Upload folder using huggingface_hub

145_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": [
+    "gate_proj",
+    "k_proj",
+    "down_proj",
+    "o_proj",
+    "v_proj",
+    "q_proj",
+    "up_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_rslora": false
+}

145_128_e3_3e-5/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:54a2917b8185a1eaf3b7d3733f42c7cf5043a73b4b746e06de1e0ad586f49e6b
+size 791751704

145_128_e3_3e-5/latest

@@ -0,0 +1 @@
+global_step1389

145_128_e3_3e-5/rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:313a9485a5543a5b513786f3467fb84825478ab72d2be0b0d481b38d62a2196e
+size 16389

145_128_e3_3e-5/rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5c410633b43cb9557403de8cccee1b1d2ef8b2fa332b6e37a3396fee4cdddf35
+size 16389

145_128_e3_3e-5/rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ac54ef537e001c5b017c039e056249e249fc1f0c8d23a1efa57a4187afab78dd
+size 16389

145_128_e3_3e-5/rng_state_3.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2c1584a59c9f92bd8055377e85254e5cce63639a200c4c1381aeb5dd3227cfc4
+size 16389

145_128_e3_3e-5/rng_state_4.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:85b500ecad92414b9401507207ef15a0755ee7cc5b579e7d8a56b72df80b3f03
+size 16389

145_128_e3_3e-5/rng_state_5.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f23f72797939153fd68c14a3bf8c991c1aaa3705027dc59383bddd2e9ea0c8b4
+size 16389

145_128_e3_3e-5/rng_state_6.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ce95a9fbf692e2444db6ce4804e06b1d2a26b6a648b151d3fec952edd890d0e1
+size 16389

145_128_e3_3e-5/rng_state_7.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6d2d97c917331ca75ebc7833d66019749b8ea36640f03e25e11b8412ba290fbe
+size 16389

145_128_e3_3e-5/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:66f686e50365d0f568aea8d64994797158507fdb5e629297959268dd055eef1a
+size 1401

