Update from wendaxu

added_tokens.json

@@ -0,0 +1,3 @@
+{
+  "[PAD]": 32000
+}

config.json

@@ -0,0 +1,24 @@
+{
+  "_name_or_path": "decapoda-research/llama-7b-hf",
+  "architectures": [
+    "LlamaForCausalLM"
+  ],
+  "bos_token_id": 0,
+  "eos_token_id": 1,
+  "hidden_act": "silu",
+  "hidden_size": 4096,
+  "initializer_range": 0.02,
+  "intermediate_size": 11008,
+  "max_position_embeddings": 2048,
+  "max_sequence_length": 2048,
+  "model_type": "llama",
+  "num_attention_heads": 32,
+  "num_hidden_layers": 32,
+  "pad_token_id": -1,
+  "rms_norm_eps": 1e-06,
+  "tie_word_embeddings": false,
+  "torch_dtype": "float16",
+  "transformers_version": "4.29.2",
+  "use_cache": true,
+  "vocab_size": 32001
+}

generation_config.json

@@ -0,0 +1,7 @@
+{
+  "_from_model_config": true,
+  "bos_token_id": 0,
+  "eos_token_id": 1,
+  "pad_token_id": 0,
+  "transformers_version": "4.29.2"
+}

latest

@@ -0,0 +1 @@
+global_step462

pytorch_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:decafc0d45ea734f4032aac74c3cef832313fec38fbe5f1710b9d8292f7d8ac8
+size 13476958625

rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4c1d4aab692d41c18c919a673e506a5264ad9de9ff132b0de5dab52510157c75
+size 21687

rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:72575f2b4522908157bba855533d15354ced6598ecdc09ca8ce5417fa99c9b92
+size 21687

rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:eabb43b1fac99cd8e34e58482a942b31a1855cd53120560e1428e8a2aae185e9
+size 21687

rng_state_3.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cfd9568c2d39f428f86eb5a99beb6f19ac57f3f024ec4494b8cadd107d148298
+size 21687

rng_state_4.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:21ab6cb42d703108752ec781af0a5227b721321c411bb7f7bca00b6ec4e15324
+size 21687

rng_state_5.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c9fe828193c13a4a1adccaf24735d85428cb925bfa6204603f3a0a4a34c8cec9
+size 21687

rng_state_6.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:36e15c4dcb706b97a74d0f8259d6466dd2e46e60fd71dcb2c6d99e1e9ff2f204
+size 21687

rng_state_7.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7a467ddda52bc75625e56adc924121f89adb1e65a5733eec05f6284586a9a926
+size 21687

special_tokens_map.json

@@ -0,0 +1,6 @@
+{
+  "bos_token": "</s>",
+  "eos_token": "</s>",
+  "pad_token": "[PAD]",
+  "unk_token": "</s>"
+}

tokenizer.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9e556afd44213b6bd1be2b850ebbbd98f5481437a8021afaf58ee7fb1818d347
+size 499723

tokenizer_config.json

@@ -0,0 +1,34 @@
+{
+  "add_bos_token": true,
+  "add_eos_token": false,
+  "bos_token": {
+    "__type": "AddedToken",
+    "content": "",
+    "lstrip": false,
+    "normalized": true,
+    "rstrip": false,
+    "single_word": false
+  },
+  "clean_up_tokenization_spaces": false,
+  "eos_token": {
+    "__type": "AddedToken",
+    "content": "",
+    "lstrip": false,
+    "normalized": true,
+    "rstrip": false,
+    "single_word": false
+  },
+  "model_max_length": 720,
+  "pad_token": null,
+  "padding_side": "right",
+  "sp_model_kwargs": {},
+  "tokenizer_class": "LlamaTokenizer",
+  "unk_token": {
+    "__type": "AddedToken",
+    "content": "",
+    "lstrip": false,
+    "normalized": true,
+    "rstrip": false,
+    "single_word": false
+  }
+}

