See axolotl config

axolotl version: 0.4.1

adapter: lora
base_model: Qwen/Qwen2-0.5B-Instruct
bf16: true
chat_template: llama3
dataset_prepared_path: null
datasets:
- data_files:
  - a23a73113e5bcd49_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/a23a73113e5bcd49_train_data.json
  type:
    field_instruction: title
    field_output: content
    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/8783abb4-4863-4d35-96d0-38591b2d8490
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: 0.1
lora_fan_in_fan_out: null
lora_model_dir: null
lora_r: 64
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: 5376
micro_batch_size: 4
mlflow_experiment_name: /tmp/a23a73113e5bcd49_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: 0.04787437763309077
wandb_entity: null
wandb_mode: online
wandb_name: 032a2798-2e45-4672-ba51-e28a8f853060
wandb_project: Gradients-On-Demand
wandb_run: your_name
wandb_runid: 032a2798-2e45-4672-ba51-e28a8f853060
warmup_steps: 10
weight_decay: 0.0
xformers_attention: null

8783abb4-4863-4d35-96d0-38591b2d8490

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

• Loss: 1.7221

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: 5376

Training results

Training Loss	Epoch	Step	Validation Loss
2.7661	0.0003	1	2.7781
2.4133	0.0322	100	2.3111
2.3796	0.0644	200	2.2424
2.1234	0.0965	300	2.2007
2.2836	0.1287	400	2.1683
2.1882	0.1609	500	2.1389
2.1209	0.1931	600	2.1168
1.9807	0.2253	700	2.0961
2.0003	0.2574	800	2.0819
1.8976	0.2896	900	2.0609
2.2535	0.3218	1000	2.0465
2.0813	0.3540	1100	2.0284
2.0837	0.3862	1200	2.0134
1.8759	0.4183	1300	2.0016
2.0767	0.4505	1400	1.9873
2.0557	0.4827	1500	1.9745
1.708	0.5149	1600	1.9607
2.0113	0.5471	1700	1.9480
2.0235	0.5792	1800	1.9328
1.9351	0.6114	1900	1.9223
1.796	0.6436	2000	1.9164
1.8088	0.6758	2100	1.9011
1.6286	0.7080	2200	1.8900
1.9361	0.7401	2300	1.8786
1.636	0.7723	2400	1.8669
1.8454	0.8045	2500	1.8579
1.9547	0.8367	2600	1.8472
1.5638	0.8689	2700	1.8384
1.611	0.9010	2800	1.8275
1.7571	0.9332	2900	1.8143
1.9847	0.9654	3000	1.8063
1.9508	0.9976	3100	1.7982
1.3798	1.0298	3200	1.7971
1.5009	1.0619	3300	1.7864
1.5614	1.0941	3400	1.7795
1.7675	1.1263	3500	1.7721
1.2293	1.1585	3600	1.7665
1.7646	1.1907	3700	1.7610
1.4301	1.2228	3800	1.7582
1.2718	1.2550	3900	1.7507
1.6806	1.2872	4000	1.7457
1.5779	1.3194	4100	1.7413
1.3326	1.3516	4200	1.7374
1.6239	1.3837	4300	1.7350
1.713	1.4159	4400	1.7322
1.5364	1.4481	4500	1.7300
1.2966	1.4803	4600	1.7277
1.5881	1.5125	4700	1.7258
1.2867	1.5447	4800	1.7244
1.6141	1.5768	4900	1.7233
1.6782	1.6090	5000	1.7233
1.7381	1.6412	5100	1.7224
1.6309	1.6734	5200	1.7222
1.6214	1.7056	5300	1.7221

Framework versions

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