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:
  - 87f3c9b2656146bb_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/87f3c9b2656146bb_train_data.json
  type:
    field_input: statements
    field_instruction: quiz
    field_output: solution_text
    format: '{instruction} {input}'
    no_input_format: '{instruction}'
    system_format: '{system}'
    system_prompt: ''
debug: null
deepspeed: null
early_stopping_patience: 4
eval_max_new_tokens: 128
eval_steps: 150
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: Romain-XV/c9a25fed-0ee9-46e3-a648-22b9ee8890c4
hub_repo: null
hub_strategy: checkpoint
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: 0.3
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
lr_scheduler: cosine
max_grad_norm: 1.0
max_steps: 7200
micro_batch_size: 2
mlflow_experiment_name: /tmp/87f3c9b2656146bb_train_data.json
model_type: AutoModelForCausalLM
num_epochs: 3
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: 150
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: 146c3328-3a36-49c9-a68a-1b82be7dcc55
wandb_project: Gradients-On-Demand
wandb_run: your_name
wandb_runid: 146c3328-3a36-49c9-a68a-1b82be7dcc55
warmup_steps: 10
weight_decay: 0.0
xformers_attention: null

c9a25fed-0ee9-46e3-a648-22b9ee8890c4

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: 0.0335

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: 2
• eval_batch_size: 2
• seed: 42
• gradient_accumulation_steps: 4
• total_train_batch_size: 8
• 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: 7200

Training results

Training Loss	Epoch	Step	Validation Loss
1.2132	0.0002	1	1.2143
0.1208	0.0340	150	0.1146
0.0996	0.0681	300	0.1296
0.0851	0.1021	450	0.1126
0.0921	0.1362	600	0.1001
0.0925	0.1702	750	0.0954
0.0789	0.2043	900	0.0919
0.1317	0.2383	1050	0.1077
0.1018	0.2724	1200	0.0936
0.0898	0.3064	1350	0.0863
0.0781	0.3405	1500	0.0785
0.0727	0.3745	1650	0.0763
0.0688	0.4086	1800	0.0736
0.1039	0.4426	1950	0.0808
0.0773	0.4766	2100	0.0675
0.0955	0.5107	2250	0.0773
0.0815	0.5447	2400	0.0620
0.0946	0.5788	2550	0.0672
0.0616	0.6128	2700	0.0616
0.0579	0.6469	2850	0.0568
0.0489	0.6809	3000	0.0561
0.0753	0.7150	3150	0.0568
0.0656	0.7490	3300	0.0520
0.0514	0.7831	3450	0.0592
0.0472	0.8171	3600	0.0507
0.046	0.8512	3750	0.0487
0.0742	0.8852	3900	0.0466
0.0426	0.9193	4050	0.0473
0.0707	0.9533	4200	0.0443
0.0726	0.9873	4350	0.0446
0.0427	1.0214	4500	0.0432
0.0649	1.0555	4650	0.0419
0.0727	1.0895	4800	0.0430
0.0696	1.1236	4950	0.0404
0.041	1.1576	5100	0.0403
0.0501	1.1917	5250	0.0393
0.0474	1.2257	5400	0.0377
0.0281	1.2598	5550	0.0380
0.0397	1.2938	5700	0.0366
0.0329	1.3279	5850	0.0373
0.0405	1.3619	6000	0.0352
0.0335	1.3960	6150	0.0346
0.0543	1.4300	6300	0.0345
0.0129	1.4641	6450	0.0340
0.0288	1.4981	6600	0.0338
0.0308	1.5321	6750	0.0337
0.022	1.5662	6900	0.0336
0.0284	1.6002	7050	0.0335
0.0338	1.6343	7200	0.0335

Framework versions

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