trainer_state.json

@@ -0,0 +1,2788 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 2.986666666666667,
+  "global_step": 462,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.01,
+      "learning_rate": 0,
+      "loss": 11.2781,
+      "step": 1
+    },
+    {
+      "epoch": 0.01,
+      "learning_rate": 0,
+      "loss": 11.4127,
+      "step": 2
+    },
+    {
+      "epoch": 0.02,
+      "learning_rate": 0,
+      "loss": 11.5953,
+      "step": 3
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0,
+      "loss": 11.382,
+      "step": 4
+    },
+    {
+      "epoch": 0.03,
+      "learning_rate": 0,
+      "loss": 11.0073,
+      "step": 5
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 0,
+      "loss": 11.1143,
+      "step": 6
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0,
+      "loss": 11.2086,
+      "step": 7
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0,
+      "loss": 10.3722,
+      "step": 8
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 0.0,
+      "loss": 10.2908,
+      "step": 9
+    },
+    {
+      "epoch": 0.06,
+      "learning_rate": 2e-05,
+      "loss": 10.135,
+      "step": 10
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 2e-05,
+      "loss": 10.3976,
+      "step": 11
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2e-05,
+      "loss": 9.583,
+      "step": 12
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 2e-05,
+      "loss": 9.2926,
+      "step": 13
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2e-05,
+      "loss": 9.1564,
+      "step": 14
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2e-05,
+      "loss": 8.9126,
+      "step": 15
+    },
+    {
+      "epoch": 0.1,
+      "learning_rate": 2e-05,
+      "loss": 8.4103,
+      "step": 16
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 2e-05,
+      "loss": 8.4349,
+      "step": 17
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2e-05,
+      "loss": 8.6729,
+      "step": 18
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 2e-05,
+      "loss": 8.5234,
+      "step": 19
+    },
+    {
+      "epoch": 0.13,
+      "learning_rate": 2e-05,
+      "loss": 8.2188,
+      "step": 20
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2e-05,
+      "loss": 8.3895,
+      "step": 21
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 2e-05,
+      "loss": 8.3381,
+      "step": 22
+    },
+    {
+      "epoch": 0.15,
+      "learning_rate": 2e-05,
+      "loss": 7.7987,
+      "step": 23
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2e-05,
+      "loss": 7.9355,
+      "step": 24
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 2e-05,
+      "loss": 7.8777,
+      "step": 25
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2e-05,
+      "loss": 7.3534,
+      "step": 26
+    },
+    {
+      "epoch": 0.17,
+      "learning_rate": 2e-05,
+      "loss": 7.2853,
+      "step": 27
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 2e-05,
+      "loss": 7.3128,
+      "step": 28
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2e-05,
+      "loss": 7.1632,
+      "step": 29
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 2e-05,
+      "loss": 7.0285,
+      "step": 30
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 2e-05,
+      "loss": 6.8622,
+      "step": 31
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2e-05,
+      "loss": 6.4805,
+      "step": 32
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 2e-05,
+      "loss": 6.2725,
+      "step": 33
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 2e-05,
+      "loss": 6.1005,
+      "step": 34
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2e-05,
+      "loss": 6.1369,
+      "step": 35
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 2e-05,
+      "loss": 5.7429,
+      "step": 36
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 2e-05,
+      "loss": 5.5371,
+      "step": 37
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2e-05,
+      "loss": 5.4108,
+      "step": 38
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 2e-05,
+      "loss": 5.0907,
+      "step": 39
+    },
+    {
+      "epoch": 0.26,
+      "learning_rate": 2e-05,
+      "loss": 5.0173,
+      "step": 40
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2e-05,
+      "loss": 4.6879,
+      "step": 41
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2e-05,
+      "loss": 4.584,
+      "step": 42
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2e-05,
+      "loss": 4.368,
+      "step": 43
+    },
+    {
+      "epoch": 0.28,
+      "learning_rate": 2e-05,
+      "loss": 4.0775,
+      "step": 44
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 2e-05,
+      "loss": 3.9747,
+      "step": 45
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2e-05,
+      "loss": 3.8133,
+      "step": 46
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 2e-05,
+      "loss": 3.578,
+      "step": 47
+    },
+    {
+      "epoch": 0.31,
+      "learning_rate": 2e-05,
+      "loss": 3.1794,
+      "step": 48
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2e-05,
+      "loss": 3.1259,
+      "step": 49
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 2e-05,
+      "loss": 2.8025,
+      "step": 50
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 2e-05,
+      "loss": 2.7038,
+      "step": 51
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2e-05,
+      "loss": 2.3681,
+      "step": 52
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 2e-05,
+      "loss": 2.0786,
+      "step": 53
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 2e-05,
+      "loss": 1.9681,
+      "step": 54
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2e-05,
+      "loss": 1.7678,
+      "step": 55
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 2e-05,
+      "loss": 1.5556,
+      "step": 56
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2e-05,
+      "loss": 1.369,
+      "step": 57
+    },
+    {
+      "epoch": 0.37,
+      "learning_rate": 2e-05,
+      "loss": 1.2035,
+      "step": 58
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 2e-05,
+      "loss": 1.0476,
+      "step": 59
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2e-05,
+      "loss": 0.9029,
+      "step": 60
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 2e-05,
+      "loss": 0.7663,
+      "step": 61
+    },
+    {
+      "epoch": 0.4,
+      "learning_rate": 2e-05,
+      "loss": 0.6331,
+      "step": 62
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2e-05,
+      "loss": 0.5272,
+      "step": 63
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 2e-05,
+      "loss": 0.4351,
+      "step": 64
+    },
+    {
+      "epoch": 0.42,
+      "learning_rate": 2e-05,
+      "loss": 0.371,
+      "step": 65
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2e-05,
+      "loss": 0.3029,
+      "step": 66
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 2e-05,
+      "loss": 0.2533,
+      "step": 67
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 2e-05,
+      "loss": 0.2196,
+      "step": 68
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2e-05,
+      "loss": 0.205,
+      "step": 69
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 2e-05,
+      "loss": 0.1685,
+      "step": 70
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 2e-05,
+      "loss": 0.1476,
+      "step": 71
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2e-05,
+      "loss": 0.1499,
+      "step": 72
+    },
+    {
+      "epoch": 0.47,
+      "learning_rate": 2e-05,
+      "loss": 0.1443,
+      "step": 73
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2e-05,
+      "loss": 0.1291,
+      "step": 74
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 2e-05,
+      "loss": 0.1138,
+      "step": 75
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 2e-05,
+      "loss": 0.0969,
+      "step": 76
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2e-05,
+      "loss": 0.0956,
