Upload folder using huggingface_hub

133_128_e3_3e-5/adapter_config.json

@@ -0,0 +1,39 @@
+{
+  "alpha_pattern": {},
+  "auto_mapping": null,
+  "base_model_name_or_path": "ibm-granite/granite-3.3-8b-base",
+  "bias": "none",
+  "corda_config": null,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 256,
+  "lora_bias": false,
+  "lora_dropout": 0.05,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "r": 128,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "q_proj",
+    "down_proj",
+    "k_proj",
+    "o_proj",
+    "gate_proj",
+    "v_proj",
+    "up_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_rslora": false
+}

133_128_e3_3e-5/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5d5569bc557ef600417ef96a2d7a06af88f13671fdf9083bd31417271e9db8b4
+size 791751704

133_128_e3_3e-5/latest

@@ -0,0 +1 @@
+global_step1707

133_128_e3_3e-5/rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:521ce323a8dac4748e809d14d37389a96721baf84cc34c874caad05cc1d43e0c
+size 16389

133_128_e3_3e-5/rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d2c1541a9474d03f1ad9b1cb85d0d937bc73e118bf7ddb681de015a74366034d
+size 16389

133_128_e3_3e-5/rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fa53a153e0ac08ced3ff338ca46b83c1bb1ac05db982a6839dfaedc4f8352dd3
+size 16389

133_128_e3_3e-5/rng_state_3.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:804e7c0472c810eb0f07a2eb98c995e8d1fe7aff8a6796665cd4b761f27587bb
+size 16389

133_128_e3_3e-5/rng_state_4.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:65348a6e9709a9b9e9f258b844cbebf2e5e2aa73bf9617553bd8d572fd0eaf75
+size 16389

133_128_e3_3e-5/rng_state_5.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5f5b7ed194b7966d26a5decdfd7686a4d4744e4bb7ae926a70ae84e4c809a123
+size 16389

133_128_e3_3e-5/rng_state_6.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4d9bdbae06ef9eb7be8ad89ca9bfec734d4e054c6f1c5e6eb8f5c5ff57688d34
+size 16389

133_128_e3_3e-5/rng_state_7.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9556e38dab56bafa49cdc374b19641c85659fa00795b8987b39104e9ead0164d
+size 16389

133_128_e3_3e-5/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cf2cfc5400b08077594747910498ba6b9203a5c938c0873e5ac4fa0f9805248b
+size 1401

133_128_e3_3e-5/special_tokens_map.json

@@ -0,0 +1,45 @@
+{
+  "additional_special_tokens": [
+    "<|endoftext|>",
+    "<fim_prefix>",
+    "<fim_middle>",
+    "<fim_suffix>",
+    "<fim_pad>",
+    "<filename>",
+    "<gh_stars>",
+    "<issue_start>",
+    "<issue_comment>",
+    "<issue_closed>",
+    "<jupyter_start>",
+    "<jupyter_text>",
+    "<jupyter_code>",
+    "<jupyter_output>",
+    "<empty_output>",
+    "<commit_before>",
+    "<commit_msg>",
+    "<commit_after>",
+    "<reponame>"
+  ],
+  "bos_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "eos_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": "<reponame>",
+  "unk_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}

133_128_e3_3e-5/tokenizer_config.json

@@ -0,0 +1,188 @@
+{
+  "add_prefix_space": false,
+  "added_tokens_decoder": {
+    "0": {
+      "content": "<|endoftext|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "1": {
+      "content": "<fim_prefix>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "2": {
+      "content": "<fim_middle>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "3": {
+      "content": "<fim_suffix>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "4": {
+      "content": "<fim_pad>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "5": {
+      "content": "<filename>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "6": {
+      "content": "<gh_stars>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "7": {
+      "content": "<issue_start>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "8": {
+      "content": "<issue_comment>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "9": {
+      "content": "<issue_closed>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "10": {
+      "content": "<jupyter_start>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "11": {
+      "content": "<jupyter_text>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "12": {
+      "content": "<jupyter_code>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "13": {
+      "content": "<jupyter_output>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "14": {
+      "content": "<empty_output>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "15": {
+      "content": "<commit_before>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "16": {
+      "content": "<commit_msg>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "17": {
+      "content": "<commit_after>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "18": {
+      "content": "<reponame>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    }
+  },
+  "additional_special_tokens": [
+    "<|endoftext|>",
+    "<fim_prefix>",
+    "<fim_middle>",
+    "<fim_suffix>",
+    "<fim_pad>",
+    "<filename>",
+    "<gh_stars>",
+    "<issue_start>",
+    "<issue_comment>",
+    "<issue_closed>",
+    "<jupyter_start>",
+    "<jupyter_text>",
+    "<jupyter_code>",
+    "<jupyter_output>",
+    "<empty_output>",
+    "<commit_before>",
+    "<commit_msg>",
+    "<commit_after>",
+    "<reponame>"
+  ],
+  "bos_token": "<|endoftext|>",
+  "clean_up_tokenization_spaces": true,
+  "eos_token": "<|endoftext|>",
+  "extra_special_tokens": {},
+  "model_max_length": 8192,
+  "pad_token": "<reponame>",
+  "padding_side": "left",
+  "tokenizer_class": "GPT2Tokenizer",
+  "unk_token": "<|endoftext|>",
+  "vocab_size": 49152
+}

