See axolotl config

axolotl version: 0.4.1

adapter: lora
base_model: unsloth/Qwen2.5-Math-7B-Instruct
bf16: true
chat_template: llama3
dataset_prepared_path: null
datasets:
- data_files:
  - 0c268a24a189e736_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/0c268a24a189e736_train_data.json
  type:
    field_instruction: sentence1
    field_output: sentence2
    format: '{instruction}'
    no_input_format: '{instruction}'
    system_format: '{system}'
    system_prompt: ''
debug: null
deepspeed: null
early_stopping_patience: 2
eval_max_new_tokens: 128
eval_steps: 100
eval_table_size: null
flash_attention: true
fp16: null
fsdp: null
fsdp_config: null
gradient_accumulation_steps: 4
gradient_checkpointing: true
group_by_length: false
hub_model_id: romainnn/66dfe5fa-fd69-4293-abab-78e89c0f92d3
hub_repo: null
hub_strategy: checkpoint
hub_token: null
learning_rate: 0.0001
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
lr_scheduler: cosine
max_steps: 1836
micro_batch_size: 4
mlflow_experiment_name: /tmp/0c268a24a189e736_train_data.json
model_type: AutoModelForCausalLM
num_epochs: 1
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: 0.05
wandb_entity: null
wandb_mode: online
wandb_name: 3fb659b5-4304-4b9e-8d6e-5af1f430fc24
wandb_project: Gradients-On-Demand
wandb_run: your_name
wandb_runid: 3fb659b5-4304-4b9e-8d6e-5af1f430fc24
warmup_steps: 10
weight_decay: 0.0
xformers_attention: null

66dfe5fa-fd69-4293-abab-78e89c0f92d3

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

• Loss: 1.4781

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: 0.0001
• train_batch_size: 4
• eval_batch_size: 4
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 16
• 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: 1836

Training results

Training Loss	Epoch	Step	Validation Loss
6.4697	0.0002	1	6.1452
2.0781	0.0190	100	2.0017
1.9594	0.0380	200	1.8549
1.5803	0.0570	300	1.7733
1.6428	0.0760	400	1.7145
1.5254	0.0950	500	1.6687
2.2101	0.1140	600	1.6367
2.0177	0.1329	700	1.6152
1.39	0.1519	800	1.5810
1.6422	0.1709	900	1.5632
1.136	0.1899	1000	1.5430
1.6514	0.2089	1100	1.5243
1.6233	0.2279	1200	1.5086
1.0544	0.2469	1300	1.4982
1.2562	0.2659	1400	1.4917
1.4276	0.2849	1500	1.4858
1.4498	0.3039	1600	1.4806
1.2587	0.3229	1700	1.4786
1.232	0.3419	1800	1.4781

Framework versions

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