+      "step": 77
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 2e-05,
+      "loss": 0.1048,
+      "step": 78
+    },
+    {
+      "epoch": 0.51,
+      "learning_rate": 2e-05,
+      "loss": 0.0915,
+      "step": 79
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2e-05,
+      "loss": 0.0985,
+      "step": 80
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 2e-05,
+      "loss": 0.1121,
+      "step": 81
+    },
+    {
+      "epoch": 0.53,
+      "learning_rate": 2e-05,
+      "loss": 0.1009,
+      "step": 82
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2e-05,
+      "loss": 0.1013,
+      "step": 83
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 2e-05,
+      "loss": 0.0991,
+      "step": 84
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 2e-05,
+      "loss": 0.0958,
+      "step": 85
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2e-05,
+      "loss": 0.0892,
+      "step": 86
+    },
+    {
+      "epoch": 0.56,
+      "learning_rate": 2e-05,
+      "loss": 0.0774,
+      "step": 87
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 2e-05,
+      "loss": 0.072,
+      "step": 88
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2e-05,
+      "loss": 0.0864,
+      "step": 89
+    },
+    {
+      "epoch": 0.58,
+      "learning_rate": 2e-05,
+      "loss": 0.0764,
+      "step": 90
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2e-05,
+      "loss": 0.086,
+      "step": 91
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 2e-05,
+      "loss": 0.0828,
+      "step": 92
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 2e-05,
+      "loss": 0.0702,
+      "step": 93
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2e-05,
+      "loss": 0.0835,
+      "step": 94
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 2e-05,
+      "loss": 0.0776,
+      "step": 95
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 2e-05,
+      "loss": 0.0802,
+      "step": 96
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2e-05,
+      "loss": 0.0798,
+      "step": 97
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2e-05,
+      "loss": 0.0695,
+      "step": 98
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 2e-05,
+      "loss": 0.0788,
+      "step": 99
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2e-05,
+      "loss": 0.0625,
+      "step": 100
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 2e-05,
+      "loss": 0.0742,
+      "step": 101
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 2e-05,
+      "loss": 0.0687,
+      "step": 102
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2e-05,
+      "loss": 0.0863,
+      "step": 103
+    },
+    {
+      "epoch": 0.67,
+      "learning_rate": 2e-05,
+      "loss": 0.0828,
+      "step": 104
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 2e-05,
+      "loss": 0.0698,
+      "step": 105
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2e-05,
+      "loss": 0.0728,
+      "step": 106
+    },
+    {
+      "epoch": 0.69,
+      "learning_rate": 2e-05,
+      "loss": 0.065,
+      "step": 107
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2e-05,
+      "loss": 0.0503,
+      "step": 108
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 2e-05,
+      "loss": 0.0714,
+      "step": 109
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 2e-05,
+      "loss": 0.0669,
+      "step": 110
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2e-05,
+      "loss": 0.0836,
+      "step": 111
+    },
+    {
+      "epoch": 0.72,
+      "learning_rate": 2e-05,
+      "loss": 0.0634,
+      "step": 112
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 2e-05,
+      "loss": 0.0705,
+      "step": 113
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2e-05,
+      "loss": 0.074,
+      "step": 114
+    },
+    {
+      "epoch": 0.74,
+      "learning_rate": 2e-05,
+      "loss": 0.0783,
+      "step": 115
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 2e-05,
+      "loss": 0.0756,
+      "step": 116
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2e-05,
+      "loss": 0.0498,
+      "step": 117
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 2e-05,
+      "loss": 0.0743,
+      "step": 118
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 2e-05,
+      "loss": 0.0813,
+      "step": 119
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2e-05,
+      "loss": 0.0648,
+      "step": 120
+    },
+    {
+      "epoch": 0.78,
+      "learning_rate": 2e-05,
+      "loss": 0.0611,
+      "step": 121
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 2e-05,
+      "loss": 0.0694,
+      "step": 122
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2e-05,
+      "loss": 0.0676,
+      "step": 123
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 2e-05,
+      "loss": 0.0662,
+      "step": 124
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2e-05,
+      "loss": 0.0656,
+      "step": 125
+    },
+    {
+      "epoch": 0.81,
+      "learning_rate": 2e-05,
+      "loss": 0.065,
+      "step": 126
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 2e-05,
+      "loss": 0.0744,
+      "step": 127
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2e-05,
+      "loss": 0.0583,
+      "step": 128
+    },
+    {
+      "epoch": 0.83,
+      "learning_rate": 2e-05,
+      "loss": 0.0681,
+      "step": 129
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 2e-05,
+      "loss": 0.0604,
+      "step": 130
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2e-05,
+      "loss": 0.0591,
+      "step": 131
+    },
+    {
+      "epoch": 0.85,
+      "learning_rate": 2e-05,
+      "loss": 0.057,
+      "step": 132
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 2e-05,
+      "loss": 0.0826,
+      "step": 133
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2e-05,
+      "loss": 0.0708,
+      "step": 134
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 2e-05,
+      "loss": 0.0701,
+      "step": 135
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 2e-05,
+      "loss": 0.0724,
+      "step": 136
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2e-05,
+      "loss": 0.0713,
+      "step": 137
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 2e-05,
+      "loss": 0.0721,
+      "step": 138
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 2e-05,
+      "loss": 0.077,
+      "step": 139
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2e-05,
+      "loss": 0.0637,
+      "step": 140
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 2e-05,
+      "loss": 0.0707,
+      "step": 141
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2e-05,
+      "loss": 0.0685,
+      "step": 142
+    },
+    {
+      "epoch": 0.92,
+      "learning_rate": 2e-05,
+      "loss": 0.0646,
+      "step": 143
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 2e-05,
+      "loss": 0.0778,
+      "step": 144
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2e-05,
+      "loss": 0.0726,
+      "step": 145
+    },
+    {
+      "epoch": 0.94,
+      "learning_rate": 2e-05,
+      "loss": 0.0711,
+      "step": 146
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 2e-05,
+      "loss": 0.073,
+      "step": 147
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2e-05,
+      "loss": 0.0704,
+      "step": 148
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 2e-05,
+      "loss": 0.0618,
+      "step": 149
+    },
+    {
+      "epoch": 0.97,
+      "learning_rate": 2e-05,
+      "loss": 0.0565,
+      "step": 150
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2e-05,
+      "loss": 0.0673,
+      "step": 151
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 2e-05,
+      "loss": 0.064,
+      "step": 152
+    },
+    {
+      "epoch": 0.99,
+      "learning_rate": 2e-05,
+      "loss": 0.0681,
+      "step": 153
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2e-05,
+      "loss": 0.0599,
+      "step": 154
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 2e-05,
