Upload folder using huggingface_hub

.gitattributes

@@ -243,3 +243,12 @@ bpe_on_all_boundless/0000100000/__5_0.distcp filter=lfs diff=lfs merge=lfs -text
 bpe_on_all_boundless/0000100000/__6_0.distcp filter=lfs diff=lfs merge=lfs -text
 bpe_on_all_boundless/0000100000/__7_0.distcp filter=lfs diff=lfs merge=lfs -text
 bpe_on_all_gpt4_323k/metrics.jsonl filter=lfs diff=lfs merge=lfs -text
+bpe_on_all_gpt4_323k/0000100000/.metadata filter=lfs diff=lfs merge=lfs -text
+bpe_on_all_gpt4_323k/0000100000/__0_0.distcp filter=lfs diff=lfs merge=lfs -text
+bpe_on_all_gpt4_323k/0000100000/__1_0.distcp filter=lfs diff=lfs merge=lfs -text
+bpe_on_all_gpt4_323k/0000100000/__2_0.distcp filter=lfs diff=lfs merge=lfs -text
+bpe_on_all_gpt4_323k/0000100000/__3_0.distcp filter=lfs diff=lfs merge=lfs -text
+bpe_on_all_gpt4_323k/0000100000/__4_0.distcp filter=lfs diff=lfs merge=lfs -text
+bpe_on_all_gpt4_323k/0000100000/__5_0.distcp filter=lfs diff=lfs merge=lfs -text
+bpe_on_all_gpt4_323k/0000100000/__6_0.distcp filter=lfs diff=lfs merge=lfs -text
+bpe_on_all_gpt4_323k/0000100000/__7_0.distcp filter=lfs diff=lfs merge=lfs -text

bpe_on_all_gpt4_323k/0000100000/.metadata

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5913437e6fbaf08a03db1f3c650b698f0e11fedbe265264691d132a1e9f04d88
+size 1148646

bpe_on_all_gpt4_323k/0000100000/__0_0.distcp

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:871cbcfd9706c721482452e43a6e526fe5954b82fa90acca245c6fb90daf388b
+size 3912283344

bpe_on_all_gpt4_323k/0000100000/__1_0.distcp

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:189139667c0f37dabeb7a12be1bbcdc1823ac91f13456c20a5c5ad64d7f09447
+size 3912322284

bpe_on_all_gpt4_323k/0000100000/__2_0.distcp

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4b74710ed5fce70076d1a808c0aab55755e573d7853ee1a93b6963511d64d7f5
+size 3912322284

bpe_on_all_gpt4_323k/0000100000/__3_0.distcp

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:42646f70ae3e775d325b9b3d608322a188ecb31a719c2d16967b959662c102fe
+size 3912322284

bpe_on_all_gpt4_323k/0000100000/__4_0.distcp

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8780446dde8909164772c3977b99e4ddb9d11840c4eece54f56e656631c2c02f
+size 3912322284

bpe_on_all_gpt4_323k/0000100000/__5_0.distcp

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e2998c32c15958109cff2063c191b9b0404f83a41bb7acda90fc1ec95170aa0a
+size 3912324560

bpe_on_all_gpt4_323k/0000100000/__6_0.distcp

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ee7a91fbe397d5778b5820b4482e1fe3dec24a10aadcc3b28548ac4439867d70
+size 3912324560

bpe_on_all_gpt4_323k/0000100000/__7_0.distcp

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9fc94ee1c903017c6915a6205612dfa35bebc6d5d7043c99820d15bcee95f451
+size 3912136208

bpe_on_all_gpt4_323k/0000100000/params.json

@@ -0,0 +1 @@
+{"name": "bpe_on_all_gpt4_323k", "dump_dir": "/fsx/craffel/lingua_logs/bpe_on_all_gpt4_323k", "seed": 777, "grad_acc_steps": 8, "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": "huggingface", "path": "flexitok/bpe_gpt4_AllL_323000", "tokenizers": null, "load_supermapping": false, "dropout": 0.0, "seed": 42, "superset_code_name": "super_vocab", "n_words": null}, "routing": {"source_to_tokenizer": {}, "task_to_tokenizer": {}, "suitable_tokenizer_probability": 1.0}}, "optim": {"lr": 0.001, "weight_decay": 0.1, "epsilon": 1e-08, "beta1": 0.9, "beta2": 0.95, "clip": 1.0, "scheduler": "cosine", "warmup": 2000, "lr_min_ratio": 1e-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": 1e-05, "rope_theta": 10000.0, "init_base_std": null, "init_std_factor": "disabled", "max_seqlen": 4096, "seed": 42, "vocab_size": 323004, "weight_tying": false, "sliding_window": null, "use_factorized_embeddings": false, "factorized_embedding_dim": 0}, "distributed": {"dp_shard": 1, "dp_replicate": 8, "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": 10000, "keep": -1}, "eval": {"every": 10000, "keep": -1}, "path": "/fsx/craffel/lingua_logs/checkpoints/bpe_on_all_gpt4_323k/", "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": "bpe_on_all_gpt4_323k", "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": 8, "eval": {"harness": {"tasks": ["hellaswag", "piqa", "arc_easy", "arc_challenge", "include_base_44_arabic", "include_base_44_chinese", "include_base_44_german", "include_base_44_greek", "include_base_44_persian", "include_base_44_french", "include_base_44_hungarian", "include_base_44_indonesian", "include_base_44_italian", "include_base_44_japanese", "include_base_44_dutch", "include_base_44_polish", "include_base_44_portuguese", "include_base_44_russian", "include_base_44_spanish", "include_base_44_turkish", "include_base_44_vietnamese", "belebele_arb_Arab", "belebele_ces_Latn", "belebele_zho_Hans", "belebele_dan_Latn", "belebele_deu_Latn", "belebele_ell_Grek", "belebele_pes_Arab", "belebele_fra_Latn", "belebele_hun_Latn", "belebele_ind_Latn", "belebele_ita_Latn", "belebele_jpn_Jpan", "belebele_nld_Latn", "belebele_pol_Latn", "belebele_por_Latn", "belebele_rus_Cyrl", "belebele_spa_Latn", "belebele_swe_Latn", "belebele_tur_Latn", "belebele_vie_Latn", "belebele_eng_Latn", "xnli_ar", "xnli_zh", "xnli_de", "xnli_el", "xnli_en", "xnli_es", "xnli_fr", "xnli_ru", "xnli_tr", "xnli_vi"]}, "generator": {"max_tokens": 16384, "dtype": "bf16", "add_bos": false}}}

