Upload folder using huggingface_hub

.gitattributes

@@ -33,3 +33,5 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+checkpoint-390/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+tokenizer.json filter=lfs diff=lfs merge=lfs -text

checkpoint-390/config.json

@@ -0,0 +1,28 @@
+{
+  "_name_or_path": "/share/chaofan/code/IR-Studio-up/online_test/finetune_result/medical_huatuo/retriever_round2",
+  "architectures": [
+    "XLMRobertaModel"
+  ],
+  "attention_probs_dropout_prob": 0.1,
+  "bos_token_id": 0,
+  "classifier_dropout": null,
+  "eos_token_id": 2,
+  "hidden_act": "gelu",
+  "hidden_dropout_prob": 0.1,
+  "hidden_size": 1024,
+  "initializer_range": 0.02,
+  "intermediate_size": 4096,
+  "layer_norm_eps": 1e-05,
+  "max_position_embeddings": 8194,
+  "model_type": "xlm-roberta",
+  "num_attention_heads": 16,
+  "num_hidden_layers": 24,
+  "output_past": true,
+  "pad_token_id": 1,
+  "position_embedding_type": "absolute",
+  "torch_dtype": "float16",
+  "transformers_version": "4.44.2",
+  "type_vocab_size": 1,
+  "use_cache": true,
+  "vocab_size": 250002
+}

checkpoint-390/latest

@@ -0,0 +1 @@
+global_step390

checkpoint-390/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:855b4280f2d306c83c3ecfddf1ddbaae1d9032d4b20f51853faaa0d6a16c0f77
+size 1135554344

checkpoint-390/rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d450ca4e8a4d23fa4a961f741ce98adf42ebc214a8d43159f209a432e702f2bb
+size 15920

checkpoint-390/rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d5aefb075d1611d454cce415a89a3e8758a14668167e62ee223c940f18466f91
+size 15920

checkpoint-390/rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5dcbfd1b600592bb0373238657fc34e638f3e5ebb520ba7db76f02c937d90106
+size 15920

checkpoint-390/rng_state_3.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:296318f56f3d5adaa653c100d65f4bb29d9606f913baf047f89d721e68fd1569
+size 15920

checkpoint-390/rng_state_4.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fc3c7b79f449b60a26fd9e32b3de05b3ee74517ca49e3df6c65fbbc771a17468
+size 15920

checkpoint-390/rng_state_5.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e53f54f8a626791d2d224a0053ec939ecf47b34d107fa7ae18122ca133a03eba
+size 15920

checkpoint-390/rng_state_6.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2e38621271f3c4435c03fbc26b4480a7e4696df44d864f711718dd711df370c0
+size 15920

checkpoint-390/rng_state_7.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1dc421f4a40233aeeabdf5e226ccd9b8cc265fb6c63945e94ce3b532cb661f5a
+size 15920

checkpoint-390/sentencepiece.bpe.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cfc8146abe2a0488e9e2a0c56de7952f7c11ab059eca145a0a727afce0db2865
+size 5069051

checkpoint-390/special_tokens_map.json

@@ -0,0 +1,51 @@
+{
+  "bos_token": {
+    "content": "<s>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "cls_token": {
+    "content": "<s>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "eos_token": {
+    "content": "</s>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "mask_token": {
+    "content": "<mask>",
+    "lstrip": true,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": {
+    "content": "<pad>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "sep_token": {
+    "content": "</s>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "unk_token": {
+    "content": "<unk>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}

checkpoint-390/tokenizer.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:33cd99e33ce09bdd8a6136fddfe90a1c47f85bafedf7309d0eecc19012d43586
+size 17082897

checkpoint-390/tokenizer_config.json

@@ -0,0 +1,59 @@
+{
+  "added_tokens_decoder": {
+    "0": {
+      "content": "<s>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "1": {
+      "content": "<pad>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "2": {
+      "content": "</s>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "3": {
+      "content": "<unk>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "250001": {
+      "content": "<mask>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    }
+  },
+  "bos_token": "<s>",
+  "clean_up_tokenization_spaces": true,
+  "cls_token": "<s>",
+  "eos_token": "</s>",
+  "mask_token": "<mask>",
+  "max_length": 512,
+  "model_max_length": 8192,
+  "pad_token": "<pad>",
+  "sep_token": "</s>",
+  "sp_model_kwargs": {},
+  "stride": 0,
+  "tokenizer_class": "XLMRobertaTokenizer",
+  "truncation_side": "right",
+  "truncation_strategy": "longest_first",
+  "unk_token": "<unk>"
+}

