| 2025-08-30 18:41:06 - pico-train - INFO - Step 0 -- 📊 Evaluation Results | |
| 2025-08-30 18:41:06 - pico-train - INFO - └── paloma: inf | |
| 2025-08-30 18:41:06 - pico-train - INFO - ================================================== | |
| 2025-08-30 18:41:06 - pico-train - INFO - ✨ Training Configuration | |
| 2025-08-30 18:41:06 - pico-train - INFO - ================================================== | |
| 2025-08-30 18:41:06 - pico-train - INFO - ╭─────────────────────────────────────────────────────╮ | |
| 2025-08-30 18:41:06 - pico-train - INFO - │ checkpointing: │ | |
| 2025-08-30 18:41:06 - pico-train - INFO - │ checkpoints_dir: checkpoints │ | |
| 2025-08-30 18:41:06 - pico-train - INFO - │ evaluation: │ | |
| 2025-08-30 18:41:06 - pico-train - INFO - │ eval_results_dir: eval_results │ | |
| 2025-08-30 18:41:06 - pico-train - INFO - │ fabric_checkpoint_dir: fabric_state │ | |
| 2025-08-30 18:41:06 - pico-train - INFO - │ fabric_checkpoint_filename: checkpoint.pt │ | |
| 2025-08-30 18:41:06 - pico-train - INFO - │ hf_checkpoint: │ | |
| 2025-08-30 18:41:06 - pico-train - INFO - │ collection_slug: null │ | |
| 2025-08-30 18:41:06 - pico-train - INFO - │ repo_id: ThomasTheMaker/pico-decoder-tiny │ | |
| 2025-08-30 18:41:06 - pico-train - INFO - │ learning_dynamics: │ | |
| 2025-08-30 18:41:06 - pico-train - INFO - │ batch_size: 1 │ | |
| 2025-08-30 18:41:06 - pico-train - INFO - │ eval_data: null │ | |
| 2025-08-30 18:41:06 - pico-train - INFO - │ layer_suffixes: │ | |
| 2025-08-30 18:41:06 - pico-train - INFO - │ - attention.v_proj │ | |
| 2025-08-30 18:41:06 - pico-train - INFO - │ - attention.o_proj │ | |
| 2025-08-30 18:41:06 - pico-train - INFO - │ - swiglu.w_2 │ | |
| 2025-08-30 18:41:06 - pico-train - INFO - │ sequence_idx: -1 │ | |
| 2025-08-30 18:41:06 - pico-train - INFO - │ learning_dynamics_dir: learning_dynamics │ | |
| 2025-08-30 18:41:06 - pico-train - INFO - │ logs_dir: logs │ | |
| 2025-08-30 18:41:06 - pico-train - INFO - │ run_name: pico-decoder-tiny-wikipedia_en-v1 │ | |
| 2025-08-30 18:41:06 - pico-train - INFO - │ runs_dir: runs │ | |
| 2025-08-30 18:41:06 - pico-train - INFO - │ save_every_n_steps: 2000 │ | |
| 2025-08-30 18:41:06 - pico-train - INFO - │ save_to_hf: false │ | |
| 2025-08-30 18:41:06 - pico-train - INFO - │ training: │ | |
| 2025-08-30 18:41:06 - pico-train - INFO - │ auto_resume: true │ | |
| 2025-08-30 18:41:06 - pico-train - INFO - │ data: │ | |
| 2025-08-30 18:41:06 - pico-train - INFO - │ dataloader: │ | |
| 2025-08-30 18:41:06 - pico-train - INFO - │ batch_size: 16 │ | |
| 2025-08-30 18:41:06 - pico-train - INFO - │ dataset: │ | |
| 2025-08-30 18:41:06 - pico-train - INFO - │ name: ThomasTheMaker/pretokenized_wiki_en │ | |
| 2025-08-30 18:41:06 - pico-train - INFO - │ tokenizer: │ | |
| 2025-08-30 18:41:06 - pico-train - INFO - │ name: allenai/OLMo-7B-0724-hf │ | |
| 2025-08-30 18:41:06 - pico-train - INFO - │ vocab_size: 50304 │ | |
| 2025-08-30 18:41:06 - pico-train - INFO - │ evaluation: │ | |
| 2025-08-30 18:41:06 - pico-train - INFO - │ metrics: │ | |
| 2025-08-30 18:41:06 - pico-train - INFO - │ - paloma │ | |
| 2025-08-30 18:41:06 - pico-train - INFO - │ paloma: │ | |
| 2025-08-30 18:41:06 - pico-train - INFO - │ batch_size: 1 │ | |
| 2025-08-30 18:41:06 - pico-train - INFO - │ dataset_name: pico-lm/pretokenized-paloma-tinsy │ | |
| 2025-08-30 18:41:06 - pico-train - INFO - │ dataset_split: val │ | |
| 2025-08-30 18:41:06 - pico-train - INFO - │ max_length: 2048 │ | |
| 2025-08-30 18:41:06 - pico-train - INFO - │ model: │ | |
| 2025-08-30 18:41:06 - pico-train - INFO - │ activation_hidden_dim: 384 │ | |
| 2025-08-30 18:41:06 - pico-train - INFO - │ attention_n_heads: 12 │ | |
| 2025-08-30 18:41:06 - pico-train - INFO - │ attention_n_kv_heads: 4 │ | |
