See axolotl config

axolotl version: 0.4.1

adapter: lora
base_model: Qwen/Qwen2.5-Math-7B-Instruct
bf16: true
chat_template: llama3
dataset_prepared_path: null
datasets:
- data_files:
  - 62a9e2878ae88813_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/62a9e2878ae88813_train_data.json
  type:
    field_instruction: question
    field_output: answer
    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/b2f97e89-fe1c-4a1d-a135-f54c9da4f682
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: 3168
micro_batch_size: 4
mlflow_experiment_name: /tmp/62a9e2878ae88813_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: 0.023880026745629956
wandb_entity: null
wandb_mode: online
wandb_name: 4c2648cd-2f83-40bf-8de7-aebad4780f8e
wandb_project: Gradients-On-Demand
wandb_run: your_name
wandb_runid: 4c2648cd-2f83-40bf-8de7-aebad4780f8e
warmup_steps: 10
weight_decay: 0.0
xformers_attention: null

b2f97e89-fe1c-4a1d-a135-f54c9da4f682

This model is a fine-tuned version of Qwen/Qwen2.5-Math-7B-Instruct on the None dataset. It achieves the following results on the evaluation set:

• Loss: 1.8621

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: 3168

Training results

Training Loss	Epoch	Step	Validation Loss
3.713	0.0002	1	3.6150
2.2302	0.0157	100	2.3311
2.1697	0.0313	200	2.2208
2.0789	0.0470	300	2.1619
2.1861	0.0626	400	2.1230
2.1874	0.0783	500	2.0914
2.1201	0.0939	600	2.0676
2.0033	0.1096	700	2.0460
1.8952	0.1253	800	2.0263
2.0334	0.1409	900	2.0115
2.0062	0.1566	1000	1.9978
1.9782	0.1722	1100	1.9826
1.8561	0.1879	1200	1.9696
1.9684	0.2035	1300	1.9595
1.8611	0.2192	1400	1.9485
2.0116	0.2349	1500	1.9382
1.8545	0.2505	1600	1.9299
2.032	0.2662	1700	1.9192
1.8863	0.2818	1800	1.9116
1.9535	0.2975	1900	1.9042
1.895	0.3131	2000	1.8972
1.8298	0.3288	2100	1.8903
1.9696	0.3445	2200	1.8843
1.7887	0.3601	2300	1.8794
1.8385	0.3758	2400	1.8751
1.8666	0.3914	2500	1.8709
1.8953	0.4071	2600	1.8681
1.8449	0.4227	2700	1.8657
1.7412	0.4384	2800	1.8639
1.9192	0.4541	2900	1.8629
1.8781	0.4697	3000	1.8623
1.7919	0.4854	3100	1.8621

Framework versions

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