Training in progress, epoch 1

.gitignore

@@ -0,0 +1 @@
+checkpoint-*/

config.json

@@ -0,0 +1,24 @@
+{
+  "architectures": [
+    "BertForMaskedLM"
+  ],
+  "attention_probs_dropout_prob": 0.1,
+  "classifier_dropout": null,
+  "hidden_act": "gelu",
+  "hidden_dropout_prob": 0.1,
+  "hidden_size": 512,
+  "initializer_range": 0.02,
+  "intermediate_size": 2048,
+  "layer_norm_eps": 1e-12,
+  "max_position_embeddings": 512,
+  "model_type": "bert",
+  "num_attention_heads": 8,
+  "num_hidden_layers": 4,
+  "pad_token_id": 0,
+  "position_embedding_type": "absolute",
+  "torch_dtype": "float16",
+  "transformers_version": "4.19.2",
+  "type_vocab_size": 2,
+  "use_cache": true,
+  "vocab_size": 32000
+}

last-checkpoint/config.json

@@ -0,0 +1,24 @@
+{
+  "architectures": [
+    "BertForMaskedLM"
+  ],
+  "attention_probs_dropout_prob": 0.1,
+  "classifier_dropout": null,
+  "hidden_act": "gelu",
+  "hidden_dropout_prob": 0.1,
+  "hidden_size": 512,
+  "initializer_range": 0.02,
+  "intermediate_size": 2048,
+  "layer_norm_eps": 1e-12,
+  "max_position_embeddings": 512,
+  "model_type": "bert",
+  "num_attention_heads": 8,
+  "num_hidden_layers": 4,
+  "pad_token_id": 0,
+  "position_embedding_type": "absolute",
+  "torch_dtype": "float16",
+  "transformers_version": "4.19.2",
+  "type_vocab_size": 2,
+  "use_cache": true,
+  "vocab_size": 32000
+}

last-checkpoint/global_step69473/mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2e0050713b381dc3ba13434465c1d278d0ab5cadc86d4b70311ef30a660d2275
+size 59134503

last-checkpoint/global_step69473/zero_pp_rank_0_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cd6363f2eea53b7241ae383f12fb281c2963ee7c89a67d88e11acee9e9724fec
+size 118216675

last-checkpoint/global_step69473/zero_pp_rank_1_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4037d0ac9c29daf2f01103e710e3a7c88a7a367ae8d382d498157ed65496fba0
+size 118217955

last-checkpoint/global_step69473/zero_pp_rank_2_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1bc5b79dabb04bd744c6dc6035eb6f76051655e5732b24ecb018e0c9a68202d3
+size 118221091

last-checkpoint/latest

@@ -0,0 +1 @@
+global_step69473

last-checkpoint/pytorch_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8ea8f19b9cfae5ac1f4aea53be2d9bc355af72960cde88e7effe294d0e4c658a
+size 59121639

last-checkpoint/rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a80824c9c80bca9b568638795ecb7ff1fafcdbd4e21352551ee3d13f947d45ef
+size 14503

last-checkpoint/rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cb80857bf4eca9a59ff8597441b8c91a7ea9b5632f862c81cad8c50b6099f5f8
+size 14503

last-checkpoint/rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1593b05733d15d785bb94bec329a81cf0c61504cea8453fd684516fb9b5dbf83
+size 14503

last-checkpoint/special_tokens_map.json

@@ -0,0 +1 @@
+{"unk_token": "[UNK]", "sep_token": "[SEP]", "pad_token": "[PAD]", "cls_token": "[CLS]", "mask_token": "[MASK]"}

last-checkpoint/tokenizer_config.json

@@ -0,0 +1 @@
+{"cls_token": "[CLS]", "mask_token": "[MASK]", "model_max_length": 128, "pad_token": "[PAD]", "padding_side": "right", "sep_token": "[SEP]", "truncation_side": "right", "unk_token": "[UNK]", "special_tokens_map_file": "pretrained_tokenizers/UnidicBpe2/special_tokens_map.json", "name_or_path": "pretrained_tokenizers/UnidicBpe2", "tokenizer_class": "PreTrainedTokenizerFast"}