+      "loss": 0.0666,
+      "step": 155
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 2e-05,
+      "loss": 0.0525,
+      "step": 156
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 2e-05,
+      "loss": 0.0543,
+      "step": 157
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 2e-05,
+      "loss": 0.0582,
+      "step": 158
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 2e-05,
+      "loss": 0.0651,
+      "step": 159
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 2e-05,
+      "loss": 0.0478,
+      "step": 160
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 2e-05,
+      "loss": 0.0516,
+      "step": 161
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 2e-05,
+      "loss": 0.048,
+      "step": 162
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 2e-05,
+      "loss": 0.0582,
+      "step": 163
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 2e-05,
+      "loss": 0.0479,
+      "step": 164
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 2e-05,
+      "loss": 0.0527,
+      "step": 165
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 2e-05,
+      "loss": 0.0444,
+      "step": 166
+    },
+    {
+      "epoch": 1.08,
+      "learning_rate": 2e-05,
+      "loss": 0.048,
+      "step": 167
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 2e-05,
+      "loss": 0.0552,
+      "step": 168
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 2e-05,
+      "loss": 0.054,
+      "step": 169
+    },
+    {
+      "epoch": 1.1,
+      "learning_rate": 2e-05,
+      "loss": 0.0552,
+      "step": 170
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 2e-05,
+      "loss": 0.0507,
+      "step": 171
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 2e-05,
+      "loss": 0.0503,
+      "step": 172
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 2e-05,
+      "loss": 0.054,
+      "step": 173
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 2e-05,
+      "loss": 0.0508,
+      "step": 174
+    },
+    {
+      "epoch": 1.13,
+      "learning_rate": 2e-05,
+      "loss": 0.053,
+      "step": 175
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 2e-05,
+      "loss": 0.0412,
+      "step": 176
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 2e-05,
+      "loss": 0.0504,
+      "step": 177
+    },
+    {
+      "epoch": 1.15,
+      "learning_rate": 2e-05,
+      "loss": 0.0589,
+      "step": 178
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 2e-05,
+      "loss": 0.0492,
+      "step": 179
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 2e-05,
+      "loss": 0.0532,
+      "step": 180
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 2e-05,
+      "loss": 0.0568,
+      "step": 181
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 2e-05,
+      "loss": 0.0634,
+      "step": 182
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 2e-05,
+      "loss": 0.0591,
+      "step": 183
+    },
+    {
+      "epoch": 1.19,
+      "learning_rate": 2e-05,
+      "loss": 0.0404,
+      "step": 184
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 2e-05,
+      "loss": 0.0565,
+      "step": 185
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 2e-05,
+      "loss": 0.0472,
+      "step": 186
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 2e-05,
+      "loss": 0.041,
+      "step": 187
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 2e-05,
+      "loss": 0.0506,
+      "step": 188
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 2e-05,
+      "loss": 0.0487,
+      "step": 189
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 2e-05,
+      "loss": 0.0536,
+      "step": 190
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 2e-05,
+      "loss": 0.0459,
+      "step": 191
+    },
+    {
+      "epoch": 1.24,
+      "learning_rate": 2e-05,
+      "loss": 0.0484,
+      "step": 192
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 2e-05,
+      "loss": 0.0454,
+      "step": 193
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 2e-05,
+      "loss": 0.062,
+      "step": 194
+    },
+    {
+      "epoch": 1.26,
+      "learning_rate": 2e-05,
+      "loss": 0.0509,
+      "step": 195
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 2e-05,
+      "loss": 0.055,
+      "step": 196
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 2e-05,
+      "loss": 0.0578,
+      "step": 197
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 2e-05,
+      "loss": 0.0516,
+      "step": 198
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 2e-05,
+      "loss": 0.0545,
+      "step": 199
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 2e-05,
+      "loss": 0.0597,
+      "step": 200
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 2e-05,
+      "loss": 0.0523,
+      "step": 201
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 2e-05,
+      "loss": 0.0457,
+      "step": 202
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 2e-05,
+      "loss": 0.0523,
+      "step": 203
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 2e-05,
+      "loss": 0.0578,
+      "step": 204
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 2e-05,
+      "loss": 0.0475,
+      "step": 205
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 2e-05,
+      "loss": 0.0513,
+      "step": 206
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 2e-05,
+      "loss": 0.0485,
+      "step": 207
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 2e-05,
+      "loss": 0.0421,
+      "step": 208
+    },
+    {
+      "epoch": 1.35,
+      "learning_rate": 2e-05,
+      "loss": 0.0451,
+      "step": 209
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 2e-05,
+      "loss": 0.0461,
+      "step": 210
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 2e-05,
+      "loss": 0.0474,
+      "step": 211
+    },
+    {
+      "epoch": 1.37,
+      "learning_rate": 2e-05,
+      "loss": 0.0565,
+      "step": 212
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 2e-05,
+      "loss": 0.0582,
+      "step": 213
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 2e-05,
+      "loss": 0.0499,
+      "step": 214
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 2e-05,
+      "loss": 0.0604,
+      "step": 215
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2e-05,
+      "loss": 0.0517,
+      "step": 216
+    },
+    {
+      "epoch": 1.4,
+      "learning_rate": 2e-05,
+      "loss": 0.0545,
+      "step": 217
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 2e-05,
+      "loss": 0.0493,
+      "step": 218
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2e-05,
+      "loss": 0.0497,
+      "step": 219
+    },
+    {
+      "epoch": 1.42,
+      "learning_rate": 2e-05,
+      "loss": 0.0533,
+      "step": 220
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 2e-05,
+      "loss": 0.0492,
+      "step": 221
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2e-05,
+      "loss": 0.0507,
+      "step": 222
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 2e-05,
+      "loss": 0.0563,
+      "step": 223
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2e-05,
+      "loss": 0.0479,
+      "step": 224
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 2e-05,
+      "loss": 0.0511,
+      "step": 225
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 2e-05,
+      "loss": 0.0484,
+      "step": 226
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2e-05,
+      "loss": 0.0445,
+      "step": 227
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 2e-05,
+      "loss": 0.0545,
+      "step": 228
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 2e-05,
+      "loss": 0.052,
+      "step": 229
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2e-05,
+      "loss": 0.0548,
+      "step": 230
+    },
+    {
+      "epoch": 1.49,
+      "learning_rate": 2e-05,
+      "loss": 0.051,
+      "step": 231
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 2e-05,
