Buckets:
| name: safe_v2_2nodes | |
| dump_dir: /fsx/craffel/lingua_logs/safe_v2_2nodes | |
| seed: 777 | |
| grad_acc_steps: 4 | |
| gc_collect_freq: 1000 | |
| probe_freq: null | |
| steps: 100000 | |
| data: | |
| root_dir: /scratch/craffel/lingua/data/flexitok/ | |
| sources: | |
| fw_edu: 0.4 | |
| dan_Latn: 0.0216582869670702 | |
| swe_Latn: 0.0216359765418466 | |
| vie_Latn: 0.0197485510268674 | |
| hun_Latn: 0.0247194573562308 | |
| fas_Arab: 0.0205634624231076 | |
| tur_Latn: 0.0235455794841729 | |
| ces_Latn: 0.0248024455266208 | |
| arb_Arab: 0.0234323706569333 | |
| ell_Grek: 0.0233670886888026 | |
| ind_Latn: 0.0269322054593488 | |
| nld_Latn: 0.0277796326621489 | |
| pol_Latn: 0.0294120104572311 | |
| por_Latn: 0.0301413168306825 | |
| ita_Latn: 0.0324056371021865 | |
| jpn_Jpan: 0.03553104151369 | |
| fra_Latn: 0.0381835560678536 | |
| spa_Latn: 0.0387222793083669 | |
| deu_Latn: 0.0419925340453022 | |
| cmn_Hani: 0.0454067521384114 | |
| rus_Cyrl: 0.0500198157431261 | |
| batch_size: 4 | |
| seq_len: 4096 | |
| n_views: 2 | |
| seed: 42 | |
| add_bos: true | |
| add_eos: true | |
| load_async: true | |
| prefetch_size: 1024 | |
| tokenizer: | |
| name: supertokenizer | |
| path: meta-llama/Llama-3.2-1B | |
| tokenizers: | |
| - name: huggingface | |
| path: flexitok/reindexed-flexitok-bpe_arb_Arab_8000 | |
| load_supermapping: false | |
| - name: huggingface | |
| path: flexitok/reindexed-flexitok-bpe_ces_Latn_8000 | |
| load_supermapping: false | |
| - name: huggingface | |
| path: flexitok/reindexed-flexitok-bpe_ltr_cmn_Hani_8000_v2 | |
| load_supermapping: false | |
| - name: huggingface | |
| path: flexitok/reindexed-flexitok-bpe_dan_Latn_8000 | |
| load_supermapping: false | |
| - name: huggingface | |
| path: flexitok/reindexed-flexitok-bpe_deu_Latn_8000 | |
| load_supermapping: false | |
| - name: huggingface | |
| path: flexitok/reindexed-flexitok-bpe_ell_Grek_8000 | |
| load_supermapping: false | |
| - name: huggingface | |
| path: flexitok/reindexed-flexitok-bpe_fas_Arab_8000 | |
| load_supermapping: false | |
| - name: huggingface | |
| path: flexitok/reindexed-flexitok-bpe_fra_Latn_8000 | |
| load_supermapping: false | |
| - name: huggingface | |
| path: flexitok/reindexed-flexitok-bpe_fw_edu_8000 | |
| load_supermapping: false | |
| - name: huggingface | |
| path: flexitok/reindexed-flexitok-bpe_hun_Latn_8000 | |
| load_supermapping: false | |
| - name: huggingface | |
| path: flexitok/reindexed-flexitok-bpe_ind_Latn_8000 | |
| load_supermapping: false | |
| - name: huggingface | |
| path: flexitok/reindexed-flexitok-bpe_ita_Latn_8000 | |
| load_supermapping: false | |
| - name: huggingface | |
| path: flexitok/reindexed-flexitok-bpe_ltr_jpn_Jpan_8000_v2 | |
| load_supermapping: false | |
| - name: huggingface | |