last-checkpoint/trainer_state.json

@@ -0,0 +1,853 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.0,
+  "global_step": 69473,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.01,
+      "learning_rate": 6.763399204600845e-05,
+      "loss": 6.9832,
+      "step": 500
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 7.520427275451058e-05,
+      "loss": 5.5271,
+      "step": 1000
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 7.961254711978457e-05,
+      "loss": 4.6913,
+      "step": 1500
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 8.275270773909965e-05,
+      "loss": 4.2907,
+      "step": 2000
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 8.517839539271491e-05,
+      "loss": 4.0302,
+      "step": 2500
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 8.716825914421595e-05,
+      "loss": 3.8379,
+      "step": 3000
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 8.884385574179297e-05,
+      "loss": 3.6906,
+      "step": 3500
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.030114272368873e-05,
+      "loss": 3.5642,
+      "step": 4000
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 9.158138540964539e-05,
+      "loss": 3.4704,
+      "step": 4500
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 9.273119601765861e-05,
+      "loss": 3.3821,
+      "step": 5000
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 9.376714826181023e-05,
+      "loss": 3.3064,
+      "step": 5500
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.471669412880503e-05,
+      "loss": 3.2476,
+      "step": 6000
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 9.558668751136338e-05,
+      "loss": 3.1892,
+      "step": 6500
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 9.639540886232853e-05,
+      "loss": 3.1299,
+      "step": 7000
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 9.714529240377096e-05,
+      "loss": 3.0869,
+      "step": 7500
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.784957770827781e-05,
+      "loss": 3.0405,
+      "step": 8000
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 9.850850144881638e-05,
+      "loss": 2.9989,
+      "step": 8500
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 9.91322455339213e-05,
+      "loss": 2.9583,
+      "step": 9000
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.97198937413398e-05,
+      "loss": 2.923,
+      "step": 9500
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 9.997382892215664e-05,
+      "loss": 2.8859,
+      "step": 10000
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 9.99221098873709e-05,
+      "loss": 2.8536,
+      "step": 10500
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 9.987018314561817e-05,
+      "loss": 2.8222,
+      "step": 11000
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.981846411083245e-05,
+      "loss": 2.7972,
+      "step": 11500
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 9.97665373690797e-05,
+      "loss": 2.7657,
+      "step": 12000
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 9.971481833429398e-05,
+      "loss": 2.7461,
+      "step": 12500
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.966289159254125e-05,
+      "loss": 2.721,
+      "step": 13000
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 9.961117255775553e-05,
+      "loss": 2.7012,
+      "step": 13500
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 9.95592458160028e-05,
+      "loss": 2.6758,
+      "step": 14000
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 9.950752678121706e-05,
+      "loss": 2.666,
+      "step": 14500
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.945560003946433e-05,
+      "loss": 2.6412,
+      "step": 15000
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 9.94038810046786e-05,
+      "loss": 2.6264,
+      "step": 15500
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 9.935195426292587e-05,
+      "loss": 2.6089,
+      "step": 16000
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.930023522814015e-05,
+      "loss": 2.5915,
+      "step": 16500
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 9.92483084863874e-05,
+      "loss": 2.5783,
+      "step": 17000
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 9.919658945160168e-05,
+      "loss": 2.558,
+      "step": 17500
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 9.914466270984895e-05,
+      "loss": 2.5481,
+      "step": 18000
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.909294367506321e-05,
+      "loss": 2.5376,
+      "step": 18500
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 9.90410169333105e-05,
+      "loss": 2.5236,
+      "step": 19000
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 9.898929789852477e-05,
+      "loss": 2.5129,
+      "step": 19500
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 9.893737115677203e-05,
+      "loss": 2.5024,
+      "step": 20000
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.88856521219863e-05,
+      "loss": 2.4841,
+      "step": 20500
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 9.883372538023356e-05,
+      "loss": 2.4802,
+      "step": 21000
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 9.878200634544784e-05,
+      "loss": 2.4725,
+      "step": 21500
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.873007960369511e-05,
+      "loss": 2.4581,
+      "step": 22000
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 9.867836056890939e-05,
+      "loss": 2.4493,
+      "step": 22500
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 9.862643382715665e-05,
+      "loss": 2.4399,
+      "step": 23000
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 9.857471479237093e-05,