checkpoint-390/trainer_state.json

@@ -0,0 +1,2763 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.0,
+  "eval_steps": 500,
+  "global_step": 390,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.002564102564102564,
+      "grad_norm": 1.828752040863037,
+      "learning_rate": 0.0,
+      "loss": 1.0674,
+      "step": 1
+    },
+    {
+      "epoch": 0.005128205128205128,
+      "grad_norm": 2.0600810050964355,
+      "learning_rate": 1e-05,
+      "loss": 1.0127,
+      "step": 2
+    },
+    {
+      "epoch": 0.007692307692307693,
+      "grad_norm": 1.7609747648239136,
+      "learning_rate": 1e-05,
+      "loss": 0.8945,
+      "step": 3
+    },
+    {
+      "epoch": 0.010256410256410256,
+      "grad_norm": 2.0786192417144775,
+      "learning_rate": 9.974226804123713e-06,
+      "loss": 1.0947,
+      "step": 4
+    },
+    {
+      "epoch": 0.01282051282051282,
+      "grad_norm": 1.8740049600601196,
+      "learning_rate": 9.948453608247423e-06,
+      "loss": 1.0859,
+      "step": 5
+    },
+    {
+      "epoch": 0.015384615384615385,
+      "grad_norm": 1.8995977640151978,
+      "learning_rate": 9.922680412371136e-06,
+      "loss": 0.9326,
+      "step": 6
+    },
+    {
+      "epoch": 0.017948717948717947,
+      "grad_norm": 1.807106375694275,
+      "learning_rate": 9.896907216494846e-06,
+      "loss": 0.9839,
+      "step": 7
+    },
+    {
+      "epoch": 0.020512820512820513,
+      "grad_norm": 2.0719048976898193,
+      "learning_rate": 9.871134020618558e-06,
+      "loss": 1.1914,
+      "step": 8
+    },
+    {
+      "epoch": 0.023076923076923078,
+      "grad_norm": 1.714697241783142,
+      "learning_rate": 9.84536082474227e-06,
+      "loss": 0.9263,
+      "step": 9
+    },
+    {
+      "epoch": 0.02564102564102564,
+      "grad_norm": 1.7930278778076172,
+      "learning_rate": 9.819587628865979e-06,
+      "loss": 0.9546,
+      "step": 10
+    },
+    {
+      "epoch": 0.028205128205128206,
+      "grad_norm": 1.8702346086502075,
+      "learning_rate": 9.793814432989691e-06,
+      "loss": 1.0645,
+      "step": 11
+    },
+    {
+      "epoch": 0.03076923076923077,
+      "grad_norm": 1.7376536130905151,
+      "learning_rate": 9.768041237113403e-06,
+      "loss": 1.0166,
+      "step": 12
+    },
+    {
+      "epoch": 0.03333333333333333,
+      "grad_norm": 1.804457426071167,
+      "learning_rate": 9.742268041237114e-06,
+      "loss": 0.8545,
+      "step": 13
+    },
+    {
+      "epoch": 0.035897435897435895,
+      "grad_norm": 1.8418951034545898,
+      "learning_rate": 9.716494845360826e-06,
+      "loss": 1.1855,
+      "step": 14
+    },
+    {
+      "epoch": 0.038461538461538464,
+      "grad_norm": 3.295741558074951,
+      "learning_rate": 9.690721649484536e-06,
+      "loss": 1.1631,
+      "step": 15
+    },
+    {
+      "epoch": 0.041025641025641026,
+      "grad_norm": 2.6767563819885254,
+      "learning_rate": 9.664948453608248e-06,
+      "loss": 1.0205,
+      "step": 16
+    },
+    {
+      "epoch": 0.04358974358974359,
+      "grad_norm": 2.088998556137085,
+      "learning_rate": 9.63917525773196e-06,
+      "loss": 0.9512,
+      "step": 17
+    },
+    {
+      "epoch": 0.046153846153846156,
+      "grad_norm": 3.8526885509490967,
+      "learning_rate": 9.613402061855671e-06,
+      "loss": 1.2227,
+      "step": 18
+    },
+    {
+      "epoch": 0.04871794871794872,
+      "grad_norm": 2.5801403522491455,
+      "learning_rate": 9.587628865979383e-06,
+      "loss": 0.9424,
+      "step": 19
+    },
+    {
+      "epoch": 0.05128205128205128,
+      "grad_norm": 2.2087137699127197,
+      "learning_rate": 9.561855670103093e-06,
+      "loss": 1.3262,
+      "step": 20
+    },
+    {
+      "epoch": 0.05384615384615385,
+      "grad_norm": 1.9106372594833374,
+      "learning_rate": 9.536082474226806e-06,
+      "loss": 0.9004,
+      "step": 21
+    },
+    {
+      "epoch": 0.05641025641025641,
+      "grad_norm": 1.7588039636611938,
+      "learning_rate": 9.510309278350516e-06,
+      "loss": 0.9482,
+      "step": 22
+    },
+    {
+      "epoch": 0.05897435897435897,
+      "grad_norm": 2.0292000770568848,
+      "learning_rate": 9.484536082474226e-06,
+      "loss": 1.0371,
+      "step": 23
+    },
+    {
+      "epoch": 0.06153846153846154,
+      "grad_norm": 1.8534419536590576,
+      "learning_rate": 9.458762886597939e-06,
+      "loss": 1.0303,
+      "step": 24
+    },
+    {
+      "epoch": 0.0641025641025641,
+      "grad_norm": 2.4608726501464844,
+      "learning_rate": 9.43298969072165e-06,
+      "loss": 1.4395,
+      "step": 25
+    },
+    {
+      "epoch": 0.06666666666666667,
+      "grad_norm": 2.119417905807495,
+      "learning_rate": 9.407216494845361e-06,
+      "loss": 1.1143,
+      "step": 26
+    },
+    {
+      "epoch": 0.06923076923076923,
+      "grad_norm": 2.145531177520752,
+      "learning_rate": 9.381443298969073e-06,
+      "loss": 0.9443,
+      "step": 27
+    },
+    {
+      "epoch": 0.07179487179487179,
+      "grad_norm": 4.710203170776367,
+      "learning_rate": 9.355670103092784e-06,
+      "loss": 1.3311,
+      "step": 28
+    },
+    {
+      "epoch": 0.07435897435897436,
+      "grad_norm": 3.2553186416625977,
+      "learning_rate": 9.329896907216496e-06,
+      "loss": 1.123,
+      "step": 29
+    },
+    {
+      "epoch": 0.07692307692307693,
+      "grad_norm": 2.38273024559021,
+      "learning_rate": 9.304123711340208e-06,
+      "loss": 1.124,
+      "step": 30
+    },
+    {
+      "epoch": 0.07948717948717948,
+      "grad_norm": 3.6473400592803955,
+      "learning_rate": 9.278350515463918e-06,
+      "loss": 1.2568,
+      "step": 31
+    },
+    {
+      "epoch": 0.08205128205128205,
+      "grad_norm": 2.2762739658355713,
+      "learning_rate": 9.25257731958763e-06,
+      "loss": 1.0898,
+      "step": 32
+    },
+    {
+      "epoch": 0.08461538461538462,
+      "grad_norm": 1.870347499847412,
+      "learning_rate": 9.226804123711341e-06,
+      "loss": 0.8887,
+      "step": 33
+    },
+    {
+      "epoch": 0.08717948717948718,
+      "grad_norm": 3.9631059169769287,
+      "learning_rate": 9.201030927835051e-06,
+      "loss": 0.937,
+      "step": 34
+    },
+    {
+      "epoch": 0.08974358974358974,
+      "grad_norm": 2.2279770374298096,
+      "learning_rate": 9.175257731958764e-06,
+      "loss": 0.9785,
+      "step": 35
+    },
+    {
+      "epoch": 0.09230769230769231,
+      "grad_norm": 2.2554802894592285,
+      "learning_rate": 9.149484536082474e-06,
+      "loss": 1.0762,
+      "step": 36
+    },
+    {
+      "epoch": 0.09487179487179487,
+      "grad_norm": 2.188344955444336,
+      "learning_rate": 9.123711340206186e-06,
+      "loss": 1.1318,
+      "step": 37
+    },
+    {
+      "epoch": 0.09743589743589744,
+      "grad_norm": 2.1649510860443115,
+      "learning_rate": 9.097938144329898e-06,
+      "loss": 0.7827,
+      "step": 38
+    },
+    {
+      "epoch": 0.1,
+      "grad_norm": 2.4596219062805176,
+      "learning_rate": 9.072164948453609e-06,
+      "loss": 0.8247,
+      "step": 39
+    },
+    {
+      "epoch": 0.10256410256410256,
+      "grad_norm": 2.487870931625366,
+      "learning_rate": 9.04639175257732e-06,
+      "loss": 1.0635,
+      "step": 40
+    },
+    {
+      "epoch": 0.10512820512820513,
+      "grad_norm": 1.9071369171142578,
+      "learning_rate": 9.020618556701031e-06,
+      "loss": 0.9946,
+      "step": 41
+    },
+    {
+      "epoch": 0.1076923076923077,
+      "grad_norm": 2.219597101211548,
+      "learning_rate": 8.994845360824743e-06,
+      "loss": 1.1309,
+      "step": 42
+    },
+    {
+      "epoch": 0.11025641025641025,
+      "grad_norm": 2.2724695205688477,
+      "learning_rate": 8.969072164948455e-06,
+      "loss": 1.2637,
+      "step": 43
+    },
+    {
+      "epoch": 0.11282051282051282,
+      "grad_norm": 2.331315755844116,
+      "learning_rate": 8.943298969072166e-06,
+      "loss": 1.1611,
+      "step": 44
+    },
+    {
+      "epoch": 0.11538461538461539,
+      "grad_norm": 2.036611318588257,
+      "learning_rate": 8.917525773195878e-06,
+      "loss": 0.9233,
+      "step": 45
+    },
+    {
+      "epoch": 0.11794871794871795,
+      "grad_norm": 1.9779893159866333,
+      "learning_rate": 8.891752577319588e-06,
+      "loss": 0.8071,
+      "step": 46
+    },
+    {
+      "epoch": 0.12051282051282051,
+      "grad_norm": 2.511869192123413,
+      "learning_rate": 8.865979381443299e-06,
+      "loss": 0.9165,
+      "step": 47
+    },
+    {
+      "epoch": 0.12307692307692308,
+      "grad_norm": 1.9824575185775757,
+      "learning_rate": 8.840206185567011e-06,
+      "loss": 0.9712,
+      "step": 48
+    },
+    {
+      "epoch": 0.12564102564102564,
+      "grad_norm": 2.224874973297119,
+      "learning_rate": 8.814432989690721e-06,
+      "loss": 0.8857,
+      "step": 49
+    },
+    {
+      "epoch": 0.1282051282051282,
+      "grad_norm": 2.290484666824341,
+      "learning_rate": 8.788659793814434e-06,
+      "loss": 1.0908,
+      "step": 50
+    },
+    {
+      "epoch": 0.13076923076923078,
+      "grad_norm": 2.009584903717041,
+      "learning_rate": 8.762886597938146e-06,
+      "loss": 0.833,
+      "step": 51
+    },
+    {
+      "epoch": 0.13333333333333333,
+      "grad_norm": 4.70264196395874,
+      "learning_rate": 8.737113402061856e-06,
+      "loss": 0.998,
+      "step": 52
+    },
+    {
+      "epoch": 0.1358974358974359,
+      "grad_norm": 3.110806465148926,
+      "learning_rate": 8.711340206185568e-06,
+      "loss": 0.9189,
+      "step": 53
+    },
+    {
+      "epoch": 0.13846153846153847,
+      "grad_norm": 1.902529239654541,
+      "learning_rate": 8.685567010309279e-06,
+      "loss": 1.0293,
+      "step": 54
+    },
+    {
+      "epoch": 0.14102564102564102,
+      "grad_norm": 2.1648287773132324,
+      "learning_rate": 8.65979381443299e-06,
+      "loss": 1.0488,
+      "step": 55
+    },
+    {
+      "epoch": 0.14358974358974358,
+      "grad_norm": 1.7489367723464966,
+      "learning_rate": 8.634020618556703e-06,
+      "loss": 0.8091,
+      "step": 56
+    },
+    {
+      "epoch": 0.14615384615384616,
+      "grad_norm": 1.9431246519088745,
+      "learning_rate": 8.608247422680413e-06,
+      "loss": 1.0049,
+      "step": 57
+    },
+    {
+      "epoch": 0.14871794871794872,
+      "grad_norm": 2.2493724822998047,
+      "learning_rate": 8.582474226804124e-06,
+      "loss": 0.9321,
+      "step": 58
+    },
+    {
+      "epoch": 0.15128205128205127,
+      "grad_norm": 2.1866893768310547,
+      "learning_rate": 8.556701030927836e-06,
+      "loss": 1.2656,
+      "step": 59
+    },
+    {
+      "epoch": 0.15384615384615385,
+      "grad_norm": 2.024313449859619,
+      "learning_rate": 8.530927835051546e-06,
+      "loss": 1.0801,
+      "step": 60
+    },
+    {
+      "epoch": 0.1564102564102564,
+      "grad_norm": 1.77311110496521,
+      "learning_rate": 8.505154639175259e-06,
+      "loss": 0.9946,
+      "step": 61
+    },
+    {
+      "epoch": 0.15897435897435896,
+      "grad_norm": 1.6627540588378906,
+      "learning_rate": 8.479381443298969e-06,
+      "loss": 0.9702,
+      "step": 62
+    },
+    {
+      "epoch": 0.16153846153846155,
+      "grad_norm": 1.9997308254241943,
+      "learning_rate": 8.453608247422681e-06,
+      "loss": 0.9487,
+      "step": 63
+    },
+    {
+      "epoch": 0.1641025641025641,
+      "grad_norm": 1.6049851179122925,
+      "learning_rate": 8.427835051546393e-06,
+      "loss": 0.9512,
+      "step": 64
+    },
+    {
+      "epoch": 0.16666666666666666,
+      "grad_norm": 1.8791325092315674,
+      "learning_rate": 8.402061855670104e-06,