133_128_e3_3e-5/trainer_state.json

@@ -0,0 +1,2421 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.0,
+  "eval_steps": 500,
+  "global_step": 1707,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.008795074758135445,
+      "grad_norm": 0.973562479019165,
+      "learning_rate": 1.3953488372093023e-06,
+      "loss": 1.4576,
+      "step": 5
+    },
+    {
+      "epoch": 0.01759014951627089,
+      "grad_norm": 0.835927426815033,
+      "learning_rate": 3.1395348837209307e-06,
+      "loss": 1.4288,
+      "step": 10
+    },
+    {
+      "epoch": 0.026385224274406333,
+      "grad_norm": 0.7772343158721924,
+      "learning_rate": 4.883720930232559e-06,
+      "loss": 1.4353,
+      "step": 15
+    },
+    {
+      "epoch": 0.03518029903254178,
+      "grad_norm": 0.5898579955101013,
+      "learning_rate": 6.627906976744187e-06,
+      "loss": 1.4051,
+      "step": 20
+    },
+    {
+      "epoch": 0.04397537379067722,
+      "grad_norm": 0.48407605290412903,
+      "learning_rate": 8.372093023255815e-06,
+      "loss": 1.3949,
+      "step": 25
+    },
+    {
+      "epoch": 0.052770448548812667,
+      "grad_norm": 0.509476363658905,
+      "learning_rate": 1.0116279069767442e-05,
+      "loss": 1.3775,
+      "step": 30
+    },
+    {
+      "epoch": 0.06156552330694811,
+      "grad_norm": 0.5475961565971375,
+      "learning_rate": 1.1860465116279069e-05,
+      "loss": 1.3501,
+      "step": 35
+    },
+    {
+      "epoch": 0.07036059806508356,
+      "grad_norm": 0.515484094619751,
+      "learning_rate": 1.3604651162790698e-05,
+      "loss": 1.4133,
+      "step": 40
+    },
+    {
+      "epoch": 0.079155672823219,
+      "grad_norm": 0.5684940814971924,
+      "learning_rate": 1.5348837209302328e-05,
+      "loss": 1.2915,
+      "step": 45
+    },
+    {
+      "epoch": 0.08795074758135445,
+      "grad_norm": 0.508490264415741,
+      "learning_rate": 1.7093023255813955e-05,
+      "loss": 1.3068,
+      "step": 50
+    },
+    {
+      "epoch": 0.09674582233948989,
+      "grad_norm": 0.606148898601532,
+      "learning_rate": 1.8837209302325582e-05,
+      "loss": 1.2869,
+      "step": 55
+    },
+    {
+      "epoch": 0.10554089709762533,
+      "grad_norm": 0.700124204158783,
+      "learning_rate": 2.058139534883721e-05,
+      "loss": 1.3159,
+      "step": 60
+    },
+    {
+      "epoch": 0.11433597185576078,
+      "grad_norm": 0.6885865926742554,
+      "learning_rate": 2.2325581395348837e-05,
+      "loss": 1.2799,
+      "step": 65
+    },
+    {
+      "epoch": 0.12313104661389622,
+      "grad_norm": 0.5658232569694519,
+      "learning_rate": 2.4069767441860464e-05,
+      "loss": 1.2554,
+      "step": 70
+    },
+    {
+      "epoch": 0.13192612137203166,
+      "grad_norm": 0.5208723545074463,
+      "learning_rate": 2.5813953488372094e-05,
+      "loss": 1.3309,
+      "step": 75
+    },
+    {
+      "epoch": 0.14072119613016712,
+      "grad_norm": 0.5213214755058289,
+      "learning_rate": 2.755813953488372e-05,
+      "loss": 1.2778,
+      "step": 80
+    },
+    {
+      "epoch": 0.14951627088830255,
+      "grad_norm": 0.5797815322875977,
+      "learning_rate": 2.930232558139535e-05,
+      "loss": 1.2451,
+      "step": 85
+    },
+    {
+      "epoch": 0.158311345646438,
+      "grad_norm": 0.5916420221328735,
+      "learning_rate": 2.9999746465966867e-05,
+      "loss": 1.248,
+      "step": 90
+    },
+    {
+      "epoch": 0.16710642040457344,
+      "grad_norm": 0.565865695476532,
+      "learning_rate": 2.999819712235739e-05,
+      "loss": 1.2018,
+      "step": 95
+    },
+    {
+      "epoch": 0.1759014951627089,
+      "grad_norm": 0.6843475103378296,
+      "learning_rate": 2.9995239432687672e-05,
+      "loss": 1.26,
+      "step": 100
+    },
+    {
+      "epoch": 0.18469656992084432,
+      "grad_norm": 0.7608978748321533,
+      "learning_rate": 2.999087367468779e-05,
+      "loss": 1.2149,
+      "step": 105
+    },
+    {
+      "epoch": 0.19349164467897978,
+      "grad_norm": 0.7299783229827881,
+      "learning_rate": 2.9985100258306897e-05,
+      "loss": 1.1746,
+      "step": 110
+    },
+    {
+      "epoch": 0.2022867194371152,
+      "grad_norm": 0.6360028982162476,
+      "learning_rate": 2.9977919725674697e-05,
+      "loss": 1.1463,
+      "step": 115
+    },
+    {
+      "epoch": 0.21108179419525067,
+      "grad_norm": 0.749293327331543,
+      "learning_rate": 2.996933275105055e-05,
+      "loss": 1.1934,
+      "step": 120
+    },
+    {
+      "epoch": 0.2198768689533861,
+      "grad_norm": 0.7878008484840393,
+      "learning_rate": 2.995934014076017e-05,
+      "loss": 1.143,
+      "step": 125
+    },
+    {
+      "epoch": 0.22867194371152155,
+      "grad_norm": 0.6791767477989197,
+      "learning_rate": 2.9947942833119882e-05,
+      "loss": 1.1571,
+      "step": 130
+    },
+    {
+      "epoch": 0.23746701846965698,
+      "grad_norm": 0.6570654511451721,
+      "learning_rate": 2.993514189834855e-05,
+      "loss": 1.0513,
+      "step": 135
+    },
+    {
+      "epoch": 0.24626209322779244,
+      "grad_norm": 0.7771459221839905,
+      "learning_rate": 2.992093853846704e-05,
+      "loss": 1.0587,
+      "step": 140
+    },
+    {
+      "epoch": 0.25505716798592787,
+      "grad_norm": 0.7587772607803345,
+      "learning_rate": 2.9905334087185382e-05,
+      "loss": 1.0841,
+      "step": 145
+    },
+    {
+      "epoch": 0.2638522427440633,
+      "grad_norm": 0.8423944711685181,
+      "learning_rate": 2.9888330009777523e-05,
+      "loss": 1.0439,
+      "step": 150
+    },
+    {
+      "epoch": 0.2726473175021988,
+      "grad_norm": 0.9509164094924927,
+      "learning_rate": 2.9869927902943743e-05,
+      "loss": 1.0285,
+      "step": 155
+    },
+    {
+      "epoch": 0.28144239226033424,
+      "grad_norm": 0.8729593753814697,
+      "learning_rate": 2.9850129494660702e-05,
+      "loss": 1.0192,
+      "step": 160
+    },
+    {
+      "epoch": 0.29023746701846964,
+      "grad_norm": 0.9015172123908997,
+      "learning_rate": 2.9828936644019197e-05,
+      "loss": 1.0017,
+      "step": 165
+    },
+    {
+      "epoch": 0.2990325417766051,
+      "grad_norm": 0.9222761988639832,
+      "learning_rate": 2.9806351341049602e-05,
+      "loss": 0.9988,
+      "step": 170
+    },
+    {
+      "epoch": 0.30782761653474056,
+      "grad_norm": 0.8090435266494751,
+      "learning_rate": 2.9782375706534982e-05,
+      "loss": 0.969,
+      "step": 175
+    },
+    {
+      "epoch": 0.316622691292876,
+      "grad_norm": 0.8706099987030029,
+      "learning_rate": 2.9757011991811945e-05,
+      "loss": 0.9655,
+      "step": 180
+    },
+    {
+      "epoch": 0.3254177660510114,
+      "grad_norm": 1.0139667987823486,
+      "learning_rate": 2.9730262578559273e-05,
+      "loss": 0.9748,
+      "step": 185
+    },
+    {
+      "epoch": 0.33421284080914687,
+      "grad_norm": 1.0537158250808716,
+      "learning_rate": 2.9702129978574248e-05,
+      "loss": 0.9515,
+      "step": 190
+    },
+    {
+      "epoch": 0.34300791556728233,
+      "grad_norm": 0.9278015494346619,
+      "learning_rate": 2.96726168335368e-05,
+      "loss": 0.9355,
+      "step": 195
+    },
+    {
+      "epoch": 0.3518029903254178,
+      "grad_norm": 0.8674100041389465,
+      "learning_rate": 2.9641725914761447e-05,
+      "loss": 0.9109,
+      "step": 200
+    },
+    {
+      "epoch": 0.3605980650835532,
+      "grad_norm": 1.1447285413742065,
+      "learning_rate": 2.960946012293709e-05,
+      "loss": 0.9148,
+      "step": 205
+    },
+    {
+      "epoch": 0.36939313984168864,
+      "grad_norm": 1.1224510669708252,
+      "learning_rate": 2.9575822487854602e-05,
+      "loss": 0.8869,
+      "step": 210
+    },
+    {
+      "epoch": 0.3781882145998241,
+      "grad_norm": 1.1465867757797241,
+      "learning_rate": 2.954081616812234e-05,
+      "loss": 0.9277,
+      "step": 215
+    },
+    {
+      "epoch": 0.38698328935795956,
+      "grad_norm": 0.9554082751274109,
+      "learning_rate": 2.950444445086956e-05,
+      "loss": 0.8384,
+      "step": 220
+    },
+    {
+      "epoch": 0.39577836411609496,
+      "grad_norm": 1.095732569694519,
+      "learning_rate": 2.9466710751437738e-05,
+      "loss": 0.8707,
+      "step": 225
+    },
+    {
+      "epoch": 0.4045734388742304,
+      "grad_norm": 1.1254558563232422,
+      "learning_rate": 2.9427618613059867e-05,
+      "loss": 0.8421,
+      "step": 230
+    },
+    {
+      "epoch": 0.4133685136323659,
+      "grad_norm": 1.1809462308883667,
+      "learning_rate": 2.938717170652775e-05,
+      "loss": 0.8706,
+      "step": 235
+    },
+    {
+      "epoch": 0.42216358839050133,
+      "grad_norm": 1.0442577600479126,
+      "learning_rate": 2.9345373829847318e-05,
+      "loss": 0.8578,
+      "step": 240
+    },
+    {
+      "epoch": 0.4309586631486368,