+      "loss": 0.0509,
+      "step": 232
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2e-05,
+      "loss": 0.044,
+      "step": 233
+    },
+    {
+      "epoch": 1.51,
+      "learning_rate": 2e-05,
+      "loss": 0.0561,
+      "step": 234
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 2e-05,
+      "loss": 0.0468,
+      "step": 235
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2e-05,
+      "loss": 0.0425,
+      "step": 236
+    },
+    {
+      "epoch": 1.53,
+      "learning_rate": 2e-05,
+      "loss": 0.0496,
+      "step": 237
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 2e-05,
+      "loss": 0.0502,
+      "step": 238
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2e-05,
+      "loss": 0.0514,
+      "step": 239
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 2e-05,
+      "loss": 0.0437,
+      "step": 240
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2e-05,
+      "loss": 0.0551,
+      "step": 241
+    },
+    {
+      "epoch": 1.56,
+      "learning_rate": 2e-05,
+      "loss": 0.0513,
+      "step": 242
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 2e-05,
+      "loss": 0.0503,
+      "step": 243
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2e-05,
+      "loss": 0.0443,
+      "step": 244
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 2e-05,
+      "loss": 0.0408,
+      "step": 245
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 2e-05,
+      "loss": 0.0506,
+      "step": 246
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2e-05,
+      "loss": 0.0586,
+      "step": 247
+    },
+    {
+      "epoch": 1.6,
+      "learning_rate": 2e-05,
+      "loss": 0.0514,
+      "step": 248
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 2e-05,
+      "loss": 0.0495,
+      "step": 249
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 2e-05,
+      "loss": 0.0606,
+      "step": 250
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 2e-05,
+      "loss": 0.0471,
+      "step": 251
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 2e-05,
+      "loss": 0.0615,
+      "step": 252
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 2e-05,
+      "loss": 0.0507,
+      "step": 253
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 2e-05,
+      "loss": 0.0545,
+      "step": 254
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 2e-05,
+      "loss": 0.0395,
+      "step": 255
+    },
+    {
+      "epoch": 1.65,
+      "learning_rate": 2e-05,
+      "loss": 0.0524,
+      "step": 256
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 2e-05,
+      "loss": 0.0486,
+      "step": 257
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 2e-05,
+      "loss": 0.0532,
+      "step": 258
+    },
+    {
+      "epoch": 1.67,
+      "learning_rate": 2e-05,
+      "loss": 0.0449,
+      "step": 259
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 2e-05,
+      "loss": 0.0459,
+      "step": 260
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 2e-05,
+      "loss": 0.0575,
+      "step": 261
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 2e-05,
+      "loss": 0.0569,
+      "step": 262
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 2e-05,
+      "loss": 0.0451,
+      "step": 263
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2e-05,
+      "loss": 0.061,
+      "step": 264
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 2e-05,
+      "loss": 0.0573,
+      "step": 265
+    },
+    {
+      "epoch": 1.72,
+      "learning_rate": 2e-05,
+      "loss": 0.0611,
+      "step": 266
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2e-05,
+      "loss": 0.0522,
+      "step": 267
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 2e-05,
+      "loss": 0.0457,
+      "step": 268
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 2e-05,
+      "loss": 0.0399,
+      "step": 269
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2e-05,
+      "loss": 0.0512,
+      "step": 270
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 2e-05,
+      "loss": 0.0513,
+      "step": 271
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2e-05,
+      "loss": 0.0441,
+      "step": 272
+    },
+    {
+      "epoch": 1.76,
+      "learning_rate": 2e-05,
+      "loss": 0.0463,
+      "step": 273
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 2e-05,
+      "loss": 0.0517,
+      "step": 274
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2e-05,
+      "loss": 0.0507,
+      "step": 275
+    },
+    {
+      "epoch": 1.78,
+      "learning_rate": 2e-05,
+      "loss": 0.0527,
+      "step": 276
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 2e-05,
+      "loss": 0.0456,
+      "step": 277
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2e-05,
+      "loss": 0.0472,
+      "step": 278
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 2e-05,
+      "loss": 0.0532,
+      "step": 279
+    },
+    {
+      "epoch": 1.81,
+      "learning_rate": 2e-05,
+      "loss": 0.0456,
+      "step": 280
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2e-05,
+      "loss": 0.0497,
+      "step": 281
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 2e-05,
+      "loss": 0.0444,
+      "step": 282
+    },
+    {
+      "epoch": 1.83,
+      "learning_rate": 2e-05,
+      "loss": 0.0492,
+      "step": 283
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2e-05,
+      "loss": 0.0513,
+      "step": 284
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 2e-05,
+      "loss": 0.0565,
+      "step": 285
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 2e-05,
+      "loss": 0.0517,
+      "step": 286
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2e-05,
+      "loss": 0.0441,
+      "step": 287
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 2e-05,
+      "loss": 0.0613,
+      "step": 288
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2e-05,
+      "loss": 0.0411,
+      "step": 289
+    },
+    {
+      "epoch": 1.87,
+      "learning_rate": 2e-05,
+      "loss": 0.0483,
+      "step": 290
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 2e-05,
+      "loss": 0.0518,
+      "step": 291
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 2e-05,
+      "loss": 0.0498,
+      "step": 292
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 2e-05,
+      "loss": 0.0572,
+      "step": 293
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 2e-05,
+      "loss": 0.0475,
+      "step": 294
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 2e-05,
+      "loss": 0.0632,
+      "step": 295
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 2e-05,
+      "loss": 0.0549,
+      "step": 296
+    },
+    {
+      "epoch": 1.92,
+      "learning_rate": 2e-05,
+      "loss": 0.0501,
+      "step": 297
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 2e-05,
+      "loss": 0.054,
+      "step": 298
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 2e-05,
+      "loss": 0.0535,
+      "step": 299
+    },
+    {
+      "epoch": 1.94,
+      "learning_rate": 2e-05,
+      "loss": 0.0489,
+      "step": 300
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2e-05,
+      "loss": 0.0422,
+      "step": 301
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 2e-05,
+      "loss": 0.0475,
+      "step": 302
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 2e-05,
+      "loss": 0.0526,
+      "step": 303
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 2e-05,
+      "loss": 0.0488,
+      "step": 304
+    },
+    {
+      "epoch": 1.97,
+      "learning_rate": 2e-05,
+      "loss": 0.0512,
+      "step": 305
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 2e-05,
+      "loss": 0.0476,
+      "step": 306
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 2e-05,
+      "loss": 0.0533,
+      "step": 307
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 2e-05,
+      "loss": 0.0489,
+      "step": 308
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 2e-05,
+      "loss": 0.0461,
