• Model creator: https://huggingface.co/mfirth/
• Original model: https://huggingface.co/mfirth/l3t-bad/

Original model description:

• generated_from_trainer datasets:
• axolotl_test.json model-index:
• name: models/llama_wm_test results: []

---

See axolotl config

axolotl version: 0.5.3.dev44+g5bef1906

yaml

1base_model: meta-llama/Llama-3.2-3B-Instruct 2 3datasets: 4 - path: axolotl_test.json 5 type: input_output 6dataset_prepared_path: last_run_prepared 7 8output_dir: ./models/llama_wm_test 9sequence_length: 4096 10 11wandb_project: agent-v0 12wandb_name: llama-3b_wm_test 13 14train_on_inputs: false 15gradient_checkpointing: true 16gradient_accumulation_steps: 2 17micro_batch_size: 1 18num_epochs: 5 19optimizer: adamw_torch 20learning_rate: 2e-5 21flash_attention: true 22 23logging_steps: 5 24 25warmup_steps: 10 26saves_per_epoch: 1 27weight_decay: 0.0 28 29deepspeed: axolotl/deepspeed_configs/zero3_bf16_cpuoffload_all.json 30 31special_tokens: 32 pad_token: <|end_of_text|>

models/llama_wm_test

Model description

Intended uses & limitations

Training and evaluation data

Training procedure

Training hyperparameters

• learning_rate: 2e-05
• train_batch_size: 1
• eval_batch_size: 1
• seed: 42
• distributed_type: multi-GPU
• num_devices: 8
• gradient_accumulation_steps: 2
• total_train_batch_size: 16
• total_eval_batch_size: 8
• optimizer: Use OptimizerNames.ADAMW_TORCH 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
• num_epochs: 5

Training results

Framework versions

• Transformers 4.47.0
• Pytorch 2.5.1+cu124
• Datasets 3.1.0
• Tokenizers 0.21.0