+      "grad_norm": 1.1527485847473145,
+      "learning_rate": 2.9302228907881956e-05,
+      "loss": 0.8321,
+      "step": 245
+    },
+    {
+      "epoch": 0.4397537379067722,
+      "grad_norm": 1.1622178554534912,
+      "learning_rate": 2.9257740991984007e-05,
+      "loss": 0.835,
+      "step": 250
+    },
+    {
+      "epoch": 0.44854881266490765,
+      "grad_norm": 1.1275705099105835,
+      "learning_rate": 2.92119142596143e-05,
+      "loss": 0.8301,
+      "step": 255
+    },
+    {
+      "epoch": 0.4573438874230431,
+      "grad_norm": 1.075764536857605,
+      "learning_rate": 2.9164753013949908e-05,
+      "loss": 0.8736,
+      "step": 260
+    },
+    {
+      "epoch": 0.46613896218117856,
+      "grad_norm": 1.220468521118164,
+      "learning_rate": 2.911626168348007e-05,
+      "loss": 0.8054,
+      "step": 265
+    },
+    {
+      "epoch": 0.47493403693931396,
+      "grad_norm": 1.2002390623092651,
+      "learning_rate": 2.9066444821590345e-05,
+      "loss": 0.8284,
+      "step": 270
+    },
+    {
+      "epoch": 0.4837291116974494,
+      "grad_norm": 1.1792253255844116,
+      "learning_rate": 2.9015307106135053e-05,
+      "loss": 0.8127,
+      "step": 275
+    },
+    {
+      "epoch": 0.4925241864555849,
+      "grad_norm": 1.1782032251358032,
+      "learning_rate": 2.896285333899802e-05,
+      "loss": 0.7998,
+      "step": 280
+    },
+    {
+      "epoch": 0.5013192612137203,
+      "grad_norm": 1.1237090826034546,
+      "learning_rate": 2.8909088445641653e-05,
+      "loss": 0.7966,
+      "step": 285
+    },
+    {
+      "epoch": 0.5101143359718557,
+      "grad_norm": 1.3340306282043457,
+      "learning_rate": 2.885401747464447e-05,
+      "loss": 0.7898,
+      "step": 290
+    },
+    {
+      "epoch": 0.5189094107299912,
+      "grad_norm": 1.1375333070755005,
+      "learning_rate": 2.8797645597227017e-05,
+      "loss": 0.7156,
+      "step": 295
+    },
+    {
+      "epoch": 0.5277044854881267,
+      "grad_norm": 1.519955039024353,
+      "learning_rate": 2.873997810676628e-05,
+      "loss": 0.7534,
+      "step": 300
+    },
+    {
+      "epoch": 0.5364995602462621,
+      "grad_norm": 1.2972135543823242,
+      "learning_rate": 2.8681020418298632e-05,
+      "loss": 0.6989,
+      "step": 305
+    },
+    {
+      "epoch": 0.5452946350043976,
+      "grad_norm": 1.4623491764068604,
+      "learning_rate": 2.862077806801136e-05,
+      "loss": 0.7611,
+      "step": 310
+    },
+    {
+      "epoch": 0.554089709762533,
+      "grad_norm": 1.15494966506958,
+      "learning_rate": 2.8559256712722818e-05,
+      "loss": 0.7484,
+      "step": 315
+    },
+    {
+      "epoch": 0.5628847845206685,
+      "grad_norm": 1.213477611541748,
+      "learning_rate": 2.8496462129351225e-05,
+      "loss": 0.6869,
+      "step": 320
+    },
+    {
+      "epoch": 0.5716798592788038,
+      "grad_norm": 1.1656783819198608,
+      "learning_rate": 2.8432400214372244e-05,
+      "loss": 0.738,
+      "step": 325
+    },
+    {
+      "epoch": 0.5804749340369393,
+      "grad_norm": 1.1367297172546387,
+      "learning_rate": 2.8367076983265247e-05,
+      "loss": 0.6771,
+      "step": 330
+    },
+    {
+      "epoch": 0.5892700087950747,
+      "grad_norm": 1.3425236940383911,
+      "learning_rate": 2.8300498569948493e-05,
+      "loss": 0.6607,
+      "step": 335
+    },
+    {
+      "epoch": 0.5980650835532102,
+      "grad_norm": 1.2506884336471558,
+      "learning_rate": 2.8232671226203143e-05,
+      "loss": 0.7093,
+      "step": 340
+    },
+    {
+      "epoch": 0.6068601583113457,
+      "grad_norm": 1.2948859930038452,
+      "learning_rate": 2.816360132108618e-05,
+      "loss": 0.7,
+      "step": 345
+    },
+    {
+      "epoch": 0.6156552330694811,
+      "grad_norm": 1.3027925491333008,
+      "learning_rate": 2.80932953403324e-05,
+      "loss": 0.6803,
+      "step": 350
+    },
+    {
+      "epoch": 0.6244503078276166,
+      "grad_norm": 1.1062813997268677,
+      "learning_rate": 2.802175988574535e-05,
+      "loss": 0.658,
+      "step": 355
+    },
+    {
+      "epoch": 0.633245382585752,
+      "grad_norm": 1.2913652658462524,
+      "learning_rate": 2.7949001674577424e-05,
+      "loss": 0.7403,
+      "step": 360
+    },
+    {
+      "epoch": 0.6420404573438874,
+      "grad_norm": 1.1860114336013794,
+      "learning_rate": 2.787502753889913e-05,
+      "loss": 0.6092,
+      "step": 365
+    },
+    {
+      "epoch": 0.6508355321020228,
+      "grad_norm": 1.1381458044052124,
+      "learning_rate": 2.779984442495751e-05,
+      "loss": 0.6458,
+      "step": 370
+    },
+    {
+      "epoch": 0.6596306068601583,
+      "grad_norm": 1.31987726688385,
+      "learning_rate": 2.7723459392523912e-05,
+      "loss": 0.6426,
+      "step": 375
+    },
+    {
+      "epoch": 0.6684256816182937,
+      "grad_norm": 1.1997981071472168,
+      "learning_rate": 2.764587961423107e-05,
+      "loss": 0.6654,
+      "step": 380
+    },
+    {
+      "epoch": 0.6772207563764292,
+      "grad_norm": 1.2875627279281616,
+      "learning_rate": 2.7567112374899563e-05,
+      "loss": 0.6251,
+      "step": 385
+    },
+    {
+      "epoch": 0.6860158311345647,
+      "grad_norm": 1.2959027290344238,
+      "learning_rate": 2.748716507085377e-05,
+      "loss": 0.5587,
+      "step": 390
+    },
+    {
+      "epoch": 0.6948109058927001,
+      "grad_norm": 1.3617380857467651,
+      "learning_rate": 2.740604520922738e-05,
+      "loss": 0.6266,
+      "step": 395
+    },
+    {
+      "epoch": 0.7036059806508356,
+      "grad_norm": 1.2740793228149414,
+      "learning_rate": 2.7323760407258426e-05,
+      "loss": 0.642,
+      "step": 400
+    },
+    {
+      "epoch": 0.712401055408971,
+      "grad_norm": 1.1810040473937988,
+      "learning_rate": 2.724031839157402e-05,
+      "loss": 0.5944,
+      "step": 405
+    },
+    {
+      "epoch": 0.7211961301671064,
+      "grad_norm": 1.4832911491394043,
+      "learning_rate": 2.7155726997464843e-05,
+      "loss": 0.5935,
+      "step": 410
+    },
+    {
+      "epoch": 0.7299912049252418,
+      "grad_norm": 1.1355174779891968,
+      "learning_rate": 2.706999416814938e-05,
+      "loss": 0.5707,
+      "step": 415
+    },
+    {
+      "epoch": 0.7387862796833773,
+      "grad_norm": 1.3322994709014893,
+      "learning_rate": 2.698312795402804e-05,
+      "loss": 0.6021,
+      "step": 420
+    },
+    {
+      "epoch": 0.7475813544415127,
+      "grad_norm": 1.3649718761444092,
+      "learning_rate": 2.6895136511927253e-05,
+      "loss": 0.569,
+      "step": 425
+    },
+    {
+      "epoch": 0.7563764291996482,
+      "grad_norm": 1.2992080450057983,
+      "learning_rate": 2.680602810433348e-05,
+      "loss": 0.6118,
+      "step": 430
+    },
+    {
+      "epoch": 0.7651715039577837,
+      "grad_norm": 1.3155995607376099,
+      "learning_rate": 2.6715811098617392e-05,
+      "loss": 0.5819,
+      "step": 435
+    },
+    {
+      "epoch": 0.7739665787159191,
+      "grad_norm": 1.3068788051605225,
+      "learning_rate": 2.6624493966248172e-05,
+      "loss": 0.5512,
+      "step": 440
+    },
+    {
+      "epoch": 0.7827616534740546,
+      "grad_norm": 1.310539722442627,
+      "learning_rate": 2.6532085281998002e-05,
+      "loss": 0.5719,
+      "step": 445
+    },
+    {
+      "epoch": 0.7915567282321899,
+      "grad_norm": 1.2950756549835205,
+      "learning_rate": 2.6438593723136922e-05,
+      "loss": 0.5356,
+      "step": 450
+    },
+    {
+      "epoch": 0.8003518029903254,
+      "grad_norm": 1.4679863452911377,
+      "learning_rate": 2.6344028068617984e-05,
+      "loss": 0.5435,
+      "step": 455
+    },
+    {
+      "epoch": 0.8091468777484608,
+      "grad_norm": 1.267077088356018,
+      "learning_rate": 2.6248397198252943e-05,
+      "loss": 0.5457,
+      "step": 460
+    },
+    {
+      "epoch": 0.8179419525065963,
+      "grad_norm": 1.4413871765136719,
+      "learning_rate": 2.6151710091878395e-05,
+      "loss": 0.5163,
+      "step": 465
+    },
+    {
+      "epoch": 0.8267370272647317,
+      "grad_norm": 1.3673219680786133,
+      "learning_rate": 2.60539758285126e-05,
+      "loss": 0.5439,
+      "step": 470
+    },
+    {
+      "epoch": 0.8355321020228672,
+      "grad_norm": 1.4192862510681152,
+      "learning_rate": 2.5955203585502902e-05,
+      "loss": 0.5286,
+      "step": 475
+    },
+    {
+      "epoch": 0.8443271767810027,
+      "grad_norm": 1.2355852127075195,
+      "learning_rate": 2.5855402637664026e-05,
+      "loss": 0.5117,
+      "step": 480
+    },
+    {
+      "epoch": 0.8531222515391381,
+      "grad_norm": 1.3679404258728027,
+      "learning_rate": 2.575458235640711e-05,
+      "loss": 0.4703,
+      "step": 485
+    },
+    {
+      "epoch": 0.8619173262972736,
+      "grad_norm": 1.2575867176055908,
+      "learning_rate": 2.5652752208859753e-05,
