| just_score_model: false |
| tau: null |
| data_start_step: null |
| sft_dataset: null |
| load_checkpoint_type: null |
| run_name: olmo_11816275_1 |
| seed: 0 |
| epoch: null |
| dry_run: false |
| model: |
| d_model: 1024 |
| n_heads: 16 |
| n_kv_heads: null |
| clip_qkv: null |
| n_layers: 12 |
| mlp_ratio: 4 |
| mlp_hidden_size: 4096 |
| activation_type: gelu |
| block_type: sequential |
| block_group_size: 1 |
| alibi: false |
| alibi_bias_max: 8.0 |
| rope: true |
| rope_full_precision: true |
| flash_attention: false |
| attention_dropout: 0.0 |
| multi_query_attention: false |
| attention_layer_norm: true |
| residual_dropout: 0.0 |
| embedding_dropout: 0.0 |
| layer_norm_type: default |
| layer_norm_with_affine: true |
| attention_layer_norm_with_affine: true |
| max_sequence_length: 512 |
| include_bias: false |
| bias_for_layer_norm: false |
| scale_logits: false |
| vocab_size: 50280 |
| embedding_size: 50304 |
| weight_tying: false |
| eos_token_id: 50279 |
| pad_token_id: 1 |
| init_device: meta |
| init_fn: mitchell |
| init_std: 0.02 |
| init_cutoff_factor: null |
| precision: amp_bf16 |
| optimizer: |
| name: adamw |
| learning_rate: 0.003 |
| weight_decay: 0.033 |
| betas: |
| - 0.9 |
| - 0.95 |
| eps: 1.0e-15 |
| no_decay_norm_and_bias: null |
| decay_norm_and_bias: false |
| decay_embeddings: false |
| metrics_log_interval: 100 |
| scheduler: |
| name: cosine_with_warmup |
| units: steps |
| t_warmup: 1000 |
| t_max: null |
| alpha_f: 0.1 |
| grad_clip_warmup_steps: null |
| grad_clip_warmup_factor: null |
| warmup_min_lr: null |
| data: |
| paths: finewebedu-top3B |
| datasets: null |
| label_mask_paths: null |
| pad_direction: right |
| generate_attention_mask: true |
| num_workers: 16 |
| drop_last: true |
| pin_memory: true |
| prefetch_factor: 16 |
| persistent_workers: true |
| timeout: 0 |
| seed: null |
| memmap_dtype: uint16 |
| extra_data_paths: null |
| extra_data_key: null |
| load_extra_data_to_ram: false |
| index_path: null |
| restore_dataloader: true |
| fast_forward_batches: null |
| evaluators: |
| - label: finewebedu-val |
| type: lm |
| data: |
| paths: finewebedu-val |
| datasets: null |
| label_mask_paths: null |
| pad_direction: right |
| generate_attention_mask: true |
| num_workers: 0 |
| drop_last: true |
| pin_memory: false |
| prefetch_factor: null |
| persistent_workers: false |
| timeout: 0 |
| seed: null |
| memmap_dtype: uint16 |
| extra_data_paths: null |
| extra_data_key: null |
| load_extra_data_to_ram: false |
| index_path: null |
| device_eval_batch_size: null |
| subset_num_batches: null |
| sft_use_label: false |
| sft: false |
| - label: piqa_train |
| type: downstream |
| data: |
| paths: null |
| datasets: null |
| label_mask_paths: null |
| pad_direction: right |
| generate_attention_mask: true |
| num_workers: 0 |
| drop_last: false |
| pin_memory: false |
| prefetch_factor: null |
| persistent_workers: false |
| timeout: 0 |
| seed: null |
| memmap_dtype: uint16 |
| extra_data_paths: null |
| extra_data_key: null |
| load_extra_data_to_ram: false |
| index_path: null |
| device_eval_batch_size: null |
| subset_num_batches: null |
| sft_use_label: false |
| sft: false |
| - label: openbook_qa_train |
| type: downstream |
| data: |
| paths: null |
| datasets: null |
| label_mask_paths: null |
| pad_direction: right |
| generate_attention_mask: true |