bpe_on_all_gpt4_323k/0000100000/train_state_00000.json

@@ -0,0 +1 @@
+{"step": 100000, "acc_step": 0, "data_loader_state": {"it_state": {"start_token": 4558, "it_state": {"it_state": {"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}, "source_to_state": {"fw_edu": {"file_path": "/scratch/craffel/lingua/data/flexitok/fw_edu/fineweb_edu_100bt.chunk.00.jsonl", "position": 30873665159, "block_size": 1, "offset": 0, "current_iter": 0}, "dan_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/dan_Latn/fineweb_2_hq.dan_Latn.chunk.00.jsonl", "position": 1413769622, "block_size": 1, "offset": 0, "current_iter": 0}, "swe_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/swe_Latn/fineweb_2_hq.swe_Latn.chunk.00.jsonl", "position": 1331284590, "block_size": 1, "offset": 0, "current_iter": 0}, "vie_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/vie_Latn/fineweb_2_hq.vie_Latn.chunk.00.jsonl", "position": 1420998189, "block_size": 1, "offset": 0, "current_iter": 0}, "hun_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/hun_Latn/fineweb_2_hq.hun_Latn.chunk.00.jsonl", "position": 1887743706, "block_size": 1, "offset": 0, "current_iter": 0}, "fas_Arab": {"file_path": "/scratch/craffel/lingua/data/flexitok/fas_Arab/fineweb_2_hq.fas_Arab.chunk.00.jsonl", "position": 1587799986, "block_size": 1, "offset": 0, "current_iter": 0}, "tur_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/tur_Latn/fineweb_2_hq.tur_Latn.chunk.00.jsonl", "position": 1206324533, "block_size": 1, "offset": 0, "current_iter": 0}, "ces_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/ces_Latn/fineweb_2_hq.ces_Latn.chunk.00.jsonl", "position": 1849225410, "block_size": 1, "offset": 0, "current_iter": 0}, "arb_Arab": {"file_path": "/scratch/craffel/lingua/data/flexitok/arb_Arab/fineweb_2_hq.arb_Arab.chunk.00.jsonl", "position": 2081853145, "block_size": 1, "offset": 0, "current_iter": 0}, "ell_Grek": {"file_path": "/scratch/craffel/lingua/data/flexitok/ell_Grek/fineweb_2_hq.ell_Grek.chunk.00.jsonl", "position": 2586188392, "block_size": 1, "offset": 0, "current_iter": 0}, "ind_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/ind_Latn/fineweb_2_hq.ind_Latn.chunk.00.jsonl", "position": 1837980591, "block_size": 1, "offset": 0, "current_iter": 0}, "nld_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/nld_Latn/fineweb_2_hq.nld_Latn.chunk.00.jsonl", "position": 1460997923, "block_size": 1, "offset": 0, "current_iter": 0}, "pol_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/pol_Latn/fineweb_2_hq.pol_Latn.chunk.00.jsonl", "position": 1538241097, "block_size": 1, "offset": 0, "current_iter": 0}, "por_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/por_Latn/fineweb_2_hq.por_Latn.chunk.00.jsonl", "position": 1593300151, "block_size": 1, "offset": 0, "current_iter": 0}, "ita_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/ita_Latn/fineweb_2_hq.ita_Latn.chunk.00.jsonl", "position": 1784042325, "block_size": 1, "offset": 0, "current_iter": 0}, "jpn_Jpan": {"file_path": "/scratch/craffel/lingua/data/flexitok/jpn_Jpan/fineweb_2_hq.jpn_Jpan.chunk.00.jsonl", "position": 1868640801, "block_size": 1, "offset": 0, "current_iter": 0}, "fra_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/fra_Latn/fineweb_2_hq.fra_Latn.chunk.00.jsonl", "position": 2478386814, "block_size": 1, "offset": 0, "current_iter": 0}, "spa_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/spa_Latn/fineweb_2_hq.spa_Latn.chunk.00.jsonl", "position": 2219527004, "block_size": 1, "offset": 0, "current_iter": 0}, "deu_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/deu_Latn/fineweb_2_hq.deu_Latn.chunk.00.jsonl", "position": 2704859897, "block_size": 1, "offset": 0, "current_iter": 0}, "cmn_Hani": {"file_path": "/scratch/craffel/lingua/data/flexitok/cmn_Hani/fineweb_2_hq.cmn_Hani.chunk.00.jsonl", "position": 2865415525, "block_size": 1, "offset": 0, "current_iter": 0}, "rus_Cyrl": {"file_path": "/scratch/craffel/lingua/data/flexitok/rus_Cyrl/fineweb_2_hq.rus_Cyrl.chunk.00.jsonl", "position": 6728112330, "block_size": 1, "offset": 0, "current_iter": 0}}, "rng_state": {"bit_generator": "PCG64", "state": {"state": 320801901391341645039653501565138710593, "inc": 252101603063402394885084957393789173453}, "has_uint32": 0, "uinteger": 0}}, "add_bos": true, "add_eos": true, "name": "huggingface", "path": "flexitok/bpe_gpt4_AllL_323000", "tokenizers": null, "dropout": 0.0, "rng_state": null, "seed": 42, "superset_code_name": "super_vocab", "n_words": null, "routing": {"source_to_tokenizer": {}, "suitable_tokenizer_probability": 1.0}}, "output_seq_len": 4096, "n_views": 2}, "seq_idx": 256, "rng_state": {"bit_generator": "PCG64", "state": {"state": 300569569296678341640414112158566886944, "inc": 257317082376085721142933171929815648017}, "has_uint32": 1, "uinteger": 630457105}, "batch_size": 4, "prefetch_size": 1024}, "scheduler": {"base_lrs": [0.001], "last_epoch": 100000, "verbose": false, "_step_count": 100001, "_get_lr_called_within_step": false, "_last_lr": [0.001], "lr_lambdas": [{}]}}

