See axolotl config

axolotl version: 0.4.1

base_model: NousResearch/Llama-3.2-1B
batch_size: 32
bf16: true
chat_template: tokenizer_default_fallback_alpaca
datasets:
- data_files:
  - f51beb4c568b9128_train_data.json
  ds_type: json
  format: custom
  path: /workspace/input_data/f51beb4c568b9128_train_data.json
  type:
    field_input: keywords
    field_instruction: idea
    field_output: full_response
    format: '{instruction} {input}'
    no_input_format: '{instruction}'
    system_format: '{system}'
    system_prompt: ''
eval_steps: 20
flash_attention: true
gpu_memory_limit: 80GiB
gradient_checkpointing: true
group_by_length: true
hub_model_id: willtensora/0c2649cc-2fe7-4e88-b672-6da1fee4001f
hub_strategy: checkpoint
learning_rate: 0.0002
logging_steps: 10
lr_scheduler: cosine
max_steps: 2500
micro_batch_size: 4
model_type: AutoModelForCausalLM
optimizer: adamw_bnb_8bit
output_dir: /workspace/axolotl/configs
pad_to_sequence_len: true
resize_token_embeddings_to_32x: false
sample_packing: false
save_steps: 40
save_total_limit: 1
sequence_len: 2048
special_tokens:
  pad_token: <|end_of_text|>
tokenizer_type: PreTrainedTokenizerFast
train_on_inputs: false
trust_remote_code: true
val_set_size: 0.1
wandb_entity: ''
wandb_mode: online
wandb_name: NousResearch/Llama-3.2-1B-/workspace/input_data/f51beb4c568b9128_train_data.json
wandb_project: Gradients-On-Demand
wandb_run: your_name
wandb_runid: default
warmup_ratio: 0.05
xformers_attention: true

0c2649cc-2fe7-4e88-b672-6da1fee4001f

This model is a fine-tuned version of NousResearch/Llama-3.2-1B on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.0849

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
• distributed_type: multi-GPU
• num_devices: 8
• total_train_batch_size: 32
• total_eval_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: 12
• training_steps: 258

Training results

Training Loss	Epoch	Step	Validation Loss
No log	0.0005	1	0.2074
0.5472	0.0097	20	0.1746
0.3199	0.0194	40	0.2036
0.2013	0.0291	60	0.1772
0.0903	0.0388	80	0.1702
0.0875	0.0485	100	0.2040
0.1425	0.0582	120	0.1392
0.1982	0.0679	140	0.1194
0.1372	0.0776	160	0.1014
0.0278	0.0873	180	0.0952
0.0248	0.0970	200	0.0893
0.1051	0.1067	220	0.0875
0.0649	0.1164	240	0.0849

Framework versions

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