+      "loss": 2.4297,
+      "step": 23500
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.85227880506182e-05,
+      "loss": 2.4259,
+      "step": 24000
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 9.847106901583248e-05,
+      "loss": 2.4199,
+      "step": 24500
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 9.841914227407973e-05,
+      "loss": 2.4068,
+      "step": 25000
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.836742323929401e-05,
+      "loss": 2.4023,
+      "step": 25500
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 9.831549649754126e-05,
+      "loss": 2.3903,
+      "step": 26000
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 9.826377746275554e-05,
+      "loss": 2.3849,
+      "step": 26500
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 9.821185072100282e-05,
+      "loss": 2.3787,
+      "step": 27000
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.816013168621709e-05,
+      "loss": 2.3722,
+      "step": 27500
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 9.810820494446436e-05,
+      "loss": 2.3646,
+      "step": 28000
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.805648590967863e-05,
+      "loss": 2.3537,
+      "step": 28500
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.800455916792589e-05,
+      "loss": 2.353,
+      "step": 29000
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 9.795284013314017e-05,
+      "loss": 2.3468,
+      "step": 29500
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 9.790091339138744e-05,
+      "loss": 2.3371,
+      "step": 30000
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 9.784919435660171e-05,
+      "loss": 2.3346,
+      "step": 30500
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.779726761484898e-05,
+      "loss": 2.3244,
+      "step": 31000
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 9.774554858006325e-05,
+      "loss": 2.3233,
+      "step": 31500
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 9.769362183831051e-05,
+      "loss": 2.3214,
+      "step": 32000
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 9.764190280352479e-05,
+      "loss": 2.312,
+      "step": 32500
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.758997606177206e-05,
+      "loss": 2.3099,
+      "step": 33000
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 9.753825702698634e-05,
+      "loss": 2.3025,
+      "step": 33500
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 9.748633028523359e-05,
+      "loss": 2.2935,
+      "step": 34000
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.743461125044787e-05,
+      "loss": 2.2924,
+      "step": 34500
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 9.738268450869514e-05,
+      "loss": 2.2895,
+      "step": 35000
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 9.73309654739094e-05,
+      "loss": 2.2806,
+      "step": 35500
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 9.727903873215668e-05,
+      "loss": 2.2815,
+      "step": 36000
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.722731969737096e-05,
+      "loss": 2.2735,
+      "step": 36500
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 9.717539295561822e-05,
+      "loss": 2.2697,
+      "step": 37000
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 9.71236739208325e-05,
+      "loss": 2.2686,
+      "step": 37500
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.707174717907976e-05,
+      "loss": 2.2617,
+      "step": 38000
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 9.702002814429404e-05,
+      "loss": 2.2561,
+      "step": 38500
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 9.69681014025413e-05,
+      "loss": 2.2533,
+      "step": 39000
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 9.691638236775557e-05,
+      "loss": 2.2484,
+      "step": 39500
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.686445562600284e-05,
+      "loss": 2.2471,
+      "step": 40000
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 9.681273659121712e-05,
+      "loss": 2.2443,
+      "step": 40500
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 9.676080984946439e-05,
+      "loss": 2.2331,
+      "step": 41000
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.670909081467867e-05,
+      "loss": 2.2349,
+      "step": 41500
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 9.665716407292592e-05,
+      "loss": 2.2296,
+      "step": 42000
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 9.66054450381402e-05,
+      "loss": 2.2296,
+      "step": 42500
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 9.655351829638745e-05,
+      "loss": 2.2238,
+      "step": 43000
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.650179926160173e-05,
+      "loss": 2.2193,
+      "step": 43500
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 9.6449872519849e-05,
+      "loss": 2.2185,
+      "step": 44000
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 9.639815348506328e-05,
+      "loss": 2.2157,
+      "step": 44500
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.634622674331054e-05,
+      "loss": 2.2092,
+      "step": 45000
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 9.629450770852482e-05,
+      "loss": 2.2096,
+      "step": 45500
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 9.624258096677208e-05,
+      "loss": 2.2026,
+      "step": 46000
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 9.619086193198635e-05,
+      "loss": 2.1971,
+      "step": 46500
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.613893519023362e-05,
+      "loss": 2.1962,
+      "step": 47000