bpe_on_all_gpt4_323k/0000100000/train_state_00001.json

@@ -0,0 +1 @@
+{"step": 100000, "acc_step": 0, "data_loader_state": {"it_state": {"start_token": 216, "it_state": {"it_state": {"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}, "source_to_state": {"fw_edu": {"file_path": "/scratch/craffel/lingua/data/flexitok/fw_edu/fineweb_edu_100bt.chunk.01.jsonl", "position": 30874218619, "block_size": 1, "offset": 0, "current_iter": 0}, "dan_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/dan_Latn/fineweb_2_hq.dan_Latn.chunk.01.jsonl", "position": 1412252032, "block_size": 1, "offset": 0, "current_iter": 0}, "swe_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/swe_Latn/fineweb_2_hq.swe_Latn.chunk.01.jsonl", "position": 1343453190, "block_size": 1, "offset": 0, "current_iter": 0}, "vie_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/vie_Latn/fineweb_2_hq.vie_Latn.chunk.01.jsonl", "position": 1421438048, "block_size": 1, "offset": 0, "current_iter": 0}, "hun_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/hun_Latn/fineweb_2_hq.hun_Latn.chunk.01.jsonl", "position": 1878912804, "block_size": 1, "offset": 0, "current_iter": 0}, "fas_Arab": {"file_path": "/scratch/craffel/lingua/data/flexitok/fas_Arab/fineweb_2_hq.fas_Arab.chunk.01.jsonl", "position": 1592367737, "block_size": 1, "offset": 0, "current_iter": 0}, "tur_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/tur_Latn/fineweb_2_hq.tur_Latn.chunk.01.jsonl", "position": 1199601856, "block_size": 1, "offset": 0, "current_iter": 0}, "ces_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/ces_Latn/fineweb_2_hq.ces_Latn.chunk.01.jsonl", "position": 1863966396, "block_size": 1, "offset": 0, "current_iter": 0}, "arb_Arab": {"file_path": "/scratch/craffel/lingua/data/flexitok/arb_Arab/fineweb_2_hq.arb_Arab.chunk.01.jsonl", "position": 2082801885, "block_size": 1, "offset": 0, "current_iter": 0}, "ell_Grek": {"file_path": "/scratch/craffel/lingua/data/flexitok/ell_Grek/fineweb_2_hq.ell_Grek.chunk.01.jsonl", "position": 2600936144, "block_size": 1, "offset": 0, "current_iter": 0}, "ind_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/ind_Latn/fineweb_2_hq.ind_Latn.chunk.01.jsonl", "position": 1824839411, "block_size": 1, "offset": 0, "current_iter": 0}, "nld_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/nld_Latn/fineweb_2_hq.nld_Latn.chunk.01.jsonl", "position": 1461150483, "block_size": 1, "offset": 0, "current_iter": 0}, "pol_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/pol_Latn/fineweb_2_hq.pol_Latn.chunk.01.jsonl", "position": 1541443044, "block_size": 1, "offset": 0, "current_iter": 0}, "por_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/por_Latn/fineweb_2_hq.por_Latn.chunk.01.jsonl", "position": 1606946898, "block_size": 1, "offset": 0, "current_iter": 0}, "ita_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/ita_Latn/fineweb_2_hq.ita_Latn.chunk.01.jsonl", "position": 1783509219, "block_size": 1, "offset": 0, "current_iter": 0}, "jpn_Jpan": {"file_path": "/scratch/craffel/lingua/data/flexitok/jpn_Jpan/fineweb_2_hq.jpn_Jpan.chunk.01.jsonl", "position": 1861452573, "block_size": 1, "offset": 0, "current_iter": 0}, "fra_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/fra_Latn/fineweb_2_hq.fra_Latn.chunk.01.jsonl", "position": 2480683313, "block_size": 1, "offset": 0, "current_iter": 0}, "spa_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/spa_Latn/fineweb_2_hq.spa_Latn.chunk.01.jsonl", "position": 2241739182, "block_size": 1, "offset": 0, "current_iter": 0}, "deu_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/deu_Latn/fineweb_2_hq.deu_Latn.chunk.01.jsonl", "position": 2699139173, "block_size": 1, "offset": 0, "current_iter": 0}, "cmn_Hani": {"file_path": "/scratch/craffel/lingua/data/flexitok/cmn_Hani/fineweb_2_hq.cmn_Hani.chunk.01.jsonl", "position": 2866238226, "block_size": 1, "offset": 0, "current_iter": 0}, "rus_Cyrl": {"file_path": "/scratch/craffel/lingua/data/flexitok/rus_Cyrl/fineweb_2_hq.rus_Cyrl.chunk.01.jsonl", "position": 6675742332, "block_size": 1, "offset": 0, "current_iter": 0}}, "rng_state": {"bit_generator": "PCG64", "state": {"state": 287379513783510908174434403585955888849, "inc": 246509925186285949978196491240064802315}, "has_uint32": 0, "uinteger": 0}}, "add_bos": true, "add_eos": true, "name": "huggingface", "path": "flexitok/bpe_gpt4_AllL_323000", "tokenizers": null, "dropout": 0.0, "rng_state": null, "seed": 42, "superset_code_name": "super_vocab", "n_words": null, "routing": {"source_to_tokenizer": {}, "suitable_tokenizer_probability": 1.0}}, "output_seq_len": 4096, "n_views": 2}, "seq_idx": 256, "rng_state": {"bit_generator": "PCG64", "state": {"state": 298609680158430271867266436931522339460, "inc": 173555323965545256606922338259303677603}, "has_uint32": 1, "uinteger": 1182378492}, "batch_size": 4, "prefetch_size": 1024}, "scheduler": {"base_lrs": [0.001], "last_epoch": 100000, "verbose": false, "_step_count": 100001, "_get_lr_called_within_step": false, "_last_lr": [0.001], "lr_lambdas": [{}]}}