+      "loss": 0.9355,
+      "step": 65
+    },
+    {
+      "epoch": 0.16923076923076924,
+      "grad_norm": 2.173393964767456,
+      "learning_rate": 8.376288659793816e-06,
+      "loss": 1.1992,
+      "step": 66
+    },
+    {
+      "epoch": 0.1717948717948718,
+      "grad_norm": 2.0367043018341064,
+      "learning_rate": 8.350515463917526e-06,
+      "loss": 1.1426,
+      "step": 67
+    },
+    {
+      "epoch": 0.17435897435897435,
+      "grad_norm": 1.7792794704437256,
+      "learning_rate": 8.324742268041238e-06,
+      "loss": 0.9097,
+      "step": 68
+    },
+    {
+      "epoch": 0.17692307692307693,
+      "grad_norm": 1.9103853702545166,
+      "learning_rate": 8.29896907216495e-06,
+      "loss": 1.0225,
+      "step": 69
+    },
+    {
+      "epoch": 0.1794871794871795,
+      "grad_norm": 2.0614025592803955,
+      "learning_rate": 8.27319587628866e-06,
+      "loss": 1.2012,
+      "step": 70
+    },
+    {
+      "epoch": 0.18205128205128204,
+      "grad_norm": 1.5830662250518799,
+      "learning_rate": 8.247422680412371e-06,
+      "loss": 0.7656,
+      "step": 71
+    },
+    {
+      "epoch": 0.18461538461538463,
+      "grad_norm": 1.8411104679107666,
+      "learning_rate": 8.221649484536083e-06,
+      "loss": 1.1162,
+      "step": 72
+    },
+    {
+      "epoch": 0.18717948717948718,
+      "grad_norm": 1.954744815826416,
+      "learning_rate": 8.195876288659794e-06,
+      "loss": 1.2031,
+      "step": 73
+    },
+    {
+      "epoch": 0.18974358974358974,
+      "grad_norm": 2.1274709701538086,
+      "learning_rate": 8.170103092783506e-06,
+      "loss": 1.1641,
+      "step": 74
+    },
+    {
+      "epoch": 0.19230769230769232,
+      "grad_norm": 1.8388986587524414,
+      "learning_rate": 8.144329896907216e-06,
+      "loss": 0.9517,
+      "step": 75
+    },
+    {
+      "epoch": 0.19487179487179487,
+      "grad_norm": 2.0911478996276855,
+      "learning_rate": 8.118556701030929e-06,
+      "loss": 1.0801,
+      "step": 76
+    },
+    {
+      "epoch": 0.19743589743589743,
+      "grad_norm": 1.9565143585205078,
+      "learning_rate": 8.09278350515464e-06,
+      "loss": 0.9478,
+      "step": 77
+    },
+    {
+      "epoch": 0.2,
+      "grad_norm": 1.8666242361068726,
+      "learning_rate": 8.067010309278351e-06,
+      "loss": 1.3125,
+      "step": 78
+    },
+    {
+      "epoch": 0.20256410256410257,
+      "grad_norm": 1.9826719760894775,
+      "learning_rate": 8.041237113402063e-06,
+      "loss": 0.9492,
+      "step": 79
+    },
+    {
+      "epoch": 0.20512820512820512,
+      "grad_norm": 2.062222719192505,
+      "learning_rate": 8.015463917525774e-06,
+      "loss": 1.0195,
+      "step": 80
+    },
+    {
+      "epoch": 0.2076923076923077,
+      "grad_norm": 2.2730824947357178,
+      "learning_rate": 7.989690721649486e-06,
+      "loss": 0.9248,
+      "step": 81
+    },
+    {
+      "epoch": 0.21025641025641026,
+      "grad_norm": 3.432387113571167,
+      "learning_rate": 7.963917525773196e-06,
+      "loss": 0.9678,
+      "step": 82
+    },
+    {
+      "epoch": 0.2128205128205128,
+      "grad_norm": 2.1514482498168945,
+      "learning_rate": 7.938144329896907e-06,
+      "loss": 0.7881,
+      "step": 83
+    },
+    {
+      "epoch": 0.2153846153846154,
+      "grad_norm": 1.6954137086868286,
+      "learning_rate": 7.912371134020619e-06,
+      "loss": 0.8833,
+      "step": 84
+    },
+    {
+      "epoch": 0.21794871794871795,
+      "grad_norm": 1.9222341775894165,
+      "learning_rate": 7.886597938144331e-06,
+      "loss": 0.9917,
+      "step": 85
+    },
+    {
+      "epoch": 0.2205128205128205,
+      "grad_norm": 1.9066567420959473,
+      "learning_rate": 7.860824742268041e-06,
+      "loss": 0.9507,
+      "step": 86
+    },
+    {
+      "epoch": 0.2230769230769231,
+      "grad_norm": 1.9370355606079102,
+      "learning_rate": 7.835051546391754e-06,
+      "loss": 0.9814,
+      "step": 87
+    },
+    {
+      "epoch": 0.22564102564102564,
+      "grad_norm": 1.8217931985855103,
+      "learning_rate": 7.809278350515464e-06,
+      "loss": 1.0186,
+      "step": 88
+    },
+    {
+      "epoch": 0.2282051282051282,
+      "grad_norm": 1.6907788515090942,
+      "learning_rate": 7.783505154639176e-06,
+      "loss": 0.9624,
+      "step": 89
+    },
+    {
+      "epoch": 0.23076923076923078,
+      "grad_norm": 1.6538673639297485,
+      "learning_rate": 7.757731958762888e-06,
+      "loss": 0.8237,
+      "step": 90
+    },
+    {
+      "epoch": 0.23333333333333334,
+      "grad_norm": 1.9128988981246948,
+      "learning_rate": 7.731958762886599e-06,
+      "loss": 0.9175,
+      "step": 91
+    },
+    {
+      "epoch": 0.2358974358974359,
+      "grad_norm": 3.1747217178344727,
+      "learning_rate": 7.70618556701031e-06,
+      "loss": 0.9775,
+      "step": 92
+    },
+    {
+      "epoch": 0.23846153846153847,
+      "grad_norm": 1.6946772336959839,
+      "learning_rate": 7.680412371134021e-06,
+      "loss": 0.8823,
+      "step": 93
+    },
+    {
+      "epoch": 0.24102564102564103,
+      "grad_norm": 2.1718826293945312,
+      "learning_rate": 7.654639175257732e-06,
+      "loss": 1.1543,
+      "step": 94
+    },
+    {
+      "epoch": 0.24358974358974358,
+      "grad_norm": 2.0842125415802,
+      "learning_rate": 7.628865979381444e-06,
+      "loss": 0.9058,
+      "step": 95
+    },
+    {
+      "epoch": 0.24615384615384617,
+      "grad_norm": 1.8005638122558594,
+      "learning_rate": 7.603092783505155e-06,
+      "loss": 0.9146,
+      "step": 96
+    },
+    {
+      "epoch": 0.24871794871794872,
+      "grad_norm": 1.6228526830673218,
+      "learning_rate": 7.577319587628866e-06,
+      "loss": 0.7095,
+      "step": 97
+    },
+    {
+      "epoch": 0.2512820512820513,
+      "grad_norm": 2.022739887237549,
+      "learning_rate": 7.551546391752578e-06,
+      "loss": 0.9229,
+      "step": 98
+    },
+    {
+      "epoch": 0.25384615384615383,
+      "grad_norm": 1.7935630083084106,
+      "learning_rate": 7.525773195876289e-06,
+      "loss": 0.7896,
+      "step": 99
+    },
+    {
+      "epoch": 0.2564102564102564,
+      "grad_norm": 1.747321367263794,
+      "learning_rate": 7.500000000000001e-06,
+      "loss": 0.9551,
+      "step": 100
+    },
+    {
+      "epoch": 0.258974358974359,
+      "grad_norm": 1.8076218366622925,
+      "learning_rate": 7.474226804123712e-06,
+      "loss": 0.8394,
+      "step": 101
+    },
+    {
+      "epoch": 0.26153846153846155,
+      "grad_norm": 1.730443000793457,
+      "learning_rate": 7.448453608247424e-06,
+      "loss": 0.9062,
+      "step": 102
+    },
+    {
+      "epoch": 0.2641025641025641,
+      "grad_norm": 2.290266275405884,
+      "learning_rate": 7.422680412371135e-06,
+      "loss": 1.248,
+      "step": 103
+    },
+    {
+      "epoch": 0.26666666666666666,
+      "grad_norm": 1.772620677947998,
+      "learning_rate": 7.396907216494846e-06,
+      "loss": 0.8296,
+      "step": 104
+    },
+    {
+      "epoch": 0.2692307692307692,
+      "grad_norm": 1.7270923852920532,
+      "learning_rate": 7.3711340206185574e-06,
+      "loss": 0.8735,
+      "step": 105
+    },
+    {
+      "epoch": 0.2717948717948718,
+      "grad_norm": 3.0547115802764893,
+      "learning_rate": 7.3453608247422696e-06,
+      "loss": 0.9385,
+      "step": 106
+    },
+    {
+      "epoch": 0.2743589743589744,
+      "grad_norm": 2.1224873065948486,
+      "learning_rate": 7.319587628865979e-06,
+      "loss": 1.0918,
+      "step": 107
+    },
+    {
+      "epoch": 0.27692307692307694,
+      "grad_norm": 2.175182819366455,
+      "learning_rate": 7.293814432989691e-06,
+      "loss": 0.9707,
+      "step": 108
+    },
+    {
+      "epoch": 0.2794871794871795,
+      "grad_norm": 1.8396246433258057,
+      "learning_rate": 7.2680412371134026e-06,
+      "loss": 0.9609,
+      "step": 109
+    },
+    {
+      "epoch": 0.28205128205128205,
+      "grad_norm": 1.8551801443099976,
+      "learning_rate": 7.242268041237114e-06,
+      "loss": 0.8916,
+      "step": 110
+    },
+    {
+      "epoch": 0.2846153846153846,
+      "grad_norm": 1.960680603981018,
+      "learning_rate": 7.216494845360825e-06,
+      "loss": 1.0352,
+      "step": 111
+    },
+    {
+      "epoch": 0.28717948717948716,
+      "grad_norm": 1.8434001207351685,
+      "learning_rate": 7.190721649484536e-06,
+      "loss": 1.0576,
+      "step": 112
+    },
+    {
+      "epoch": 0.28974358974358977,
+      "grad_norm": 2.5577752590179443,
+      "learning_rate": 7.164948453608248e-06,
+      "loss": 1.0762,
+      "step": 113
+    },
+    {
+      "epoch": 0.2923076923076923,
+      "grad_norm": 1.8725932836532593,
+      "learning_rate": 7.13917525773196e-06,
+      "loss": 0.79,
+      "step": 114
+    },
+    {
+      "epoch": 0.2948717948717949,
+      "grad_norm": 2.0246224403381348,
+      "learning_rate": 7.113402061855671e-06,
+      "loss": 0.9058,
+      "step": 115
+    },
+    {
+      "epoch": 0.29743589743589743,
+      "grad_norm": 1.8787579536437988,
+      "learning_rate": 7.087628865979382e-06,
+      "loss": 0.8569,
+      "step": 116
+    },
+    {
+      "epoch": 0.3,
+      "grad_norm": 1.6573154926300049,
+      "learning_rate": 7.061855670103094e-06,
+      "loss": 0.8091,
+      "step": 117
+    },
+    {
+      "epoch": 0.30256410256410254,
+      "grad_norm": 1.7937979698181152,
+      "learning_rate": 7.036082474226805e-06,
+      "loss": 0.978,
+      "step": 118
+    },
+    {
+      "epoch": 0.30512820512820515,
+      "grad_norm": 1.9054224491119385,
+      "learning_rate": 7.010309278350515e-06,
+      "loss": 1.0635,
+      "step": 119
+    },
+    {
+      "epoch": 0.3076923076923077,
+      "grad_norm": 1.6725414991378784,
+      "learning_rate": 6.984536082474227e-06,
+      "loss": 1.0312,
+      "step": 120
+    },
+    {
+      "epoch": 0.31025641025641026,
+      "grad_norm": 1.8967286348342896,
+      "learning_rate": 6.958762886597939e-06,
+      "loss": 0.9746,
+      "step": 121
+    },
+    {
+      "epoch": 0.3128205128205128,
+      "grad_norm": 2.158597230911255,
+      "learning_rate": 6.93298969072165e-06,
+      "loss": 1.1309,
+      "step": 122
+    },
+    {
+      "epoch": 0.3153846153846154,
+      "grad_norm": 2.2662479877471924,
+      "learning_rate": 6.907216494845361e-06,
+      "loss": 1.0674,
+      "step": 123
+    },
+    {
+      "epoch": 0.31794871794871793,
+      "grad_norm": 1.9628630876541138,
+      "learning_rate": 6.881443298969073e-06,
+      "loss": 0.9644,
+      "step": 124
+    },
+    {
+      "epoch": 0.32051282051282054,
+      "grad_norm": 1.7974278926849365,
+      "learning_rate": 6.855670103092784e-06,
+      "loss": 0.8877,
+      "step": 125
+    },
+    {
+      "epoch": 0.3230769230769231,
+      "grad_norm": 1.7548365592956543,
+      "learning_rate": 6.829896907216495e-06,
+      "loss": 0.9204,
+      "step": 126
+    },
+    {
+      "epoch": 0.32564102564102565,
+      "grad_norm": 1.9516572952270508,
+      "learning_rate": 6.804123711340207e-06,
+      "loss": 1.1602,
+      "step": 127
+    },
+    {
+      "epoch": 0.3282051282051282,
+      "grad_norm": 2.1989433765411377,
+      "learning_rate": 6.778350515463919e-06,
+      "loss": 1.1133,
+      "step": 128
+    },
+    {
+      "epoch": 0.33076923076923076,
+      "grad_norm": 1.800710916519165,
+      "learning_rate": 6.75257731958763e-06,
+      "loss": 0.791,
+      "step": 129
+    },
+    {
+      "epoch": 0.3333333333333333,
+      "grad_norm": 3.385741710662842,
+      "learning_rate": 6.726804123711341e-06,
+      "loss": 1.1758,
+      "step": 130
+    },
+    {
+      "epoch": 0.33589743589743587,