| 2025-08-30 18:41:06 - pico-train - INFO - │ batch_size: 1024 │ | |
| 2025-08-30 18:41:06 - pico-train - INFO - │ d_model: 96 │ | |
| 2025-08-30 18:41:06 - pico-train - INFO - │ max_seq_len: 2048 │ | |
| 2025-08-30 18:41:06 - pico-train - INFO - │ model_type: pico_decoder │ | |
| 2025-08-30 18:41:06 - pico-train - INFO - │ n_layers: 12 │ | |
| 2025-08-30 18:41:06 - pico-train - INFO - │ norm_eps: 1.0e-06 │ | |
| 2025-08-30 18:41:06 - pico-train - INFO - │ position_emb_theta: 10000.0 │ | |
| 2025-08-30 18:41:06 - pico-train - INFO - │ vocab_size: 50304 │ | |
| 2025-08-30 18:41:06 - pico-train - INFO - │ monitoring: │ | |
| 2025-08-30 18:41:06 - pico-train - INFO - │ logging: │ | |
| 2025-08-30 18:41:06 - pico-train - INFO - │ log_every_n_steps: 100 │ | |
| 2025-08-30 18:41:06 - pico-train - INFO - │ log_level: INFO │ | |
| 2025-08-30 18:41:06 - pico-train - INFO - │ save_to_wandb: false │ | |
| 2025-08-30 18:41:06 - pico-train - INFO - │ wandb: │ | |
| 2025-08-30 18:41:06 - pico-train - INFO - │ entity: boymyc │ | |
| 2025-08-30 18:41:06 - pico-train - INFO - │ project: pico-decoder-tiny │ | |
| 2025-08-30 18:41:06 - pico-train - INFO - │ training: │ | |
| 2025-08-30 18:41:06 - pico-train - INFO - │ fabric: │ | |
| 2025-08-30 18:41:06 - pico-train - INFO - │ accelerator: cuda │ | |
| 2025-08-30 18:41:06 - pico-train - INFO - │ num_devices: 1 │ | |
| 2025-08-30 18:41:06 - pico-train - INFO - │ num_nodes: 1 │ | |
| 2025-08-30 18:41:06 - pico-train - INFO - │ precision: bf16-mixed │ | |
| 2025-08-30 18:41:06 - pico-train - INFO - │ max_steps: 100000 │ | |
| 2025-08-30 18:41:06 - pico-train - INFO - │ optimization: │ | |
| 2025-08-30 18:41:06 - pico-train - INFO - │ gradient_accumulation_steps: 1 │ | |
| 2025-08-30 18:41:06 - pico-train - INFO - │ lr: 0.0002 │ | |
| 2025-08-30 18:41:06 - pico-train - INFO - │ lr_scheduler: cosine │ | |
| 2025-08-30 18:41:06 - pico-train - INFO - │ lr_warmup_steps: 2000 │ | |
| 2025-08-30 18:41:06 - pico-train - INFO - │ optimizer: adamw │ | |
| 2025-08-30 18:41:06 - pico-train - INFO - │ │ | |
| 2025-08-30 18:41:06 - pico-train - INFO - ╰─────────────────────────────────────────────────────╯ | |
| 2025-08-30 18:41:06 - pico-train - INFO - ================================================== | |
| 2025-08-30 18:41:06 - pico-train - INFO - ⛭ Runtime Summary: | |
| 2025-08-30 18:41:06 - pico-train - INFO - ================================================== | |
| 2025-08-30 18:41:06 - pico-train - INFO - Starting from step: 0 | |
| 2025-08-30 18:41:06 - pico-train - INFO - Model Setup: | |
| 2025-08-30 18:41:06 - pico-train - INFO - └─ Total Parameters: 11,282,784 | |
| 2025-08-30 18:41:06 - pico-train - INFO - └─ Trainable Parameters: 11,282,784 | |
| 2025-08-30 18:41:06 - pico-train - INFO - Distributed Setup: | |
| 2025-08-30 18:41:06 - pico-train - INFO - └─ Number of Devices: 1 | |
| 2025-08-30 18:41:06 - pico-train - INFO - └─ Device Type: NVIDIA GeForce RTX 5090 | |
| 2025-08-30 18:41:06 - pico-train - INFO - └─ Available Memory: 33.68 GB | |
| 2025-08-30 18:41:06 - pico-train - INFO - Software Setup: | |
| 2025-08-30 18:41:06 - pico-train - INFO - └─ Python Version: 3.10.12 | |
| 2025-08-30 18:41:06 - pico-train - INFO - └─ PyTorch Version: 2.8.0+cu128 | |
| 2025-08-30 18:41:06 - pico-train - INFO - └─ CUDA Version: 12.8 | |
| 2025-08-30 18:41:06 - pico-train - INFO - └─ Operating System: Linux 6.8.0-63-generic | |
| 2025-08-30 18:41:06 - pico-train - INFO - Batch Size Configuration: | |
| 2025-08-30 18:41:06 - pico-train - INFO - └─ Global Batch Size: 16 | |
| 2025-08-30 18:41:06 - pico-train - INFO - └─ Per Device Batch Size: 16 | |
| 2025-08-30 18:41:06 - pico-train - INFO - └─ Gradient Accumulation Steps: 1 | |
| 2025-08-30 18:41:06 - pico-train - INFO - ================================================== | |