+      "loss": 0.5181,
+      "step": 490
+    },
+    {
+      "epoch": 0.8707124010554089,
+      "grad_norm": 1.5026286840438843,
+      "learning_rate": 2.5549921756977036e-05,
+      "loss": 0.5164,
+      "step": 495
+    },
+    {
+      "epoch": 0.8795074758135444,
+      "grad_norm": 1.680506706237793,
+      "learning_rate": 2.5446100656643638e-05,
+      "loss": 0.4697,
+      "step": 500
+    },
+    {
+      "epoch": 0.8883025505716798,
+      "grad_norm": 1.3909192085266113,
+      "learning_rate": 2.5341298656767123e-05,
+      "loss": 0.4966,
+      "step": 505
+    },
+    {
+      "epoch": 0.8970976253298153,
+      "grad_norm": 1.2556389570236206,
+      "learning_rate": 2.5235525598362564e-05,
+      "loss": 0.4687,
+      "step": 510
+    },
+    {
+      "epoch": 0.9058927000879508,
+      "grad_norm": 1.3569135665893555,
+      "learning_rate": 2.5128791413628395e-05,
+      "loss": 0.4793,
+      "step": 515
+    },
+    {
+      "epoch": 0.9146877748460862,
+      "grad_norm": 1.348023533821106,
+      "learning_rate": 2.50211061250138e-05,
+      "loss": 0.46,
+      "step": 520
+    },
+    {
+      "epoch": 0.9234828496042217,
+      "grad_norm": 1.341158151626587,
+      "learning_rate": 2.4912479844277595e-05,
+      "loss": 0.4795,
+      "step": 525
+    },
+    {
+      "epoch": 0.9322779243623571,
+      "grad_norm": 1.45597243309021,
+      "learning_rate": 2.480292277153872e-05,
+      "loss": 0.4606,
+      "step": 530
+    },
+    {
+      "epoch": 0.9410729991204925,
+      "grad_norm": 1.3844624757766724,
+      "learning_rate": 2.4692445194318442e-05,
+      "loss": 0.434,
+      "step": 535
+    },
+    {
+      "epoch": 0.9498680738786279,
+      "grad_norm": 1.2690527439117432,
+      "learning_rate": 2.4581057486574322e-05,
+      "loss": 0.4361,
+      "step": 540
+    },
+    {
+      "epoch": 0.9586631486367634,
+      "grad_norm": 1.4192856550216675,
+      "learning_rate": 2.446877010772613e-05,
+      "loss": 0.4211,
+      "step": 545
+    },
+    {
+      "epoch": 0.9674582233948988,
+      "grad_norm": 1.5629619359970093,
+      "learning_rate": 2.435559360167366e-05,
+      "loss": 0.4429,
+      "step": 550
+    },
+    {
+      "epoch": 0.9762532981530343,
+      "grad_norm": 1.3519330024719238,
+      "learning_rate": 2.424153859580667e-05,
+      "loss": 0.4404,
+      "step": 555
+    },
+    {
+      "epoch": 0.9850483729111698,
+      "grad_norm": 1.4851571321487427,
+      "learning_rate": 2.412661580000694e-05,
+      "loss": 0.443,
+      "step": 560
+    },
+    {
+      "epoch": 0.9938434476693052,
+      "grad_norm": 1.4712505340576172,
+      "learning_rate": 2.4010836005642617e-05,
+      "loss": 0.4359,
+      "step": 565
+    },
+    {
+      "epoch": 1.0017590149516271,
+      "grad_norm": 1.2725117206573486,
+      "learning_rate": 2.3894210084554894e-05,
+      "loss": 0.4127,
+      "step": 570
+    },
+    {
+      "epoch": 1.0105540897097625,
+      "grad_norm": 1.312313437461853,
+      "learning_rate": 2.3776748988037126e-05,
+      "loss": 0.4013,
+      "step": 575
+    },
+    {
+      "epoch": 1.019349164467898,
+      "grad_norm": 1.361315369606018,
+      "learning_rate": 2.3658463745806495e-05,
+      "loss": 0.3824,
+      "step": 580
+    },
+    {
+      "epoch": 1.0281442392260334,
+      "grad_norm": 1.4416288137435913,
+      "learning_rate": 2.353936546496831e-05,
+      "loss": 0.3521,
+      "step": 585
+    },
+    {
+      "epoch": 1.036939313984169,
+      "grad_norm": 1.256807565689087,
+      "learning_rate": 2.3419465328973034e-05,
+      "loss": 0.3328,
+      "step": 590
+    },
+    {
+      "epoch": 1.0457343887423043,
+      "grad_norm": 1.2563031911849976,
+      "learning_rate": 2.329877459656616e-05,
+      "loss": 0.3771,
+      "step": 595
+    },
+    {
+      "epoch": 1.0545294635004399,
+      "grad_norm": 1.5360652208328247,
+      "learning_rate": 2.317730460073098e-05,
+      "loss": 0.3398,
+      "step": 600
+    },
+    {
+      "epoch": 1.0633245382585752,
+      "grad_norm": 1.3718454837799072,
+      "learning_rate": 2.3055066747624428e-05,
+      "loss": 0.3047,
+      "step": 605
+    },
+    {
+      "epoch": 1.0721196130167105,
+      "grad_norm": 1.2570449113845825,
+      "learning_rate": 2.293207251550602e-05,
+      "loss": 0.3122,
+      "step": 610
+    },
+    {
+      "epoch": 1.0809146877748461,
+      "grad_norm": 1.4068742990493774,
+      "learning_rate": 2.2808333453660035e-05,
+      "loss": 0.37,
+      "step": 615
+    },
+    {
+      "epoch": 1.0897097625329815,
+      "grad_norm": 1.552269697189331,
+      "learning_rate": 2.268386118131103e-05,
+      "loss": 0.3197,
+      "step": 620
+    },
+    {
+      "epoch": 1.098504837291117,
+      "grad_norm": 1.4864991903305054,
+      "learning_rate": 2.2558667386532767e-05,
+      "loss": 0.3324,
+      "step": 625
+    },
+    {
+      "epoch": 1.1072999120492524,
+      "grad_norm": 1.5119129419326782,
+      "learning_rate": 2.243276382515071e-05,
+      "loss": 0.3479,
+      "step": 630
+    },
+    {
+      "epoch": 1.116094986807388,
+      "grad_norm": 1.3214279413223267,
+      "learning_rate": 2.230616231963813e-05,
+      "loss": 0.3102,
+      "step": 635
+    },
+    {
+      "epoch": 1.1248900615655233,
+      "grad_norm": 1.4211530685424805,
+      "learning_rate": 2.2178874758005987e-05,
+      "loss": 0.3316,
+      "step": 640
+    },
+    {
+      "epoch": 1.1336851363236589,
+      "grad_norm": 1.2067914009094238,
+      "learning_rate": 2.2050913092686585e-05,
+      "loss": 0.3403,
+      "step": 645
+    },
+    {
+      "epoch": 1.1424802110817942,
+      "grad_norm": 1.2802555561065674,
+      "learning_rate": 2.1922289339411254e-05,
+      "loss": 0.3259,
+      "step": 650
+    },
+    {
+      "epoch": 1.1512752858399296,
+      "grad_norm": 1.5187349319458008,
+      "learning_rate": 2.1793015576082087e-05,
+      "loss": 0.3397,
+      "step": 655
+    },
+    {
+      "epoch": 1.1600703605980651,
+      "grad_norm": 1.4011081457138062,
+      "learning_rate": 2.1663103941637763e-05,
+      "loss": 0.3003,
+      "step": 660
+    },
+    {
+      "epoch": 1.1688654353562005,
+      "grad_norm": 1.4166609048843384,
+      "learning_rate": 2.1532566634913717e-05,
+      "loss": 0.3439,
+      "step": 665
+    },
+    {
+      "epoch": 1.177660510114336,
+      "grad_norm": 1.5283876657485962,
+      "learning_rate": 2.1401415913496667e-05,
+      "loss": 0.3185,
+      "step": 670
+    },
+    {
+      "epoch": 1.1864555848724714,
+      "grad_norm": 1.4254592657089233,
+      "learning_rate": 2.1269664092573568e-05,
+      "loss": 0.3144,
+      "step": 675
+    },
+    {
+      "epoch": 1.195250659630607,
+      "grad_norm": 1.398084044456482,
+      "learning_rate": 2.113732354377526e-05,
+      "loss": 0.3077,
+      "step": 680
+    },
+    {
+      "epoch": 1.2040457343887423,
+      "grad_norm": 1.5452698469161987,
+      "learning_rate": 2.1004406694014722e-05,
+      "loss": 0.2995,
+      "step": 685
+    },
+    {
+      "epoch": 1.2128408091468779,
+      "grad_norm": 1.2492598295211792,
+      "learning_rate": 2.087092602432018e-05,
+      "loss": 0.3046,
+      "step": 690
+    },
+    {
+      "epoch": 1.2216358839050132,
+      "grad_norm": 1.3515474796295166,
+      "learning_rate": 2.0736894068663145e-05,
+      "loss": 0.3353,
+      "step": 695
+    },
+    {
+      "epoch": 1.2304309586631486,
+      "grad_norm": 1.3467352390289307,
+      "learning_rate": 2.060232341278143e-05,
+      "loss": 0.297,
+      "step": 700
+    },
+    {
+      "epoch": 1.2392260334212841,
+      "grad_norm": 1.244985818862915,
+      "learning_rate": 2.0467226692997347e-05,
+      "loss": 0.3307,
+      "step": 705
+    },
+    {
+      "epoch": 1.2480211081794195,
+      "grad_norm": 1.4247772693634033,
+      "learning_rate": 2.0331616595031156e-05,
+      "loss": 0.3101,
+      "step": 710
+    },
+    {
+      "epoch": 1.256816182937555,
+      "grad_norm": 1.4809576272964478,
+      "learning_rate": 2.0195505852809855e-05,
+      "loss": 0.3083,
+      "step": 715
+    },
+    {
+      "epoch": 1.2656112576956904,
+      "grad_norm": 1.3201909065246582,
+      "learning_rate": 2.0058907247271446e-05,
+      "loss": 0.2825,
+      "step": 720
+    },
+    {
+      "epoch": 1.2744063324538257,
+      "grad_norm": 1.5269622802734375,
+      "learning_rate": 1.9921833605164786e-05,
+      "loss": 0.2972,
+      "step": 725
+    },
+    {
+      "epoch": 1.2832014072119613,
+      "grad_norm": 1.3980199098587036,
+      "learning_rate": 1.9784297797845176e-05,
+      "loss": 0.2849,
+      "step": 730
+    },
+    {
+      "epoch": 1.2919964819700969,
+      "grad_norm": 1.453404426574707,
+      "learning_rate": 1.9646312740065684e-05,
+      "loss": 0.2912,
+      "step": 735
+    },
+    {