+      "step": 309
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 2e-05,
+      "loss": 0.0358,
+      "step": 310
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 2e-05,
+      "loss": 0.0284,
+      "step": 311
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 2e-05,
+      "loss": 0.0301,
+      "step": 312
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 2e-05,
+      "loss": 0.0271,
+      "step": 313
+    },
+    {
+      "epoch": 2.03,
+      "learning_rate": 2e-05,
+      "loss": 0.0245,
+      "step": 314
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 2e-05,
+      "loss": 0.0264,
+      "step": 315
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 2e-05,
+      "loss": 0.0292,
+      "step": 316
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 2e-05,
+      "loss": 0.0278,
+      "step": 317
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 2e-05,
+      "loss": 0.0301,
+      "step": 318
+    },
+    {
+      "epoch": 2.06,
+      "learning_rate": 2e-05,
+      "loss": 0.0284,
+      "step": 319
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 2e-05,
+      "loss": 0.0205,
+      "step": 320
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 2e-05,
+      "loss": 0.0286,
+      "step": 321
+    },
+    {
+      "epoch": 2.08,
+      "learning_rate": 2e-05,
+      "loss": 0.0255,
+      "step": 322
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 2e-05,
+      "loss": 0.0267,
+      "step": 323
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 2e-05,
+      "loss": 0.0282,
+      "step": 324
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 2e-05,
+      "loss": 0.0263,
+      "step": 325
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 2e-05,
+      "loss": 0.0299,
+      "step": 326
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 2e-05,
+      "loss": 0.0254,
+      "step": 327
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 2e-05,
+      "loss": 0.0266,
+      "step": 328
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 2e-05,
+      "loss": 0.0213,
+      "step": 329
+    },
+    {
+      "epoch": 2.13,
+      "learning_rate": 2e-05,
+      "loss": 0.0254,
+      "step": 330
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 2e-05,
+      "loss": 0.0301,
+      "step": 331
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 2e-05,
+      "loss": 0.0247,
+      "step": 332
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 2e-05,
+      "loss": 0.0244,
+      "step": 333
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 2e-05,
+      "loss": 0.0308,
+      "step": 334
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 2e-05,
+      "loss": 0.03,
+      "step": 335
+    },
+    {
+      "epoch": 2.17,
+      "learning_rate": 2e-05,
+      "loss": 0.0285,
+      "step": 336
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 2e-05,
+      "loss": 0.0281,
+      "step": 337
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 2e-05,
+      "loss": 0.0258,
+      "step": 338
+    },
+    {
+      "epoch": 2.19,
+      "learning_rate": 2e-05,
+      "loss": 0.0224,
+      "step": 339
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 2e-05,
+      "loss": 0.0284,
+      "step": 340
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 2e-05,
+      "loss": 0.0221,
+      "step": 341
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 2e-05,
+      "loss": 0.0269,
+      "step": 342
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 2e-05,
+      "loss": 0.0279,
+      "step": 343
+    },
+    {
+      "epoch": 2.22,
+      "learning_rate": 2e-05,
+      "loss": 0.0296,
+      "step": 344
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 2e-05,
+      "loss": 0.0239,
+      "step": 345
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2e-05,
+      "loss": 0.026,
+      "step": 346
+    },
+    {
+      "epoch": 2.24,
+      "learning_rate": 2e-05,
+      "loss": 0.026,
+      "step": 347
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2e-05,
+      "loss": 0.0231,
+      "step": 348
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2e-05,
+      "loss": 0.029,
+      "step": 349
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 2e-05,
+      "loss": 0.0239,
+      "step": 350
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2e-05,
+      "loss": 0.0235,
+      "step": 351
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2e-05,
+      "loss": 0.0302,
+      "step": 352
+    },
+    {
+      "epoch": 2.28,
+      "learning_rate": 2e-05,
+      "loss": 0.0283,
+      "step": 353
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2e-05,
+      "loss": 0.0253,
+      "step": 354
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2e-05,
+      "loss": 0.0254,
+      "step": 355
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2e-05,
+      "loss": 0.0262,
+      "step": 356
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2e-05,
+      "loss": 0.0246,
+      "step": 357
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 2e-05,
+      "loss": 0.0338,
+      "step": 358
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2e-05,
+      "loss": 0.0314,
+      "step": 359
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2e-05,
+      "loss": 0.0294,
+      "step": 360
+    },
+    {
+      "epoch": 2.33,
+      "learning_rate": 2e-05,
+      "loss": 0.0258,
+      "step": 361
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2e-05,
+      "loss": 0.0233,
+      "step": 362
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2e-05,
+      "loss": 0.0298,
+      "step": 363
+    },
+    {
+      "epoch": 2.35,
+      "learning_rate": 2e-05,
+      "loss": 0.0283,
+      "step": 364
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2e-05,
+      "loss": 0.0245,
+      "step": 365
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2e-05,
+      "loss": 0.0258,
+      "step": 366
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 2e-05,
+      "loss": 0.0273,
+      "step": 367
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2e-05,
+      "loss": 0.0291,
+      "step": 368
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2e-05,
+      "loss": 0.0354,
+      "step": 369
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2e-05,
+      "loss": 0.0238,
+      "step": 370
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2e-05,
+      "loss": 0.0314,
+      "step": 371
+    },
+    {
+      "epoch": 2.4,
+      "learning_rate": 2e-05,
+      "loss": 0.0325,
+      "step": 372
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2e-05,
+      "loss": 0.0286,
+      "step": 373
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2e-05,
+      "loss": 0.0258,
+      "step": 374
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 2e-05,
+      "loss": 0.0238,
+      "step": 375
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2e-05,
+      "loss": 0.0251,
+      "step": 376
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2e-05,
+      "loss": 0.0263,
+      "step": 377
+    },
+    {
+      "epoch": 2.44,
+      "learning_rate": 2e-05,
+      "loss": 0.0259,
+      "step": 378
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2e-05,
+      "loss": 0.0282,
+      "step": 379
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2e-05,
+      "loss": 0.0217,
+      "step": 380
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2e-05,
+      "loss": 0.024,
+      "step": 381
+    },
+    {
+      "epoch": 2.47,
+      "learning_rate": 2e-05,
+      "loss": 0.0244,
+      "step": 382
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2e-05,
+      "loss": 0.0286,
+      "step": 383
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2e-05,
+      "loss": 0.029,
+      "step": 384
+    },
+    {
+      "epoch": 2.49,
+      "learning_rate": 2e-05,
+      "loss": 0.0278,
+      "step": 385
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2e-05,
+      "loss": 0.0293,
+      "step": 386
