See axolotl config

axolotl version: 0.4.1

adapter: lora
base_model: unsloth/Llama-3.2-3B
bf16: true
chat_template: llama3
dataset_prepared_path: null
datasets:
- data_files:
  - b10b004d99069455_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/b10b004d99069455_train_data.json
  type:
    field_instruction: startphrase
    field_output: gold-ending
    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/e7a8531a-f667-43d7-96f2-07ed1b116e7e
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: 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
- o_proj
- down_proj
- up_proj
lr_scheduler: cosine
max_grad_norm: 1.0
max_steps: 2346
micro_batch_size: 4
mlflow_experiment_name: /tmp/b10b004d99069455_train_data.json
model_type: AutoModelForCausalLM
num_epochs: 2
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: 2048
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: 2d26c05a-cd41-4443-90b8-af6e34d0351b
wandb_project: Gradients-On-Demand
wandb_run: your_name
wandb_runid: 2d26c05a-cd41-4443-90b8-af6e34d0351b
warmup_steps: 10
weight_decay: 0.0
xformers_attention: null

e7a8531a-f667-43d7-96f2-07ed1b116e7e

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

• Loss: 2.2395

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: 2346

Training results

Training Loss	Epoch	Step	Validation Loss
4.4298	0.0004	1	4.2132
2.6515	0.0364	100	2.4191
2.224	0.0729	200	2.3802
2.7286	0.1093	300	2.3568
2.4288	0.1457	400	2.3433
2.2517	0.1822	500	2.3338
2.3638	0.2186	600	2.3226
2.4078	0.2550	700	2.3116
2.4365	0.2915	800	2.3066
2.3669	0.3279	900	2.3001
2.363	0.3643	1000	2.2917
2.2615	0.4008	1100	2.2847
2.3083	0.4372	1200	2.2774
2.1599	0.4736	1300	2.2743
2.1445	0.5101	1400	2.2687
2.3314	0.5465	1500	2.2618
2.2573	0.5829	1600	2.2566
2.4727	0.6194	1700	2.2512
2.1696	0.6558	1800	2.2482
2.3054	0.6922	1900	2.2447
2.312	0.7287	2000	2.2424
2.2064	0.7651	2100	2.2407
2.4036	0.8015	2200	2.2398
2.2385	0.8380	2300	2.2395

Framework versions

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