| path: flexitok/reindexed-flexitok-bpe_nld_Latn_8000 | |
| load_supermapping: false | |
| - name: huggingface | |
| path: flexitok/reindexed-flexitok-bpe_pol_Latn_8000 | |
| load_supermapping: false | |
| - name: huggingface | |
| path: flexitok/reindexed-flexitok-bpe_por_Latn_8000 | |
| load_supermapping: false | |
| - name: huggingface | |
| path: flexitok/reindexed-flexitok-bpe_rus_Cyrl_8000 | |
| load_supermapping: false | |
| - name: huggingface | |
| path: flexitok/reindexed-flexitok-bpe_spa_Latn_8000 | |
| load_supermapping: false | |
| - name: huggingface | |
| path: flexitok/reindexed-flexitok-bpe_swe_Latn_8000 | |
| load_supermapping: false | |
| - name: huggingface | |
| path: flexitok/reindexed-flexitok-bpe_ltr_tur_Latn_8000_v2 | |
| load_supermapping: false | |
| - name: huggingface | |
| path: flexitok/reindexed-flexitok-bpe_vie_Latn_8000 | |
| load_supermapping: false | |
| load_supermapping: false | |
| dropout: 0.0 | |
| seed: 42 | |
| superset_code_name: safe_v2 | |
| n_words: 122554 | |
| routing: | |
| source_to_tokenizer: | |
| arb_Arab: flexitok/reindexed-flexitok-bpe_arb_Arab_8000 | |
| ces_Latn: flexitok/reindexed-flexitok-bpe_ces_Latn_8000 | |
| cmn_Hani: flexitok/reindexed-flexitok-bpe_ltr_cmn_Hani_8000 | |
| dan_Latn: flexitok/reindexed-flexitok-bpe_dan_Latn_8000 | |
| deu_Latn: flexitok/reindexed-flexitok-bpe_deu_Latn_8000 | |
| ell_Grek: flexitok/reindexed-flexitok-bpe_ell_Grek_8000 | |
| fas_Arab: flexitok/reindexed-flexitok-bpe_fas_Arab_8000 | |
| fra_Latn: flexitok/reindexed-flexitok-bpe_fra_Latn_8000 | |
| fw_edu: flexitok/reindexed-flexitok-bpe_fw_edu_8000 | |
| hun_Latn: flexitok/reindexed-flexitok-bpe_hun_Latn_8000 | |
| ind_Latn: flexitok/reindexed-flexitok-bpe_ind_Latn_8000 | |
| ita_Latn: flexitok/reindexed-flexitok-bpe_ita_Latn_8000 | |
| jpn_Jpan: flexitok/reindexed-flexitok-bpe_ltr_jpn_Jpan_8000_v2 | |
| nld_Latn: flexitok/reindexed-flexitok-bpe_nld_Latn_8000 | |
| pol_Latn: flexitok/reindexed-flexitok-bpe_pol_Latn_8000 | |
| por_Latn: flexitok/reindexed-flexitok-bpe_por_Latn_8000 | |
| rus_Cyrl: flexitok/reindexed-flexitok-bpe_rus_Cyrl_8000 | |
| spa_Latn: flexitok/reindexed-flexitok-bpe_spa_Latn_8000 | |
| swe_Latn: flexitok/reindexed-flexitok-bpe_swe_Latn_8000 | |
| tur_Latn: flexitok/reindexed-flexitok-bpe_ltr_tur_Latn_8000_v2 | |
| vie_Latn: flexitok/reindexed-flexitok-bpe_vie_Latn_8000 | |
| task_to_tokenizer: {} | |
| suitable_tokenizer_probability: 1.0 | |
| optim: | |
| lr: 0.001 | |
| weight_decay: 0.1 | |
| epsilon: 1.0e-08 | |
| beta1: 0.9 | |
| beta2: 0.95 | |
| clip: 1.0 | |
| scheduler: cosine | |
| warmup: 2000 | |
| lr_min_ratio: 1.0e-06 | |
| cycle_length: 1.0 | |
| cosine_theta: 1.0 | |
| annealing_step: 1000 | |
| decay_fraction: 0.1 | |
| exp_factor: 0.5 | |
| model: | |
| dim: 2048 | |
| n_layers: 25 | |