+      "epoch": 1.3007915567282322,
+      "grad_norm": 1.3130892515182495,
+      "learning_rate": 1.9507891388764452e-05,
+      "loss": 0.2837,
+      "step": 740
+    },
+    {
+      "epoch": 1.3095866314863676,
+      "grad_norm": 1.3477885723114014,
+      "learning_rate": 1.9369046741848053e-05,
+      "loss": 0.2595,
+      "step": 745
+    },
+    {
+      "epoch": 1.3183817062445031,
+      "grad_norm": 1.5332351922988892,
+      "learning_rate": 1.9229791836970937e-05,
+      "loss": 0.2665,
+      "step": 750
+    },
+    {
+      "epoch": 1.3271767810026385,
+      "grad_norm": 1.2507930994033813,
+      "learning_rate": 1.9090139750311185e-05,
+      "loss": 0.2501,
+      "step": 755
+    },
+    {
+      "epoch": 1.335971855760774,
+      "grad_norm": 1.3676016330718994,
+      "learning_rate": 1.8950103595342685e-05,
+      "loss": 0.284,
+      "step": 760
+    },
+    {
+      "epoch": 1.3447669305189094,
+      "grad_norm": 1.3517746925354004,
+      "learning_rate": 1.8809696521603698e-05,
+      "loss": 0.2805,
+      "step": 765
+    },
+    {
+      "epoch": 1.3535620052770447,
+      "grad_norm": 1.2311815023422241,
+      "learning_rate": 1.866893171346216e-05,
+      "loss": 0.2509,
+      "step": 770
+    },
+    {
+      "epoch": 1.3623570800351803,
+      "grad_norm": 1.3325191736221313,
+      "learning_rate": 1.8527822388877623e-05,
+      "loss": 0.2842,
+      "step": 775
+    },
+    {
+      "epoch": 1.3711521547933159,
+      "grad_norm": 1.6356605291366577,
+      "learning_rate": 1.838638179816009e-05,
+      "loss": 0.2441,
+      "step": 780
+    },
+    {
+      "epoch": 1.3799472295514512,
+      "grad_norm": 1.4159555435180664,
+      "learning_rate": 1.8244623222725797e-05,
+      "loss": 0.2378,
+      "step": 785
+    },
+    {
+      "epoch": 1.3887423043095866,
+      "grad_norm": 1.4356229305267334,
+      "learning_rate": 1.810255997385006e-05,
+      "loss": 0.2552,
+      "step": 790
+    },
+    {
+      "epoch": 1.3975373790677221,
+      "grad_norm": 1.3625457286834717,
+      "learning_rate": 1.7960205391417345e-05,
+      "loss": 0.2456,
+      "step": 795
+    },
+    {
+      "epoch": 1.4063324538258575,
+      "grad_norm": 1.426663875579834,
+      "learning_rate": 1.7817572842668648e-05,
+      "loss": 0.239,
+      "step": 800
+    },
+    {
+      "epoch": 1.415127528583993,
+      "grad_norm": 1.3557077646255493,
+      "learning_rate": 1.7674675720946276e-05,
+      "loss": 0.2582,
+      "step": 805
+    },
+    {
+      "epoch": 1.4239226033421284,
+      "grad_norm": 1.3964656591415405,
+      "learning_rate": 1.7531527444436216e-05,
+      "loss": 0.252,
+      "step": 810
+    },
+    {
+      "epoch": 1.4327176781002637,
+      "grad_norm": 1.4174063205718994,
+      "learning_rate": 1.738814145490813e-05,
+      "loss": 0.2282,
+      "step": 815
+    },
+    {
+      "epoch": 1.4415127528583993,
+      "grad_norm": 1.3315582275390625,
+      "learning_rate": 1.7244531216453204e-05,
+      "loss": 0.2249,
+      "step": 820
+    },
+    {
+      "epoch": 1.4503078276165349,
+      "grad_norm": 1.5227763652801514,
+      "learning_rate": 1.7100710214219804e-05,
+      "loss": 0.2632,
+      "step": 825
+    },
+    {
+      "epoch": 1.4591029023746702,
+      "grad_norm": 1.3635790348052979,
+      "learning_rate": 1.695669195314723e-05,
+      "loss": 0.2244,
+      "step": 830
+    },
+    {
+      "epoch": 1.4678979771328056,
+      "grad_norm": 1.554488182067871,
+      "learning_rate": 1.681248995669762e-05,
+      "loss": 0.244,
+      "step": 835
+    },
+    {
+      "epoch": 1.4766930518909411,
+      "grad_norm": 1.3058902025222778,
+      "learning_rate": 1.6668117765586013e-05,
+      "loss": 0.223,
+      "step": 840
+    },
+    {
+      "epoch": 1.4854881266490765,
+      "grad_norm": 1.4751936197280884,
+      "learning_rate": 1.6523588936508916e-05,
+      "loss": 0.2213,
+      "step": 845
+    },
+    {
+      "epoch": 1.494283201407212,
+      "grad_norm": 1.254716157913208,
+      "learning_rate": 1.6378917040871315e-05,
+      "loss": 0.248,
+      "step": 850
+    },
+    {
+      "epoch": 1.5030782761653474,
+      "grad_norm": 1.335308313369751,
+      "learning_rate": 1.623411566351227e-05,
+      "loss": 0.2218,
+      "step": 855
+    },
+    {
+      "epoch": 1.5118733509234827,
+      "grad_norm": 1.3555564880371094,
+      "learning_rate": 1.608919840142932e-05,
+      "loss": 0.2325,
+      "step": 860
+    },
+    {
+      "epoch": 1.5206684256816183,
+      "grad_norm": 1.2356692552566528,
+      "learning_rate": 1.5944178862501692e-05,
+      "loss": 0.2095,
+      "step": 865
+    },
+    {
+      "epoch": 1.5294635004397539,
+      "grad_norm": 1.3401411771774292,
+      "learning_rate": 1.5799070664212528e-05,
+      "loss": 0.2201,
+      "step": 870
+    },
+    {
+      "epoch": 1.5382585751978892,
+      "grad_norm": 1.2337650060653687,
+      "learning_rate": 1.565388743237015e-05,
+      "loss": 0.2074,
+      "step": 875
+    },
+    {
+      "epoch": 1.5470536499560246,
+      "grad_norm": 1.2010562419891357,
+      "learning_rate": 1.550864279982863e-05,
+      "loss": 0.1964,
+      "step": 880
+    },
+    {
+      "epoch": 1.55584872471416,
+      "grad_norm": 1.2470057010650635,
+      "learning_rate": 1.5363350405207615e-05,
+      "loss": 0.2159,
+      "step": 885
+    },
+    {
+      "epoch": 1.5646437994722955,
+      "grad_norm": 1.2449960708618164,
+      "learning_rate": 1.5218023891611668e-05,
+      "loss": 0.194,
+      "step": 890
+    },
+    {
+      "epoch": 1.573438874230431,
+      "grad_norm": 1.3210610151290894,
+      "learning_rate": 1.5072676905349155e-05,
+      "loss": 0.1984,
+      "step": 895
+    },
+    {
+      "epoch": 1.5822339489885664,
+      "grad_norm": 1.4236935377120972,
+      "learning_rate": 1.492732309465085e-05,
+      "loss": 0.2121,
+      "step": 900
+    },
+    {
+      "epoch": 1.5910290237467017,
+      "grad_norm": 1.132291316986084,
+      "learning_rate": 1.478197610838833e-05,
+      "loss": 0.2056,
+      "step": 905
+    },
+    {
+      "epoch": 1.5998240985048373,
+      "grad_norm": 1.2488665580749512,
+      "learning_rate": 1.4636649594792386e-05,
+      "loss": 0.1905,
+      "step": 910
+    },
+    {
+      "epoch": 1.6086191732629729,
+      "grad_norm": 1.2800037860870361,
+      "learning_rate": 1.4491357200171374e-05,
+      "loss": 0.1806,
+      "step": 915
+    },
+    {
+      "epoch": 1.6174142480211082,
+      "grad_norm": 1.0714298486709595,
+      "learning_rate": 1.4346112567629849e-05,
+      "loss": 0.2006,
+      "step": 920
+    },
+    {
+      "epoch": 1.6262093227792436,
+      "grad_norm": 1.3651036024093628,
+      "learning_rate": 1.4200929335787475e-05,
+      "loss": 0.1821,
+      "step": 925
+    },
+    {
+      "epoch": 1.635004397537379,
+      "grad_norm": 1.2874627113342285,
+      "learning_rate": 1.4055821137498309e-05,
+      "loss": 0.1967,
+      "step": 930
+    },
+    {
+      "epoch": 1.6437994722955145,
+      "grad_norm": 1.3940821886062622,
+      "learning_rate": 1.391080159857068e-05,
+      "loss": 0.2013,
+      "step": 935
+    },
+    {
+      "epoch": 1.65259454705365,
+      "grad_norm": 1.3354371786117554,
+      "learning_rate": 1.3765884336487732e-05,
+      "loss": 0.1817,
+      "step": 940
+    },
+    {
+      "epoch": 1.6613896218117854,
+      "grad_norm": 1.3363789319992065,
+      "learning_rate": 1.362108295912869e-05,
+      "loss": 0.1868,
+      "step": 945
+    },
+    {
+      "epoch": 1.6701846965699207,
+      "grad_norm": 1.2558488845825195,
+      "learning_rate": 1.3476411063491081e-05,
+      "loss": 0.1705,
+      "step": 950
+    },
+    {
+      "epoch": 1.6789797713280563,
+      "grad_norm": 1.2083759307861328,
+      "learning_rate": 1.3331882234413989e-05,
+      "loss": 0.1748,
+      "step": 955
+    },
+    {
+      "epoch": 1.6877748460861919,
+      "grad_norm": 1.203922152519226,
+      "learning_rate": 1.3187510043302384e-05,
+      "loss": 0.1782,
+      "step": 960
+    },
+    {
+      "epoch": 1.6965699208443272,
+      "grad_norm": 1.237924337387085,
+      "learning_rate": 1.3043308046852765e-05,
+      "loss": 0.1934,
+      "step": 965
+    },
+    {
+      "epoch": 1.7053649956024626,
+      "grad_norm": 1.131487488746643,
+      "learning_rate": 1.28992897857802e-05,
+      "loss": 0.1623,
+      "step": 970
+    },
+    {
+      "epoch": 1.714160070360598,
+      "grad_norm": 1.179271936416626,
+      "learning_rate": 1.27554687835468e-05,
+      "loss": 0.1817,
+      "step": 975
+    },
+    {
+      "epoch": 1.7229551451187335,
+      "grad_norm": 1.1134639978408813,
+      "learning_rate": 1.261185854509187e-05,
+      "loss": 0.1678,
+      "step": 980
+    },
+    {
+      "epoch": 1.731750219876869,
+      "grad_norm": 1.1828957796096802,