bpe_on_all_gpt4_323k/0000100000/train_state_00002.json

@@ -0,0 +1 @@
+{"step": 100000, "acc_step": 0, "data_loader_state": {"it_state": {"start_token": 390, "it_state": {"it_state": {"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}, "source_to_state": {"fw_edu": {"file_path": "/scratch/craffel/lingua/data/flexitok/fw_edu/fineweb_edu_100bt.chunk.02.jsonl", "position": 30877240713, "block_size": 1, "offset": 0, "current_iter": 0}, "dan_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/dan_Latn/fineweb_2_hq.dan_Latn.chunk.02.jsonl", "position": 1418620808, "block_size": 1, "offset": 0, "current_iter": 0}, "swe_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/swe_Latn/fineweb_2_hq.swe_Latn.chunk.02.jsonl", "position": 1331774914, "block_size": 1, "offset": 0, "current_iter": 0}, "vie_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/vie_Latn/fineweb_2_hq.vie_Latn.chunk.02.jsonl", "position": 1422288587, "block_size": 1, "offset": 0, "current_iter": 0}, "hun_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/hun_Latn/fineweb_2_hq.hun_Latn.chunk.02.jsonl", "position": 1876769750, "block_size": 1, "offset": 0, "current_iter": 0}, "fas_Arab": {"file_path": "/scratch/craffel/lingua/data/flexitok/fas_Arab/fineweb_2_hq.fas_Arab.chunk.02.jsonl", "position": 1586171911, "block_size": 1, "offset": 0, "current_iter": 0}, "tur_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/tur_Latn/fineweb_2_hq.tur_Latn.chunk.02.jsonl", "position": 1208895855, "block_size": 1, "offset": 0, "current_iter": 0}, "ces_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/ces_Latn/fineweb_2_hq.ces_Latn.chunk.02.jsonl", "position": 1839724833, "block_size": 1, "offset": 0, "current_iter": 0}, "arb_Arab": {"file_path": "/scratch/craffel/lingua/data/flexitok/arb_Arab/fineweb_2_hq.arb_Arab.chunk.02.jsonl", "position": 2094243575, "block_size": 1, "offset": 0, "current_iter": 0}, "ell_Grek": {"file_path": "/scratch/craffel/lingua/data/flexitok/ell_Grek/fineweb_2_hq.ell_Grek.chunk.02.jsonl", "position": 2586905001, "block_size": 1, "offset": 0, "current_iter": 0}, "ind_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/ind_Latn/fineweb_2_hq.ind_Latn.chunk.02.jsonl", "position": 1844205087, "block_size": 1, "offset": 0, "current_iter": 0}, "nld_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/nld_Latn/fineweb_2_hq.nld_Latn.chunk.02.jsonl", "position": 1467682017, "block_size": 1, "offset": 0, "current_iter": 0}, "pol_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/pol_Latn/fineweb_2_hq.pol_Latn.chunk.02.jsonl", "position": 1531876306, "block_size": 1, "offset": 0, "current_iter": 0}, "por_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/por_Latn/fineweb_2_hq.por_Latn.chunk.02.jsonl", "position": 1606511781, "block_size": 1, "offset": 0, "current_iter": 0}, "ita_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/ita_Latn/fineweb_2_hq.ita_Latn.chunk.02.jsonl", "position": 1789939227, "block_size": 1, "offset": 0, "current_iter": 0}, "jpn_Jpan": {"file_path": "/scratch/craffel/lingua/data/flexitok/jpn_Jpan/fineweb_2_hq.jpn_Jpan.chunk.02.jsonl", "position": 1873083737, "block_size": 1, "offset": 0, "current_iter": 0}, "fra_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/fra_Latn/fineweb_2_hq.fra_Latn.chunk.02.jsonl", "position": 2473403375, "block_size": 1, "offset": 0, "current_iter": 0}, "spa_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/spa_Latn/fineweb_2_hq.spa_Latn.chunk.02.jsonl", "position": 2216330113, "block_size": 1, "offset": 0, "current_iter": 0}, "deu_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/deu_Latn/fineweb_2_hq.deu_Latn.chunk.02.jsonl", "position": 2699495345, "block_size": 1, "offset": 0, "current_iter": 0}, "cmn_Hani": {"file_path": "/scratch/craffel/lingua/data/flexitok/cmn_Hani/fineweb_2_hq.cmn_Hani.chunk.02.jsonl", "position": 2864795412, "block_size": 1, "offset": 0, "current_iter": 0}, "rus_Cyrl": {"file_path": "/scratch/craffel/lingua/data/flexitok/rus_Cyrl/fineweb_2_hq.rus_Cyrl.chunk.02.jsonl", "position": 6694078305, "block_size": 1, "offset": 0, "current_iter": 0}}, "rng_state": {"bit_generator": "PCG64", "state": {"state": 224377842048921656361223266741412677900, "inc": 234358335530849485425064040311006256713}, "has_uint32": 0, "uinteger": 0}}, "add_bos": true, "add_eos": true, "name": "huggingface", "path": "flexitok/bpe_gpt4_AllL_323000", "tokenizers": null, "dropout": 0.0, "rng_state": null, "seed": 42, "superset_code_name": "super_vocab", "n_words": null, "routing": {"source_to_tokenizer": {}, "suitable_tokenizer_probability": 1.0}}, "output_seq_len": 4096, "n_views": 2}, "seq_idx": 256, "rng_state": {"bit_generator": "PCG64", "state": {"state": 172206642884315098528897268843500314535, "inc": 319170006889470250209362588441616495209}, "has_uint32": 1, "uinteger": 2183176397}, "batch_size": 4, "prefetch_size": 1024}, "scheduler": {"base_lrs": [0.001], "last_epoch": 100000, "verbose": false, "_step_count": 100001, "_get_lr_called_within_step": false, "_last_lr": [0.001], "lr_lambdas": [{}]}}