+      "grad_norm": 2.0922465324401855,
+      "learning_rate": 6.701030927835052e-06,
+      "loss": 0.9214,
+      "step": 131
+    },
+    {
+      "epoch": 0.3384615384615385,
+      "grad_norm": 1.573805332183838,
+      "learning_rate": 6.675257731958763e-06,
+      "loss": 0.7715,
+      "step": 132
+    },
+    {
+      "epoch": 0.34102564102564104,
+      "grad_norm": 1.7901231050491333,
+      "learning_rate": 6.649484536082474e-06,
+      "loss": 0.9443,
+      "step": 133
+    },
+    {
+      "epoch": 0.3435897435897436,
+      "grad_norm": 1.6376028060913086,
+      "learning_rate": 6.623711340206186e-06,
+      "loss": 0.9854,
+      "step": 134
+    },
+    {
+      "epoch": 0.34615384615384615,
+      "grad_norm": 2.0153403282165527,
+      "learning_rate": 6.597938144329898e-06,
+      "loss": 1.2373,
+      "step": 135
+    },
+    {
+      "epoch": 0.3487179487179487,
+      "grad_norm": 1.819011926651001,
+      "learning_rate": 6.572164948453609e-06,
+      "loss": 0.9048,
+      "step": 136
+    },
+    {
+      "epoch": 0.35128205128205126,
+      "grad_norm": 1.623148798942566,
+      "learning_rate": 6.54639175257732e-06,
+      "loss": 0.8262,
+      "step": 137
+    },
+    {
+      "epoch": 0.35384615384615387,
+      "grad_norm": 2.2499585151672363,
+      "learning_rate": 6.520618556701031e-06,
+      "loss": 1.1152,
+      "step": 138
+    },
+    {
+      "epoch": 0.3564102564102564,
+      "grad_norm": 1.9235867261886597,
+      "learning_rate": 6.494845360824743e-06,
+      "loss": 0.9067,
+      "step": 139
+    },
+    {
+      "epoch": 0.358974358974359,
+      "grad_norm": 1.9163473844528198,
+      "learning_rate": 6.469072164948455e-06,
+      "loss": 1.0469,
+      "step": 140
+    },
+    {
+      "epoch": 0.36153846153846153,
+      "grad_norm": 2.092963457107544,
+      "learning_rate": 6.443298969072166e-06,
+      "loss": 1.0352,
+      "step": 141
+    },
+    {
+      "epoch": 0.3641025641025641,
+      "grad_norm": 9.061576843261719,
+      "learning_rate": 6.417525773195877e-06,
+      "loss": 0.833,
+      "step": 142
+    },
+    {
+      "epoch": 0.36666666666666664,
+      "grad_norm": 2.0472512245178223,
+      "learning_rate": 6.391752577319588e-06,
+      "loss": 1.043,
+      "step": 143
+    },
+    {
+      "epoch": 0.36923076923076925,
+      "grad_norm": 1.9477477073669434,
+      "learning_rate": 6.365979381443299e-06,
+      "loss": 0.9644,
+      "step": 144
+    },
+    {
+      "epoch": 0.3717948717948718,
+      "grad_norm": 1.9295361042022705,
+      "learning_rate": 6.34020618556701e-06,
+      "loss": 0.9746,
+      "step": 145
+    },
+    {
+      "epoch": 0.37435897435897436,
+      "grad_norm": 1.9540655612945557,
+      "learning_rate": 6.314432989690722e-06,
+      "loss": 0.9375,
+      "step": 146
+    },
+    {
+      "epoch": 0.3769230769230769,
+      "grad_norm": 1.9495689868927002,
+      "learning_rate": 6.288659793814433e-06,
+      "loss": 1.123,
+      "step": 147
+    },
+    {
+      "epoch": 0.37948717948717947,
+      "grad_norm": 1.8979523181915283,
+      "learning_rate": 6.262886597938145e-06,
+      "loss": 1.0762,
+      "step": 148
+    },
+    {
+      "epoch": 0.382051282051282,
+      "grad_norm": 2.086167573928833,
+      "learning_rate": 6.237113402061856e-06,
+      "loss": 0.8975,
+      "step": 149
+    },
+    {
+      "epoch": 0.38461538461538464,
+      "grad_norm": 1.9562528133392334,
+      "learning_rate": 6.211340206185568e-06,
+      "loss": 0.9429,
+      "step": 150
+    },
+    {
+      "epoch": 0.3871794871794872,
+      "grad_norm": 2.3442628383636475,
+      "learning_rate": 6.185567010309279e-06,
+      "loss": 0.9844,
+      "step": 151
+    },
+    {
+      "epoch": 0.38974358974358975,
+      "grad_norm": 1.8352185487747192,
+      "learning_rate": 6.15979381443299e-06,
+      "loss": 1.1045,
+      "step": 152
+    },
+    {
+      "epoch": 0.3923076923076923,
+      "grad_norm": 3.1961944103240967,
+      "learning_rate": 6.134020618556702e-06,
+      "loss": 1.0596,
+      "step": 153
+    },
+    {
+      "epoch": 0.39487179487179486,
+      "grad_norm": 1.8429200649261475,
+      "learning_rate": 6.108247422680414e-06,
+      "loss": 0.9365,
+      "step": 154
+    },
+    {
+      "epoch": 0.3974358974358974,
+      "grad_norm": 1.929801106452942,
+      "learning_rate": 6.082474226804124e-06,
+      "loss": 0.9883,
+      "step": 155
+    },
+    {
+      "epoch": 0.4,
+      "grad_norm": 1.670444369316101,
+      "learning_rate": 6.056701030927835e-06,
+      "loss": 0.7524,
+      "step": 156
+    },
+    {
+      "epoch": 0.4025641025641026,
+      "grad_norm": 2.022891044616699,
+      "learning_rate": 6.030927835051547e-06,
+      "loss": 1.2197,
+      "step": 157
+    },
+    {
+      "epoch": 0.40512820512820513,
+      "grad_norm": 1.7056248188018799,
+      "learning_rate": 6.005154639175258e-06,
+      "loss": 0.9121,
+      "step": 158
+    },
+    {
+      "epoch": 0.4076923076923077,
+      "grad_norm": 2.0686631202697754,
+      "learning_rate": 5.979381443298969e-06,
+      "loss": 1.1865,
+      "step": 159
+    },
+    {
+      "epoch": 0.41025641025641024,
+      "grad_norm": 1.8868600130081177,
+      "learning_rate": 5.9536082474226805e-06,
+      "loss": 0.9658,
+      "step": 160
+    },
+    {
+      "epoch": 0.4128205128205128,
+      "grad_norm": 2.3580541610717773,
+      "learning_rate": 5.927835051546393e-06,
+      "loss": 0.9297,
+      "step": 161
+    },
+    {
+      "epoch": 0.4153846153846154,
+      "grad_norm": 2.071708917617798,
+      "learning_rate": 5.902061855670104e-06,
+      "loss": 1.0693,
+      "step": 162
+    },
+    {
+      "epoch": 0.41794871794871796,
+      "grad_norm": 1.8569782972335815,
+      "learning_rate": 5.876288659793815e-06,
+      "loss": 0.9438,
+      "step": 163
+    },
+    {
+      "epoch": 0.4205128205128205,
+      "grad_norm": 1.7496881484985352,
+      "learning_rate": 5.8505154639175264e-06,
+      "loss": 0.8574,
+      "step": 164
+    },
+    {
+      "epoch": 0.4230769230769231,
+      "grad_norm": 1.825770378112793,
+      "learning_rate": 5.824742268041238e-06,
+      "loss": 0.9678,
+      "step": 165
+    },
+    {
+      "epoch": 0.4256410256410256,
+      "grad_norm": 2.156632423400879,
+      "learning_rate": 5.79896907216495e-06,
+      "loss": 1.2031,
+      "step": 166
+    },
+    {
+      "epoch": 0.4282051282051282,
+      "grad_norm": 1.6969801187515259,
+      "learning_rate": 5.7731958762886594e-06,
+      "loss": 0.8359,
+      "step": 167
+    },
+    {
+      "epoch": 0.4307692307692308,
+      "grad_norm": 2.4445745944976807,
+      "learning_rate": 5.7474226804123716e-06,
+      "loss": 1.0293,
+      "step": 168
+    },
+    {
+      "epoch": 0.43333333333333335,
+      "grad_norm": 1.9905025959014893,
+      "learning_rate": 5.721649484536083e-06,
+      "loss": 1.0791,
+      "step": 169
+    },
+    {
+      "epoch": 0.4358974358974359,
+      "grad_norm": 1.9443804025650024,
+      "learning_rate": 5.695876288659794e-06,
+      "loss": 0.9702,
+      "step": 170
+    },
+    {
+      "epoch": 0.43846153846153846,
+      "grad_norm": 2.2694649696350098,
+      "learning_rate": 5.670103092783505e-06,
+      "loss": 1.1914,
+      "step": 171
+    },
+    {
+      "epoch": 0.441025641025641,
+      "grad_norm": 2.1340649127960205,
+      "learning_rate": 5.644329896907217e-06,
+      "loss": 1.0371,
+      "step": 172
+    },
+    {
+      "epoch": 0.44358974358974357,
+      "grad_norm": 1.9536010026931763,
+      "learning_rate": 5.618556701030928e-06,
+      "loss": 0.9888,
+      "step": 173
+    },
+    {
+      "epoch": 0.4461538461538462,
+      "grad_norm": 1.837241530418396,
+      "learning_rate": 5.59278350515464e-06,
+      "loss": 0.9033,
+      "step": 174
+    },
+    {
+      "epoch": 0.44871794871794873,
+      "grad_norm": 1.8256818056106567,
+      "learning_rate": 5.567010309278351e-06,
+      "loss": 1.0117,
+      "step": 175
+    },
+    {
+      "epoch": 0.4512820512820513,
+      "grad_norm": 2.1817359924316406,
+      "learning_rate": 5.541237113402063e-06,
+      "loss": 0.8506,
+      "step": 176
+    },
+    {
+      "epoch": 0.45384615384615384,
+      "grad_norm": 1.7470033168792725,
+      "learning_rate": 5.515463917525774e-06,
+      "loss": 0.9473,
+      "step": 177
+    },
+    {
+      "epoch": 0.4564102564102564,
+      "grad_norm": 1.8381420373916626,
+      "learning_rate": 5.489690721649485e-06,
+      "loss": 0.8828,
+      "step": 178
+    },
+    {
+      "epoch": 0.45897435897435895,
+      "grad_norm": 1.9407833814620972,
+      "learning_rate": 5.463917525773196e-06,
+      "loss": 1.3193,
+      "step": 179
+    },
+    {
+      "epoch": 0.46153846153846156,
+      "grad_norm": 1.9037673473358154,
+      "learning_rate": 5.438144329896907e-06,
+      "loss": 0.916,
+      "step": 180
+    },
+    {
+      "epoch": 0.4641025641025641,
+      "grad_norm": 1.8702125549316406,
+      "learning_rate": 5.412371134020619e-06,
+      "loss": 1.0039,
+      "step": 181
+    },
+    {
+      "epoch": 0.4666666666666667,
+      "grad_norm": 2.0988097190856934,
+      "learning_rate": 5.38659793814433e-06,
+      "loss": 1.0225,
+      "step": 182
+    },
+    {
+      "epoch": 0.46923076923076923,
+      "grad_norm": 1.9512522220611572,
+      "learning_rate": 5.360824742268042e-06,
+      "loss": 0.9854,
+      "step": 183
+    },
+    {
+      "epoch": 0.4717948717948718,
+      "grad_norm": 1.8114992380142212,
+      "learning_rate": 5.335051546391753e-06,
+      "loss": 0.8008,
+      "step": 184
+    },
+    {
+      "epoch": 0.47435897435897434,
+      "grad_norm": 1.7766282558441162,
+      "learning_rate": 5.309278350515464e-06,
+      "loss": 0.9795,
+      "step": 185
+    },
+    {
+      "epoch": 0.47692307692307695,
+      "grad_norm": 2.319395065307617,
+      "learning_rate": 5.2835051546391755e-06,
+      "loss": 1.1572,
+      "step": 186
+    },
+    {
+      "epoch": 0.4794871794871795,
+      "grad_norm": 1.7959656715393066,
+      "learning_rate": 5.257731958762888e-06,
+      "loss": 0.8994,
+      "step": 187
+    },
+    {
+      "epoch": 0.48205128205128206,
+      "grad_norm": 1.7262598276138306,
+      "learning_rate": 5.231958762886599e-06,
+      "loss": 1.0264,
+      "step": 188
+    },
+    {
+      "epoch": 0.4846153846153846,
+      "grad_norm": 1.9442336559295654,
+      "learning_rate": 5.20618556701031e-06,
+      "loss": 1.0674,
+      "step": 189
+    },
+    {
+      "epoch": 0.48717948717948717,
+      "grad_norm": 1.7376888990402222,
+      "learning_rate": 5.1804123711340214e-06,
+      "loss": 0.8032,
+      "step": 190
+    },
+    {
+      "epoch": 0.4897435897435897,
+      "grad_norm": 1.5488858222961426,
+      "learning_rate": 5.154639175257732e-06,
+      "loss": 0.8101,
+      "step": 191
+    },
+    {
+      "epoch": 0.49230769230769234,
+      "grad_norm": 1.9175901412963867,
+      "learning_rate": 5.128865979381443e-06,
+      "loss": 0.8418,
+      "step": 192
+    },
+    {
+      "epoch": 0.4948717948717949,
+      "grad_norm": 2.069321393966675,
+      "learning_rate": 5.1030927835051544e-06,
+      "loss": 1.1826,
+      "step": 193
+    },
+    {
+      "epoch": 0.49743589743589745,
+      "grad_norm": 1.918543815612793,
+      "learning_rate": 5.077319587628866e-06,
+      "loss": 1.0059,
+      "step": 194
+    },
+    {
+      "epoch": 0.5,
+      "grad_norm": 1.7345309257507324,
+      "learning_rate": 5.051546391752578e-06,
+      "loss": 0.9028,
+      "step": 195
+    },
+    {
+      "epoch": 0.5025641025641026,
+      "grad_norm": 1.7581024169921875,