+      "learning_rate": 1.2468472555563788e-05,
+      "loss": 0.1788,
+      "step": 985
+    },
+    {
+      "epoch": 1.7405452946350044,
+      "grad_norm": 1.1047483682632446,
+      "learning_rate": 1.232532427905373e-05,
+      "loss": 0.1622,
+      "step": 990
+    },
+    {
+      "epoch": 1.7493403693931397,
+      "grad_norm": 1.6455639600753784,
+      "learning_rate": 1.2182427157331351e-05,
+      "loss": 0.1972,
+      "step": 995
+    },
+    {
+      "epoch": 1.7581354441512753,
+      "grad_norm": 1.2293169498443604,
+      "learning_rate": 1.2039794608582659e-05,
+      "loss": 0.1654,
+      "step": 1000
+    },
+    {
+      "epoch": 1.7669305189094109,
+      "grad_norm": 1.2946192026138306,
+      "learning_rate": 1.1897440026149948e-05,
+      "loss": 0.1785,
+      "step": 1005
+    },
+    {
+      "epoch": 1.7757255936675462,
+      "grad_norm": 1.311557650566101,
+      "learning_rate": 1.1755376777274202e-05,
+      "loss": 0.1716,
+      "step": 1010
+    },
+    {
+      "epoch": 1.7845206684256816,
+      "grad_norm": 1.3270156383514404,
+      "learning_rate": 1.1613618201839912e-05,
+      "loss": 0.1669,
+      "step": 1015
+    },
+    {
+      "epoch": 1.793315743183817,
+      "grad_norm": 1.1812915802001953,
+      "learning_rate": 1.1472177611122381e-05,
+      "loss": 0.1518,
+      "step": 1020
+    },
+    {
+      "epoch": 1.8021108179419525,
+      "grad_norm": 1.1074270009994507,
+      "learning_rate": 1.1331068286537844e-05,
+      "loss": 0.1759,
+      "step": 1025
+    },
+    {
+      "epoch": 1.810905892700088,
+      "grad_norm": 1.1673787832260132,
+      "learning_rate": 1.1190303478396304e-05,
+      "loss": 0.1723,
+      "step": 1030
+    },
+    {
+      "epoch": 1.8197009674582234,
+      "grad_norm": 1.3313769102096558,
+      "learning_rate": 1.104989640465732e-05,
+      "loss": 0.1591,
+      "step": 1035
+    },
+    {
+      "epoch": 1.8284960422163588,
+      "grad_norm": 1.1003154516220093,
+      "learning_rate": 1.0909860249688812e-05,
+      "loss": 0.1392,
+      "step": 1040
+    },
+    {
+      "epoch": 1.8372911169744943,
+      "grad_norm": 1.2614023685455322,
+      "learning_rate": 1.0770208163029066e-05,
+      "loss": 0.1733,
+      "step": 1045
+    },
+    {
+      "epoch": 1.8460861917326299,
+      "grad_norm": 1.2860573530197144,
+      "learning_rate": 1.0630953258151946e-05,
+      "loss": 0.1492,
+      "step": 1050
+    },
+    {
+      "epoch": 1.8548812664907652,
+      "grad_norm": 1.5209264755249023,
+      "learning_rate": 1.0492108611235545e-05,
+      "loss": 0.1809,
+      "step": 1055
+    },
+    {
+      "epoch": 1.8636763412489006,
+      "grad_norm": 1.2931969165802002,
+      "learning_rate": 1.0353687259934322e-05,
+      "loss": 0.1522,
+      "step": 1060
+    },
+    {
+      "epoch": 1.872471416007036,
+      "grad_norm": 1.1818463802337646,
+      "learning_rate": 1.0215702202154828e-05,
+      "loss": 0.1729,
+      "step": 1065
+    },
+    {
+      "epoch": 1.8812664907651715,
+      "grad_norm": 1.2603001594543457,
+      "learning_rate": 1.0078166394835213e-05,
+      "loss": 0.1516,
+      "step": 1070
+    },
+    {
+      "epoch": 1.890061565523307,
+      "grad_norm": 1.4248186349868774,
+      "learning_rate": 9.94109275272856e-06,
+      "loss": 0.1461,
+      "step": 1075
+    },
+    {
+      "epoch": 1.8988566402814424,
+      "grad_norm": 1.1855427026748657,
+      "learning_rate": 9.80449414719015e-06,
+      "loss": 0.1526,
+      "step": 1080
+    },
+    {
+      "epoch": 1.9076517150395778,
+      "grad_norm": 1.4520982503890991,
+      "learning_rate": 9.668383404968845e-06,
+      "loss": 0.1387,
+      "step": 1085
+    },
+    {
+      "epoch": 1.9164467897977133,
+      "grad_norm": 1.1324946880340576,
+      "learning_rate": 9.532773307002659e-06,
+      "loss": 0.1425,
+      "step": 1090
+    },
+    {
+      "epoch": 1.9252418645558487,
+      "grad_norm": 1.2053685188293457,
+      "learning_rate": 9.397676587218577e-06,
+      "loss": 0.137,
+      "step": 1095
+    },
+    {
+      "epoch": 1.9340369393139842,
+      "grad_norm": 1.1318391561508179,
+      "learning_rate": 9.263105931336854e-06,
+      "loss": 0.1458,
+      "step": 1100
+    },
+    {
+      "epoch": 1.9428320140721196,
+      "grad_norm": 1.193960189819336,
+      "learning_rate": 9.129073975679822e-06,
+      "loss": 0.1383,
+      "step": 1105
+    },
+    {
+      "epoch": 1.951627088830255,
+      "grad_norm": 1.046067714691162,
+      "learning_rate": 8.995593305985284e-06,
+      "loss": 0.1488,
+      "step": 1110
+    },
+    {
+      "epoch": 1.9604221635883905,
+      "grad_norm": 1.0187636613845825,
+      "learning_rate": 8.862676456224744e-06,
+      "loss": 0.1345,
+      "step": 1115
+    },
+    {
+      "epoch": 1.969217238346526,
+      "grad_norm": 1.1997153759002686,
+      "learning_rate": 8.730335907426436e-06,
+      "loss": 0.1289,
+      "step": 1120
+    },
+    {
+      "epoch": 1.9780123131046614,
+      "grad_norm": 1.0095492601394653,
+      "learning_rate": 8.598584086503342e-06,
+      "loss": 0.1301,
+      "step": 1125
+    },
+    {
+      "epoch": 1.9868073878627968,
+      "grad_norm": 1.2394682168960571,
+      "learning_rate": 8.467433365086277e-06,
+      "loss": 0.1471,
+      "step": 1130
+    },
+    {
+      "epoch": 1.9956024626209323,
+      "grad_norm": 1.0819083452224731,
+      "learning_rate": 8.336896058362238e-06,
+      "loss": 0.1411,
+      "step": 1135
+    },
+    {
+      "epoch": 2.0035180299032542,
+      "grad_norm": 0.8920767307281494,
+      "learning_rate": 8.206984423917919e-06,
+      "loss": 0.1162,
+      "step": 1140
+    },
+    {
+      "epoch": 2.0123131046613896,
+      "grad_norm": 1.2982041835784912,
+      "learning_rate": 8.077710660588749e-06,
+      "loss": 0.1077,
+      "step": 1145
+    },
+    {
+      "epoch": 2.021108179419525,
+      "grad_norm": 1.179733157157898,
+      "learning_rate": 7.949086907313419e-06,
+      "loss": 0.1159,
+      "step": 1150
+    },
+    {
+      "epoch": 2.0299032541776607,
+      "grad_norm": 1.0822230577468872,
+      "learning_rate": 7.821125241994017e-06,
+      "loss": 0.1111,
+      "step": 1155
+    },
+    {
+      "epoch": 2.038698328935796,
+      "grad_norm": 1.1669689416885376,
+      "learning_rate": 7.693837680361868e-06,
+      "loss": 0.1095,
+      "step": 1160
+    },
+    {
+      "epoch": 2.0474934036939314,
+      "grad_norm": 1.1375830173492432,
+      "learning_rate": 7.567236174849293e-06,
+      "loss": 0.1108,
+      "step": 1165
+    },
+    {
+      "epoch": 2.0562884784520667,
+      "grad_norm": 1.2458380460739136,
+      "learning_rate": 7.441332613467242e-06,
+      "loss": 0.1044,
+      "step": 1170
+    },
+    {
+      "epoch": 2.065083553210202,
+      "grad_norm": 1.3224544525146484,
+      "learning_rate": 7.316138818688976e-06,
+      "loss": 0.1086,
+      "step": 1175
+    },
+    {
+      "epoch": 2.073878627968338,
+      "grad_norm": 1.1177361011505127,
+      "learning_rate": 7.191666546339965e-06,
+      "loss": 0.1094,
+      "step": 1180
+    },
+    {
+      "epoch": 2.0826737027264732,
+      "grad_norm": 0.9296045303344727,
+      "learning_rate": 7.067927484493984e-06,
+      "loss": 0.1031,
+      "step": 1185
+    },
+    {
+      "epoch": 2.0914687774846086,
+      "grad_norm": 0.9930837154388428,
+      "learning_rate": 6.944933252375575e-06,
+      "loss": 0.1102,
+      "step": 1190
+    },
+    {
+      "epoch": 2.100263852242744,
+      "grad_norm": 1.159774661064148,
+      "learning_rate": 6.822695399269022e-06,
+      "loss": 0.1038,
+      "step": 1195
+    },
+    {
+      "epoch": 2.1090589270008797,
+      "grad_norm": 1.079761266708374,
+      "learning_rate": 6.701225403433846e-06,
+      "loss": 0.1103,
+      "step": 1200
+    },
+    {
+      "epoch": 2.117854001759015,
+      "grad_norm": 0.9876915216445923,
+      "learning_rate": 6.5805346710269695e-06,
+      "loss": 0.1005,
+      "step": 1205
+    },
+    {
+      "epoch": 2.1266490765171504,
+      "grad_norm": 1.025357961654663,
+      "learning_rate": 6.4606345350316935e-06,
+      "loss": 0.1033,
+      "step": 1210
+    },
+    {
+      "epoch": 2.1354441512752858,
+      "grad_norm": 1.040122151374817,
+      "learning_rate": 6.34153625419351e-06,
+      "loss": 0.1072,
+      "step": 1215
+    },
+    {
+      "epoch": 2.144239226033421,
+      "grad_norm": 1.1730767488479614,
+      "learning_rate": 6.223251011962876e-06,
+      "loss": 0.1058,
+      "step": 1220
+    },
+    {
+      "epoch": 2.153034300791557,
+      "grad_norm": 1.0713180303573608,
+      "learning_rate": 6.105789915445112e-06,
+      "loss": 0.1022,
+      "step": 1225
+    },
+    {
+      "epoch": 2.1618293755496922,
+      "grad_norm": 0.8331393003463745,
+      "learning_rate": 5.989163994357386e-06,