145_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
+  }
+}

145_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
+}

145_128_e3_3e-5/trainer_state.json

@@ -0,0 +1,1973 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.0,
+  "eval_steps": 500,
+  "global_step": 1389,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.01079913606911447,
+      "grad_norm": 0.8999239802360535,
+      "learning_rate": 1.7142857142857143e-06,
+      "loss": 1.4709,
+      "step": 5
+    },
+    {
+      "epoch": 0.02159827213822894,
+      "grad_norm": 0.9575327038764954,
+      "learning_rate": 3.857142857142857e-06,
+      "loss": 1.4094,
+      "step": 10
+    },
+    {
+      "epoch": 0.032397408207343416,
+      "grad_norm": 0.7031283378601074,
+      "learning_rate": 6e-06,
+      "loss": 1.3747,
+      "step": 15
+    },
+    {
+      "epoch": 0.04319654427645788,
+      "grad_norm": 0.6973682641983032,
+      "learning_rate": 8.142857142857142e-06,
+      "loss": 1.4418,
+      "step": 20
+    },
+    {
+      "epoch": 0.05399568034557235,
+      "grad_norm": 0.6896507740020752,
+      "learning_rate": 1.0285714285714286e-05,
+      "loss": 1.3597,
+      "step": 25
+    },
+    {
+      "epoch": 0.06479481641468683,
+      "grad_norm": 0.5128210186958313,
+      "learning_rate": 1.242857142857143e-05,
+      "loss": 1.3681,
+      "step": 30
+    },
+    {
+      "epoch": 0.0755939524838013,
+      "grad_norm": 0.5501849055290222,
+      "learning_rate": 1.4571428571428571e-05,
+      "loss": 1.3629,
+      "step": 35
+    },
+    {
+      "epoch": 0.08639308855291576,
+      "grad_norm": 0.5030866861343384,
+      "learning_rate": 1.6714285714285716e-05,
+      "loss": 1.3765,
+      "step": 40
+    },
+    {
+      "epoch": 0.09719222462203024,
+      "grad_norm": 0.4635452330112457,
+      "learning_rate": 1.8857142857142856e-05,
+      "loss": 1.3284,
+      "step": 45
+    },
+    {
+      "epoch": 0.1079913606911447,
+      "grad_norm": 0.5290427803993225,
+      "learning_rate": 2.1e-05,
+      "loss": 1.3383,
+      "step": 50
+    },
+    {
+      "epoch": 0.11879049676025918,
+      "grad_norm": 0.4894343912601471,
+      "learning_rate": 2.3142857142857145e-05,
+      "loss": 1.3159,
+      "step": 55
+    },
+    {
+      "epoch": 0.12958963282937366,
+      "grad_norm": 0.5415617227554321,
+      "learning_rate": 2.5285714285714285e-05,
+      "loss": 1.2917,
+      "step": 60
+    },
+    {
+      "epoch": 0.14038876889848811,
+      "grad_norm": 0.5164010524749756,
+      "learning_rate": 2.7428571428571428e-05,
+      "loss": 1.2813,
+      "step": 65
+    },
+    {
+      "epoch": 0.1511879049676026,
+      "grad_norm": 0.6710526347160339,
+      "learning_rate": 2.9571428571428575e-05,
+      "loss": 1.266,
+      "step": 70
+    },
+    {
+      "epoch": 0.16198704103671707,
+      "grad_norm": 0.5664787292480469,
+      "learning_rate": 2.999931924927058e-05,
+      "loss": 1.2965,
+      "step": 75
+    },
+    {
+      "epoch": 0.17278617710583152,
+      "grad_norm": 0.6182973980903625,
+      "learning_rate": 2.999655380533103e-05,
+      "loss": 1.2033,
+      "step": 80
+    },
+    {
+      "epoch": 0.183585313174946,
+      "grad_norm": 0.5756394863128662,
+      "learning_rate": 2.999166151316113e-05,
+      "loss": 1.1868,
+      "step": 85
+    },
+    {
+      "epoch": 0.19438444924406048,
+      "grad_norm": 0.7031417489051819,
+      "learning_rate": 2.9984643066597815e-05,
+      "loss": 1.1748,
+      "step": 90
+    },
+    {
+      "epoch": 0.20518358531317496,
+      "grad_norm": 0.7027549147605896,
+      "learning_rate": 2.997549946101448e-05,
+      "loss": 1.1635,
+      "step": 95
+    },
+    {
+      "epoch": 0.2159827213822894,
+      "grad_norm": 0.7605417370796204,