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2e-05,
+      "loss": 0.0293,
+      "step": 387
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2e-05,
+      "loss": 0.0295,
+      "step": 388
+    },
+    {
+      "epoch": 2.51,
+      "learning_rate": 2e-05,
+      "loss": 0.0266,
+      "step": 389
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2e-05,
+      "loss": 0.0285,
+      "step": 390
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2e-05,
+      "loss": 0.0334,
+      "step": 391
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 2e-05,
+      "loss": 0.0243,
+      "step": 392
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2e-05,
+      "loss": 0.0224,
+      "step": 393
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2e-05,
+      "loss": 0.0257,
+      "step": 394
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2e-05,
+      "loss": 0.0304,
+      "step": 395
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2e-05,
+      "loss": 0.0277,
+      "step": 396
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2e-05,
+      "loss": 0.03,
+      "step": 397
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2e-05,
+      "loss": 0.0279,
+      "step": 398
+    },
+    {
+      "epoch": 2.58,
+      "learning_rate": 2e-05,
+      "loss": 0.0234,
+      "step": 399
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2e-05,
+      "loss": 0.0306,
+      "step": 400
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2e-05,
+      "loss": 0.0261,
+      "step": 401
+    },
+    {
+      "epoch": 2.6,
+      "learning_rate": 2e-05,
+      "loss": 0.0277,
+      "step": 402
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2e-05,
+      "loss": 0.0277,
+      "step": 403
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2e-05,
+      "loss": 0.0298,
+      "step": 404
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2e-05,
+      "loss": 0.0214,
+      "step": 405
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2e-05,
+      "loss": 0.0272,
+      "step": 406
+    },
+    {
+      "epoch": 2.63,
+      "learning_rate": 2e-05,
+      "loss": 0.0283,
+      "step": 407
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2e-05,
+      "loss": 0.0267,
+      "step": 408
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2e-05,
+      "loss": 0.0248,
+      "step": 409
+    },
+    {
+      "epoch": 2.65,
+      "learning_rate": 2e-05,
+      "loss": 0.025,
+      "step": 410
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2e-05,
+      "loss": 0.0267,
+      "step": 411
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2e-05,
+      "loss": 0.025,
+      "step": 412
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2e-05,
+      "loss": 0.0293,
+      "step": 413
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2e-05,
+      "loss": 0.0279,
+      "step": 414
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2e-05,
+      "loss": 0.0257,
+      "step": 415
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 2e-05,
+      "loss": 0.027,
+      "step": 416
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2e-05,
+      "loss": 0.0329,
+      "step": 417
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2e-05,
+      "loss": 0.029,
+      "step": 418
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 2e-05,
+      "loss": 0.0204,
+      "step": 419
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 2e-05,
+      "loss": 0.0302,
+      "step": 420
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 2e-05,
+      "loss": 0.0298,
+      "step": 421
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 2e-05,
+      "loss": 0.0316,
+      "step": 422
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 2e-05,
+      "loss": 0.0293,
+      "step": 423
+    },
+    {
+      "epoch": 2.74,
+      "learning_rate": 2e-05,
+      "loss": 0.0292,
+      "step": 424
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 2e-05,
+      "loss": 0.0314,
+      "step": 425
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 2e-05,
+      "loss": 0.0264,
+      "step": 426
+    },
+    {
+      "epoch": 2.76,
+      "learning_rate": 2e-05,
+      "loss": 0.0225,
+      "step": 427
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2e-05,
+      "loss": 0.0312,
+      "step": 428
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 2e-05,
+      "loss": 0.0251,
+      "step": 429
+    },
+    {
+      "epoch": 2.78,
+      "learning_rate": 2e-05,
+      "loss": 0.0275,
+      "step": 430
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2e-05,
+      "loss": 0.0292,
+      "step": 431
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 2e-05,
+      "loss": 0.0253,
+      "step": 432
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 2e-05,
+      "loss": 0.032,
+      "step": 433
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2e-05,
+      "loss": 0.024,
+      "step": 434
+    },
+    {
+      "epoch": 2.81,
+      "learning_rate": 2e-05,
+      "loss": 0.0297,
+      "step": 435
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 2e-05,
+      "loss": 0.0249,
+      "step": 436
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 2e-05,
+      "loss": 0.029,
+      "step": 437
+    },
+    {
+      "epoch": 2.83,
+      "learning_rate": 2e-05,
+      "loss": 0.0314,
+      "step": 438
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 2e-05,
+      "loss": 0.0226,
+      "step": 439
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 2e-05,
+      "loss": 0.0283,
+      "step": 440
+    },
+    {
+      "epoch": 2.85,
+      "learning_rate": 2e-05,
+      "loss": 0.0317,
+      "step": 441
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 2e-05,
+      "loss": 0.0294,
+      "step": 442
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 2e-05,
+      "loss": 0.0278,
+      "step": 443
+    },
+    {
+      "epoch": 2.87,
+      "learning_rate": 2e-05,
+      "loss": 0.0319,
+      "step": 444
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 2e-05,
+      "loss": 0.0301,
+      "step": 445
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 2e-05,
+      "loss": 0.0233,
+      "step": 446
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 2e-05,
+      "loss": 0.0373,
+      "step": 447
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 2e-05,
+      "loss": 0.0248,
+      "step": 448
+    },
+    {
+      "epoch": 2.9,
+      "learning_rate": 2e-05,
+      "loss": 0.0268,
+      "step": 449
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 2e-05,
+      "loss": 0.0303,
+      "step": 450
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 2e-05,
+      "loss": 0.0271,
+      "step": 451
+    },
+    {
+      "epoch": 2.92,
+      "learning_rate": 2e-05,
+      "loss": 0.0263,
+      "step": 452
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2e-05,
+      "loss": 0.0276,
+      "step": 453
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 2e-05,
+      "loss": 0.0324,
+      "step": 454
+    },
+    {
+      "epoch": 2.94,
+      "learning_rate": 2e-05,
+      "loss": 0.0258,
+      "step": 455
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2e-05,
+      "loss": 0.0297,
+      "step": 456
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 2e-05,
+      "loss": 0.0305,
+      "step": 457
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 2e-05,
+      "loss": 0.0302,
+      "step": 458
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 2e-05,
+      "loss": 0.0279,
+      "step": 459
+    },
+    {
+      "epoch": 2.97,
+      "learning_rate": 2e-05,
+      "loss": 0.0264,
+      "step": 460
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 2e-05,
+      "loss": 0.0296,
+      "step": 461
+    },
+    {
+      "epoch": 2.99,
+      "learning_rate": 2e-05,
+      "loss": 0.0298,
+      "step": 462
+    }
+  ],
+  "max_steps": 462,
+  "num_train_epochs": 3,
+  "total_flos": 68006471270400.0,
+  "trial_name": null,
+  "trial_params": null
+}