+      "loss": 0.1037,
+      "step": 1230
+    },
+    {
+      "epoch": 2.1706244503078276,
+      "grad_norm": 0.9630013108253479,
+      "learning_rate": 5.873384199993063e-06,
+      "loss": 0.1038,
+      "step": 1235
+    },
+    {
+      "epoch": 2.179419525065963,
+      "grad_norm": 1.0274111032485962,
+      "learning_rate": 5.758461404193335e-06,
+      "loss": 0.1025,
+      "step": 1240
+    },
+    {
+      "epoch": 2.1882145998240983,
+      "grad_norm": 0.8318763375282288,
+      "learning_rate": 5.644406398326341e-06,
+      "loss": 0.0958,
+      "step": 1245
+    },
+    {
+      "epoch": 2.197009674582234,
+      "grad_norm": 0.9395310878753662,
+      "learning_rate": 5.531229892273871e-06,
+      "loss": 0.1055,
+      "step": 1250
+    },
+    {
+      "epoch": 2.2058047493403694,
+      "grad_norm": 0.8520766496658325,
+      "learning_rate": 5.418942513425682e-06,
+      "loss": 0.0876,
+      "step": 1255
+    },
+    {
+      "epoch": 2.2145998240985048,
+      "grad_norm": 1.1263083219528198,
+      "learning_rate": 5.307554805681562e-06,
+      "loss": 0.1015,
+      "step": 1260
+    },
+    {
+      "epoch": 2.22339489885664,
+      "grad_norm": 0.8551110029220581,
+      "learning_rate": 5.197077228461279e-06,
+      "loss": 0.0903,
+      "step": 1265
+    },
+    {
+      "epoch": 2.232189973614776,
+      "grad_norm": 0.9229046106338501,
+      "learning_rate": 5.0875201557224085e-06,
+      "loss": 0.0927,
+      "step": 1270
+    },
+    {
+      "epoch": 2.2409850483729112,
+      "grad_norm": 0.7913192510604858,
+      "learning_rate": 4.9788938749862025e-06,
+      "loss": 0.0967,
+      "step": 1275
+    },
+    {
+      "epoch": 2.2497801231310466,
+      "grad_norm": 0.7556148171424866,
+      "learning_rate": 4.8712085863716065e-06,
+      "loss": 0.0947,
+      "step": 1280
+    },
+    {
+      "epoch": 2.258575197889182,
+      "grad_norm": 1.0492786169052124,
+      "learning_rate": 4.7644744016374385e-06,
+      "loss": 0.1015,
+      "step": 1285
+    },
+    {
+      "epoch": 2.2673702726473177,
+      "grad_norm": 0.9224637746810913,
+      "learning_rate": 4.658701343232876e-06,
+      "loss": 0.1059,
+      "step": 1290
+    },
+    {
+      "epoch": 2.276165347405453,
+      "grad_norm": 0.9385767579078674,
+      "learning_rate": 4.553899343356367e-06,
+      "loss": 0.1004,
+      "step": 1295
+    },
+    {
+      "epoch": 2.2849604221635884,
+      "grad_norm": 0.9364350438117981,
+      "learning_rate": 4.450078243022967e-06,
+      "loss": 0.1036,
+      "step": 1300
+    },
+    {
+      "epoch": 2.2937554969217238,
+      "grad_norm": 0.9326781630516052,
+      "learning_rate": 4.347247791140247e-06,
+      "loss": 0.0976,
+      "step": 1305
+    },
+    {
+      "epoch": 2.302550571679859,
+      "grad_norm": 1.0285263061523438,
+      "learning_rate": 4.24541764359289e-06,
+      "loss": 0.1023,
+      "step": 1310
+    },
+    {
+      "epoch": 2.311345646437995,
+      "grad_norm": 0.9973950386047363,
+      "learning_rate": 4.144597362335977e-06,
+      "loss": 0.1071,
+      "step": 1315
+    },
+    {
+      "epoch": 2.3201407211961302,
+      "grad_norm": 0.9997219443321228,
+      "learning_rate": 4.044796414497097e-06,
+      "loss": 0.0935,
+      "step": 1320
+    },
+    {
+      "epoch": 2.3289357959542656,
+      "grad_norm": 0.916297972202301,
+      "learning_rate": 3.946024171487407e-06,
+      "loss": 0.0939,
+      "step": 1325
+    },
+    {
+      "epoch": 2.337730870712401,
+      "grad_norm": 0.8260137438774109,
+      "learning_rate": 3.848289908121605e-06,
+      "loss": 0.0869,
+      "step": 1330
+    },
+    {
+      "epoch": 2.3465259454705363,
+      "grad_norm": 0.8840376138687134,
+      "learning_rate": 3.75160280174706e-06,
+      "loss": 0.0773,
+      "step": 1335
+    },
+    {
+      "epoch": 2.355321020228672,
+      "grad_norm": 1.0563877820968628,
+      "learning_rate": 3.6559719313820205e-06,
+      "loss": 0.0954,
+      "step": 1340
+    },
+    {
+      "epoch": 2.3641160949868074,
+      "grad_norm": 0.8720323443412781,
+      "learning_rate": 3.5614062768630827e-06,
+      "loss": 0.0822,
+      "step": 1345
+    },
+    {
+      "epoch": 2.3729111697449428,
+      "grad_norm": 0.8708292245864868,
+      "learning_rate": 3.467914718001996e-06,
+      "loss": 0.0918,
+      "step": 1350
+    },
+    {
+      "epoch": 2.381706244503078,
+      "grad_norm": 0.8966621160507202,
+      "learning_rate": 3.3755060337518314e-06,
+      "loss": 0.0915,
+      "step": 1355
+    },
+    {
+      "epoch": 2.390501319261214,
+      "grad_norm": 0.9872451424598694,
+      "learning_rate": 3.2841889013826074e-06,
+      "loss": 0.0917,
+      "step": 1360
+    },
+    {
+      "epoch": 2.3992963940193492,
+      "grad_norm": 1.0449951887130737,
+      "learning_rate": 3.193971895666521e-06,
+      "loss": 0.0894,
+      "step": 1365
+    },
+    {
+      "epoch": 2.4080914687774846,
+      "grad_norm": 0.719258189201355,
+      "learning_rate": 3.104863488072749e-06,
+      "loss": 0.0936,
+      "step": 1370
+    },
+    {
+      "epoch": 2.41688654353562,
+      "grad_norm": 0.7323921918869019,
+      "learning_rate": 3.016872045971958e-06,
+      "loss": 0.0921,
+      "step": 1375
+    },
+    {
+      "epoch": 2.4256816182937557,
+      "grad_norm": 0.901992917060852,
+      "learning_rate": 2.930005831850623e-06,
+      "loss": 0.0857,
+      "step": 1380
+    },
+    {
+      "epoch": 2.434476693051891,
+      "grad_norm": 0.8706481456756592,
+      "learning_rate": 2.8442730025351598e-06,
+      "loss": 0.0827,
+      "step": 1385
+    },
+    {
+      "epoch": 2.4432717678100264,
+      "grad_norm": 0.9546716213226318,
+      "learning_rate": 2.7596816084259803e-06,
+      "loss": 0.083,
+      "step": 1390
+    },
+    {
+      "epoch": 2.4520668425681618,
+      "grad_norm": 0.7810299396514893,
+      "learning_rate": 2.676239592741575e-06,
+      "loss": 0.0932,
+      "step": 1395
+    },
+    {
+      "epoch": 2.460861917326297,
+      "grad_norm": 0.7423774600028992,
+      "learning_rate": 2.593954790772619e-06,
+      "loss": 0.0928,
+      "step": 1400
+    },
+    {
+      "epoch": 2.469656992084433,
+      "grad_norm": 0.7438393235206604,
+      "learning_rate": 2.512834929146232e-06,
+      "loss": 0.0842,
+      "step": 1405
+    },
+    {
+      "epoch": 2.4784520668425682,
+      "grad_norm": 0.8357313871383667,
+      "learning_rate": 2.4328876251004394e-06,
+      "loss": 0.0807,
+      "step": 1410
+    },
+    {
+      "epoch": 2.4872471416007036,
+      "grad_norm": 0.8226934671401978,
+      "learning_rate": 2.3541203857689308e-06,
+      "loss": 0.0919,
+      "step": 1415
+    },
+    {
+      "epoch": 2.496042216358839,
+      "grad_norm": 0.8223289251327515,
+      "learning_rate": 2.276540607476089e-06,
+      "loss": 0.075,
+      "step": 1420
+    },
+    {
+      "epoch": 2.5048372911169743,
+      "grad_norm": 0.8878283500671387,
+      "learning_rate": 2.2001555750424897e-06,
+      "loss": 0.0865,
+      "step": 1425
+    },
+    {
+      "epoch": 2.51363236587511,
+      "grad_norm": 0.633674681186676,
+      "learning_rate": 2.1249724611008714e-06,
+      "loss": 0.0815,
+      "step": 1430
+    },
+    {
+      "epoch": 2.5224274406332454,
+      "grad_norm": 0.8103867173194885,
+      "learning_rate": 2.0509983254225772e-06,
+      "loss": 0.0895,
+      "step": 1435
+    },
+    {
+      "epoch": 2.5312225153913808,
+      "grad_norm": 0.7639410495758057,
+      "learning_rate": 1.978240114254653e-06,
+      "loss": 0.0792,
+      "step": 1440
+    },
+    {
+      "epoch": 2.540017590149516,
+      "grad_norm": 0.7429800629615784,
+      "learning_rate": 1.9067046596676013e-06,
+      "loss": 0.0807,
+      "step": 1445
+    },
+    {
+      "epoch": 2.5488126649076515,
+      "grad_norm": 0.840923011302948,
+      "learning_rate": 1.8363986789138215e-06,
+      "loss": 0.0744,
+      "step": 1450
+    },
+    {
+      "epoch": 2.5576077396657872,
+      "grad_norm": 0.6393332481384277,
+      "learning_rate": 1.7673287737968624e-06,
+      "loss": 0.0797,
+      "step": 1455
+    },
+    {
+      "epoch": 2.5664028144239226,
+      "grad_norm": 0.7780035138130188,
+      "learning_rate": 1.6995014300515067e-06,
+      "loss": 0.083,
+      "step": 1460
+    },
+    {
+      "epoch": 2.575197889182058,
+      "grad_norm": 0.8644776344299316,
+      "learning_rate": 1.6329230167347559e-06,
+      "loss": 0.0909,
+      "step": 1465
+    },
+    {
+      "epoch": 2.5839929639401937,
+      "grad_norm": 0.9181379675865173,
+      "learning_rate": 1.567599785627758e-06,
+      "loss": 0.0904,
+      "step": 1470
+    },
+    {
+      "epoch": 2.592788038698329,
+      "grad_norm": 0.8803585171699524,
+      "learning_rate": 1.5035378706487712e-06,
+      "loss": 0.0959,
+      "step": 1475