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 9.60872161554479e-05,
+      "loss": 2.1918,
+      "step": 47500
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 9.603528941369516e-05,
+      "loss": 2.1932,
+      "step": 48000
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 9.598357037890943e-05,
+      "loss": 2.1844,
+      "step": 48500
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 9.59316436371567e-05,
+      "loss": 2.1872,
+      "step": 49000
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 9.587992460237098e-05,
+      "loss": 2.1781,
+      "step": 49500
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 9.582799786061825e-05,
+      "loss": 2.1797,
+      "step": 50000
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 9.577627882583253e-05,
+      "loss": 2.1793,
+      "step": 50500
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 9.572435208407978e-05,
+      "loss": 2.1736,
+      "step": 51000
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 9.567263304929406e-05,
+      "loss": 2.1723,
+      "step": 51500
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 9.562070630754133e-05,
+      "loss": 2.1732,
+      "step": 52000
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 9.55689872727556e-05,
+      "loss": 2.1654,
+      "step": 52500
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 9.551706053100287e-05,
+      "loss": 2.1678,
+      "step": 53000
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 9.546534149621714e-05,
+      "loss": 2.1632,
+      "step": 53500
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.54134147544644e-05,
+      "loss": 2.158,
+      "step": 54000
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 9.536169571967868e-05,
+      "loss": 2.1551,
+      "step": 54500
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 9.530976897792595e-05,
+      "loss": 2.1553,
+      "step": 55000
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 9.525804994314023e-05,
+      "loss": 2.1469,
+      "step": 55500
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.520612320138748e-05,
+      "loss": 2.1517,
+      "step": 56000
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 9.515440416660176e-05,
+      "loss": 2.1499,
+      "step": 56500
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 9.510247742484903e-05,
+      "loss": 2.144,
+      "step": 57000
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 9.50507583900633e-05,
+      "loss": 2.1478,
+      "step": 57500
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 9.499883164831058e-05,
+      "loss": 2.1369,
+      "step": 58000
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 9.494711261352485e-05,
+      "loss": 2.1394,
+      "step": 58500
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 9.489518587177211e-05,
+      "loss": 2.1391,
+      "step": 59000
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.484346683698639e-05,
+      "loss": 2.1344,
+      "step": 59500
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 9.479154009523364e-05,
+      "loss": 2.1371,
+      "step": 60000
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 9.473982106044792e-05,
+      "loss": 2.1289,
+      "step": 60500
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 9.468789431869519e-05,
+      "loss": 2.1271,
+      "step": 61000
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.463617528390946e-05,
+      "loss": 2.1276,
+      "step": 61500
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 9.458424854215673e-05,
+      "loss": 2.1255,
+      "step": 62000
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 9.453252950737101e-05,
+      "loss": 2.1192,
+      "step": 62500
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.448060276561828e-05,
+      "loss": 2.1166,
+      "step": 63000
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 9.442888373083256e-05,
+      "loss": 2.1171,
+      "step": 63500
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 9.437695698907981e-05,
+      "loss": 2.1127,
+      "step": 64000
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 9.432523795429409e-05,
+      "loss": 2.1116,
+      "step": 64500
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.427331121254134e-05,
+      "loss": 2.1121,
+      "step": 65000
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 9.422159217775562e-05,
+      "loss": 2.1093,
+      "step": 65500
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 9.41696654360029e-05,
+      "loss": 2.1103,
+      "step": 66000
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.411794640121717e-05,
+      "loss": 2.1073,
+      "step": 66500
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 9.406601965946444e-05,
+      "loss": 2.1057,
+      "step": 67000
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 9.401430062467871e-05,
+      "loss": 2.1012,
+      "step": 67500
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 9.396237388292597e-05,
+      "loss": 2.1012,
+      "step": 68000
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.391065484814025e-05,
+      "loss": 2.1041,
+      "step": 68500
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 9.385872810638751e-05,
+      "loss": 2.0974,
+      "step": 69000
+    },
+    {
+      "epoch": 1.0,
+      "eval_accuracy": 0.6071486253641302,
+      "eval_loss": 1.974609375,
+      "eval_runtime": 652.8749,
+      "eval_samples_per_second": 825.483,
+      "eval_steps_per_second": 34.396,
+      "step": 69473
+    }
+  ],
+  "max_steps": 972622,
+  "num_train_epochs": 14,
+  "total_flos": 4.979851725289554e+17,
+  "trial_name": null,
+  "trial_params": null
+}