| num_workers: 0 |
| drop_last: false |
| pin_memory: false |
| prefetch_factor: null |
| persistent_workers: false |
| timeout: 0 |
| seed: null |
| memmap_dtype: uint16 |
| extra_data_paths: null |
| extra_data_key: null |
| load_extra_data_to_ram: false |
| index_path: null |
| device_eval_batch_size: null |
| subset_num_batches: null |
| sft_use_label: false |
| sft: false |
| - label: hellaswag_train |
| type: downstream |
| data: |
| paths: null |
| datasets: null |
| label_mask_paths: null |
| pad_direction: right |
| generate_attention_mask: true |
| num_workers: 0 |
| drop_last: false |
| pin_memory: false |
| prefetch_factor: null |
| persistent_workers: false |
| timeout: 0 |
| seed: null |
| memmap_dtype: uint16 |
| extra_data_paths: null |
| extra_data_key: null |
| load_extra_data_to_ram: false |
| index_path: null |
| device_eval_batch_size: null |
| subset_num_batches: null |
| sft_use_label: false |
| sft: false |
| - label: winogrande_train |
| type: downstream |
| data: |
| paths: null |
| datasets: null |
| label_mask_paths: null |
| pad_direction: right |
| generate_attention_mask: true |
| num_workers: 0 |
| drop_last: false |
| pin_memory: false |
| prefetch_factor: null |
| persistent_workers: false |
| timeout: 0 |
| seed: null |
| memmap_dtype: uint16 |
| extra_data_paths: null |
| extra_data_key: null |
| load_extra_data_to_ram: false |
| index_path: null |
| device_eval_batch_size: null |
| subset_num_batches: null |
| sft_use_label: false |
| sft: false |
| - label: arc_easy_train |
| type: downstream |
| data: |
| paths: null |
| datasets: null |
| label_mask_paths: null |
| pad_direction: right |
| generate_attention_mask: true |
| num_workers: 0 |
| drop_last: false |
| pin_memory: false |
| prefetch_factor: null |
| persistent_workers: false |
| timeout: 0 |
| seed: null |
| memmap_dtype: uint16 |
| extra_data_paths: null |
| extra_data_key: null |
| load_extra_data_to_ram: false |
| index_path: null |
| device_eval_batch_size: null |
| subset_num_batches: null |
| sft_use_label: false |
| sft: false |
| - label: arc_challenge_train |
| type: downstream |
| data: |
| paths: null |
| datasets: null |
| label_mask_paths: null |
| pad_direction: right |
| generate_attention_mask: true |
| num_workers: 0 |
| drop_last: false |
| pin_memory: false |
| prefetch_factor: null |
| persistent_workers: false |
| timeout: 0 |
| seed: null |
| memmap_dtype: uint16 |
| extra_data_paths: null |
| extra_data_key: null |
| load_extra_data_to_ram: false |
| index_path: null |
| device_eval_batch_size: null |
| subset_num_batches: null |
| sft_use_label: false |
| sft: false |
| - label: boolq_train |
| type: downstream |
| data: |
| paths: null |
| datasets: null |
| label_mask_paths: null |
| pad_direction: right |
| generate_attention_mask: true |
| num_workers: 0 |
| drop_last: false |
| pin_memory: false |
| prefetch_factor: null |
| persistent_workers: false |
| timeout: 0 |
| seed: null |
| memmap_dtype: uint16 |
| extra_data_paths: null |
| extra_data_key: null |
| load_extra_data_to_ram: false |
| index_path: null |
| device_eval_batch_size: null |
| subset_num_batches: null |
| sft_use_label: false |
| sft: false |
| - label: sciq_train |
| type: downstream |
| data: |
| paths: null |
| datasets: null |
| label_mask_paths: null |
| pad_direction: right |
