See axolotl config

axolotl version: 0.4.1

adapter: lora
base_model: unsloth/llama-3-8b-Instruct
bf16: true
chat_template: llama3
dataset_prepared_path: null
datasets:
- data_files:
  - c0ec6108c83afe61_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/c0ec6108c83afe61_train_data.json
  type:
    field_input: input
    field_instruction: instruction
    field_output: chosen
    format: '{instruction} {input}'
    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/aa4f337e-4dc1-4577-8c3f-5f20fe95556a
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: 1218
micro_batch_size: 4
mlflow_experiment_name: /tmp/c0ec6108c83afe61_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: 0.047493754571273875
wandb_entity: null
wandb_mode: online
wandb_name: aa6026b9-2848-4f1c-9f66-f064de659c63
wandb_project: Gradients-On-Demand
wandb_run: your_name
wandb_runid: aa6026b9-2848-4f1c-9f66-f064de659c63
warmup_steps: 10
weight_decay: 0.0
xformers_attention: null

aa4f337e-4dc1-4577-8c3f-5f20fe95556a

This model is a fine-tuned version of unsloth/llama-3-8b-Instruct on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.1383

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: 1218

Training results

Training Loss	Epoch	Step	Validation Loss
0.6968	0.0003	1	0.6846
0.2434	0.0319	100	0.2253
0.2147	0.0638	200	0.2074
0.2058	0.0957	300	0.1914
0.19	0.1276	400	0.1808
0.1718	0.1596	500	0.1709
0.1899	0.1915	600	0.1628
0.1424	0.2234	700	0.1549
0.1678	0.2553	800	0.1484
0.151	0.2872	900	0.1433
0.1485	0.3191	1000	0.1403
0.1541	0.3510	1100	0.1387
0.1446	0.3829	1200	0.1383

Framework versions

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