training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:51b5c4e4aedb747c30399017adfddedbdc3ef91d69285e23c8199901ca59c243
+size 5371

zero_to_fp32.py

@@ -0,0 +1,578 @@
+#!/usr/bin/env python
+
+# Copyright (c) Microsoft Corporation.
+# SPDX-License-Identifier: Apache-2.0
+
+# DeepSpeed Team
+
+# 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
+from dataclasses import dataclass
+
+# 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,
+                                            FROZEN_PARAM_SHAPES, FROZEN_PARAM_FRAGMENTS)
+
+
+@dataclass
+class zero_model_state:
+    buffers: dict()
+    param_shapes: dict()
+    shared_params: list
+    ds_version: int
+    frozen_param_shapes: dict()
+    frozen_param_fragments: dict()
+
+
+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_checkpoint_files(checkpoint_dir, glob_pattern):
+    # XXX: need to test that this simple glob rule works for multi-node setup too
+    ckpt_files = sorted(glob.glob(os.path.join(checkpoint_dir, glob_pattern)), key=natural_keys)
+
+    if len(ckpt_files) == 0:
+        raise FileNotFoundError(f"can't find {glob_pattern} files in directory '{checkpoint_dir}'")
+
+    return ckpt_files
+
+
+def get_optim_files(checkpoint_dir):
+    return get_checkpoint_files(checkpoint_dir, "*_optim_states.pt")
+
+
+def get_model_state_files(checkpoint_dir):
+    return get_checkpoint_files(checkpoint_dir, "*_model_states.pt")
+
+
+def parse_model_states(files):
+    zero_model_states = []
+    for file in files:
+        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]
+
+        # collect parameters that are included in param_shapes
+        param_names = []
+        for s in param_shapes:
+            for name in s.keys():
+                param_names.append(name)
+
+        # update with frozen parameters
+        frozen_param_shapes = state_dict.get(FROZEN_PARAM_SHAPES, None)
+        if frozen_param_shapes is not None:
+            if debug:
+                print(f"Found frozen_param_shapes: {frozen_param_shapes}")
+            param_names += list(frozen_param_shapes.keys())
+
+        # handle shared params
+        shared_params = [[k, v] for k, v in state_dict["shared_params"].items()]
+
+        ds_version = state_dict.get(DS_VERSION, None)
+
+        frozen_param_fragments = state_dict.get(FROZEN_PARAM_FRAGMENTS, None)
+
+        z_model_state = zero_model_state(buffers=buffers,
+                                         param_shapes=param_shapes,
+                                         shared_params=shared_params,
+                                         ds_version=ds_version,
+                                         frozen_param_shapes=frozen_param_shapes,
+                                         frozen_param_fragments=frozen_param_fragments)
+        zero_model_states.append(z_model_state)
+
+    return zero_model_states
+
+
+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_files = get_model_state_files(ds_checkpoint_dir)
+
+    zero_model_states = parse_model_states(model_files)
+    print(f'Parsing checkpoint created by deepspeed=={zero_model_states[0].ds_version}')
+
+    if zero_stage == 2:
+        return _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states)
+    elif zero_stage == 3:
+        return _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states)
+
+
+def _zero2_merge_frozen_params(state_dict, zero_model_states):
+    if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
+        return
+
+    frozen_param_shapes = zero_model_states[0].frozen_param_shapes
+    frozen_param_fragments = zero_model_states[0].frozen_param_fragments
+
+    if debug:
+        num_elem = sum(s.numel() for s in frozen_param_shapes.values())
+        print(f'rank 0: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
+
+        wanted_params = len(frozen_param_shapes)
+        wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
+        avail_numel = sum([p.numel() for p in frozen_param_fragments.values()])
+        print(f'Frozen params: Have {avail_numel} numels to process.')
+        print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
+
+    total_params = 0
+    total_numel = 0
+    for name, shape in frozen_param_shapes.items():
+        total_params += 1
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+
+        state_dict[name] = frozen_param_fragments[name]
+
+        if debug:
+            print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
+
+    print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
+    param_shapes = zero_model_states[0].param_shapes
+
+    # 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.")
+
+    # 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")
+
+
+def _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    _zero2_merge_frozen_params(state_dict, zero_model_states)
+
+    _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
+
+    # recover shared parameters
+    for pair in zero_model_states[0].shared_params:
+        if pair[1] in state_dict:
+            state_dict[pair[0]] = state_dict[pair[1]]
+
+    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 _zero3_merge_frozen_params(state_dict, world_size, zero_model_states):
+    if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
+        return
+
+    if debug:
+        for i in range(world_size):
+            num_elem = sum(s.numel() for s in zero_model_states[i].frozen_param_fragments.values())
+            print(f'rank {i}: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
+
+        frozen_param_shapes = zero_model_states[0].frozen_param_shapes
+        wanted_params = len(frozen_param_shapes)
+        wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
+        avail_numel = sum([p.numel() for p in zero_model_states[0].frozen_param_fragments.values()]) * world_size
+        print(f'Frozen params: Have {avail_numel} numels to process.')
+        print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
+
+    total_params = 0
+    total_numel = 0
+    for name, shape in zero_model_states[0].frozen_param_shapes.items():
+        total_params += 1
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+
+        param_frags = tuple(model_state.frozen_param_fragments[name] for model_state in zero_model_states)
+        state_dict[name] = torch.cat(param_frags, 0).narrow(0, 0, unpartitioned_numel).view(shape)
+
+        partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
+
+        if debug:
+            print(
+                f"Frozen params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
+            )
+
+    print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
+    param_shapes = zero_model_states[0].param_shapes
+    avail_numel = fp32_flat_groups[0].numel() * world_size
+    # 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
+
+    # 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
+        avail_numel = fp32_flat_groups[0].numel() * world_size
+        print(f"Trainable params: Have {avail_numel} numels to process.")
+        print(f"Trainable params: Need {wanted_numel} numels in {wanted_params} params.")
+
+    # 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"Trainable params: {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 Trainable fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    _zero3_merge_frozen_params(state_dict, world_size, zero_model_states)
+
+    _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
+
+    # recover shared parameters
+    for pair in zero_model_states[0].shared_params:
+        if pair[1] in state_dict:
+            state_dict[pair[0]] = state_dict[pair[1]]
+
+    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)