last-checkpoint/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bd54042cd89784f24506b5912f250d73ea83faa93bfe5b987bbcaed0572116e8
+size 4335

last-checkpoint/zero_to_fp32.py

@@ -0,0 +1,482 @@
+#!/usr/bin/env python
+
+# This script extracts fp32 consolidated weights from a zero 2 and 3 DeepSpeed checkpoints. It gets
+# copied into the top level checkpoint dir, so the user can easily do the conversion at any point in
+# the future. Once extracted, the weights don't require DeepSpeed and can be used in any
+# application.
+#
+# example: python zero_to_fp32.py . pytorch_model.bin
+
+import argparse
+import torch
+import glob
+import math
+import os
+import re
+from collections import OrderedDict
+
+# while this script doesn't use deepspeed to recover data, since the checkpoints are pickled with
+# DeepSpeed data structures it has to be available in the current python environment.
+from deepspeed.utils import logger
+from deepspeed.checkpoint.constants import (DS_VERSION,
+                                            OPTIMIZER_STATE_DICT,
+                                            SINGLE_PARTITION_OF_FP32_GROUPS,
+                                            FP32_FLAT_GROUPS,
+                                            ZERO_STAGE,
+                                            PARTITION_COUNT,
+                                            PARAM_SHAPES,
+                                            BUFFER_NAMES)
+
+debug = 0
+
+# load to cpu
+device = torch.device('cpu')
+
+
+def atoi(text):
+    return int(text) if text.isdigit() else text
+
+
+def natural_keys(text):
+    '''
+    alist.sort(key=natural_keys) sorts in human order
+    http://nedbatchelder.com/blog/200712/human_sorting.html
+    (See Toothy's implementation in the comments)
+    '''
+    return [atoi(c) for c in re.split(r'(\d+)', text)]
+
+
+def get_model_state_file(checkpoint_dir, zero_stage):
+    if not os.path.isdir(checkpoint_dir):
+        raise FileNotFoundError(f"Directory '{checkpoint_dir}' doesn't exist")
+
+    # there should be only one file
+    if zero_stage == 2:
+        file = os.path.join(checkpoint_dir, "mp_rank_00_model_states.pt")
+    elif zero_stage == 3:
+        file = os.path.join(checkpoint_dir, "zero_pp_rank_0_mp_rank_00_model_states.pt")
+
+    if not os.path.exists(file):
+        raise FileNotFoundError(f"can't find model states file at '{file}'")
+
+    return file
+
+
+def get_optim_files(checkpoint_dir):
+    # XXX: need to test that this simple glob rule works for multi-node setup too
+    optim_files = sorted(glob.glob(os.path.join(checkpoint_dir,
+                                                "*_optim_states.pt")),
+                         key=natural_keys)
+
+    if len(optim_files) == 0:
+        raise FileNotFoundError(
+            f"can't find '*_optim_states.pt' files in directory '{checkpoint_dir}'")
+
+    return optim_files
+
+
+def parse_model_state(file):
+    state_dict = torch.load(file, map_location=device)
+
+    if BUFFER_NAMES not in state_dict:
+        raise ValueError(f"{file} is not a model state checkpoint")
+    buffer_names = state_dict[BUFFER_NAMES]
+    if debug:
+        print("Found buffers:", buffer_names)
+
+    # recover just the buffers while restoring them to fp32 if they were saved in fp16
+    buffers = {
+        k: v.float()
+        for k,
+        v in state_dict["module"].items() if k in buffer_names
+    }
+    param_shapes = state_dict[PARAM_SHAPES]
+
+    ds_version = state_dict.get(DS_VERSION, None)
+
+    return buffers, param_shapes, ds_version
+
+
+def parse_optim_states(files, ds_checkpoint_dir):
+
+    total_files = len(files)
+    state_dicts = []
+    for f in files:
+        state_dicts.append(torch.load(f, map_location=device))
+
+    if not ZERO_STAGE in state_dicts[0][OPTIMIZER_STATE_DICT]:
+        raise ValueError(f"{files[0]} is not a zero checkpoint")
+    zero_stage = state_dicts[0][OPTIMIZER_STATE_DICT][ZERO_STAGE]
+    world_size = state_dicts[0][OPTIMIZER_STATE_DICT][PARTITION_COUNT]
+
+    # For ZeRO-2 each param group can have different partition_count as data parallelism for expert
+    # parameters can be different from data parallelism for non-expert parameters. So we can just
+    # use the max of the partition_count to get the dp world_size.
+
+    if type(world_size) is list:
+        world_size = max(world_size)
+
+    if world_size != total_files:
+        raise ValueError(
+            f"Expected {world_size} of '*_optim_states.pt' under '{ds_checkpoint_dir}' but found {total_files} files. "
+            "Possibly due to an overwrite of an old checkpoint, or a checkpoint didn't get saved by one or more processes."
+        )
+
+    # the groups are named differently in each stage
+    if zero_stage == 2:
+        fp32_groups_key = SINGLE_PARTITION_OF_FP32_GROUPS
+    elif zero_stage == 3:
+        fp32_groups_key = FP32_FLAT_GROUPS
+    else:
+        raise ValueError(f"unknown zero stage {zero_stage}")
+
+    if zero_stage == 2:
+        fp32_flat_groups = [
+            state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key]
+            for i in range(len(state_dicts))
+        ]
+    elif zero_stage == 3:
+        # if there is more than one param group, there will be multiple flattened tensors - one
+        # flattened tensor per group - for simplicity merge them into a single tensor
+        #
+        # XXX: could make the script more memory efficient for when there are multiple groups - it
+        # will require matching the sub-lists of param_shapes for each param group flattened tensor
+
+        fp32_flat_groups = [
+            torch.cat(state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key],
+                      0) for i in range(len(state_dicts))
+        ]
+
+    return zero_stage, world_size, fp32_flat_groups
+
+
+def _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir):
+    """
+    Returns fp32 state_dict reconstructed from ds checkpoint
+
+    Args:
+        - ``ds_checkpoint_dir``: path to the deepspeed checkpoint folder (where the optimizer files are)
+
+    """
+    print(f"Processing zero checkpoint '{ds_checkpoint_dir}'")
+
+    optim_files = get_optim_files(ds_checkpoint_dir)