bpe_on_all_gpt4_323k/0000100000/train_state_00003.json

@@ -0,0 +1 @@
+{"step": 100000, "acc_step": 0, "data_loader_state": {"it_state": {"start_token": 324, "it_state": {"it_state": {"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}, "source_to_state": {"fw_edu": {"file_path": "/scratch/craffel/lingua/data/flexitok/fw_edu/fineweb_edu_100bt.chunk.03.jsonl", "position": 30862590493, "block_size": 1, "offset": 0, "current_iter": 0}, "dan_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/dan_Latn/fineweb_2_hq.dan_Latn.chunk.03.jsonl", "position": 1412674229, "block_size": 1, "offset": 0, "current_iter": 0}, "swe_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/swe_Latn/fineweb_2_hq.swe_Latn.chunk.03.jsonl", "position": 1328301445, "block_size": 1, "offset": 0, "current_iter": 0}, "vie_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/vie_Latn/fineweb_2_hq.vie_Latn.chunk.03.jsonl", "position": 1426865970, "block_size": 1, "offset": 0, "current_iter": 0}, "hun_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/hun_Latn/fineweb_2_hq.hun_Latn.chunk.03.jsonl", "position": 1898562062, "block_size": 1, "offset": 0, "current_iter": 0}, "fas_Arab": {"file_path": "/scratch/craffel/lingua/data/flexitok/fas_Arab/fineweb_2_hq.fas_Arab.chunk.03.jsonl", "position": 1582617606, "block_size": 1, "offset": 0, "current_iter": 0}, "tur_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/tur_Latn/fineweb_2_hq.tur_Latn.chunk.03.jsonl", "position": 1211061168, "block_size": 1, "offset": 0, "current_iter": 0}, "ces_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/ces_Latn/fineweb_2_hq.ces_Latn.chunk.03.jsonl", "position": 1846828408, "block_size": 1, "offset": 0, "current_iter": 0}, "arb_Arab": {"file_path": "/scratch/craffel/lingua/data/flexitok/arb_Arab/fineweb_2_hq.arb_Arab.chunk.03.jsonl", "position": 2086068979, "block_size": 1, "offset": 0, "current_iter": 0}, "ell_Grek": {"file_path": "/scratch/craffel/lingua/data/flexitok/ell_Grek/fineweb_2_hq.ell_Grek.chunk.03.jsonl", "position": 2599955546, "block_size": 1, "offset": 0, "current_iter": 0}, "ind_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/ind_Latn/fineweb_2_hq.ind_Latn.chunk.03.jsonl", "position": 1830161142, "block_size": 1, "offset": 0, "current_iter": 0}, "nld_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/nld_Latn/fineweb_2_hq.nld_Latn.chunk.03.jsonl", "position": 1473130564, "block_size": 1, "offset": 0, "current_iter": 0}, "pol_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/pol_Latn/fineweb_2_hq.pol_Latn.chunk.03.jsonl", "position": 1537982997, "block_size": 1, "offset": 0, "current_iter": 0}, "por_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/por_Latn/fineweb_2_hq.por_Latn.chunk.03.jsonl", "position": 1597185315, "block_size": 1, "offset": 0, "current_iter": 0}, "ita_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/ita_Latn/fineweb_2_hq.ita_Latn.chunk.03.jsonl", "position": 1787500280, "block_size": 1, "offset": 0, "current_iter": 0}, "jpn_Jpan": {"file_path": "/scratch/craffel/lingua/data/flexitok/jpn_Jpan/fineweb_2_hq.jpn_Jpan.chunk.03.jsonl", "position": 1853620179, "block_size": 1, "offset": 0, "current_iter": 0}, "fra_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/fra_Latn/fineweb_2_hq.fra_Latn.chunk.03.jsonl", "position": 2470915319, "block_size": 1, "offset": 0, "current_iter": 0}, "spa_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/spa_Latn/fineweb_2_hq.spa_Latn.chunk.03.jsonl", "position": 2236795996, "block_size": 1, "offset": 0, "current_iter": 0}, "deu_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/deu_Latn/fineweb_2_hq.deu_Latn.chunk.03.jsonl", "position": 2704936707, "block_size": 1, "offset": 0, "current_iter": 0}, "cmn_Hani": {"file_path": "/scratch/craffel/lingua/data/flexitok/cmn_Hani/fineweb_2_hq.cmn_Hani.chunk.03.jsonl", "position": 2860542214, "block_size": 1, "offset": 0, "current_iter": 0}, "rus_Cyrl": {"file_path": "/scratch/craffel/lingua/data/flexitok/rus_Cyrl/fineweb_2_hq.rus_Cyrl.chunk.03.jsonl", "position": 6715187193, "block_size": 1, "offset": 0, "current_iter": 0}}, "rng_state": {"bit_generator": "PCG64", "state": {"state": 103567042630829093593610576402782212662, "inc": 148211758571781046255077612135386035203}, "has_uint32": 0, "uinteger": 0}}, "add_bos": true, "add_eos": true, "name": "huggingface", "path": "flexitok/bpe_gpt4_AllL_323000", "tokenizers": null, "dropout": 0.0, "rng_state": null, "seed": 42, "superset_code_name": "super_vocab", "n_words": null, "routing": {"source_to_tokenizer": {}, "suitable_tokenizer_probability": 1.0}}, "output_seq_len": 4096, "n_views": 2}, "seq_idx": 256, "rng_state": {"bit_generator": "PCG64", "state": {"state": 54691439170624420224489263842165716902, "inc": 115810872492597857501795428972873905393}, "has_uint32": 1, "uinteger": 1237320779}, "batch_size": 4, "prefetch_size": 1024}, "scheduler": {"base_lrs": [0.001], "last_epoch": 100000, "verbose": false, "_step_count": 100001, "_get_lr_called_within_step": false, "_last_lr": [0.001], "lr_lambdas": [{}]}}

bpe_on_all_gpt4_323k/0000100000/train_state_00004.json

@@ -0,0 +1 @@
+{"step": 100000, "acc_step": 0, "data_loader_state": {"it_state": {"start_token": 300, "it_state": {"it_state": {"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}, "source_to_state": {"fw_edu": {"file_path": "/scratch/craffel/lingua/data/flexitok/fw_edu/fineweb_edu_100bt.chunk.04.jsonl", "position": 30860328582, "block_size": 1, "offset": 0, "current_iter": 0}, "dan_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/dan_Latn/fineweb_2_hq.dan_Latn.chunk.04.jsonl", "position": 1423708711, "block_size": 1, "offset": 0, "current_iter": 0}, "swe_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/swe_Latn/fineweb_2_hq.swe_Latn.chunk.04.jsonl", "position": 1320410030, "block_size": 1, "offset": 0, "current_iter": 0}, "vie_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/vie_Latn/fineweb_2_hq.vie_Latn.chunk.04.jsonl", "position": 1421129563, "block_size": 1, "offset": 0, "current_iter": 0}, "hun_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/hun_Latn/fineweb_2_hq.hun_Latn.chunk.04.jsonl", "position": 1893937127, "block_size": 1, "offset": 0, "current_iter": 0}, "fas_Arab": {"file_path": "/scratch/craffel/lingua/data/flexitok/fas_Arab/fineweb_2_hq.fas_Arab.chunk.04.jsonl", "position": 1581295588, "block_size": 1, "offset": 0, "current_iter": 0}, "tur_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/tur_Latn/fineweb_2_hq.tur_Latn.chunk.04.jsonl", "position": 1209467502, "block_size": 1, "offset": 0, "current_iter": 0}, "ces_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/ces_Latn/fineweb_2_hq.ces_Latn.chunk.04.jsonl", "position": 1856350493, "block_size": 1, "offset": 0, "current_iter": 0}, "arb_Arab": {"file_path": "/scratch/craffel/lingua/data/flexitok/arb_Arab/fineweb_2_hq.arb_Arab.chunk.04.jsonl", "position": 2087062210, "block_size": 1, "offset": 0, "current_iter": 0}, "ell_Grek": {"file_path": "/scratch/craffel/lingua/data/flexitok/ell_Grek/fineweb_2_hq.ell_Grek.chunk.04.jsonl", "position": 2609958419, "block_size": 1, "offset": 0, "current_iter": 0}, "ind_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/ind_Latn/fineweb_2_hq.ind_Latn.chunk.04.jsonl", "position": 1829464197, "block_size": 1, "offset": 0, "current_iter": 0}, "nld_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/nld_Latn/fineweb_2_hq.nld_Latn.chunk.04.jsonl", "position": 1466648551, "block_size": 1, "offset": 0, "current_iter": 0}, "pol_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/pol_Latn/fineweb_2_hq.pol_Latn.chunk.04.jsonl", "position": 1536061915, "block_size": 1, "offset": 0, "current_iter": 0}, "por_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/por_Latn/fineweb_2_hq.por_Latn.chunk.04.jsonl", "position": 1606812971, "block_size": 1, "offset": 0, "current_iter": 0}, "ita_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/ita_Latn/fineweb_2_hq.ita_Latn.chunk.04.jsonl", "position": 1774231485, "block_size": 1, "offset": 0, "current_iter": 0}, "jpn_Jpan": {"file_path": "/scratch/craffel/lingua/data/flexitok/jpn_Jpan/fineweb_2_hq.jpn_Jpan.chunk.04.jsonl", "position": 1874768554, "block_size": 1, "offset": 0, "current_iter": 0}, "fra_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/fra_Latn/fineweb_2_hq.fra_Latn.chunk.04.jsonl", "position": 2485836265, "block_size": 1, "offset": 0, "current_iter": 0}, "spa_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/spa_Latn/fineweb_2_hq.spa_Latn.chunk.04.jsonl", "position": 2219709447, "block_size": 1, "offset": 0, "current_iter": 0}, "deu_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/deu_Latn/fineweb_2_hq.deu_Latn.chunk.04.jsonl", "position": 2698841601, "block_size": 1, "offset": 0, "current_iter": 0}, "cmn_Hani": {"file_path": "/scratch/craffel/lingua/data/flexitok/cmn_Hani/fineweb_2_hq.cmn_Hani.chunk.04.jsonl", "position": 2859282622, "block_size": 1, "offset": 0, "current_iter": 0}, "rus_Cyrl": {"file_path": "/scratch/craffel/lingua/data/flexitok/rus_Cyrl/fineweb_2_hq.rus_Cyrl.chunk.04.jsonl", "position": 6713362766, "block_size": 1, "offset": 0, "current_iter": 0}}, "rng_state": {"bit_generator": "PCG64", "state": {"state": 285569262280395580031062987212793845269, "inc": 186633262021180533256729114674950595327}, "has_uint32": 0, "uinteger": 0}}, "add_bos": true, "add_eos": true, "name": "huggingface", "path": "flexitok/bpe_gpt4_AllL_323000", "tokenizers": null, "dropout": 0.0, "rng_state": null, "seed": 42, "superset_code_name": "super_vocab", "n_words": null, "routing": {"source_to_tokenizer": {}, "suitable_tokenizer_probability": 1.0}}, "output_seq_len": 4096, "n_views": 2}, "seq_idx": 256, "rng_state": {"bit_generator": "PCG64", "state": {"state": 330736315247521707224292038935126153445, "inc": 303111205818808944921858206842105131807}, "has_uint32": 1, "uinteger": 211256137}, "batch_size": 4, "prefetch_size": 1024}, "scheduler": {"base_lrs": [0.001], "last_epoch": 100000, "verbose": false, "_step_count": 100001, "_get_lr_called_within_step": false, "_last_lr": [0.001], "lr_lambdas": [{}]}}

