| tokenizer: | |
| _component_: torchtune.models.gemma.gemma_tokenizer | |
| path: ./target/gemma-2b/tokenizer.model | |
| dataset: | |
| _component_: torchtune.datasets.alpaca_dataset | |
| seed: null | |
| shuffle: true | |
| model: | |
| _component_: torchtune.models.gemma.lora_gemma_2b | |
| lora_attn_modules: | |
| - q_proj | |
| - k_proj | |
| - v_proj | |
| apply_lora_to_mlp: true | |
| lora_rank: 64 | |
| lora_alpha: 128 | |
| lora_dropout: 0.0 | |
| checkpointer: | |
| _component_: torchtune.training.FullModelHFCheckpointer | |
| checkpoint_dir: ./target/gemma-2b/ | |
| checkpoint_files: | |
| - model-00001-of-00002.safetensors | |
| - model-00002-of-00002.safetensors | |
| recipe_checkpoint: null | |
| output_dir: ${output_dir}/weights | |
| model_type: GEMMA | |
| resume_from_checkpoint: false | |
| save_adapter_weights_only: false | |
| optimizer: | |
| _component_: torch.optim.AdamW | |
| fused: false | |
| lr: 2.0e-05 | |
| lr_scheduler: | |
| _component_: torchtune.modules.get_cosine_schedule_with_warmup | |
| num_warmup_steps: 10 | |
| loss: | |
| _component_: torchtune.modules.loss.CEWithChunkedOutputLoss | |
| batch_size: 4 | |
| epochs: 1 | |
| max_steps_per_epoch: 10 | |
| gradient_accumulation_steps: 4 | |
| compile: false | |
| device: mps | |
| enable_activation_checkpointing: true | |
| enable_activation_offloading: false | |
| dtype: bf16 | |
| metric_logger: | |
| _component_: torchtune.training.metric_logging.DiskLogger | |
| log_dir: ${output_dir} | |
| output_dir: ./target/gemma_tmp | |
| log_every_n_steps: 1 | |
| log_peak_memory_stats: false | |
| profiler: | |
| _component_: torchtune.training.setup_torch_profiler | |
| enabled: false | |
| output_dir: ${output_dir}/profiling_outputs | |
| cpu: true | |
| cuda: true | |
| profile_memory: false | |
| with_stack: false | |
| record_shapes: true | |
| with_flops: false | |
| wait_steps: 5 | |
| warmup_steps: 5 | |
| active_steps: 2 | |
| num_cycles: 1 | |