: 1
lora_alpha: 128
lora_dropout: 0.04
lora_fan_in_fan_out: null
lora_model_dir: null
lora_r: 64
lora_target_linear: true
lr_scheduler: cosine
lr_scheduler_warmup_steps: 50
max_grad_norm: 1.0
max_steps: 2850
micro_batch_size: 4
mlflow_experiment_name: /tmp/ad15d34493388077_train_data.json
model_type: AutoModelForCausalLM
num_epochs: 3
optim_args:
  adam_beta1: 0.9
  adam_beta2: 0.95
  adam_epsilon: 1e-8
optimizer: adamw_torch
output_dir: miner_id_24
pad_to_sequence_len: true
resume_from_checkpoint: null
s2_attention: null
sample_packing: false
save_steps: 150
saves_per_epoch: null
seed: 17333
sequence_len: 1024
strict: false
tf32: true
tokenizer_type: AutoTokenizer
total_train_batch_size: 16
train_batch_size: 16
train_on_inputs: false
trust_remote_code: true
val_set_size: 0.05
wandb_entity: null
wandb_mode: online
wandb_name: c14c8534-b36a-469b-aa5d-bef0fc990a0b
wandb_project: Gradients-On-Demand
wandb_run: your_name
wandb_runid: c14c8534-b36a-469b-aa5d-bef0fc990a0b
warmup_steps: 10
weight_decay: 0.0
xformers_attention: null

ec60756e-f241-4bf6-b4fc-27db961e8a33

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

• Loss: 0.8035

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: 8e-05
• train_batch_size: 4
• eval_batch_size: 8
• seed: 17333
• distributed_type: multi-GPU
• num_devices: 2
• gradient_accumulation_steps: 4
• total_train_batch_size: 32
• total_eval_batch_size: 16
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=adam_beta1=0.9,adam_beta2=0.95,adam_epsilon=1e-8
• lr_scheduler_type: cosine
• lr_scheduler_warmup_steps: 10
• training_steps: 2850

Training results

Training Loss	Epoch	Step	Validation Loss
0.7784	0.0006	1	1.4497
1.7915	0.0906	150	1.0063
1.609	0.1813	300	0.9535
1.5236	0.2719	450	0.9209
1.5994	0.3625	600	0.8973
1.4063	0.4532	750	0.8783
1.319	0.5438	900	0.8662
1.6625	0.6344	1050	0.8540
1.722	0.7251	1200	0.8401
1.7885	0.8157	1350	0.8311
1.6219	0.9063	1500	0.8221
1.4787	0.9970	1650	0.8151
0.7906	1.0876	1800	0.8175
0.818	1.1782	1950	0.8140
0.8121	1.2689	2100	0.8089
0.8442	1.3595	2250	0.8085
0.7529	1.4502	2400	0.8056
0.8249	1.5408	2550	0.8048
0.8182	1.6314	2700	0.8035
0.7966	1.7221	2850	0.8035

Framework versions

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