+      "learning_rate": 5.025773195876289e-06,
+      "loss": 1.0,
+      "step": 196
+    },
+    {
+      "epoch": 0.5051282051282051,
+      "grad_norm": 2.021634817123413,
+      "learning_rate": 5e-06,
+      "loss": 0.8677,
+      "step": 197
+    },
+    {
+      "epoch": 0.5076923076923077,
+      "grad_norm": 2.0879619121551514,
+      "learning_rate": 4.974226804123712e-06,
+      "loss": 1.1279,
+      "step": 198
+    },
+    {
+      "epoch": 0.5102564102564102,
+      "grad_norm": 2.040804386138916,
+      "learning_rate": 4.948453608247423e-06,
+      "loss": 1.2178,
+      "step": 199
+    },
+    {
+      "epoch": 0.5128205128205128,
+      "grad_norm": 2.0652830600738525,
+      "learning_rate": 4.922680412371135e-06,
+      "loss": 0.873,
+      "step": 200
+    },
+    {
+      "epoch": 0.5153846153846153,
+      "grad_norm": 1.8770358562469482,
+      "learning_rate": 4.8969072164948455e-06,
+      "loss": 0.7212,
+      "step": 201
+    },
+    {
+      "epoch": 0.517948717948718,
+      "grad_norm": 1.8745349645614624,
+      "learning_rate": 4.871134020618557e-06,
+      "loss": 1.126,
+      "step": 202
+    },
+    {
+      "epoch": 0.5205128205128206,
+      "grad_norm": 1.8858857154846191,
+      "learning_rate": 4.845360824742268e-06,
+      "loss": 1.0771,
+      "step": 203
+    },
+    {
+      "epoch": 0.5230769230769231,
+      "grad_norm": 2.110069513320923,
+      "learning_rate": 4.81958762886598e-06,
+      "loss": 1.2754,
+      "step": 204
+    },
+    {
+      "epoch": 0.5256410256410257,
+      "grad_norm": 4.52376127243042,
+      "learning_rate": 4.7938144329896915e-06,
+      "loss": 1.1182,
+      "step": 205
+    },
+    {
+      "epoch": 0.5282051282051282,
+      "grad_norm": 1.6591471433639526,
+      "learning_rate": 4.768041237113403e-06,
+      "loss": 0.8276,
+      "step": 206
+    },
+    {
+      "epoch": 0.5307692307692308,
+      "grad_norm": 1.9472140073776245,
+      "learning_rate": 4.742268041237113e-06,
+      "loss": 1.0273,
+      "step": 207
+    },
+    {
+      "epoch": 0.5333333333333333,
+      "grad_norm": 1.8485890626907349,
+      "learning_rate": 4.716494845360825e-06,
+      "loss": 0.813,
+      "step": 208
+    },
+    {
+      "epoch": 0.5358974358974359,
+      "grad_norm": 1.7967491149902344,
+      "learning_rate": 4.690721649484537e-06,
+      "loss": 0.9946,
+      "step": 209
+    },
+    {
+      "epoch": 0.5384615384615384,
+      "grad_norm": 1.7534973621368408,
+      "learning_rate": 4.664948453608248e-06,
+      "loss": 0.7993,
+      "step": 210
+    },
+    {
+      "epoch": 0.541025641025641,
+      "grad_norm": 1.8620507717132568,
+      "learning_rate": 4.639175257731959e-06,
+      "loss": 0.9341,
+      "step": 211
+    },
+    {
+      "epoch": 0.5435897435897435,
+      "grad_norm": 1.8512839078903198,
+      "learning_rate": 4.6134020618556705e-06,
+      "loss": 0.8013,
+      "step": 212
+    },
+    {
+      "epoch": 0.5461538461538461,
+      "grad_norm": 2.151174306869507,
+      "learning_rate": 4.587628865979382e-06,
+      "loss": 1.2969,
+      "step": 213
+    },
+    {
+      "epoch": 0.5487179487179488,
+      "grad_norm": 1.9421318769454956,
+      "learning_rate": 4.561855670103093e-06,
+      "loss": 0.9375,
+      "step": 214
+    },
+    {
+      "epoch": 0.5512820512820513,
+      "grad_norm": 1.6634801626205444,
+      "learning_rate": 4.536082474226804e-06,
+      "loss": 0.7603,
+      "step": 215
+    },
+    {
+      "epoch": 0.5538461538461539,
+      "grad_norm": 1.8529914617538452,
+      "learning_rate": 4.510309278350516e-06,
+      "loss": 0.8545,
+      "step": 216
+    },
+    {
+      "epoch": 0.5564102564102564,
+      "grad_norm": 1.8996697664260864,
+      "learning_rate": 4.484536082474228e-06,
+      "loss": 0.9492,
+      "step": 217
+    },
+    {
+      "epoch": 0.558974358974359,
+      "grad_norm": 1.793915033340454,
+      "learning_rate": 4.458762886597939e-06,
+      "loss": 0.6826,
+      "step": 218
+    },
+    {
+      "epoch": 0.5615384615384615,
+      "grad_norm": 2.973825454711914,
+      "learning_rate": 4.4329896907216494e-06,
+      "loss": 0.9023,
+      "step": 219
+    },
+    {
+      "epoch": 0.5641025641025641,
+      "grad_norm": 1.786086082458496,
+      "learning_rate": 4.407216494845361e-06,
+      "loss": 0.9502,
+      "step": 220
+    },
+    {
+      "epoch": 0.5666666666666667,
+      "grad_norm": 2.263026714324951,
+      "learning_rate": 4.381443298969073e-06,
+      "loss": 0.9326,
+      "step": 221
+    },
+    {
+      "epoch": 0.5692307692307692,
+      "grad_norm": 2.35546612739563,
+      "learning_rate": 4.355670103092784e-06,
+      "loss": 1.1348,
+      "step": 222
+    },
+    {
+      "epoch": 0.5717948717948718,
+      "grad_norm": 1.6735016107559204,
+      "learning_rate": 4.329896907216495e-06,
+      "loss": 0.9204,
+      "step": 223
+    },
+    {
+      "epoch": 0.5743589743589743,
+      "grad_norm": 2.4395010471343994,
+      "learning_rate": 4.304123711340207e-06,
+      "loss": 1.1689,
+      "step": 224
+    },
+    {
+      "epoch": 0.5769230769230769,
+      "grad_norm": 1.8484439849853516,
+      "learning_rate": 4.278350515463918e-06,
+      "loss": 0.9751,
+      "step": 225
+    },
+    {
+      "epoch": 0.5794871794871795,
+      "grad_norm": 2.61309552192688,
+      "learning_rate": 4.252577319587629e-06,
+      "loss": 0.9272,
+      "step": 226
+    },
+    {
+      "epoch": 0.5820512820512821,
+      "grad_norm": 1.8143435716629028,
+      "learning_rate": 4.2268041237113405e-06,
+      "loss": 1.0508,
+      "step": 227
+    },
+    {
+      "epoch": 0.5846153846153846,
+      "grad_norm": 1.9212270975112915,
+      "learning_rate": 4.201030927835052e-06,
+      "loss": 0.8813,
+      "step": 228
+    },
+    {
+      "epoch": 0.5871794871794872,
+      "grad_norm": 2.3576104640960693,
+      "learning_rate": 4.175257731958763e-06,
+      "loss": 0.98,
+      "step": 229
+    },
+    {
+      "epoch": 0.5897435897435898,
+      "grad_norm": 2.2258710861206055,
+      "learning_rate": 4.149484536082475e-06,
+      "loss": 0.9766,
+      "step": 230
+    },
+    {
+      "epoch": 0.5923076923076923,
+      "grad_norm": 1.8784958124160767,
+      "learning_rate": 4.123711340206186e-06,
+      "loss": 0.9575,
+      "step": 231
+    },
+    {
+      "epoch": 0.5948717948717949,
+      "grad_norm": 1.7360793352127075,
+      "learning_rate": 4.097938144329897e-06,
+      "loss": 0.9014,
+      "step": 232
+    },
+    {
+      "epoch": 0.5974358974358974,
+      "grad_norm": 2.718904733657837,
+      "learning_rate": 4.072164948453608e-06,
+      "loss": 0.8682,
+      "step": 233
+    },
+    {
+      "epoch": 0.6,
+      "grad_norm": 3.4316585063934326,
+      "learning_rate": 4.04639175257732e-06,
+      "loss": 0.8096,
+      "step": 234
+    },
+    {
+      "epoch": 0.6025641025641025,
+      "grad_norm": 1.9802138805389404,
+      "learning_rate": 4.020618556701032e-06,
+      "loss": 1.1055,
+      "step": 235
+    },
+    {
+      "epoch": 0.6051282051282051,
+      "grad_norm": 2.3054115772247314,
+      "learning_rate": 3.994845360824743e-06,
+      "loss": 0.8696,
+      "step": 236
+    },
+    {
+      "epoch": 0.6076923076923076,
+      "grad_norm": 2.1070103645324707,
+      "learning_rate": 3.969072164948453e-06,
+      "loss": 0.9907,
+      "step": 237
+    },
+    {
+      "epoch": 0.6102564102564103,
+      "grad_norm": 1.71999192237854,
+      "learning_rate": 3.9432989690721655e-06,
+      "loss": 1.0342,
+      "step": 238
+    },
+    {
+      "epoch": 0.6128205128205129,
+      "grad_norm": 1.7554974555969238,
+      "learning_rate": 3.917525773195877e-06,
+      "loss": 0.9712,
+      "step": 239
+    },
+    {
+      "epoch": 0.6153846153846154,
+      "grad_norm": 1.8095436096191406,
+      "learning_rate": 3.891752577319588e-06,
+      "loss": 1.0586,
+      "step": 240
+    },
+    {
+      "epoch": 0.617948717948718,
+      "grad_norm": 2.331573963165283,
+      "learning_rate": 3.865979381443299e-06,
+      "loss": 1.1729,
+      "step": 241
+    },
+    {
+      "epoch": 0.6205128205128205,
+      "grad_norm": 1.8929247856140137,
+      "learning_rate": 3.840206185567011e-06,
+      "loss": 1.0576,
+      "step": 242
+    },
+    {
+      "epoch": 0.6230769230769231,
+      "grad_norm": 1.6739861965179443,
+      "learning_rate": 3.814432989690722e-06,
+      "loss": 0.8008,
+      "step": 243
+    },
+    {
+      "epoch": 0.6256410256410256,
+      "grad_norm": 1.6616827249526978,
+      "learning_rate": 3.788659793814433e-06,
+      "loss": 0.876,
+      "step": 244
+    },
+    {
+      "epoch": 0.6282051282051282,
+      "grad_norm": 7.3671746253967285,
+      "learning_rate": 3.7628865979381445e-06,
+      "loss": 1.0293,
+      "step": 245
+    },
+    {
+      "epoch": 0.6307692307692307,
+      "grad_norm": 1.9617962837219238,
+      "learning_rate": 3.737113402061856e-06,
+      "loss": 1.0117,
+      "step": 246
+    },
+    {
+      "epoch": 0.6333333333333333,
+      "grad_norm": 1.3733716011047363,
+      "learning_rate": 3.7113402061855674e-06,
+      "loss": 0.6865,
+      "step": 247
+    },
+    {
+      "epoch": 0.6358974358974359,
+      "grad_norm": 1.6532529592514038,
+      "learning_rate": 3.6855670103092787e-06,
+      "loss": 0.9604,
+      "step": 248
+    },
+    {
+      "epoch": 0.6384615384615384,
+      "grad_norm": 1.8402736186981201,
+      "learning_rate": 3.6597938144329896e-06,
+      "loss": 0.916,
+      "step": 249
+    },
+    {
+      "epoch": 0.6410256410256411,
+      "grad_norm": 1.7600955963134766,
+      "learning_rate": 3.6340206185567013e-06,
+      "loss": 0.9678,
+      "step": 250
+    },
+    {
+      "epoch": 0.6435897435897436,
+      "grad_norm": 1.6891589164733887,
+      "learning_rate": 3.6082474226804126e-06,
+      "loss": 0.793,
+      "step": 251
+    },
+    {
+      "epoch": 0.6461538461538462,
+      "grad_norm": 1.8142110109329224,
+      "learning_rate": 3.582474226804124e-06,
+      "loss": 1.0518,
+      "step": 252
+    },
+    {
+      "epoch": 0.6487179487179487,
+      "grad_norm": 2.1421236991882324,
+      "learning_rate": 3.5567010309278356e-06,
+      "loss": 0.9414,
+      "step": 253
+    },
+    {
+      "epoch": 0.6512820512820513,
+      "grad_norm": 5.681818962097168,
+      "learning_rate": 3.530927835051547e-06,
+      "loss": 0.9951,
+      "step": 254
+    },
+    {
+      "epoch": 0.6538461538461539,
+      "grad_norm": 2.093968391418457,
+      "learning_rate": 3.5051546391752577e-06,
+      "loss": 0.7842,
+      "step": 255
+    },
+    {
+      "epoch": 0.6564102564102564,
+      "grad_norm": 1.8707085847854614,
+      "learning_rate": 3.4793814432989694e-06,
+      "loss": 0.9858,
+      "step": 256
+    },
+    {
+      "epoch": 0.658974358974359,
+      "grad_norm": 1.9755574464797974,
+      "learning_rate": 3.4536082474226807e-06,
+      "loss": 0.9136,
+      "step": 257
+    },
+    {
+      "epoch": 0.6615384615384615,
+      "grad_norm": 2.0634946823120117,
+      "learning_rate": 3.427835051546392e-06,
+      "loss": 0.7593,
+      "step": 258
+    },
+    {
+      "epoch": 0.6641025641025641,
+      "grad_norm": 1.842301845550537,
+      "learning_rate": 3.4020618556701037e-06,
+      "loss": 0.8062,
+      "step": 259
+    },
+    {
+      "epoch": 0.6666666666666666,
+      "grad_norm": 1.857254981994629,
+      "learning_rate": 3.376288659793815e-06,
+      "loss": 1.2021,
+      "step": 260
+    },
+    {
+      "epoch": 0.6692307692307692,
+      "grad_norm": 1.9125348329544067,
+      "learning_rate": 3.350515463917526e-06,
+      "loss": 1.084,
+      "step": 261
