See axolotl config

axolotl version: 0.4.1

adapter: lora
base_model: unsloth/Qwen2.5-1.5B-Instruct
bf16: true
chat_template: llama3
dataset_prepared_path: null
datasets:
- data_files:
  - e92167d2023abfa8_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/e92167d2023abfa8_train_data.json
  type:
    field_instruction: ja
    field_output: en
    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/f790e96a-4ceb-4271-ac22-5165f95329f6
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: 4140
micro_batch_size: 4
mlflow_experiment_name: /tmp/e92167d2023abfa8_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: 0.03406551479805963
wandb_entity: null
wandb_mode: online
wandb_name: 3d884858-79a9-4666-9f90-38fb8fe94cb0
wandb_project: Gradients-On-Demand
wandb_run: your_name
wandb_runid: 3d884858-79a9-4666-9f90-38fb8fe94cb0
warmup_steps: 10
weight_decay: 0.0
xformers_attention: null

f790e96a-4ceb-4271-ac22-5165f95329f6

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

• Loss: 0.7197

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: 4140

Training results

Training Loss	Epoch	Step	Validation Loss
4.1503	0.0002	1	4.1704
0.8996	0.0226	100	0.9047
0.9677	0.0451	200	0.8716
0.8013	0.0677	300	0.8545
0.8224	0.0903	400	0.8457
0.8158	0.1129	500	0.8424
0.8297	0.1354	600	0.8289
0.8232	0.1580	700	0.8220
1.0478	0.1806	800	0.8119
0.843	0.2031	900	0.8083
0.7756	0.2257	1000	0.7978
0.6862	0.2483	1100	0.7935
0.6653	0.2708	1200	0.7885
0.7703	0.2934	1300	0.7819
0.8533	0.3160	1400	0.7768
0.721	0.3386	1500	0.7736
0.8648	0.3611	1600	0.7678
0.8437	0.3837	1700	0.7674
0.9612	0.4063	1800	0.7648
0.7561	0.4288	1900	0.7584
0.8658	0.4514	2000	0.7533
0.9037	0.4740	2100	0.7504
0.765	0.4966	2200	0.7484
0.6422	0.5191	2300	0.7438
0.6245	0.5417	2400	0.7429
0.7151	0.5643	2500	0.7408
0.7394	0.5868	2600	0.7367
0.8205	0.6094	2700	0.7344
0.6712	0.6320	2800	0.7313
0.4926	0.6546	2900	0.7294
0.9032	0.6771	3000	0.7272
0.7309	0.6997	3100	0.7262
0.6001	0.7223	3200	0.7248
0.8966	0.7448	3300	0.7231
0.6417	0.7674	3400	0.7223
0.8726	0.7900	3500	0.7214
0.6451	0.8125	3600	0.7205
0.6828	0.8351	3700	0.7203
0.6273	0.8577	3800	0.7198
0.8253	0.8803	3900	0.7197
0.7465	0.9028	4000	0.7195
0.7324	0.9254	4100	0.7197

Framework versions

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