See axolotl config

axolotl version: 0.4.1

adapter: lora
base_model: unsloth/Qwen2.5-0.5B-Instruct
bf16: true
chat_template: llama3
dataset_prepared_path: null
datasets:
- data_files:
  - baefaea883679862_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/baefaea883679862_train_data.json
  type:
    field_input: input
    field_instruction: instruction
    field_output: output
    format: '{instruction} {input}'
    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: true
gradient_accumulation_steps: 4
gradient_checkpointing: true
group_by_length: false
hub_model_id: Alphatao/4bb1fb3b-9da9-4331-b85b-36f0c5175da8
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: 29912
micro_batch_size: 2
mlflow_experiment_name: /tmp/baefaea883679862_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: 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: 576aa853-9400-49a2-9b33-d7849e6e83c6
wandb_project: Gradients-On-Demand
wandb_run: your_name
wandb_runid: 576aa853-9400-49a2-9b33-d7849e6e83c6
warmup_steps: 10
weight_decay: 0.0
xformers_attention: null

4bb1fb3b-9da9-4331-b85b-36f0c5175da8

This model is a fine-tuned version of unsloth/Qwen2.5-0.5B-Instruct on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.9987

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: 15966

Training results

Training Loss	Epoch	Step	Validation Loss
2.3951	0.0001	1	2.4080
1.3849	0.0501	400	1.1384
1.3071	0.1002	800	1.1044
1.1328	0.1503	1200	1.0976
1.2365	0.2004	1600	1.0958
1.4147	0.2505	2000	1.0896
1.0192	0.3006	2400	1.0820
1.1322	0.3508	2800	1.0800
1.1178	0.4009	3200	1.0719
1.6549	0.4510	3600	1.0713
1.1885	0.5011	4000	1.0651
1.0638	0.5512	4400	1.0648
1.4894	0.6013	4800	1.0620
1.5195	0.6514	5200	1.0695
0.9605	0.7015	5600	1.0542
1.1562	0.7516	6000	1.0432
0.6789	0.8017	6400	1.0450
1.0593	0.8518	6800	1.0394
1.3117	0.9019	7200	1.0416
1.2763	0.9521	7600	1.0348
0.8916	1.0022	8000	1.0313
0.5844	1.0523	8400	1.0307
1.0794	1.1024	8800	1.0265
1.5881	1.1525	9200	1.0250
1.0954	1.2027	9600	1.0237
1.1153	1.2528	10000	1.0204
0.6512	1.3029	10400	1.0165
0.498	1.3530	10800	1.0149
1.048	1.4031	11200	1.0130
0.9372	1.4532	11600	1.0102
0.9142	1.5033	12000	1.0067
0.8444	1.5534	12400	1.0064
1.1215	1.6035	12800	1.0047
1.8016	1.6536	13200	1.0028
0.9739	1.7037	13600	1.0016
0.8278	1.7538	14000	1.0006
1.1798	1.8040	14400	0.9996
0.9944	1.8541	14800	0.9992
0.8047	1.9042	15200	0.9988
0.7495	1.9543	15600	0.9987

Framework versions

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