| head_dim: null | |
| n_heads: 16 | |
| n_kv_heads: null | |
| ffn_dim_multiplier: null | |
| multiple_of: 256 | |
| norm_eps: 1.0e-05 | |
| rope_theta: 10000.0 | |
| init_base_std: null | |
| init_std_factor: disabled | |
| max_seqlen: 4096 | |
| seed: 42 | |
| vocab_size: 122554 | |
| weight_tying: false | |
| sliding_window: null | |
| use_factorized_embeddings: false | |
| factorized_embedding_dim: 0 | |
| distributed: | |
| dp_shard: 1 | |
| dp_replicate: 16 | |
| tp_size: 1 | |
| selective_activation_checkpointing: false | |
| compile: true | |
| fsdp_type: full_shard | |
| model_dtype: bf16 | |
| float8_recipe: null | |
| float8_filter: layers\.[0-9]+\. | |
| matmul_allow_tf32: false | |
| detect_anomaly: false | |
| compile_cache_size_limit: 8 | |
| spawn_method: forkserver | |
| env: | |
| MKL_SERVICE_FORCE_INTEL: GNU | |
| OMP_NUM_THREADS: '1' | |
| MKL_NUM_THREADS: '1' | |
| ENABLE_INTRA_NODE_COMM: '1' | |
| TORCH_NCCL_AVOID_RECORD_STREAMS: '1' | |
| NCCL_IB_TIMEOUT: '22' | |
| NCCL_DEBUG: INFO | |
| TORCH_NCCL_ASYNC_ERROR_HANDLING: '1' | |
| checkpoint: | |
| dump: | |
| every: 2500 | |
| keep: -1 | |
| eval: | |
| every: 2500 | |
| keep: -1 | |
| path: /fsx/craffel/lingua_logs/checkpoints/safe_v2_2nodes | |
| init_ckpt_path: null | |
| load_init_optimizer_state: false | |
| save_init_ckpt: false | |
| profiling: | |
| run: true | |
| trace_folder: profiling | |
| mem_warmup: 0 | |
| mem_steps: 4 | |
| profile_warmup: 100 | |
| profile_steps: 4 | |
| logging: | |
| freq: 1 | |
| acc_freq: null | |
| wandb: | |
| job_type: null | |
| dir: null | |
| project: MoTo-Training | |
| entity: raffel-reports | |
| tags: null | |
| group: null | |
| name: safe_v2_2nodes | |
| notes: null | |
| config_exclude_keys: null | |
| config_include_keys: null | |
| anonymous: null | |
| mode: null | |
| allow_val_change: null | |
| resume: null | |
| force: null | |
| tensorboard: null | |
| sync_tensorboard: null | |
| monitor_gym: null | |
| save_code: null | |
| id: null | |
| fork_from: null | |
| resume_from: null | |
| async_eval_gpus: null | |
| eval: | |
| harness: | |
| tasks: | |
| - hellaswag | |
| - piqa | |
| - arc_easy | |
| - arc_challenge | |
| - xnli_ar | |
| - xnli_zh | |
| - xnli_de | |
| - xnli_el | |
| - xnli_en | |
| - xnli_es | |
| - xnli_fr | |
| - xnli_ru | |
| - xnli_tr | |
| - xnli_vi | |
| - toksuite_english | |
| - toksuite_chinese | |
| - toksuite_turkish | |
| - toksuite_italian | |
| - toksuite_farsi | |
| - toksuite_math | |
| - toksuite_stem | |
| - xcopa_id | |
| - xcopa_it | |
| - xcopa_tr | |
| - xcopa_vi | |
| - xcopa_zh | |
| - xstorycloze_ar | |
| - xstorycloze_en | |
| - xstorycloze_es | |
| - xstorycloze_id | |
| - xstorycloze_ru | |
| - xstorycloze_zh | |
| - paws_de | |
| - paws_en | |
| - paws_es | |
| - paws_fr | |