bpe_on_all_gpt4_323k/0000100000/train_state_00005.json

@@ -0,0 +1 @@
+{"step": 100000, "acc_step": 0, "data_loader_state": {"it_state": {"start_token": 98, "it_state": {"it_state": {"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}, "source_to_state": {"fw_edu": {"file_path": "/scratch/craffel/lingua/data/flexitok/fw_edu/fineweb_edu_100bt.chunk.05.jsonl", "position": 30889002049, "block_size": 1, "offset": 0, "current_iter": 0}, "dan_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/dan_Latn/fineweb_2_hq.dan_Latn.chunk.05.jsonl", "position": 1407585304, "block_size": 1, "offset": 0, "current_iter": 0}, "swe_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/swe_Latn/fineweb_2_hq.swe_Latn.chunk.05.jsonl", "position": 1322813929, "block_size": 1, "offset": 0, "current_iter": 0}, "vie_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/vie_Latn/fineweb_2_hq.vie_Latn.chunk.05.jsonl", "position": 1429031820, "block_size": 1, "offset": 0, "current_iter": 0}, "hun_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/hun_Latn/fineweb_2_hq.hun_Latn.chunk.05.jsonl", "position": 1875825145, "block_size": 1, "offset": 0, "current_iter": 0}, "fas_Arab": {"file_path": "/scratch/craffel/lingua/data/flexitok/fas_Arab/fineweb_2_hq.fas_Arab.chunk.05.jsonl", "position": 1579679758, "block_size": 1, "offset": 0, "current_iter": 0}, "tur_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/tur_Latn/fineweb_2_hq.tur_Latn.chunk.05.jsonl", "position": 1201647546, "block_size": 1, "offset": 0, "current_iter": 0}, "ces_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/ces_Latn/fineweb_2_hq.ces_Latn.chunk.05.jsonl", "position": 1855177425, "block_size": 1, "offset": 0, "current_iter": 0}, "arb_Arab": {"file_path": "/scratch/craffel/lingua/data/flexitok/arb_Arab/fineweb_2_hq.arb_Arab.chunk.05.jsonl", "position": 2078023748, "block_size": 1, "offset": 0, "current_iter": 0}, "ell_Grek": {"file_path": "/scratch/craffel/lingua/data/flexitok/ell_Grek/fineweb_2_hq.ell_Grek.chunk.05.jsonl", "position": 2600158831, "block_size": 1, "offset": 0, "current_iter": 0}, "ind_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/ind_Latn/fineweb_2_hq.ind_Latn.chunk.05.jsonl", "position": 1841328677, "block_size": 1, "offset": 0, "current_iter": 0}, "nld_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/nld_Latn/fineweb_2_hq.nld_Latn.chunk.05.jsonl", "position": 1451716174, "block_size": 1, "offset": 0, "current_iter": 0}, "pol_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/pol_Latn/fineweb_2_hq.pol_Latn.chunk.05.jsonl", "position": 1550033494, "block_size": 1, "offset": 0, "current_iter": 0}, "por_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/por_Latn/fineweb_2_hq.por_Latn.chunk.05.jsonl", "position": 1603607563, "block_size": 1, "offset": 0, "current_iter": 0}, "ita_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/ita_Latn/fineweb_2_hq.ita_Latn.chunk.05.jsonl", "position": 1784013456, "block_size": 1, "offset": 0, "current_iter": 0}, "jpn_Jpan": {"file_path": "/scratch/craffel/lingua/data/flexitok/jpn_Jpan/fineweb_2_hq.jpn_Jpan.chunk.05.jsonl", "position": 1864211951, "block_size": 1, "offset": 0, "current_iter": 0}, "fra_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/fra_Latn/fineweb_2_hq.fra_Latn.chunk.05.jsonl", "position": 2485330803, "block_size": 1, "offset": 0, "current_iter": 0}, "spa_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/spa_Latn/fineweb_2_hq.spa_Latn.chunk.05.jsonl", "position": 2233446914, "block_size": 1, "offset": 0, "current_iter": 0}, "deu_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/deu_Latn/fineweb_2_hq.deu_Latn.chunk.05.jsonl", "position": 2696165438, "block_size": 1, "offset": 0, "current_iter": 0}, "cmn_Hani": {"file_path": "/scratch/craffel/lingua/data/flexitok/cmn_Hani/fineweb_2_hq.cmn_Hani.chunk.05.jsonl", "position": 2862359228, "block_size": 1, "offset": 0, "current_iter": 0}, "rus_Cyrl": {"file_path": "/scratch/craffel/lingua/data/flexitok/rus_Cyrl/fineweb_2_hq.rus_Cyrl.chunk.05.jsonl", "position": 6688701019, "block_size": 1, "offset": 0, "current_iter": 0}}, "rng_state": {"bit_generator": "PCG64", "state": {"state": 17323023842230412077519032060582606213, "inc": 329233669073478483697346584247981015037}, "has_uint32": 0, "uinteger": 0}}, "add_bos": true, "add_eos": true, "name": "huggingface", "path": "flexitok/bpe_gpt4_AllL_323000", "tokenizers": null, "dropout": 0.0, "rng_state": null, "seed": 42, "superset_code_name": "super_vocab", "n_words": null, "routing": {"source_to_tokenizer": {}, "suitable_tokenizer_probability": 1.0}}, "output_seq_len": 4096, "n_views": 2}, "seq_idx": 256, "rng_state": {"bit_generator": "PCG64", "state": {"state": 152527936818944850177382785176910810558, "inc": 47382953940698287647753879262736142901}, "has_uint32": 1, "uinteger": 3675784275}, "batch_size": 4, "prefetch_size": 1024}, "scheduler": {"base_lrs": [0.001], "last_epoch": 100000, "verbose": false, "_step_count": 100001, "_get_lr_called_within_step": false, "_last_lr": [0.001], "lr_lambdas": [{}]}}

