See axolotl config

axolotl version: 0.4.1

adapter: lora
base_model: Qwen/Qwen2.5-1.5B-Instruct
bf16: true
chat_template: llama3
dataset_prepared_path: null
datasets:
- data_files:
  - 376665f92f9f4ce1_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/376665f92f9f4ce1_train_data.json
  type:
    field_instruction: query
    field_output: answers
    format: '{instruction}'
    no_input_format: '{instruction}'
    system_format: '{system}'
    system_prompt: ''
debug: null
device_map:
  ? ''
  : 0,1,2,4
early_stopping_patience: 2
eval_max_new_tokens: 128
eval_steps: 33
eval_table_size: null
flash_attention: true
gradient_accumulation_steps: 8
gradient_checkpointing: true
group_by_length: false
hub_model_id: Alphatao/e7334d33-b475-420e-bcfa-5698ede9d5ad
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: 128
lora_dropout: 0.3
lora_fan_in_fan_out: null
lora_model_dir: null
lora_r: 64
lora_target_linear: true
lr_scheduler: cosine
max_grad_norm: 1.0
max_steps: 660.0
micro_batch_size: 4
mlflow_experiment_name: /tmp/376665f92f9f4ce1_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: 33
sequence_len: 1024
strict: false
tf32: true
tokenizer_type: AutoTokenizer
train_on_inputs: false
trust_remote_code: true
val_set_size: 0.04
wandb_entity: null
wandb_mode: online
wandb_name: d55156ab-bed0-411c-9fe5-7615dc686893
wandb_project: Gradients-On-Demand
wandb_run: your_name
wandb_runid: d55156ab-bed0-411c-9fe5-7615dc686893
warmup_steps: 10
weight_decay: 0.0
xformers_attention: null

e7334d33-b475-420e-bcfa-5698ede9d5ad

This model is a fine-tuned version of Qwen/Qwen2.5-1.5B-Instruct on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.1819

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
• distributed_type: multi-GPU
• num_devices: 4
• gradient_accumulation_steps: 8
• total_train_batch_size: 128
• total_eval_batch_size: 16
• 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: 660

Training results

Training Loss	Epoch	Step	Validation Loss
1.5377	0.0027	1	1.6769
0.2408	0.0905	33	0.2099
0.2016	0.1810	66	0.2014
0.2003	0.2715	99	0.1945
0.1945	0.3620	132	0.1928
0.1899	0.4525	165	0.1902
0.2069	0.5430	198	0.1875
0.1904	0.6335	231	0.1864
0.1828	0.7240	264	0.1844
0.1861	0.8145	297	0.1837
0.1527	0.9050	330	0.1814
0.1649	0.9955	363	0.1802
0.1623	1.0874	396	0.1822
0.1421	1.1779	429	0.1819

Framework versions

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