| generate_attention_mask: true |
| num_workers: 0 |
| drop_last: false |
| pin_memory: false |
| prefetch_factor: null |
| persistent_workers: false |
| timeout: 0 |
| seed: null |
| memmap_dtype: uint16 |
| extra_data_paths: null |
| extra_data_key: null |
| load_extra_data_to_ram: false |
| index_path: null |
| device_eval_batch_size: null |
| subset_num_batches: null |
| sft_use_label: false |
| sft: false |
| - label: piqa_test |
| type: downstream |
| data: |
| paths: null |
| datasets: null |
| label_mask_paths: null |
| pad_direction: right |
| generate_attention_mask: true |
| num_workers: 0 |
| drop_last: false |
| pin_memory: false |
| prefetch_factor: null |
| persistent_workers: false |
| timeout: 0 |
| seed: null |
| memmap_dtype: uint16 |
| extra_data_paths: null |
| extra_data_key: null |
| load_extra_data_to_ram: false |
| index_path: null |
| device_eval_batch_size: null |
| subset_num_batches: 1000 |
| sft_use_label: false |
| sft: false |
| - label: openbook_qa_test |
| type: downstream |
| data: |
| paths: null |
| datasets: null |
| label_mask_paths: null |
| pad_direction: right |
| generate_attention_mask: true |
| num_workers: 0 |
| drop_last: false |
| pin_memory: false |
| prefetch_factor: null |
| persistent_workers: false |
| timeout: 0 |
| seed: null |
| memmap_dtype: uint16 |
| extra_data_paths: null |
| extra_data_key: null |
| load_extra_data_to_ram: false |
| index_path: null |
| device_eval_batch_size: null |
| subset_num_batches: 1000 |
| sft_use_label: false |
| sft: false |
| - label: hellaswag_test |
| type: downstream |
| data: |
| paths: null |
| datasets: null |
| label_mask_paths: null |
| pad_direction: right |
| generate_attention_mask: true |
| num_workers: 0 |
| drop_last: false |
| pin_memory: false |
| prefetch_factor: null |
| persistent_workers: false |
| timeout: 0 |
| seed: null |
| memmap_dtype: uint16 |
| extra_data_paths: null |
| extra_data_key: null |
| load_extra_data_to_ram: false |
| index_path: null |
| device_eval_batch_size: null |
| subset_num_batches: 1000 |
| sft_use_label: false |
| sft: false |
| - label: winogrande_test |
| type: downstream |
| data: |
| paths: null |
| datasets: null |
| label_mask_paths: null |
| pad_direction: right |
| generate_attention_mask: true |
| num_workers: 0 |
| drop_last: false |
| pin_memory: false |
| prefetch_factor: null |
| persistent_workers: false |
| timeout: 0 |
| seed: null |
| memmap_dtype: uint16 |
| extra_data_paths: null |
| extra_data_key: null |
| load_extra_data_to_ram: false |
| index_path: null |
| device_eval_batch_size: null |
| subset_num_batches: 1000 |
| sft_use_label: false |
| sft: false |
| - label: arc_easy_test |
| type: downstream |
| data: |
| paths: null |
| datasets: null |
| label_mask_paths: null |
| pad_direction: right |
| generate_attention_mask: true |
| num_workers: 0 |
| drop_last: false |
| pin_memory: false |
| prefetch_factor: null |
| persistent_workers: false |
| timeout: 0 |
| seed: null |
| memmap_dtype: uint16 |
| extra_data_paths: null |
| extra_data_key: null |
| load_extra_data_to_ram: false |
| index_path: null |
| device_eval_batch_size: null |
| subset_num_batches: 1000 |
| sft_use_label: false |
| sft: false |
| - label: arc_challenge_test |
| type: downstream |
| data: |
| paths: null |
| datasets: null |
| label_mask_paths: null |
| pad_direction: right |
