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:
  - 629ba27a4fc9a771_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/629ba27a4fc9a771_train_data.json
  type:
    field_instruction: init_prompt
    field_output: init_response
    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/38d4e54c-cfe2-48bb-8365-313dfb7b276e
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: 1980
micro_batch_size: 4
mlflow_experiment_name: /tmp/629ba27a4fc9a771_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: 0.04
wandb_entity: null
wandb_mode: online
wandb_name: 39c85264-2e28-4ba8-ba2a-9c48fbcf4c19
wandb_project: Gradients-On-Demand
wandb_run: your_name
wandb_runid: 39c85264-2e28-4ba8-ba2a-9c48fbcf4c19
warmup_steps: 10
weight_decay: 0.0
xformers_attention: null

38d4e54c-cfe2-48bb-8365-313dfb7b276e

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: 0.3069

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: 1980

Training results

Training Loss	Epoch	Step	Validation Loss
0.9532	0.0008	1	1.2198
0.5792	0.0757	100	0.5988
0.5419	0.1513	200	0.5516
0.5068	0.2270	300	0.5198
0.5463	0.3027	400	0.4919
0.5244	0.3784	500	0.4679
0.3877	0.4540	600	0.4452
0.4547	0.5297	700	0.4257
0.4216	0.6054	800	0.4028
0.4578	0.6810	900	0.3898
0.3696	0.7567	1000	0.3700
0.4146	0.8324	1100	0.3560
0.3208	0.9081	1200	0.3429
0.3825	0.9837	1300	0.3315
0.2358	1.0594	1400	0.3249
0.2111	1.1351	1500	0.3181
0.1772	1.2107	1600	0.3129
0.2073	1.2864	1700	0.3100
0.2569	1.3621	1800	0.3078
0.2343	1.4378	1900	0.3069

Framework versions

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