+    },
+    {
+      "epoch": 2.6015831134564644,
+      "grad_norm": 0.7419053912162781,
+      "learning_rate": 1.4407432872771852e-06,
+      "loss": 0.0845,
+      "step": 1480
+    },
+    {
+      "epoch": 2.6103781882145998,
+      "grad_norm": 0.7921883463859558,
+      "learning_rate": 1.3792219319886413e-06,
+      "loss": 0.0819,
+      "step": 1485
+    },
+    {
+      "epoch": 2.619173262972735,
+      "grad_norm": 0.8032109141349792,
+      "learning_rate": 1.3189795817013695e-06,
+      "loss": 0.0902,
+      "step": 1490
+    },
+    {
+      "epoch": 2.627968337730871,
+      "grad_norm": 0.7834377884864807,
+      "learning_rate": 1.2600218932337204e-06,
+      "loss": 0.0768,
+      "step": 1495
+    },
+    {
+      "epoch": 2.6367634124890063,
+      "grad_norm": 0.9693959355354309,
+      "learning_rate": 1.202354402772981e-06,
+      "loss": 0.0747,
+      "step": 1500
+    },
+    {
+      "epoch": 2.6455584872471416,
+      "grad_norm": 0.6451067328453064,
+      "learning_rate": 1.1459825253555306e-06,
+      "loss": 0.0698,
+      "step": 1505
+    },
+    {
+      "epoch": 2.654353562005277,
+      "grad_norm": 0.832350492477417,
+      "learning_rate": 1.0909115543583508e-06,
+      "loss": 0.0878,
+      "step": 1510
+    },
+    {
+      "epoch": 2.6631486367634123,
+      "grad_norm": 0.7967277765274048,
+      "learning_rate": 1.0371466610019837e-06,
+      "loss": 0.0721,
+      "step": 1515
+    },
+    {
+      "epoch": 2.671943711521548,
+      "grad_norm": 0.9016730189323425,
+      "learning_rate": 9.846928938649475e-07,
+      "loss": 0.0858,
+      "step": 1520
+    },
+    {
+      "epoch": 2.6807387862796834,
+      "grad_norm": 0.7053119540214539,
+      "learning_rate": 9.335551784096552e-07,
+      "loss": 0.0703,
+      "step": 1525
+    },
+    {
+      "epoch": 2.6895338610378188,
+      "grad_norm": 0.6417444944381714,
+      "learning_rate": 8.837383165199308e-07,
+      "loss": 0.0821,
+      "step": 1530
+    },
+    {
+      "epoch": 2.698328935795954,
+      "grad_norm": 0.6701059341430664,
+      "learning_rate": 8.35246986050095e-07,
+      "loss": 0.0762,
+      "step": 1535
+    },
+    {
+      "epoch": 2.7071240105540895,
+      "grad_norm": 0.7191208600997925,
+      "learning_rate": 7.880857403857028e-07,
+      "loss": 0.0786,
+      "step": 1540
+    },
+    {
+      "epoch": 2.7159190853122253,
+      "grad_norm": 0.6913970708847046,
+      "learning_rate": 7.422590080159947e-07,
+      "loss": 0.0764,
+      "step": 1545
+    },
+    {
+      "epoch": 2.7247141600703606,
+      "grad_norm": 0.7714747190475464,
+      "learning_rate": 6.977710921180452e-07,
+      "loss": 0.0826,
+      "step": 1550
+    },
+    {
+      "epoch": 2.733509234828496,
+      "grad_norm": 0.6777530908584595,
+      "learning_rate": 6.546261701526851e-07,
+      "loss": 0.0762,
+      "step": 1555
+    },
+    {
+      "epoch": 2.7423043095866317,
+      "grad_norm": 0.8672024011611938,
+      "learning_rate": 6.128282934722479e-07,
+      "loss": 0.0958,
+      "step": 1560
+    },
+    {
+      "epoch": 2.751099384344767,
+      "grad_norm": 0.8040085434913635,
+      "learning_rate": 5.723813869401362e-07,
+      "loss": 0.0765,
+      "step": 1565
+    },
+    {
+      "epoch": 2.7598944591029024,
+      "grad_norm": 0.7275374531745911,
+      "learning_rate": 5.332892485622648e-07,
+      "loss": 0.0718,
+      "step": 1570
+    },
+    {
+      "epoch": 2.7686895338610378,
+      "grad_norm": 0.6731064915657043,
+      "learning_rate": 4.955555491304409e-07,
+      "loss": 0.0786,
+      "step": 1575
+    },
+    {
+      "epoch": 2.777484608619173,
+      "grad_norm": 0.6927102208137512,
+      "learning_rate": 4.5918383187766143e-07,
+      "loss": 0.0896,
+      "step": 1580
+    },
+    {
+      "epoch": 2.786279683377309,
+      "grad_norm": 0.8046338558197021,
+      "learning_rate": 4.2417751214540115e-07,
+      "loss": 0.082,
+      "step": 1585
+    },
+    {
+      "epoch": 2.7950747581354443,
+      "grad_norm": 0.8124101161956787,
+      "learning_rate": 3.905398770629104e-07,
+      "loss": 0.0806,
+      "step": 1590
+    },
+    {
+      "epoch": 2.8038698328935796,
+      "grad_norm": 0.8360600471496582,
+      "learning_rate": 3.582740852385541e-07,
+      "loss": 0.0837,
+      "step": 1595
+    },
+    {
+      "epoch": 2.812664907651715,
+      "grad_norm": 0.7593919634819031,
+      "learning_rate": 3.2738316646320543e-07,
+      "loss": 0.0721,
+      "step": 1600
+    },
+    {
+      "epoch": 2.8214599824098503,
+      "grad_norm": 0.8052194714546204,
+      "learning_rate": 2.9787002142575516e-07,
+      "loss": 0.0728,
+      "step": 1605
+    },
+    {
+      "epoch": 2.830255057167986,
+      "grad_norm": 0.8136105537414551,
+      "learning_rate": 2.6973742144072775e-07,
+      "loss": 0.0754,
+      "step": 1610
+    },
+    {
+      "epoch": 2.8390501319261214,
+      "grad_norm": 0.9357880353927612,
+      "learning_rate": 2.4298800818805477e-07,
+      "loss": 0.0676,
+      "step": 1615
+    },
+    {
+      "epoch": 2.847845206684257,
+      "grad_norm": 0.7042924761772156,
+      "learning_rate": 2.1762429346502132e-07,
+      "loss": 0.0769,
+      "step": 1620
+    },
+    {
+      "epoch": 2.856640281442392,
+      "grad_norm": 0.7005944848060608,
+      "learning_rate": 1.9364865895039664e-07,
+      "loss": 0.0812,
+      "step": 1625
+    },
+    {
+      "epoch": 2.8654353562005275,
+      "grad_norm": 1.340453028678894,
+      "learning_rate": 1.7106335598080115e-07,
+      "loss": 0.0785,
+      "step": 1630
+    },
+    {
+      "epoch": 2.8742304309586633,
+      "grad_norm": 0.7811134457588196,
+      "learning_rate": 1.4987050533929902e-07,
+      "loss": 0.0778,
+      "step": 1635
+    },
+    {
+      "epoch": 2.8830255057167986,
+      "grad_norm": 0.917794942855835,
+      "learning_rate": 1.3007209705625745e-07,
+      "loss": 0.0823,
+      "step": 1640
+    },
+    {
+      "epoch": 2.891820580474934,
+      "grad_norm": 0.74846351146698,
+      "learning_rate": 1.1166999022247438e-07,
+      "loss": 0.0822,
+      "step": 1645
+    },
+    {
+      "epoch": 2.9006156552330697,
+      "grad_norm": 0.7628084421157837,
+      "learning_rate": 9.466591281461989e-08,
+      "loss": 0.0717,
+      "step": 1650
+    },
+    {
+      "epoch": 2.909410729991205,
+      "grad_norm": 0.9148209691047668,
+      "learning_rate": 7.906146153296267e-08,
+      "loss": 0.0776,
+      "step": 1655
+    },
+    {
+      "epoch": 2.9182058047493404,
+      "grad_norm": 0.8598858714103699,
+      "learning_rate": 6.485810165145156e-08,
+      "loss": 0.0873,
+      "step": 1660
+    },
+    {
+      "epoch": 2.927000879507476,
+      "grad_norm": 0.7613882422447205,
+      "learning_rate": 5.205716688011564e-08,
+      "loss": 0.0725,
+      "step": 1665
+    },
+    {
+      "epoch": 2.935795954265611,
+      "grad_norm": 0.8053538203239441,
+      "learning_rate": 4.0659859239831065e-08,
+      "loss": 0.0826,
+      "step": 1670
+    },
+    {
+      "epoch": 2.944591029023747,
+      "grad_norm": 0.8616787195205688,
+      "learning_rate": 3.0667248949449725e-08,
+      "loss": 0.0766,
+      "step": 1675
+    },
+    {
+      "epoch": 2.9533861037818823,
+      "grad_norm": 0.7508912086486816,
+      "learning_rate": 2.2080274325306237e-08,
+      "loss": 0.0806,
+      "step": 1680
+    },
+    {
+      "epoch": 2.9621811785400176,
+      "grad_norm": 0.829683244228363,
+      "learning_rate": 1.4899741693105152e-08,
+      "loss": 0.0843,
+      "step": 1685
+    },
+    {
+      "epoch": 2.970976253298153,
+      "grad_norm": 0.7025284171104431,
+      "learning_rate": 9.12632531221147e-09,
+      "loss": 0.0744,
+      "step": 1690
+    },
+    {
+      "epoch": 2.9797713280562883,
+      "grad_norm": 0.8548363447189331,
+      "learning_rate": 4.760567312331321e-09,
+      "loss": 0.0785,
+      "step": 1695
+    },
+    {
+      "epoch": 2.988566402814424,
+      "grad_norm": 0.6531686186790466,
+      "learning_rate": 1.8028776426110005e-09,
+      "loss": 0.0802,
+      "step": 1700
+    },
+    {
+      "epoch": 2.9973614775725594,
+      "grad_norm": 0.7708808183670044,
+      "learning_rate": 2.5353403313443934e-10,
+      "loss": 0.0743,
+      "step": 1705
+    }
+  ],
+  "logging_steps": 5,
+  "max_steps": 1707,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 3,
+  "save_steps": 2000,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": true
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.2514701901118833e+18,
+  "train_batch_size": 2,
+  "trial_name": null,
+  "trial_params": null
+}

133_128_e3_3e-5/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d4843001fdd68941d95ada69a1c6dfd26d665d9c1d3e34d0686f95606e4cb942
+size 8209

133_128_e3_3e-5/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)