| generate_attention_mask: true |
| num_workers: 0 |
| drop_last: false |
| pin_memory: false |
| prefetch_factor: null |
| persistent_workers: false |
| timeout: 0 |
| seed: null |
| memmap_dtype: uint16 |
| extra_data_paths: null |
| extra_data_key: null |
| load_extra_data_to_ram: false |
| index_path: null |
| device_eval_batch_size: null |
| subset_num_batches: 1000 |
| sft_use_label: false |
| sft: false |
| - label: boolq_test |
| type: downstream |
| data: |
| paths: null |
| datasets: null |
| label_mask_paths: null |
| pad_direction: right |
| generate_attention_mask: true |
| num_workers: 0 |
| drop_last: false |
| pin_memory: false |
| prefetch_factor: null |
| persistent_workers: false |
| timeout: 0 |
| seed: null |
| memmap_dtype: uint16 |
| extra_data_paths: null |
| extra_data_key: null |
| load_extra_data_to_ram: false |
| index_path: null |
| device_eval_batch_size: null |
| subset_num_batches: 1000 |
| sft_use_label: false |
| sft: false |
| - label: sciq_test |
| type: downstream |
| data: |
| paths: null |
| datasets: null |
| label_mask_paths: null |
| pad_direction: right |
| generate_attention_mask: true |
| num_workers: 0 |
| drop_last: false |
| pin_memory: false |
| prefetch_factor: null |
| persistent_workers: false |
| timeout: 0 |
| seed: null |
| memmap_dtype: uint16 |
| extra_data_paths: null |
| extra_data_key: null |
| load_extra_data_to_ram: false |
| index_path: null |
| device_eval_batch_size: null |
| subset_num_batches: 1000 |
| sft_use_label: false |
| sft: false |
| eval_interval: 500 |
| tokenizer: |
| identifier: allenai/gpt-neox-olmo-dolma-v1_5 |
| truncate_direction: right |
| save_folder: ckpts/11816275_1/ |
| remote_save_folder: null |
| canceled_check_interval: 50 |
| save_interval: 5000 |
| save_interval_unsharded: 100000 |
| save_interval_ephemeral: null |
| save_num_checkpoints_to_keep: 1 |
| save_num_unsharded_checkpoints_to_keep: 1 |
| save_overwrite: true |
| force_save_unsharded: false |
| no_pre_train_checkpoint: false |
| load_path: null |
| load_path_sharded_checkpointer: null |
| reset_optimizer_state: false |
| reset_trainer_state: false |
| sharded_checkpointer: torch_legacy |
| new_style_checkpoints: null |
| max_duration: 1ep |
| global_train_batch_size: 256 |
| device_train_batch_size: 256 |
| device_train_microbatch_size: 32 |
| device_eval_batch_size: 8 |
| eval_subset_num_batches: 100 |
| eval_on_load: false |
| device_train_grad_accum: 8 |
| max_grad_norm: 1.0 |
| max_grad_norm_ratio: null |
| precision: amp_bf16 |
| wandb: |
| project: baselines |
| entity: color-exps |
| group: null |
| name: top-3B |
| tags: |
| - watching |
| log_artifacts: false |
| rank_zero_only: true |
| log_interval: 1 |
| speed_monitor: |
| window_size: 1 |
| gpu_flops_available: null |
| console_log_interval: 10 |
| gen1_gc_interval: 1 |
| compile: |
| mode: default |
| fullgraph: false |
| backend: inductor |
| fsdp: |
| use_orig_params: true |
| sharding_strategy: FULL_SHARD |
| wrapping_strategy: null |
| precision: mixed |
| hybrid_sharding_num_model_replicas: null |
| softmax_auxiliary_loss: true |
| time_limit: null |
| extra_steps_after_cancel: 10 |
| early_stopping_factor: null |
| save_data_indices: true |
| python_profiling: false |
| torch_profiling: false |
| stop_at: null |
| stop_after: null |
| activation_checkpointing: null |
| fused_loss: null |
|
|