+    zero_stage, world_size, fp32_flat_groups = parse_optim_states(optim_files, ds_checkpoint_dir)
+    print(
+        f"Detected checkpoint of type zero stage {zero_stage}, world_size: {world_size}")
+
+    model_file = get_model_state_file(ds_checkpoint_dir, zero_stage)
+    buffers, param_shapes, ds_version = parse_model_state(model_file)
+    print(f'Parsing checkpoint created by deepspeed=={ds_version}')
+
+    if zero_stage == 2:
+        return _get_fp32_state_dict_from_zero2_checkpoint(world_size,
+                                                          param_shapes,
+                                                          fp32_flat_groups,
+                                                          buffers)
+    elif zero_stage == 3:
+        return _get_fp32_state_dict_from_zero3_checkpoint(world_size,
+                                                          param_shapes,
+                                                          fp32_flat_groups,
+                                                          buffers)
+
+
+def _get_fp32_state_dict_from_zero2_checkpoint(world_size,
+                                               param_shapes,
+                                               fp32_flat_groups,
+                                               buffers):
+
+    # Reconstruction protocol:
+    #
+    # XXX: document this
+
+    if debug:
+        for i in range(world_size):
+            for j in range(len(fp32_flat_groups[0])):
+                print(
+                    f"{FP32_FLAT_GROUPS}[{i}][{j}].shape={fp32_flat_groups[i][j].shape}")
+
+    # XXX: memory usage doubles here (zero2)
+    num_param_groups = len(fp32_flat_groups[0])
+    merged_single_partition_of_fp32_groups = []
+    for i in range(num_param_groups):
+        merged_partitions = [sd[i] for sd in fp32_flat_groups]
+        full_single_fp32_vector = torch.cat(merged_partitions, 0)
+        merged_single_partition_of_fp32_groups.append(full_single_fp32_vector)
+    avail_numel = sum([
+        full_single_fp32_vector.numel()
+        for full_single_fp32_vector in merged_single_partition_of_fp32_groups
+    ])
+
+    if debug:
+        wanted_params = sum([len(shapes) for shapes in param_shapes])
+        wanted_numel = sum(
+            [sum(shape.numel() for shape in shapes.values()) for shapes in param_shapes])
+        # not asserting if there is a mismatch due to possible padding
+        print(f"Have {avail_numel} numels to process.")
+        print(f"Need {wanted_numel} numels in {wanted_params} params.")
+
+    state_dict = OrderedDict()
+
+    # buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    # params
+    # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
+    # out-of-core computing solution
+    total_numel = 0
+    total_params = 0
+    for shapes, full_single_fp32_vector in zip(param_shapes, merged_single_partition_of_fp32_groups):
+        offset = 0
+        avail_numel = full_single_fp32_vector.numel()
+        for name, shape in shapes.items():
+
+            unpartitioned_numel = shape.numel()
+            total_numel += unpartitioned_numel
+            total_params += 1
+
+            if debug:
+                print(
+                    f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} "
+                )
+            state_dict[name] = full_single_fp32_vector.narrow(
+                0,
+                offset,
+                unpartitioned_numel).view(shape)
+            offset += unpartitioned_numel
+
+        # Z2 started to align to 2*world_size to improve nccl performance. Therefore both offset and
+        # avail_numel can differ by anywhere between 0..2*world_size. Due to two unrelated complex
+        # paddings performed in the code it's almost impossible to predict the exact numbers w/o the
+        # live optimizer object, so we are checking that the numbers are within the right range
+        align_to = 2 * world_size
+
+        def zero2_align(x):
+            return align_to * math.ceil(x / align_to)
+
+        if debug:
+            print(f"original offset={offset}, avail_numel={avail_numel}")
+
+        offset = zero2_align(offset)
+        avail_numel = zero2_align(avail_numel)
+
+        if debug:
+            print(f"aligned  offset={offset}, avail_numel={avail_numel}")
+
+        # Sanity check
+        if offset != avail_numel:
+            raise ValueError(
+                f"consumed {offset} numels out of {avail_numel} - something is wrong")
+
+    print(
+        f"Reconstructed fp32 state dict with {total_params} params {total_numel} elements"
+    )
+
+    return state_dict
+
+
+def zero3_partitioned_param_info(unpartitioned_numel, world_size):
+    remainder = unpartitioned_numel % world_size
+    padding_numel = (world_size - remainder) if remainder else 0
+    partitioned_numel = math.ceil(unpartitioned_numel / world_size)
+    return partitioned_numel, padding_numel
+
+
+def _get_fp32_state_dict_from_zero3_checkpoint(world_size,
+                                               param_shapes,
+                                               fp32_flat_groups,
+                                               buffers):
+
+    # Reconstruction protocol: For zero3 we need to zip the partitions together at boundary of each
+    # param, re-consolidating each param, while dealing with padding if any
+
+    avail_numel = fp32_flat_groups[0].numel() * world_size
+    # merge list of dicts, preserving order
+    param_shapes = {k: v for d in param_shapes for k, v in d.items()}
+
+    if debug:
+        for i in range(world_size):
+            print(f"{FP32_FLAT_GROUPS}[{i}].shape={fp32_flat_groups[i].shape}")
+
+        wanted_params = len(param_shapes)
+        wanted_numel = sum(shape.numel() for shape in param_shapes.values())
+        # not asserting if there is a mismatch due to possible padding
+        print(f"Have {avail_numel} numels to process.")
+        print(f"Need {wanted_numel} numels in {wanted_params} params.")
+
+    state_dict = OrderedDict()
+
+    # buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    # params
+    # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