bpe_on_all_gpt4_323k/0000100000/train_state_00006.json

@@ -0,0 +1 @@
+{"step": 100000, "acc_step": 0, "data_loader_state": {"it_state": {"start_token": 139, "it_state": {"it_state": {"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}, "source_to_state": {"fw_edu": {"file_path": "/scratch/craffel/lingua/data/flexitok/fw_edu/fineweb_edu_100bt.chunk.06.jsonl", "position": 30884712033, "block_size": 1, "offset": 0, "current_iter": 0}, "dan_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/dan_Latn/fineweb_2_hq.dan_Latn.chunk.06.jsonl", "position": 1424888253, "block_size": 1, "offset": 0, "current_iter": 0}, "swe_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/swe_Latn/fineweb_2_hq.swe_Latn.chunk.06.jsonl", "position": 1323850004, "block_size": 1, "offset": 0, "current_iter": 0}, "vie_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/vie_Latn/fineweb_2_hq.vie_Latn.chunk.06.jsonl", "position": 1418596916, "block_size": 1, "offset": 0, "current_iter": 0}, "hun_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/hun_Latn/fineweb_2_hq.hun_Latn.chunk.06.jsonl", "position": 1887403132, "block_size": 1, "offset": 0, "current_iter": 0}, "fas_Arab": {"file_path": "/scratch/craffel/lingua/data/flexitok/fas_Arab/fineweb_2_hq.fas_Arab.chunk.06.jsonl", "position": 1571252268, "block_size": 1, "offset": 0, "current_iter": 0}, "tur_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/tur_Latn/fineweb_2_hq.tur_Latn.chunk.06.jsonl", "position": 1211259141, "block_size": 1, "offset": 0, "current_iter": 0}, "ces_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/ces_Latn/fineweb_2_hq.ces_Latn.chunk.06.jsonl", "position": 1857088921, "block_size": 1, "offset": 0, "current_iter": 0}, "arb_Arab": {"file_path": "/scratch/craffel/lingua/data/flexitok/arb_Arab/fineweb_2_hq.arb_Arab.chunk.06.jsonl", "position": 2079485877, "block_size": 1, "offset": 0, "current_iter": 0}, "ell_Grek": {"file_path": "/scratch/craffel/lingua/data/flexitok/ell_Grek/fineweb_2_hq.ell_Grek.chunk.06.jsonl", "position": 2589794414, "block_size": 1, "offset": 0, "current_iter": 0}, "ind_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/ind_Latn/fineweb_2_hq.ind_Latn.chunk.06.jsonl", "position": 1845176098, "block_size": 1, "offset": 0, "current_iter": 0}, "nld_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/nld_Latn/fineweb_2_hq.nld_Latn.chunk.06.jsonl", "position": 1470500136, "block_size": 1, "offset": 0, "current_iter": 0}, "pol_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/pol_Latn/fineweb_2_hq.pol_Latn.chunk.06.jsonl", "position": 1550099781, "block_size": 1, "offset": 0, "current_iter": 0}, "por_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/por_Latn/fineweb_2_hq.por_Latn.chunk.06.jsonl", "position": 1600858512, "block_size": 1, "offset": 0, "current_iter": 0}, "ita_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/ita_Latn/fineweb_2_hq.ita_Latn.chunk.06.jsonl", "position": 1777722148, "block_size": 1, "offset": 0, "current_iter": 0}, "jpn_Jpan": {"file_path": "/scratch/craffel/lingua/data/flexitok/jpn_Jpan/fineweb_2_hq.jpn_Jpan.chunk.06.jsonl", "position": 1853580861, "block_size": 1, "offset": 0, "current_iter": 0}, "fra_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/fra_Latn/fineweb_2_hq.fra_Latn.chunk.06.jsonl", "position": 2477965220, "block_size": 1, "offset": 0, "current_iter": 0}, "spa_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/spa_Latn/fineweb_2_hq.spa_Latn.chunk.06.jsonl", "position": 2230429199, "block_size": 1, "offset": 0, "current_iter": 0}, "deu_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/deu_Latn/fineweb_2_hq.deu_Latn.chunk.06.jsonl", "position": 2704578338, "block_size": 1, "offset": 0, "current_iter": 0}, "cmn_Hani": {"file_path": "/scratch/craffel/lingua/data/flexitok/cmn_Hani/fineweb_2_hq.cmn_Hani.chunk.06.jsonl", "position": 2860372051, "block_size": 1, "offset": 0, "current_iter": 0}, "rus_Cyrl": {"file_path": "/scratch/craffel/lingua/data/flexitok/rus_Cyrl/fineweb_2_hq.rus_Cyrl.chunk.06.jsonl", "position": 6682943294, "block_size": 1, "offset": 0, "current_iter": 0}}, "rng_state": {"bit_generator": "PCG64", "state": {"state": 253219575850805257133041596700143910258, "inc": 95963489890761403814531195999220475639}, "has_uint32": 0, "uinteger": 0}}, "add_bos": true, "add_eos": true, "name": "huggingface", "path": "flexitok/bpe_gpt4_AllL_323000", "tokenizers": null, "dropout": 0.0, "rng_state": null, "seed": 42, "superset_code_name": "super_vocab", "n_words": null, "routing": {"source_to_tokenizer": {}, "suitable_tokenizer_probability": 1.0}}, "output_seq_len": 4096, "n_views": 2}, "seq_idx": 256, "rng_state": {"bit_generator": "PCG64", "state": {"state": 131915213571014219328634582687476804074, "inc": 72545526324180839152750112646078969085}, "has_uint32": 0, "uinteger": 2783644522}, "batch_size": 4, "prefetch_size": 1024}, "scheduler": {"base_lrs": [0.001], "last_epoch": 100000, "verbose": false, "_step_count": 100001, "_get_lr_called_within_step": false, "_last_lr": [0.001], "lr_lambdas": [{}]}}