+    },
+    {
+      "epoch": 0.6717948717948717,
+      "grad_norm": 1.8903979063034058,
+      "learning_rate": 3.324742268041237e-06,
+      "loss": 0.7666,
+      "step": 262
+    },
+    {
+      "epoch": 0.6743589743589744,
+      "grad_norm": 1.981501579284668,
+      "learning_rate": 3.298969072164949e-06,
+      "loss": 1.0508,
+      "step": 263
+    },
+    {
+      "epoch": 0.676923076923077,
+      "grad_norm": 1.5312262773513794,
+      "learning_rate": 3.27319587628866e-06,
+      "loss": 0.7231,
+      "step": 264
+    },
+    {
+      "epoch": 0.6794871794871795,
+      "grad_norm": 1.6376142501831055,
+      "learning_rate": 3.2474226804123714e-06,
+      "loss": 0.8013,
+      "step": 265
+    },
+    {
+      "epoch": 0.6820512820512821,
+      "grad_norm": 1.5917991399765015,
+      "learning_rate": 3.221649484536083e-06,
+      "loss": 0.8042,
+      "step": 266
+    },
+    {
+      "epoch": 0.6846153846153846,
+      "grad_norm": 1.497527837753296,
+      "learning_rate": 3.195876288659794e-06,
+      "loss": 0.7031,
+      "step": 267
+    },
+    {
+      "epoch": 0.6871794871794872,
+      "grad_norm": 1.692023515701294,
+      "learning_rate": 3.170103092783505e-06,
+      "loss": 0.6646,
+      "step": 268
+    },
+    {
+      "epoch": 0.6897435897435897,
+      "grad_norm": 1.862017273902893,
+      "learning_rate": 3.1443298969072165e-06,
+      "loss": 0.8765,
+      "step": 269
+    },
+    {
+      "epoch": 0.6923076923076923,
+      "grad_norm": 2.128854513168335,
+      "learning_rate": 3.118556701030928e-06,
+      "loss": 0.7183,
+      "step": 270
+    },
+    {
+      "epoch": 0.6948717948717948,
+      "grad_norm": 1.7785433530807495,
+      "learning_rate": 3.0927835051546395e-06,
+      "loss": 0.9966,
+      "step": 271
+    },
+    {
+      "epoch": 0.6974358974358974,
+      "grad_norm": 1.8306283950805664,
+      "learning_rate": 3.067010309278351e-06,
+      "loss": 0.7773,
+      "step": 272
+    },
+    {
+      "epoch": 0.7,
+      "grad_norm": 1.5251365900039673,
+      "learning_rate": 3.041237113402062e-06,
+      "loss": 0.7046,
+      "step": 273
+    },
+    {
+      "epoch": 0.7025641025641025,
+      "grad_norm": 2.0568642616271973,
+      "learning_rate": 3.0154639175257733e-06,
+      "loss": 1.0459,
+      "step": 274
+    },
+    {
+      "epoch": 0.7051282051282052,
+      "grad_norm": 1.6169660091400146,
+      "learning_rate": 2.9896907216494846e-06,
+      "loss": 0.7451,
+      "step": 275
+    },
+    {
+      "epoch": 0.7076923076923077,
+      "grad_norm": 1.6986936330795288,
+      "learning_rate": 2.9639175257731963e-06,
+      "loss": 0.8735,
+      "step": 276
+    },
+    {
+      "epoch": 0.7102564102564103,
+      "grad_norm": 1.6713733673095703,
+      "learning_rate": 2.9381443298969076e-06,
+      "loss": 0.8862,
+      "step": 277
+    },
+    {
+      "epoch": 0.7128205128205128,
+      "grad_norm": 1.7637569904327393,
+      "learning_rate": 2.912371134020619e-06,
+      "loss": 0.9609,
+      "step": 278
+    },
+    {
+      "epoch": 0.7153846153846154,
+      "grad_norm": 2.076930284500122,
+      "learning_rate": 2.8865979381443297e-06,
+      "loss": 1.1377,
+      "step": 279
+    },
+    {
+      "epoch": 0.717948717948718,
+      "grad_norm": 1.972031831741333,
+      "learning_rate": 2.8608247422680414e-06,
+      "loss": 0.9395,
+      "step": 280
+    },
+    {
+      "epoch": 0.7205128205128205,
+      "grad_norm": 1.8608795404434204,
+      "learning_rate": 2.8350515463917527e-06,
+      "loss": 1.0654,
+      "step": 281
+    },
+    {
+      "epoch": 0.7230769230769231,
+      "grad_norm": 3.467540979385376,
+      "learning_rate": 2.809278350515464e-06,
+      "loss": 0.7603,
+      "step": 282
+    },
+    {
+      "epoch": 0.7256410256410256,
+      "grad_norm": 1.765555739402771,
+      "learning_rate": 2.7835051546391757e-06,
+      "loss": 0.6758,
+      "step": 283
+    },
+    {
+      "epoch": 0.7282051282051282,
+      "grad_norm": 1.6693044900894165,
+      "learning_rate": 2.757731958762887e-06,
+      "loss": 0.8433,
+      "step": 284
+    },
+    {
+      "epoch": 0.7307692307692307,
+      "grad_norm": 1.9119174480438232,
+      "learning_rate": 2.731958762886598e-06,
+      "loss": 0.8184,
+      "step": 285
+    },
+    {
+      "epoch": 0.7333333333333333,
+      "grad_norm": 2.043612241744995,
+      "learning_rate": 2.7061855670103095e-06,
+      "loss": 1.0049,
+      "step": 286
+    },
+    {
+      "epoch": 0.735897435897436,
+      "grad_norm": 1.919756531715393,
+      "learning_rate": 2.680412371134021e-06,
+      "loss": 0.9229,
+      "step": 287
+    },
+    {
+      "epoch": 0.7384615384615385,
+      "grad_norm": 1.8900898694992065,
+      "learning_rate": 2.654639175257732e-06,
+      "loss": 1.0078,
+      "step": 288
+    },
+    {
+      "epoch": 0.7410256410256411,
+      "grad_norm": 1.793310284614563,
+      "learning_rate": 2.628865979381444e-06,
+      "loss": 1.0752,
+      "step": 289
+    },
+    {
+      "epoch": 0.7435897435897436,
+      "grad_norm": 2.732642889022827,
+      "learning_rate": 2.603092783505155e-06,
+      "loss": 0.9229,
+      "step": 290
+    },
+    {
+      "epoch": 0.7461538461538462,
+      "grad_norm": 1.845354676246643,
+      "learning_rate": 2.577319587628866e-06,
+      "loss": 0.8286,
+      "step": 291
+    },
+    {
+      "epoch": 0.7487179487179487,
+      "grad_norm": 1.9272360801696777,
+      "learning_rate": 2.5515463917525772e-06,
+      "loss": 1.0518,
+      "step": 292
+    },
+    {
+      "epoch": 0.7512820512820513,
+      "grad_norm": 2.1714913845062256,
+      "learning_rate": 2.525773195876289e-06,
+      "loss": 0.9043,
+      "step": 293
+    },
+    {
+      "epoch": 0.7538461538461538,
+      "grad_norm": 1.7473788261413574,
+      "learning_rate": 2.5e-06,
+      "loss": 0.7793,
+      "step": 294
+    },
+    {
+      "epoch": 0.7564102564102564,
+      "grad_norm": 1.9058947563171387,
+      "learning_rate": 2.4742268041237115e-06,
+      "loss": 0.8643,
+      "step": 295
+    },
+    {
+      "epoch": 0.7589743589743589,
+      "grad_norm": 2.064485788345337,
+      "learning_rate": 2.4484536082474228e-06,
+      "loss": 1.1475,
+      "step": 296
+    },
+    {
+      "epoch": 0.7615384615384615,
+      "grad_norm": 2.100177526473999,
+      "learning_rate": 2.422680412371134e-06,
+      "loss": 0.853,
+      "step": 297
+    },
+    {
+      "epoch": 0.764102564102564,
+      "grad_norm": 2.2327897548675537,
+      "learning_rate": 2.3969072164948458e-06,
+      "loss": 1.2168,
+      "step": 298
+    },
+    {
+      "epoch": 0.7666666666666667,
+      "grad_norm": 1.9748061895370483,
+      "learning_rate": 2.3711340206185566e-06,
+      "loss": 1.0713,
+      "step": 299
+    },
+    {
+      "epoch": 0.7692307692307693,
+      "grad_norm": 1.987243890762329,
+      "learning_rate": 2.3453608247422683e-06,
+      "loss": 1.0508,
+      "step": 300
+    },
+    {
+      "epoch": 0.7717948717948718,
+      "grad_norm": 2.191959857940674,
+      "learning_rate": 2.3195876288659796e-06,
+      "loss": 0.9326,
+      "step": 301
+    },
+    {
+      "epoch": 0.7743589743589744,
+      "grad_norm": 2.242733955383301,
+      "learning_rate": 2.293814432989691e-06,
+      "loss": 0.7515,
+      "step": 302
+    },
+    {
+      "epoch": 0.7769230769230769,
+      "grad_norm": 1.6882741451263428,
+      "learning_rate": 2.268041237113402e-06,
+      "loss": 0.7534,
+      "step": 303
+    },
+    {
+      "epoch": 0.7794871794871795,
+      "grad_norm": 2.7890372276306152,
+      "learning_rate": 2.242268041237114e-06,
+      "loss": 0.9082,
+      "step": 304
+    },
+    {
+      "epoch": 0.782051282051282,
+      "grad_norm": 2.7322838306427,
+      "learning_rate": 2.2164948453608247e-06,
+      "loss": 1.0332,
+      "step": 305
+    },
+    {
+      "epoch": 0.7846153846153846,
+      "grad_norm": 1.9101502895355225,
+      "learning_rate": 2.1907216494845364e-06,
+      "loss": 1.0088,
+      "step": 306
+    },
+    {
+      "epoch": 0.7871794871794872,
+      "grad_norm": 1.7795121669769287,
+      "learning_rate": 2.1649484536082477e-06,
+      "loss": 0.9082,
+      "step": 307
+    },
+    {
+      "epoch": 0.7897435897435897,
+      "grad_norm": 1.8345370292663574,
+      "learning_rate": 2.139175257731959e-06,
+      "loss": 1.0576,
+      "step": 308
+    },
+    {
+      "epoch": 0.7923076923076923,
+      "grad_norm": 2.1167471408843994,
+      "learning_rate": 2.1134020618556703e-06,
+      "loss": 0.918,
+      "step": 309
+    },
+    {
+      "epoch": 0.7948717948717948,
+      "grad_norm": 1.7261496782302856,
+      "learning_rate": 2.0876288659793816e-06,
+      "loss": 0.8975,
+      "step": 310
+    },
+    {
+      "epoch": 0.7974358974358975,
+      "grad_norm": 1.793904185295105,
+      "learning_rate": 2.061855670103093e-06,
+      "loss": 0.8916,
+      "step": 311
+    },
+    {
+      "epoch": 0.8,
+      "grad_norm": 1.9415448904037476,
+      "learning_rate": 2.036082474226804e-06,
+      "loss": 0.8379,
+      "step": 312
+    },
+    {
+      "epoch": 0.8025641025641026,
+      "grad_norm": 1.8090825080871582,
+      "learning_rate": 2.010309278350516e-06,
+      "loss": 0.8784,
+      "step": 313
+    },
+    {
+      "epoch": 0.8051282051282052,
+      "grad_norm": 1.775429368019104,
+      "learning_rate": 1.9845360824742267e-06,
+      "loss": 0.8643,
+      "step": 314
+    },
+    {
+      "epoch": 0.8076923076923077,
+      "grad_norm": 1.848243236541748,
+      "learning_rate": 1.9587628865979384e-06,
+      "loss": 1.1084,
+      "step": 315
+    },
+    {
+      "epoch": 0.8102564102564103,
+      "grad_norm": 1.7999858856201172,
+      "learning_rate": 1.9329896907216497e-06,
+      "loss": 0.9321,
+      "step": 316
+    },
+    {
+      "epoch": 0.8128205128205128,
+      "grad_norm": 2.8370208740234375,
+      "learning_rate": 1.907216494845361e-06,
+      "loss": 0.9883,
+      "step": 317
+    },
+    {
+      "epoch": 0.8153846153846154,
+      "grad_norm": 2.2633893489837646,
+      "learning_rate": 1.8814432989690722e-06,
+      "loss": 1.0166,
+      "step": 318
+    },
+    {
+      "epoch": 0.8179487179487179,
+      "grad_norm": 1.9850558042526245,
+      "learning_rate": 1.8556701030927837e-06,
+      "loss": 1.0283,
+      "step": 319
+    },
+    {
+      "epoch": 0.8205128205128205,
+      "grad_norm": 1.904843807220459,
+      "learning_rate": 1.8298969072164948e-06,
+      "loss": 0.9829,
+      "step": 320
+    },
+    {
+      "epoch": 0.823076923076923,
+      "grad_norm": 1.7778921127319336,
+      "learning_rate": 1.8041237113402063e-06,
+      "loss": 0.9229,
+      "step": 321
+    },
+    {
+      "epoch": 0.8256410256410256,
+      "grad_norm": 1.792877197265625,
+      "learning_rate": 1.7783505154639178e-06,
+      "loss": 0.876,
+      "step": 322
+    },
+    {
+      "epoch": 0.8282051282051283,
+      "grad_norm": 1.8950697183609009,
+      "learning_rate": 1.7525773195876288e-06,
+      "loss": 0.915,
+      "step": 323
+    },
+    {
+      "epoch": 0.8307692307692308,
+      "grad_norm": 1.911402940750122,
+      "learning_rate": 1.7268041237113403e-06,
+      "loss": 1.043,
+      "step": 324
+    },
+    {
+      "epoch": 0.8333333333333334,
+      "grad_norm": 2.1029365062713623,
+      "learning_rate": 1.7010309278350518e-06,
+      "loss": 0.9243,
+      "step": 325
+    },
+    {
+      "epoch": 0.8358974358974359,
+      "grad_norm": 2.052480936050415,
+      "learning_rate": 1.675257731958763e-06,
+      "loss": 1.2256,
+      "step": 326
+    },
+    {
+      "epoch": 0.8384615384615385,
+      "grad_norm": 1.7406338453292847,