| - paws_ja | |
| - paws_zh | |
| - hellaswag_ar | |
| - hellaswag_da | |
| - hellaswag_de | |
| - hellaswag_fr | |
| - hellaswag_hu | |
| - hellaswag_id | |
| - hellaswag_it | |
| - hellaswag_nl | |
| - hellaswag_pt | |
| - hellaswag_ru | |
| - hellaswag_sv | |
| - hellaswag_vi | |
| - hellaswag_es | |
| generator: | |
| max_tokens: 16384 | |
| dtype: bf16 | |
| add_bos: false | |
| routing: | |
| suitable_tokenizer_probability: 1.0 | |
| task_to_tokenizer: | |
| arc_easy: flexitok/reindexed-flexitok-bpe_fw_edu_8000 | |
| arc_challenge: flexitok/reindexed-flexitok-bpe_fw_edu_8000 | |
| hellaswag: flexitok/reindexed-flexitok-bpe_fw_edu_8000 | |
| piqa: flexitok/reindexed-flexitok-bpe_fw_edu_8000 | |
| include_base_44_arabic: flexitok/reindexed-flexitok-bpe_arb_Arab_8000 | |
| include_base_44_chinese: flexitok/reindexed-flexitok-bpe_ltr_cmn_Hani_8000 | |
| include_base_44_german: flexitok/reindexed-flexitok-bpe_deu_Latn_8000 | |
| include_base_44_greek: flexitok/reindexed-flexitok-bpe_ell_Grek_8000 | |
| include_base_44_persian: flexitok/reindexed-flexitok-bpe_fas_Arab_8000 | |
| include_base_44_french: flexitok/reindexed-flexitok-bpe_fra_Latn_8000 | |
| include_base_44_hungarian: flexitok/reindexed-flexitok-bpe_hun_Latn_8000 | |
| include_base_44_indonesian: flexitok/reindexed-flexitok-bpe_ind_Latn_8000 | |
| include_base_44_italian: flexitok/reindexed-flexitok-bpe_ita_Latn_8000 | |
| include_base_44_japanese: flexitok/reindexed-flexitok-bpe_ltr_jpn_Jpan_8000_v2 | |
| include_base_44_dutch: flexitok/reindexed-flexitok-bpe_nld_Latn_8000 | |
| include_base_44_portuguese: flexitok/reindexed-flexitok-bpe_por_Latn_8000 | |
| include_base_44_russian: flexitok/reindexed-flexitok-bpe_rus_Cyrl_8000 | |
| include_base_44_spanish: flexitok/reindexed-flexitok-bpe_spa_Latn_8000 | |
| include_base_44_turkish: flexitok/reindexed-flexitok-bpe_ltr_tur_Latn_8000_v2 | |
| include_base_44_vietnamese: flexitok/reindexed-flexitok-bpe_vie_Latn_8000 | |
| belebele_arb_Arab: flexitok/reindexed-flexitok-bpe_arb_Arab_8000 | |
| belebele_ces_Latn: flexitok/reindexed-flexitok-bpe_ces_Latn_8000 | |
| belebele_zho_Hans: flexitok/reindexed-flexitok-bpe_ltr_cmn_Hani_8000 | |
| belebele_dan_Latn: flexitok/reindexed-flexitok-bpe_dan_Latn_8000 | |
| belebele_deu_Latn: flexitok/reindexed-flexitok-bpe_deu_Latn_8000 | |
| belebele_ell_Grek: flexitok/reindexed-flexitok-bpe_ell_Grek_8000 | |
| belebele_pes_Arab: flexitok/reindexed-flexitok-bpe_fas_Arab_8000 | |
| belebele_fra_Latn: flexitok/reindexed-flexitok-bpe_fra_Latn_8000 | |
| belebele_hun_Latn: flexitok/reindexed-flexitok-bpe_hun_Latn_8000 | |
| belebele_ind_Latn: flexitok/reindexed-flexitok-bpe_ind_Latn_8000 | |
| belebele_ita_Latn: flexitok/reindexed-flexitok-bpe_ita_Latn_8000 | |
| belebele_jpn_Jpan: flexitok/reindexed-flexitok-bpe_ltr_jpn_Jpan_8000_v2 | |
