See axolotl config

axolotl version: 0.4.1

adapter: lora
base_model: unsloth/Llama-3.2-3B-Instruct
bf16: true
chat_template: llama3
dataset_prepared_path: null
datasets:
- data_files:
  - acb749d787b38378_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/acb749d787b38378_train_data.json
  type:
    field_instruction: Human
    field_output: Assistant
    format: '{instruction}'
    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/8cd86b61-a475-4058-9d06-e477ca64d232
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: 5760
micro_batch_size: 2
mlflow_experiment_name: /tmp/acb749d787b38378_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: 2048
strict: false
tf32: true
tokenizer_type: AutoTokenizer
train_on_inputs: false
trust_remote_code: true
val_set_size: 0.00833750208437552
wandb_entity: null
wandb_mode: online
wandb_name: f2e0a8b6-4518-4e1e-b98d-bf8398edaaf4
wandb_project: Gradients-On-Demand
wandb_run: your_name
wandb_runid: f2e0a8b6-4518-4e1e-b98d-bf8398edaaf4
warmup_steps: 10
weight_decay: 0.0
xformers_attention: null

8cd86b61-a475-4058-9d06-e477ca64d232

This model is a fine-tuned version of unsloth/Llama-3.2-3B-Instruct on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.9805

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: 5760

Training results

Training Loss	Epoch	Step	Validation Loss
2.3506	0.0000	1	2.2271
1.1861	0.0020	150	1.2033
1.2478	0.0040	300	1.1735
1.1161	0.0061	450	1.1780
1.1883	0.0081	600	1.1659
1.1853	0.0101	750	1.1636
1.2449	0.0121	900	1.1556
1.0864	0.0141	1050	1.1469
1.183	0.0161	1200	1.1385
1.1754	0.0182	1350	1.1398
1.3482	0.0202	1500	1.1360
1.2008	0.0222	1650	1.1291
1.0484	0.0242	1800	1.1208
0.9855	0.0262	1950	1.1180
1.0422	0.0282	2100	1.1076
1.1352	0.0303	2250	1.1020
1.3159	0.0323	2400	1.0892
1.2148	0.0343	2550	1.0849
0.9585	0.0363	2700	1.0824
1.1377	0.0383	2850	1.0702
1.0919	0.0404	3000	1.0620
1.0636	0.0424	3150	1.0557
1.062	0.0444	3300	1.0446
1.053	0.0464	3450	1.0382
0.8881	0.0484	3600	1.0311
0.9283	0.0504	3750	1.0248
1.0071	0.0525	3900	1.0165
1.1635	0.0545	4050	1.0103
1.0111	0.0565	4200	1.0052
0.9878	0.0585	4350	1.0006
1.0493	0.0605	4500	0.9971
1.0038	0.0626	4650	0.9913
1.1421	0.0646	4800	0.9880
1.007	0.0666	4950	0.9856
1.0345	0.0686	5100	0.9835
0.8716	0.0706	5250	0.9820
1.0461	0.0726	5400	0.9811
0.8566	0.0747	5550	0.9806
1.0293	0.0767	5700	0.9805

Framework versions

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