+      "learning_rate": 1.6494845360824744e-06,
+      "loss": 0.8804,
+      "step": 327
+    },
+    {
+      "epoch": 0.841025641025641,
+      "grad_norm": 1.8102054595947266,
+      "learning_rate": 1.6237113402061857e-06,
+      "loss": 0.854,
+      "step": 328
+    },
+    {
+      "epoch": 0.8435897435897436,
+      "grad_norm": 1.9868026971817017,
+      "learning_rate": 1.597938144329897e-06,
+      "loss": 0.8491,
+      "step": 329
+    },
+    {
+      "epoch": 0.8461538461538461,
+      "grad_norm": 1.774075984954834,
+      "learning_rate": 1.5721649484536082e-06,
+      "loss": 0.8457,
+      "step": 330
+    },
+    {
+      "epoch": 0.8487179487179487,
+      "grad_norm": 1.8672053813934326,
+      "learning_rate": 1.5463917525773197e-06,
+      "loss": 1.0176,
+      "step": 331
+    },
+    {
+      "epoch": 0.8512820512820513,
+      "grad_norm": 1.5850166082382202,
+      "learning_rate": 1.520618556701031e-06,
+      "loss": 0.6777,
+      "step": 332
+    },
+    {
+      "epoch": 0.8538461538461538,
+      "grad_norm": 1.6745821237564087,
+      "learning_rate": 1.4948453608247423e-06,
+      "loss": 0.939,
+      "step": 333
+    },
+    {
+      "epoch": 0.8564102564102564,
+      "grad_norm": 1.8641244173049927,
+      "learning_rate": 1.4690721649484538e-06,
+      "loss": 1.0,
+      "step": 334
+    },
+    {
+      "epoch": 0.8589743589743589,
+      "grad_norm": 1.6371408700942993,
+      "learning_rate": 1.4432989690721649e-06,
+      "loss": 0.8037,
+      "step": 335
+    },
+    {
+      "epoch": 0.8615384615384616,
+      "grad_norm": 2.0685510635375977,
+      "learning_rate": 1.4175257731958764e-06,
+      "loss": 0.9287,
+      "step": 336
+    },
+    {
+      "epoch": 0.8641025641025641,
+      "grad_norm": 1.8242462873458862,
+      "learning_rate": 1.3917525773195878e-06,
+      "loss": 0.8877,
+      "step": 337
+    },
+    {
+      "epoch": 0.8666666666666667,
+      "grad_norm": 2.243830442428589,
+      "learning_rate": 1.365979381443299e-06,
+      "loss": 0.7764,
+      "step": 338
+    },
+    {
+      "epoch": 0.8692307692307693,
+      "grad_norm": 2.188321590423584,
+      "learning_rate": 1.3402061855670104e-06,
+      "loss": 1.125,
+      "step": 339
+    },
+    {
+      "epoch": 0.8717948717948718,
+      "grad_norm": 2.412721633911133,
+      "learning_rate": 1.314432989690722e-06,
+      "loss": 1.1016,
+      "step": 340
+    },
+    {
+      "epoch": 0.8743589743589744,
+      "grad_norm": 2.149219512939453,
+      "learning_rate": 1.288659793814433e-06,
+      "loss": 1.1104,
+      "step": 341
+    },
+    {
+      "epoch": 0.8769230769230769,
+      "grad_norm": 1.603258490562439,
+      "learning_rate": 1.2628865979381445e-06,
+      "loss": 0.7148,
+      "step": 342
+    },
+    {
+      "epoch": 0.8794871794871795,
+      "grad_norm": 1.7832096815109253,
+      "learning_rate": 1.2371134020618557e-06,
+      "loss": 0.9902,
+      "step": 343
+    },
+    {
+      "epoch": 0.882051282051282,
+      "grad_norm": 1.8631786108016968,
+      "learning_rate": 1.211340206185567e-06,
+      "loss": 0.8569,
+      "step": 344
+    },
+    {
+      "epoch": 0.8846153846153846,
+      "grad_norm": 1.6944712400436401,
+      "learning_rate": 1.1855670103092783e-06,
+      "loss": 0.7397,
+      "step": 345
+    },
+    {
+      "epoch": 0.8871794871794871,
+      "grad_norm": 1.842955231666565,
+      "learning_rate": 1.1597938144329898e-06,
+      "loss": 0.8457,
+      "step": 346
+    },
+    {
+      "epoch": 0.8897435897435897,
+      "grad_norm": 2.5506865978240967,
+      "learning_rate": 1.134020618556701e-06,
+      "loss": 0.9238,
+      "step": 347
+    },
+    {
+      "epoch": 0.8923076923076924,
+      "grad_norm": 2.157381534576416,
+      "learning_rate": 1.1082474226804124e-06,
+      "loss": 1.1855,
+      "step": 348
+    },
+    {
+      "epoch": 0.8948717948717949,
+      "grad_norm": 1.835581660270691,
+      "learning_rate": 1.0824742268041239e-06,
+      "loss": 0.9375,
+      "step": 349
+    },
+    {
+      "epoch": 0.8974358974358975,
+      "grad_norm": 2.591947078704834,
+      "learning_rate": 1.0567010309278351e-06,
+      "loss": 1.0088,
+      "step": 350
+    },
+    {
+      "epoch": 0.9,
+      "grad_norm": 1.9977459907531738,
+      "learning_rate": 1.0309278350515464e-06,
+      "loss": 1.0576,
+      "step": 351
+    },
+    {
+      "epoch": 0.9025641025641026,
+      "grad_norm": 1.7029050588607788,
+      "learning_rate": 1.005154639175258e-06,
+      "loss": 0.8643,
+      "step": 352
+    },
+    {
+      "epoch": 0.9051282051282051,
+      "grad_norm": 1.6559293270111084,
+      "learning_rate": 9.793814432989692e-07,
+      "loss": 0.8652,
+      "step": 353
+    },
+    {
+      "epoch": 0.9076923076923077,
+      "grad_norm": 2.0177323818206787,
+      "learning_rate": 9.536082474226805e-07,
+      "loss": 1.0361,
+      "step": 354
+    },
+    {
+      "epoch": 0.9102564102564102,
+      "grad_norm": 1.6591455936431885,
+      "learning_rate": 9.278350515463919e-07,
+      "loss": 0.8872,
+      "step": 355
+    },
+    {
+      "epoch": 0.9128205128205128,
+      "grad_norm": 2.092712640762329,
+      "learning_rate": 9.020618556701031e-07,
+      "loss": 0.9644,
+      "step": 356
+    },
+    {
+      "epoch": 0.9153846153846154,
+      "grad_norm": 1.7917448282241821,
+      "learning_rate": 8.762886597938144e-07,
+      "loss": 0.9399,
+      "step": 357
+    },
+    {
+      "epoch": 0.9179487179487179,
+      "grad_norm": 1.9611492156982422,
+      "learning_rate": 8.505154639175259e-07,
+      "loss": 1.292,
+      "step": 358
+    },
+    {
+      "epoch": 0.9205128205128205,
+      "grad_norm": 1.7951692342758179,
+      "learning_rate": 8.247422680412372e-07,
+      "loss": 0.9551,
+      "step": 359
+    },
+    {
+      "epoch": 0.9230769230769231,
+      "grad_norm": 1.8831959962844849,
+      "learning_rate": 7.989690721649485e-07,
+      "loss": 1.1973,
+      "step": 360
+    },
+    {
+      "epoch": 0.9256410256410257,
+      "grad_norm": 1.9080712795257568,
+      "learning_rate": 7.731958762886599e-07,
+      "loss": 0.917,
+      "step": 361
+    },
+    {
+      "epoch": 0.9282051282051282,
+      "grad_norm": 2.041674852371216,
+      "learning_rate": 7.474226804123711e-07,
+      "loss": 0.9111,
+      "step": 362
+    },
+    {
+      "epoch": 0.9307692307692308,
+      "grad_norm": 1.8067349195480347,
+      "learning_rate": 7.216494845360824e-07,
+      "loss": 0.7915,
+      "step": 363
+    },
+    {
+      "epoch": 0.9333333333333333,
+      "grad_norm": 1.733258843421936,
+      "learning_rate": 6.958762886597939e-07,
+      "loss": 0.8726,
+      "step": 364
+    },
+    {
+      "epoch": 0.9358974358974359,
+      "grad_norm": 2.1133229732513428,
+      "learning_rate": 6.701030927835052e-07,
+      "loss": 0.9028,
+      "step": 365
+    },
+    {
+      "epoch": 0.9384615384615385,
+      "grad_norm": 1.8070833683013916,
+      "learning_rate": 6.443298969072165e-07,
+      "loss": 0.8071,
+      "step": 366
+    },
+    {
+      "epoch": 0.941025641025641,
+      "grad_norm": 1.610521912574768,
+      "learning_rate": 6.185567010309279e-07,
+      "loss": 0.686,
+      "step": 367
+    },
+    {
+      "epoch": 0.9435897435897436,
+      "grad_norm": 1.8273773193359375,
+      "learning_rate": 5.927835051546392e-07,
+      "loss": 1.042,
+      "step": 368
+    },
+    {
+      "epoch": 0.9461538461538461,
+      "grad_norm": 1.6408607959747314,
+      "learning_rate": 5.670103092783505e-07,
+      "loss": 0.8584,
+      "step": 369
+    },
+    {
+      "epoch": 0.9487179487179487,
+      "grad_norm": 1.950465440750122,
+      "learning_rate": 5.412371134020619e-07,
+      "loss": 0.9629,
+      "step": 370
+    },
+    {
+      "epoch": 0.9512820512820512,
+      "grad_norm": 1.6559863090515137,
+      "learning_rate": 5.154639175257732e-07,
+      "loss": 0.7119,
+      "step": 371
+    },
+    {
+      "epoch": 0.9538461538461539,
+      "grad_norm": 1.497727394104004,
+      "learning_rate": 4.896907216494846e-07,
+      "loss": 0.71,
+      "step": 372
+    },
+    {
+      "epoch": 0.9564102564102565,
+      "grad_norm": 1.7516483068466187,
+      "learning_rate": 4.6391752577319593e-07,
+      "loss": 0.7275,
+      "step": 373
+    },
+    {
+      "epoch": 0.958974358974359,
+      "grad_norm": 1.8231966495513916,
+      "learning_rate": 4.381443298969072e-07,
+      "loss": 1.0664,
+      "step": 374
+    },
+    {
+      "epoch": 0.9615384615384616,
+      "grad_norm": 2.928769826889038,
+      "learning_rate": 4.123711340206186e-07,
+      "loss": 0.7983,
+      "step": 375
+    },
+    {
+      "epoch": 0.9641025641025641,
+      "grad_norm": 1.5636661052703857,
+      "learning_rate": 3.8659793814432993e-07,
+      "loss": 0.6655,
+      "step": 376
+    },
+    {
+      "epoch": 0.9666666666666667,
+      "grad_norm": 1.7315055131912231,
+      "learning_rate": 3.608247422680412e-07,
+      "loss": 0.8994,
+      "step": 377
+    },
+    {
+      "epoch": 0.9692307692307692,
+      "grad_norm": 1.863347053527832,
+      "learning_rate": 3.350515463917526e-07,
+      "loss": 0.9229,
+      "step": 378
+    },
+    {
+      "epoch": 0.9717948717948718,
+      "grad_norm": 1.6344797611236572,
+      "learning_rate": 3.0927835051546394e-07,
+      "loss": 0.915,
+      "step": 379
+    },
+    {
+      "epoch": 0.9743589743589743,
+      "grad_norm": 1.7815521955490112,
+      "learning_rate": 2.8350515463917527e-07,
+      "loss": 1.002,
+      "step": 380
+    },
+    {
+      "epoch": 0.9769230769230769,
+      "grad_norm": 1.689065933227539,
+      "learning_rate": 2.577319587628866e-07,
+      "loss": 0.8452,
+      "step": 381
+    },
+    {
+      "epoch": 0.9794871794871794,
+      "grad_norm": 1.8653761148452759,
+      "learning_rate": 2.3195876288659797e-07,
+      "loss": 0.9282,
+      "step": 382
+    },
+    {
+      "epoch": 0.982051282051282,
+      "grad_norm": 1.7302168607711792,
+      "learning_rate": 2.061855670103093e-07,
+      "loss": 0.6318,
+      "step": 383
+    },
+    {
+      "epoch": 0.9846153846153847,
+      "grad_norm": 1.8644577264785767,
+      "learning_rate": 1.804123711340206e-07,
+      "loss": 0.8018,
+      "step": 384
+    },
+    {
+      "epoch": 0.9871794871794872,
+      "grad_norm": 1.725438117980957,
+      "learning_rate": 1.5463917525773197e-07,
+      "loss": 0.9722,
+      "step": 385
+    },
+    {
+      "epoch": 0.9897435897435898,
+      "grad_norm": 2.536639928817749,
+      "learning_rate": 1.288659793814433e-07,
+      "loss": 1.0117,
+      "step": 386
+    },
+    {
+      "epoch": 0.9923076923076923,
+      "grad_norm": 6.378511905670166,
+      "learning_rate": 1.0309278350515465e-07,
+      "loss": 1.0488,
+      "step": 387
+    },
+    {
+      "epoch": 0.9948717948717949,
+      "grad_norm": 1.8261665105819702,
+      "learning_rate": 7.731958762886598e-08,
+      "loss": 0.9814,
+      "step": 388
+    },
+    {
+      "epoch": 0.9974358974358974,
+      "grad_norm": 1.732452630996704,
+      "learning_rate": 5.1546391752577325e-08,
+      "loss": 0.7505,
+      "step": 389
+    },
+    {
+      "epoch": 1.0,
+      "grad_norm": 1.9269561767578125,
+      "learning_rate": 2.5773195876288662e-08,
+      "loss": 1.0332,
+      "step": 390
+    }
+  ],
+  "logging_steps": 1.0,
+  "max_steps": 390,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 1,
+  "save_steps": 1000,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": true
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 0.0,
+  "train_batch_size": 1,
+  "trial_name": null,
+  "trial_params": null
+}