+      "learning_rate": 2.9964231993179804e-05,
+      "loss": 1.1293,
+      "step": 100
+    },
+    {
+      "epoch": 0.2267818574514039,
+      "grad_norm": 0.685104489326477,
+      "learning_rate": 2.9950842261073867e-05,
+      "loss": 1.1369,
+      "step": 105
+    },
+    {
+      "epoch": 0.23758099352051837,
+      "grad_norm": 0.7664139866828918,
+      "learning_rate": 2.9935332163661465e-05,
+      "loss": 1.1287,
+      "step": 110
+    },
+    {
+      "epoch": 0.24838012958963282,
+      "grad_norm": 0.7763246297836304,
+      "learning_rate": 2.9917703900622867e-05,
+      "loss": 1.0942,
+      "step": 115
+    },
+    {
+      "epoch": 0.2591792656587473,
+      "grad_norm": 0.7119962573051453,
+      "learning_rate": 2.9897959972041778e-05,
+      "loss": 1.0677,
+      "step": 120
+    },
+    {
+      "epoch": 0.26997840172786175,
+      "grad_norm": 0.8102728128433228,
+      "learning_rate": 2.9876103178050833e-05,
+      "loss": 1.0625,
+      "step": 125
+    },
+    {
+      "epoch": 0.28077753779697623,
+      "grad_norm": 0.8341878056526184,
+      "learning_rate": 2.9852136618434435e-05,
+      "loss": 1.0677,
+      "step": 130
+    },
+    {
+      "epoch": 0.2915766738660907,
+      "grad_norm": 0.8508694767951965,
+      "learning_rate": 2.9826063692189152e-05,
+      "loss": 1.0316,
+      "step": 135
+    },
+    {
+      "epoch": 0.3023758099352052,
+      "grad_norm": 0.8454902768135071,
+      "learning_rate": 2.9797888097041656e-05,
+      "loss": 1.0388,
+      "step": 140
+    },
+    {
+      "epoch": 0.31317494600431967,
+      "grad_norm": 0.8273711204528809,
+      "learning_rate": 2.9767613828924316e-05,
+      "loss": 0.9911,
+      "step": 145
+    },
+    {
+      "epoch": 0.32397408207343414,
+      "grad_norm": 1.0613164901733398,
+      "learning_rate": 2.9735245181408466e-05,
+      "loss": 0.9797,
+      "step": 150
+    },
+    {
+      "epoch": 0.3347732181425486,
+      "grad_norm": 1.1455518007278442,
+      "learning_rate": 2.9700786745095483e-05,
+      "loss": 0.9431,
+      "step": 155
+    },
+    {
+      "epoch": 0.34557235421166305,
+      "grad_norm": 0.9718424081802368,
+      "learning_rate": 2.966424340696577e-05,
+      "loss": 0.9163,
+      "step": 160
+    },
+    {
+      "epoch": 0.3563714902807775,
+      "grad_norm": 1.0896577835083008,
+      "learning_rate": 2.9625620349685617e-05,
+      "loss": 0.9352,
+      "step": 165
+    },
+    {
+      "epoch": 0.367170626349892,
+      "grad_norm": 1.386668086051941,
+      "learning_rate": 2.9584923050872218e-05,
+      "loss": 0.9279,
+      "step": 170
+    },
+    {
+      "epoch": 0.3779697624190065,
+      "grad_norm": 1.1322718858718872,
+      "learning_rate": 2.9542157282316822e-05,
+      "loss": 0.9106,
+      "step": 175
+    },
+    {
+      "epoch": 0.38876889848812096,
+      "grad_norm": 1.0905592441558838,
+      "learning_rate": 2.9497329109166154e-05,
+      "loss": 0.8997,
+      "step": 180
+    },
+    {
+      "epoch": 0.39956803455723544,
+      "grad_norm": 1.0120617151260376,
+      "learning_rate": 2.9450444889062243e-05,
+      "loss": 0.8613,
+      "step": 185
+    },
+    {
+      "epoch": 0.4103671706263499,
+      "grad_norm": 1.1251927614212036,
+      "learning_rate": 2.940151127124076e-05,
+      "loss": 0.8394,
+      "step": 190
+    },
+    {
+      "epoch": 0.42116630669546434,
+      "grad_norm": 1.1108499765396118,
+      "learning_rate": 2.935053519558802e-05,
+      "loss": 0.8418,
+      "step": 195
+    },
+    {
+      "epoch": 0.4319654427645788,
+      "grad_norm": 0.9873912334442139,
+      "learning_rate": 2.929752389165673e-05,
+      "loss": 0.8319,
+      "step": 200
+    },
+    {
+      "epoch": 0.4427645788336933,
+      "grad_norm": 1.1158689260482788,
