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:
  - 56a99d4e5742b975_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/56a99d4e5742b975_train_data.json
  type:
    field_instruction: content
    field_output: title
    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: false
gradient_accumulation_steps: 8
gradient_checkpointing: true
group_by_length: false
hub_model_id: Alphatao/2f0be379-fb97-4fce-a8be-0e829c2e6278
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: 4471
micro_batch_size: 4
mlflow_experiment_name: /tmp/56a99d4e5742b975_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: 1024
strict: false
tf32: true
tokenizer_type: AutoTokenizer
train_on_inputs: false
trust_remote_code: true
val_set_size: 0.04773725415314111
wandb_entity: null
wandb_mode: online
wandb_name: ef11de7a-c9d5-45ac-b199-e622ae0b7dd3
wandb_project: Gradients-On-Demand
wandb_run: your_name
wandb_runid: ef11de7a-c9d5-45ac-b199-e622ae0b7dd3
warmup_steps: 10
weight_decay: 0.0
xformers_attention: null

2f0be379-fb97-4fce-a8be-0e829c2e6278

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: 0.7838

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: 4471

Training results

Training Loss	Epoch	Step	Validation Loss
4.0788	0.0003	1	3.8597
1.0411	0.0321	100	1.2971
1.0903	0.0642	200	1.2076
1.1281	0.0963	300	1.1715
0.8587	0.1283	400	1.1387
0.795	0.1604	500	1.1175
0.6778	0.1925	600	1.0925
0.7373	0.2246	700	1.0729
1.4486	0.2567	800	1.0459
1.1051	0.2888	900	1.0423
1.0145	0.3208	1000	1.0221
0.7398	0.3529	1100	1.0096
0.9354	0.3850	1200	0.9912
0.8828	0.4171	1300	0.9728
0.5784	0.4492	1400	0.9686
1.0613	0.4813	1500	0.9528
0.6899	0.5133	1600	0.9332
0.8255	0.5454	1700	0.9241
0.6854	0.5775	1800	0.9165
0.6808	0.6096	1900	0.8989
0.8832	0.6417	2000	0.8865
0.7159	0.6738	2100	0.8713
0.8375	0.7058	2200	0.8543
0.6961	0.7379	2300	0.8495
1.0821	0.7700	2400	0.8345
0.9692	0.8021	2500	0.8209
0.8603	0.8342	2600	0.8078
0.7592	0.8663	2700	0.7916
0.4661	0.8983	2800	0.7833
1.145	0.9304	2900	0.7754
0.7121	0.9625	3000	0.7662
0.7462	0.9946	3100	0.7582
0.5419	1.0268	3200	0.7744
0.389	1.0589	3300	0.7838

Framework versions

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