See axolotl config

axolotl version: 0.4.1

adapter: lora
base_model: Qwen/Qwen2.5-3B-Instruct
bf16: true
chat_template: llama3
dataset_prepared_path: null
datasets:
- data_files:
  - 4a9dc8a325b4a34b_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/4a9dc8a325b4a34b_train_data.json
  type:
    field_input: generated
    field_instruction: question
    field_output: answer
    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: 100
eval_table_size: null
flash_attention: true
gradient_accumulation_steps: 8
gradient_checkpointing: true
group_by_length: false
hub_model_id: Alphatao/ccfb1c8b-e93f-4f05-8188-b9f701a7c556
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: 2346
micro_batch_size: 4
mlflow_experiment_name: /tmp/4a9dc8a325b4a34b_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: 0.026075891273963744
wandb_entity: null
wandb_mode: online
wandb_name: e28ecee4-19d0-4603-a9b1-91b0efec87c4
wandb_project: Gradients-On-Demand
wandb_run: your_name
wandb_runid: e28ecee4-19d0-4603-a9b1-91b0efec87c4
warmup_steps: 10
weight_decay: 0.0
xformers_attention: null

ccfb1c8b-e93f-4f05-8188-b9f701a7c556

This model is a fine-tuned version of Qwen/Qwen2.5-3B-Instruct on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.3436

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: 2346

Training results

Training Loss	Epoch	Step	Validation Loss
0.7371	0.0002	1	0.7489
0.4927	0.0171	100	0.4624
0.4566	0.0343	200	0.4356
0.4275	0.0514	300	0.4188
0.4006	0.0685	400	0.4081
0.4142	0.0857	500	0.3993
0.4318	0.1028	600	0.3930
0.3706	0.1199	700	0.3847
0.3665	0.1371	800	0.3795
0.3891	0.1542	900	0.3743
0.3609	0.1714	1000	0.3695
0.3596	0.1885	1100	0.3655
0.391	0.2056	1200	0.3621
0.3693	0.2228	1300	0.3588
0.3525	0.2399	1400	0.3558
0.3517	0.2570	1500	0.3526
0.3139	0.2742	1600	0.3505
0.3573	0.2913	1700	0.3483
0.3452	0.3084	1800	0.3468
0.3337	0.3256	1900	0.3455
0.3444	0.3427	2000	0.3446
0.3569	0.3598	2100	0.3440
0.3404	0.3770	2200	0.3437
0.3481	0.3941	2300	0.3436

Framework versions

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