See axolotl config

axolotl version: 0.4.1

adapter: lora
base_model: unsloth/Qwen2-1.5B-Instruct
bf16: true
chat_template: llama3
dataset_prepared_path: null
datasets:
- data_files:
  - c33817b278dd1b30_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/c33817b278dd1b30_train_data.json
  type:
    field_instruction: prompt
    field_output: gold_standard_solution
    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/86eafdfb-5eb9-448a-a70a-3210c9a270e9
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: 3284
micro_batch_size: 4
mlflow_experiment_name: /tmp/c33817b278dd1b30_train_data.json
model_type: AutoModelForCausalLM
num_epochs: 10
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: 9e763106-cafb-464d-80d4-303092680580
wandb_project: Gradients-On-Demand
wandb_run: your_name
wandb_runid: 9e763106-cafb-464d-80d4-303092680580
warmup_steps: 10
weight_decay: 0.0
xformers_attention: null

86eafdfb-5eb9-448a-a70a-3210c9a270e9

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

• Loss: 1.0182

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: 3284

Training results

Training Loss	Epoch	Step	Validation Loss
4.0829	0.0006	1	4.0254
2.6836	0.0561	100	1.9840
1.7081	0.1122	200	1.8655
1.0855	0.1683	300	1.7443
0.9939	0.2244	400	1.6416
1.6035	0.2805	500	1.5873
1.9726	0.3366	600	1.5140
1.0763	0.3927	700	1.4661
1.3097	0.4488	800	1.4378
1.7479	0.5049	900	1.3994
1.3175	0.5610	1000	1.3713
1.3454	0.6171	1100	1.3504
1.0311	0.6732	1200	1.3095
0.8146	0.7293	1300	1.2796
0.8457	0.7854	1400	1.2579
1.0289	0.8415	1500	1.2365
1.4244	0.8976	1600	1.1983
0.6472	0.9536	1700	1.1770
0.9321	1.0098	1800	1.1521
1.1917	1.0659	1900	1.1388
0.6978	1.1220	2000	1.1290
0.6796	1.1781	2100	1.1184
0.9681	1.2342	2200	1.0914
0.6488	1.2903	2300	1.0788
0.9736	1.3464	2400	1.0686
0.7818	1.4025	2500	1.0503
0.4946	1.4586	2600	1.0485
0.6321	1.5147	2700	1.0376
0.9863	1.5708	2800	1.0300
0.78	1.6269	2900	1.0246
0.7719	1.6830	3000	1.0207
0.8101	1.7391	3100	1.0186
1.4023	1.7952	3200	1.0182

Framework versions

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