| belebele_nld_Latn: flexitok/reindexed-flexitok-bpe_nld_Latn_8000 | |
| belebele_pol_Latn: flexitok/reindexed-flexitok-bpe_pol_Latn_8000 | |
| belebele_por_Latn: flexitok/reindexed-flexitok-bpe_por_Latn_8000 | |
| belebele_rus_Cyrl: flexitok/reindexed-flexitok-bpe_rus_Cyrl_8000 | |
| belebele_spa_Latn: flexitok/reindexed-flexitok-bpe_spa_Latn_8000 | |
| belebele_swe_Latn: flexitok/reindexed-flexitok-bpe_swe_Latn_8000 | |
| belebele_tur_Latn: flexitok/reindexed-flexitok-bpe_ltr_tur_Latn_8000_v2 | |
| belebele_vie_Latn: flexitok/reindexed-flexitok-bpe_vie_Latn_8000 | |
| belebele_eng_Latn: flexitok/reindexed-flexitok-bpe_fw_edu_8000 | |
| xnli_ar: flexitok/reindexed-flexitok-bpe_arb_Arab_8000 | |
| xnli_zh: flexitok/reindexed-flexitok-bpe_ltr_cmn_Hani_8000 | |
| xnli_de: flexitok/reindexed-flexitok-bpe_deu_Latn_8000 | |
| xnli_el: flexitok/reindexed-flexitok-bpe_ell_Grek_8000 | |
| xnli_en: flexitok/reindexed-flexitok-bpe_fw_edu_8000 | |
| xnli_es: flexitok/reindexed-flexitok-bpe_spa_Latn_8000 | |
| xnli_fr: flexitok/reindexed-flexitok-bpe_fra_Latn_8000 | |
| xnli_ru: flexitok/reindexed-flexitok-bpe_rus_Cyrl_8000 | |
| xnli_tr: flexitok/reindexed-flexitok-bpe_ltr_tur_Latn_8000_v2 | |
| xnli_vi: flexitok/reindexed-flexitok-bpe_vie_Latn_8000 | |
| regex:toksuite_turkish*: flexitok/reindexed-flexitok-bpe_ltr_tur_Latn_8000_v2 | |
| regex:toksuite_english*: flexitok/reindexed-flexitok-bpe_fw_edu_8000 | |
| regex:toksuite_farsi*: flexitok/reindexed-flexitok-bpe_fas_Arab_8000 | |
| regex:toksuite_italian*: flexitok/reindexed-flexitok-bpe_ita_Latn_8000 | |
| regex:toksuite_chinese*: flexitok/reindexed-flexitok-bpe_ltr_cmn_Hani_8000 | |
| regex:toksuite_math*: flexitok/reindexed-flexitok-bpe_fw_edu_8000 | |
| regex:toksuite_stem*: flexitok/reindexed-flexitok-bpe_fw_edu_8000 | |
| flores200_arb_Arab: flexitok/reindexed-flexitok-bpe_arb_Arab_8000 | |
| flores200_ces_Latn: flexitok/reindexed-flexitok-bpe_ces_Latn_8000 | |
| flores200_cmn_Hani: flexitok/reindexed-flexitok-bpe_ltr_cmn_Hani_8000 | |
| flores200_dan_Latn: flexitok/reindexed-flexitok-bpe_dan_Latn_8000 | |
| flores200_deu_Latn: flexitok/reindexed-flexitok-bpe_deu_Latn_8000 | |
| flores200_ell_Grek: flexitok/reindexed-flexitok-bpe_ell_Grek_8000 | |
| flores200_fas_Arab: flexitok/reindexed-flexitok-bpe_fas_Arab_8000 | |
| flores200_fra_Latn: flexitok/reindexed-flexitok-bpe_fra_Latn_8000 | |
| flores200_fw_edu: flexitok/reindexed-flexitok-bpe_fw_edu_8000 | |
| flores200_eng_Latn: flexitok/reindexed-flexitok-bpe_fw_edu_8000 | |
| flores200_hun_Latn: flexitok/reindexed-flexitok-bpe_hun_Latn_8000 | |
| flores200_ind_Latn: flexitok/reindexed-flexitok-bpe_ind_Latn_8000 | |
| flores200_ita_Latn: flexitok/reindexed-flexitok-bpe_ita_Latn_8000 | |
| flores200_jpn_Jpan: flexitok/reindexed-flexitok-bpe_ltr_jpn_Jpan_8000_v2 | |