+      "learning_rate": 2.9242484877640706e-05,
+      "loss": 0.8877,
+      "step": 205
+    },
+    {
+      "epoch": 0.4535637149028078,
+      "grad_norm": 1.0617947578430176,
+      "learning_rate": 2.9185425959308596e-05,
+      "loss": 0.7985,
+      "step": 210
+    },
+    {
+      "epoch": 0.46436285097192226,
+      "grad_norm": 1.4247864484786987,
+      "learning_rate": 2.912635522889686e-05,
+      "loss": 0.8282,
+      "step": 215
+    },
+    {
+      "epoch": 0.47516198704103674,
+      "grad_norm": 1.0150513648986816,
+      "learning_rate": 2.9065281063962106e-05,
+      "loss": 0.8176,
+      "step": 220
+    },
+    {
+      "epoch": 0.48596112311015116,
+      "grad_norm": 1.1862380504608154,
+      "learning_rate": 2.900221212619297e-05,
+      "loss": 0.7817,
+      "step": 225
+    },
+    {
+      "epoch": 0.49676025917926564,
+      "grad_norm": 1.172380805015564,
+      "learning_rate": 2.893715736018168e-05,
+      "loss": 0.7825,
+      "step": 230
+    },
+    {
+      "epoch": 0.5075593952483801,
+      "grad_norm": 1.2643972635269165,
+      "learning_rate": 2.8870125992155527e-05,
+      "loss": 0.7535,
+      "step": 235
+    },
+    {
+      "epoch": 0.5183585313174947,
+      "grad_norm": 1.1213304996490479,
+      "learning_rate": 2.8801127528668372e-05,
+      "loss": 0.7855,
+      "step": 240
+    },
+    {
+      "epoch": 0.5291576673866091,
+      "grad_norm": 1.0870945453643799,
+      "learning_rate": 2.87301717552524e-05,
+      "loss": 0.7622,
+      "step": 245
+    },
+    {
+      "epoch": 0.5399568034557235,
+      "grad_norm": 1.3154348134994507,
+      "learning_rate": 2.8657268735030316e-05,
+      "loss": 0.7608,
+      "step": 250
+    },
+    {
+      "epoch": 0.550755939524838,
+      "grad_norm": 1.225184440612793,
+      "learning_rate": 2.8582428807288182e-05,
+      "loss": 0.7197,
+      "step": 255
+    },
+    {
+      "epoch": 0.5615550755939525,
+      "grad_norm": 1.3424427509307861,
+      "learning_rate": 2.850566258600903e-05,
+      "loss": 0.7083,
+      "step": 260
+    },
+    {
+      "epoch": 0.572354211663067,
+      "grad_norm": 1.1719582080841064,
+      "learning_rate": 2.8426980958367618e-05,
+      "loss": 0.6578,
+      "step": 265
+    },
+    {
+      "epoch": 0.5831533477321814,
+      "grad_norm": 1.3276638984680176,
+      "learning_rate": 2.8346395083186336e-05,
+      "loss": 0.7154,
+      "step": 270
+    },
+    {
+      "epoch": 0.593952483801296,
+      "grad_norm": 1.189018964767456,
+      "learning_rate": 2.8263916389352673e-05,
+      "loss": 0.6988,
+      "step": 275
+    },
+    {
+      "epoch": 0.6047516198704104,
+      "grad_norm": 1.2083779573440552,
+      "learning_rate": 2.8179556574198312e-05,
+      "loss": 0.6901,
+      "step": 280
+    },
+    {
+      "epoch": 0.6155507559395248,
+      "grad_norm": 1.1995373964309692,
+      "learning_rate": 2.80933276018402e-05,
+      "loss": 0.6874,
+      "step": 285
+    },
+    {
+      "epoch": 0.6263498920086393,
+      "grad_norm": 1.1318057775497437,
+      "learning_rate": 2.800524170148377e-05,
+      "loss": 0.6377,
+      "step": 290
+    },
+    {
+      "epoch": 0.6371490280777538,
+      "grad_norm": 1.2895493507385254,
+      "learning_rate": 2.7915311365688562e-05,
+      "loss": 0.6463,
+      "step": 295
+    },
+    {
+      "epoch": 0.6479481641468683,
+      "grad_norm": 1.2814615964889526,
+      "learning_rate": 2.78235493485965e-05,
+      "loss": 0.6752,
+      "step": 300
+    },
+    {
+      "epoch": 0.6587473002159827,
+      "grad_norm": 1.0793182849884033,
+      "learning_rate": 2.7729968664123056e-05,
+      "loss": 0.6492,
+      "step": 305
+    },
+    {
+      "epoch": 0.6695464362850972,
+      "grad_norm": 1.365412950515747,