checkpoint-390/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8d533dc63a1c358dd5f8c710590b95b45400195b129c56c2217d9b54ba19e8aa
+size 7736

checkpoint-390/zero_to_fp32.py

@@ -0,0 +1,604 @@
+#!/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 1, 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_dict = torch.load(f, map_location=device)
+        # immediately discard the potentially huge 2 optimizer states as we only care for fp32 master weights
+        # and also handle the case where it was already removed by another helper script
+        state_dict["optimizer_state_dict"].pop("optimizer_state_dict", None)
+        state_dicts.append(state_dict)
+
+    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, exclude_frozen_parameters):
+    """
+    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,
+                                                          exclude_frozen_parameters)
+    elif zero_stage == 3:
+        return _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                                          exclude_frozen_parameters)
+
+
+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 _has_callable(obj, fn):
+    attr = getattr(obj, fn, None)
+    return callable(attr)
+
+
+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() if _has_callable(shape, 'numel') else math.prod(shape)
+            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,
+                                               exclude_frozen_parameters):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    if not exclude_frozen_parameters:
+        _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,
+                                               exclude_frozen_parameters):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    if not exclude_frozen_parameters:
+        _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, exclude_frozen_parameters=False):
+    """
+    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``
+        - ``exclude_frozen_parameters``: exclude frozen parameters
+
+    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, exclude_frozen_parameters)
+
+
+def convert_zero_checkpoint_to_fp32_state_dict(checkpoint_dir, output_file, tag=None, exclude_frozen_parameters=False):
+    """
+    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``
+        - ``exclude_frozen_parameters``: exclude frozen parameters
+    """
+
+    state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag, exclude_frozen_parameters)
+    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("-t",
+                        "--tag",
+                        type=str,
+                        default=None,
+                        help="checkpoint tag used as a unique identifier for checkpoint. e.g., global_step1")
+    parser.add_argument("--exclude_frozen_parameters", action='store_true', help="exclude frozen parameters")
+    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,
+                                               tag=args.tag,
+                                               exclude_frozen_parameters=args.exclude_frozen_parameters)

config.json

@@ -0,0 +1,28 @@
+{
+  "_name_or_path": "/share/chaofan/code/IR-Studio-up/online_test/finetune_result/medical_huatuo/retriever_round2",
+  "architectures": [
+    "XLMRobertaModel"
+  ],
+  "attention_probs_dropout_prob": 0.1,
+  "bos_token_id": 0,
+  "classifier_dropout": null,
+  "eos_token_id": 2,
+  "hidden_act": "gelu",
+  "hidden_dropout_prob": 0.1,
+  "hidden_size": 1024,
+  "initializer_range": 0.02,
+  "intermediate_size": 4096,
+  "layer_norm_eps": 1e-05,
+  "max_position_embeddings": 8194,
+  "model_type": "xlm-roberta",
+  "num_attention_heads": 16,
+  "num_hidden_layers": 24,
+  "output_past": true,
+  "pad_token_id": 1,
+  "position_embedding_type": "absolute",
+  "torch_dtype": "float16",
+  "transformers_version": "4.44.2",
+  "type_vocab_size": 1,
+  "use_cache": true,
+  "vocab_size": 250002
+}

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:855b4280f2d306c83c3ecfddf1ddbaae1d9032d4b20f51853faaa0d6a16c0f77
+size 1135554344

sentencepiece.bpe.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cfc8146abe2a0488e9e2a0c56de7952f7c11ab059eca145a0a727afce0db2865
+size 5069051

special_tokens_map.json

@@ -0,0 +1,51 @@
+{
+  "bos_token": {
+    "content": "<s>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "cls_token": {
+    "content": "<s>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "eos_token": {
+    "content": "</s>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "mask_token": {
+    "content": "<mask>",
+    "lstrip": true,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": {
+    "content": "<pad>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "sep_token": {
+    "content": "</s>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "unk_token": {
+    "content": "<unk>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}

tokenizer.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:33cd99e33ce09bdd8a6136fddfe90a1c47f85bafedf7309d0eecc19012d43586
+size 17082897

tokenizer_config.json

@@ -0,0 +1,59 @@
+{
+  "added_tokens_decoder": {
+    "0": {
+      "content": "<s>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "1": {
+      "content": "<pad>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "2": {
+      "content": "</s>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "3": {
+      "content": "<unk>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "250001": {
+      "content": "<mask>",
+      "lstrip": true,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    }
+  },
+  "bos_token": "<s>",
+  "clean_up_tokenization_spaces": true,
+  "cls_token": "<s>",
+  "eos_token": "</s>",
+  "mask_token": "<mask>",
+  "max_length": 512,
+  "model_max_length": 8192,
+  "pad_token": "<pad>",
+  "sep_token": "</s>",
+  "sp_model_kwargs": {},
+  "stride": 0,
+  "tokenizer_class": "XLMRobertaTokenizer",
+  "truncation_side": "right",
+  "truncation_strategy": "longest_first",
+  "unk_token": "<unk>"
+}

training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8d533dc63a1c358dd5f8c710590b95b45400195b129c56c2217d9b54ba19e8aa
+size 7736