See axolotl config

axolotl version: 0.4.1

adapter: lora
base_model: unsloth/Qwen2.5-0.5B
bf16: true
chat_template: llama3
dataset_prepared_path: null
datasets:
- data_files:
  - ea66d0c68c983721_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/ea66d0c68c983721_train_data.json
  type:
    field_instruction: sentence1
    field_output: sentence2
    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: true
gradient_accumulation_steps: 8
gradient_checkpointing: true
group_by_length: false
hub_model_id: Alphatao/fe8d8eb3-c8bf-41c5-806b-37f7a662e395
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: 7680
micro_batch_size: 4
mlflow_experiment_name: /tmp/ea66d0c68c983721_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: 0.05
wandb_entity: null
wandb_mode: online
wandb_name: a6f83d3c-a5d6-4fb0-984f-06ddde5e8f44
wandb_project: Gradients-On-Demand
wandb_run: your_name
wandb_runid: a6f83d3c-a5d6-4fb0-984f-06ddde5e8f44
warmup_steps: 10
weight_decay: 0.0
xformers_attention: null

fe8d8eb3-c8bf-41c5-806b-37f7a662e395

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

• Loss: 1.3488

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: 5212

Training results

Training Loss	Epoch	Step	Validation Loss
4.8428	0.0004	1	4.6827
1.7175	0.0384	100	1.7035
1.4774	0.0767	200	1.6374
1.5973	0.1151	300	1.6064
1.4442	0.1535	400	1.5828
1.7331	0.1919	500	1.5556
1.4103	0.2302	600	1.5370
1.1782	0.2686	700	1.5209
1.4832	0.3070	800	1.5039
1.5766	0.3454	900	1.4915
1.3976	0.3837	1000	1.4837
1.5218	0.4221	1100	1.4667
1.5943	0.4605	1200	1.4576
1.3976	0.4989	1300	1.4535
1.3621	0.5372	1400	1.4401
1.5817	0.5756	1500	1.4286
1.0866	0.6140	1600	1.4200
1.307	0.6524	1700	1.4056
1.3197	0.6907	1800	1.4006
1.4547	0.7291	1900	1.3875
1.3411	0.7675	2000	1.3820
1.3654	0.8059	2100	1.3750
1.5037	0.8442	2200	1.3656
1.3596	0.8826	2300	1.3536
1.2466	0.9210	2400	1.3483
1.1555	0.9594	2500	1.3401
1.6812	0.9977	2600	1.3363
1.0617	1.0363	2700	1.3616
1.1065	1.0746	2800	1.3488

Framework versions

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