+    # out-of-core computing solution
+    offset = 0
+    total_numel = 0
+    total_params = 0
+    for name, shape in param_shapes.items():
+
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+        total_params += 1
+
+        partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
+
+        if debug:
+            print(
+                f"{total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
+            )
+
+        # XXX: memory usage doubles here
+        state_dict[name] = torch.cat(
+            tuple(fp32_flat_groups[i].narrow(0,
+                                             offset,
+                                             partitioned_numel)
+                  for i in range(world_size)),
+            0).narrow(0,
+                      0,
+                      unpartitioned_numel).view(shape)
+        offset += partitioned_numel
+
+    offset *= world_size
+
+    # Sanity check
+    if offset != avail_numel:
+        raise ValueError(
+            f"consumed {offset} numels out of {avail_numel} - something is wrong")
+
+    print(
+        f"Reconstructed fp32 state dict with {total_params} params {total_numel} elements"
+    )
+
+    return state_dict
+
+
+def get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag=None):
+    """
+    Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated state_dict that can be loaded with
+    ``load_state_dict()`` and used for training without DeepSpeed or shared with others, for example
+    via a model hub.
+
+    Args:
+        - ``checkpoint_dir``: path to the desired checkpoint folder
+        - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in 'latest' file. e.g., ``global_step14``
+
+    Returns:
+        - pytorch ``state_dict``
+
+    Note: this approach may not work if your application doesn't have sufficient free CPU memory and
+    you may need to use the offline approach using the ``zero_to_fp32.py`` script that is saved with
+    the checkpoint.
+
+    A typical usage might be ::
+
+        from deepspeed.utils.zero_to_fp32 import get_fp32_state_dict_from_zero_checkpoint
+        # do the training and checkpoint saving
+        state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir) # already on cpu
+        model = model.cpu() # move to cpu
+        model.load_state_dict(state_dict)
+        # submit to model hub or save the model to share with others
+
+    In this example the ``model`` will no longer be usable in the deepspeed context of the same
+    application. i.e. you will need to re-initialize the deepspeed engine, since
+    ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
+
+    If you want it all done for you, use ``load_state_dict_from_zero_checkpoint`` instead.
+
+    """
+    if tag is None:
+        latest_path = os.path.join(checkpoint_dir, 'latest')
+        if os.path.isfile(latest_path):
+            with open(latest_path, 'r') as fd:
+                tag = fd.read().strip()
+        else:
+            raise ValueError(f"Unable to find 'latest' file at {latest_path}")
+
+    ds_checkpoint_dir = os.path.join(checkpoint_dir, tag)
+
+    if not os.path.isdir(ds_checkpoint_dir):
+        raise FileNotFoundError(f"Directory '{ds_checkpoint_dir}' doesn't exist")
+
+    return _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir)
+
+
+def convert_zero_checkpoint_to_fp32_state_dict(checkpoint_dir, output_file, tag=None):
+    """
+    Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict`` file that can be
+    loaded with ``torch.load(file)`` + ``load_state_dict()`` and used for training without DeepSpeed.
+
+    Args:
+        - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
+        - ``output_file``: path to the pytorch fp32 state_dict output file (e.g. path/pytorch_model.bin)
+        - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
+    """
+
+    state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag)
+    print(f"Saving fp32 state dict to {output_file}")
+    torch.save(state_dict, output_file)
+
+
+def load_state_dict_from_zero_checkpoint(model, checkpoint_dir, tag=None):
+    """
+    1. Put the provided model to cpu
+    2. Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict``
+    3. Load it into the provided model
+
+    Args:
+        - ``model``: the model object to update
+        - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
+        - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
+
+    Returns:
+        - ``model`: modified model
+
+    Make sure you have plenty of CPU memory available before you call this function. If you don't
+    have enough use the ``zero_to_fp32.py`` utility to do the conversion. You will find it
+    conveniently placed for you in the checkpoint folder.
+
+    A typical usage might be ::
+
+        from deepspeed.utils.zero_to_fp32 import load_state_dict_from_zero_checkpoint
+        model = load_state_dict_from_zero_checkpoint(trainer.model, checkpoint_dir)
+        # submit to model hub or save the model to share with others
+
+    Note, that once this was run, the ``model`` will no longer be usable in the deepspeed context
+    of the same application. i.e. you will need to re-initialize the deepspeed engine, since
+    ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
+
+    """
+    logger.info(f"Extracting fp32 weights")
+    state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag)
+
+    logger.info(f"Overwriting model with fp32 weights")
+    model = model.cpu()
+    model.load_state_dict(state_dict, strict=False)
+
+    return model
+
+
+if __name__ == "__main__":
+
+    parser = argparse.ArgumentParser()
+    parser.add_argument(
+        "checkpoint_dir",
+        type=str,
+        help="path to the desired checkpoint folder, e.g., path/checkpoint-12")
+    parser.add_argument(
+        "output_file",
+        type=str,
+        help=
+        "path to the pytorch fp32 state_dict output file (e.g. path/checkpoint-12/pytorch_model.bin)"
+    )
+    parser.add_argument("-d", "--debug", action='store_true', help="enable debug")
+    args = parser.parse_args()
+
+    debug = args.debug
+
+    convert_zero_checkpoint_to_fp32_state_dict(args.checkpoint_dir, args.output_file)