| flores200_nld_Latn: flexitok/reindexed-flexitok-bpe_nld_Latn_8000 | |
| flores200_pol_Latn: flexitok/reindexed-flexitok-bpe_pol_Latn_8000 | |
| flores200_por_Latn: flexitok/reindexed-flexitok-bpe_por_Latn_8000 | |
| flores200_rus_Cyrl: flexitok/reindexed-flexitok-bpe_rus_Cyrl_8000 | |
| flores200_spa_Latn: flexitok/reindexed-flexitok-bpe_spa_Latn_8000 | |
| flores200_swe_Latn: flexitok/reindexed-flexitok-bpe_swe_Latn_8000 | |
| flores200_tur_Latn: flexitok/reindexed-flexitok-bpe_ltr_tur_Latn_8000_v2 | |
| flores200_vie_Latn: flexitok/reindexed-flexitok-bpe_vie_Latn_8000 | |
| xcopa_id: flexitok/reindexed-flexitok-bpe_ind_Latn_8000 | |
| xcopa_it: flexitok/reindexed-flexitok-bpe_ita_Latn_8000 | |
| xcopa_tr: flexitok/reindexed-flexitok-bpe_ltr_tur_Latn_8000_v2 | |
| xcopa_vi: flexitok/reindexed-flexitok-bpe_vie_Latn_8000 | |
| xcopa_zh: flexitok/reindexed-flexitok-bpe_ltr_cmn_Hani_8000 | |
| xstorycloze_ar: flexitok/reindexed-flexitok-bpe_arb_Arab_8000 | |
| xstorycloze_en: flexitok/reindexed-flexitok-bpe_fw_edu_8000 | |
| xstorycloze_es: flexitok/reindexed-flexitok-bpe_spa_Latn_8000 | |
| xstorycloze_id: flexitok/reindexed-flexitok-bpe_ind_Latn_8000 | |
| xstorycloze_ru: flexitok/reindexed-flexitok-bpe_rus_Cyrl_8000 | |
| xstorycloze_zh: flexitok/reindexed-flexitok-bpe_ltr_cmn_Hani_8000 | |
| paws_de: flexitok/reindexed-flexitok-bpe_deu_Latn_8000 | |
| paws_en: flexitok/reindexed-flexitok-bpe_fw_edu_8000 | |
| paws_es: flexitok/reindexed-flexitok-bpe_spa_Latn_8000 | |
| paws_fr: flexitok/reindexed-flexitok-bpe_fra_Latn_8000 | |
| paws_ja: flexitok/reindexed-flexitok-bpe_ltr_jpn_Jpan_8000_v2 | |
| paws_zh: flexitok/reindexed-flexitok-bpe_ltr_cmn_Hani_8000 | |
| hellaswag_ar: flexitok/reindexed-flexitok-bpe_arb_Arab_8000 | |
| hellaswag_da: flexitok/reindexed-flexitok-bpe_dan_Latn_8000 | |
| hellaswag_de: flexitok/reindexed-flexitok-bpe_deu_Latn_8000 | |
| hellaswag_fr: flexitok/reindexed-flexitok-bpe_fra_Latn_8000 | |
| hellaswag_hu: flexitok/reindexed-flexitok-bpe_hun_Latn_8000 | |
| hellaswag_id: flexitok/reindexed-flexitok-bpe_ind_Latn_8000 | |
| hellaswag_it: flexitok/reindexed-flexitok-bpe_ita_Latn_8000 | |
| hellaswag_nl: flexitok/reindexed-flexitok-bpe_nld_Latn_8000 | |
| hellaswag_pt: flexitok/reindexed-flexitok-bpe_por_Latn_8000 | |
| hellaswag_ru: flexitok/reindexed-flexitok-bpe_rus_Cyrl_8000 | |
| hellaswag_sv: flexitok/reindexed-flexitok-bpe_swe_Latn_8000 | |
| hellaswag_vi: flexitok/reindexed-flexitok-bpe_vie_Latn_8000 | |
| hellaswag_es: flexitok/reindexed-flexitok-bpe_spa_Latn_8000 | |
Xet Storage Details
- Size:
- 15.9 kB
- Xet hash:
- 9fe00b19970a2bbc6b59ae3dee6a681169aa22a7ce649e988cfc071ee9a50ccf
·
Xet efficiently stores files, intelligently splitting them into unique chunks and accelerating uploads and downloads. More info.