bpe_on_all_gpt4_323k/0000100000/train_state_00007.json

@@ -0,0 +1 @@
+{"step": 100000, "acc_step": 0, "data_loader_state": {"it_state": {"start_token": 2207, "it_state": {"it_state": {"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}, "source_to_state": {"fw_edu": {"file_path": "/scratch/craffel/lingua/data/flexitok/fw_edu/fineweb_edu_100bt.chunk.07.jsonl", "position": 30861750150, "block_size": 1, "offset": 0, "current_iter": 0}, "dan_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/dan_Latn/fineweb_2_hq.dan_Latn.chunk.07.jsonl", "position": 1429192794, "block_size": 1, "offset": 0, "current_iter": 0}, "swe_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/swe_Latn/fineweb_2_hq.swe_Latn.chunk.07.jsonl", "position": 1331735468, "block_size": 1, "offset": 0, "current_iter": 0}, "vie_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/vie_Latn/fineweb_2_hq.vie_Latn.chunk.07.jsonl", "position": 1418802676, "block_size": 1, "offset": 0, "current_iter": 0}, "hun_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/hun_Latn/fineweb_2_hq.hun_Latn.chunk.07.jsonl", "position": 1883690519, "block_size": 1, "offset": 0, "current_iter": 0}, "fas_Arab": {"file_path": "/scratch/craffel/lingua/data/flexitok/fas_Arab/fineweb_2_hq.fas_Arab.chunk.07.jsonl", "position": 1584729228, "block_size": 1, "offset": 0, "current_iter": 0}, "tur_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/tur_Latn/fineweb_2_hq.tur_Latn.chunk.07.jsonl", "position": 1215266381, "block_size": 1, "offset": 0, "current_iter": 0}, "ces_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/ces_Latn/fineweb_2_hq.ces_Latn.chunk.07.jsonl", "position": 1846530632, "block_size": 1, "offset": 0, "current_iter": 0}, "arb_Arab": {"file_path": "/scratch/craffel/lingua/data/flexitok/arb_Arab/fineweb_2_hq.arb_Arab.chunk.07.jsonl", "position": 2081632828, "block_size": 1, "offset": 0, "current_iter": 0}, "ell_Grek": {"file_path": "/scratch/craffel/lingua/data/flexitok/ell_Grek/fineweb_2_hq.ell_Grek.chunk.07.jsonl", "position": 2601733093, "block_size": 1, "offset": 0, "current_iter": 0}, "ind_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/ind_Latn/fineweb_2_hq.ind_Latn.chunk.07.jsonl", "position": 1829915280, "block_size": 1, "offset": 0, "current_iter": 0}, "nld_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/nld_Latn/fineweb_2_hq.nld_Latn.chunk.07.jsonl", "position": 1467441885, "block_size": 1, "offset": 0, "current_iter": 0}, "pol_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/pol_Latn/fineweb_2_hq.pol_Latn.chunk.07.jsonl", "position": 1541225133, "block_size": 1, "offset": 0, "current_iter": 0}, "por_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/por_Latn/fineweb_2_hq.por_Latn.chunk.07.jsonl", "position": 1593327441, "block_size": 1, "offset": 0, "current_iter": 0}, "ita_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/ita_Latn/fineweb_2_hq.ita_Latn.chunk.07.jsonl", "position": 1784284769, "block_size": 1, "offset": 0, "current_iter": 0}, "jpn_Jpan": {"file_path": "/scratch/craffel/lingua/data/flexitok/jpn_Jpan/fineweb_2_hq.jpn_Jpan.chunk.07.jsonl", "position": 1873512025, "block_size": 1, "offset": 0, "current_iter": 0}, "fra_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/fra_Latn/fineweb_2_hq.fra_Latn.chunk.07.jsonl", "position": 2487820550, "block_size": 1, "offset": 0, "current_iter": 0}, "spa_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/spa_Latn/fineweb_2_hq.spa_Latn.chunk.07.jsonl", "position": 2232432588, "block_size": 1, "offset": 0, "current_iter": 0}, "deu_Latn": {"file_path": "/scratch/craffel/lingua/data/flexitok/deu_Latn/fineweb_2_hq.deu_Latn.chunk.07.jsonl", "position": 2684993545, "block_size": 1, "offset": 0, "current_iter": 0}, "cmn_Hani": {"file_path": "/scratch/craffel/lingua/data/flexitok/cmn_Hani/fineweb_2_hq.cmn_Hani.chunk.07.jsonl", "position": 2860563833, "block_size": 1, "offset": 0, "current_iter": 0}, "rus_Cyrl": {"file_path": "/scratch/craffel/lingua/data/flexitok/rus_Cyrl/fineweb_2_hq.rus_Cyrl.chunk.07.jsonl", "position": 6688247066, "block_size": 1, "offset": 0, "current_iter": 0}}, "rng_state": {"bit_generator": "PCG64", "state": {"state": 157749371798247254960518411643212560116, "inc": 53245743019587277358203950863334653629}, "has_uint32": 0, "uinteger": 0}}, "add_bos": true, "add_eos": true, "name": "huggingface", "path": "flexitok/bpe_gpt4_AllL_323000", "tokenizers": null, "dropout": 0.0, "rng_state": null, "seed": 42, "superset_code_name": "super_vocab", "n_words": null, "routing": {"source_to_tokenizer": {}, "suitable_tokenizer_probability": 1.0}}, "output_seq_len": 4096, "n_views": 2}, "seq_idx": 256, "rng_state": {"bit_generator": "PCG64", "state": {"state": 239803483826676955776584746976189400951, "inc": 19761753544780285878460645500694854795}, "has_uint32": 1, "uinteger": 513301027}, "batch_size": 4, "prefetch_size": 1024}, "scheduler": {"base_lrs": [0.001], "last_epoch": 100000, "verbose": false, "_step_count": 100001, "_get_lr_called_within_step": false, "_last_lr": [0.001], "lr_lambdas": [{}]}}