pytorch_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8ea8f19b9cfae5ac1f4aea53be2d9bc355af72960cde88e7effe294d0e4c658a
+size 59121639

runs/Feb22_11-16-27_user-SYS-5049A-TR/1677032209.4666054/events.out.tfevents.1677032209.user-SYS-5049A-TR.55703.1

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:41cbcb6bee5dad543a261dde1389d4ce02265542aab50a2d33ea2155f5d50a01
+size 5216

runs/Feb22_11-16-27_user-SYS-5049A-TR/events.out.tfevents.1677032209.user-SYS-5049A-TR.55703.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1b5e6ec22f6237d4a73cb5a4cb804a2c4a4a8a5a9b3bee19852fac349088ab1e
+size 25754

special_tokens_map.json

@@ -0,0 +1 @@
+{"unk_token": "[UNK]", "sep_token": "[SEP]", "pad_token": "[PAD]", "cls_token": "[CLS]", "mask_token": "[MASK]"}

tokenizer_config.json

@@ -0,0 +1 @@
+{"cls_token": "[CLS]", "mask_token": "[MASK]", "model_max_length": 128, "pad_token": "[PAD]", "padding_side": "right", "sep_token": "[SEP]", "truncation_side": "right", "unk_token": "[UNK]", "special_tokens_map_file": "pretrained_tokenizers/UnidicBpe2/special_tokens_map.json", "name_or_path": "pretrained_tokenizers/UnidicBpe2", "tokenizer_class": "PreTrainedTokenizerFast"}

training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bd54042cd89784f24506b5912f250d73ea83faa93bfe5b987bbcaed0572116e8
+size 4335