+      "learning_rate": 2.7634582584111605e-05,
+      "loss": 0.622,
+      "step": 310
+    },
+    {
+      "epoch": 0.6803455723542117,
+      "grad_norm": 1.2403167486190796,
+      "learning_rate": 2.753740463645115e-05,
+      "loss": 0.6106,
+      "step": 315
+    },
+    {
+      "epoch": 0.6911447084233261,
+      "grad_norm": 1.3841968774795532,
+      "learning_rate": 2.7438448603157802e-05,
+      "loss": 0.6225,
+      "step": 320
+    },
+    {
+      "epoch": 0.7019438444924406,
+      "grad_norm": 1.3566217422485352,
+      "learning_rate": 2.7337728518420143e-05,
+      "loss": 0.6177,
+      "step": 325
+    },
+    {
+      "epoch": 0.712742980561555,
+      "grad_norm": 1.1903268098831177,
+      "learning_rate": 2.723525866660889e-05,
+      "loss": 0.5964,
+      "step": 330
+    },
+    {
+      "epoch": 0.7235421166306696,
+      "grad_norm": 1.5612664222717285,
+      "learning_rate": 2.713105358025104e-05,
+      "loss": 0.6102,
+      "step": 335
+    },
+    {
+      "epoch": 0.734341252699784,
+      "grad_norm": 1.2308214902877808,
+      "learning_rate": 2.7025128037968824e-05,
+      "loss": 0.5454,
+      "step": 340
+    },
+    {
+      "epoch": 0.7451403887688985,
+      "grad_norm": 1.3439821004867554,
+      "learning_rate": 2.6917497062383776e-05,
+      "loss": 0.6073,
+      "step": 345
+    },
+    {
+      "epoch": 0.755939524838013,
+      "grad_norm": 1.3285808563232422,
+      "learning_rate": 2.680817591798619e-05,
+      "loss": 0.5839,
+      "step": 350
+    },
+    {
+      "epoch": 0.7667386609071274,
+      "grad_norm": 1.5056922435760498,
+      "learning_rate": 2.669718010897026e-05,
+      "loss": 0.6062,
+      "step": 355
+    },
+    {
+      "epoch": 0.7775377969762419,
+      "grad_norm": 1.3809500932693481,
+      "learning_rate": 2.6584525377035257e-05,
+      "loss": 0.5169,
+      "step": 360
+    },
+    {
+      "epoch": 0.7883369330453563,
+      "grad_norm": 1.443048357963562,
+      "learning_rate": 2.6470227699152975e-05,
+      "loss": 0.5469,
+      "step": 365
+    },
+    {
+      "epoch": 0.7991360691144709,
+      "grad_norm": 1.4121954441070557,
+      "learning_rate": 2.635430328530186e-05,
+      "loss": 0.5299,
+      "step": 370
+    },
+    {
+      "epoch": 0.8099352051835853,
+      "grad_norm": 1.4206424951553345,
+      "learning_rate": 2.6236768576168065e-05,
+      "loss": 0.4966,
+      "step": 375
+    },
+    {
+      "epoch": 0.8207343412526998,
+      "grad_norm": 1.384053111076355,
+      "learning_rate": 2.611764024081378e-05,
+      "loss": 0.4859,
+      "step": 380
+    },
+    {
+      "epoch": 0.8315334773218143,
+      "grad_norm": 1.4241434335708618,
+      "learning_rate": 2.5996935174313215e-05,
+      "loss": 0.5185,
+      "step": 385
+    },
+    {
+      "epoch": 0.8423326133909287,
+      "grad_norm": 1.6793551445007324,
+      "learning_rate": 2.587467049535645e-05,
+      "loss": 0.5017,
+      "step": 390
+    },
+    {
+      "epoch": 0.8531317494600432,
+      "grad_norm": 1.3198827505111694,
+      "learning_rate": 2.575086354382167e-05,
+      "loss": 0.542,
+      "step": 395
+    },
+    {
+      "epoch": 0.8639308855291576,
+      "grad_norm": 1.2480366230010986,
+      "learning_rate": 2.5625531878315956e-05,
+      "loss": 0.4781,
+      "step": 400
+    },
+    {
+      "epoch": 0.8747300215982722,
+      "grad_norm": 1.4425947666168213,
+      "learning_rate": 2.5498693273685074e-05,
+      "loss": 0.4863,
+      "step": 405
+    },
+    {
+      "epoch": 0.8855291576673866,
+      "grad_norm": 1.3835172653198242,
+      "learning_rate": 2.5370365718492615e-05,
+      "loss": 0.5024,
+      "step": 410
+    },
+    {
+      "epoch": 0.896328293736501,
+      "grad_norm": 1.6777316331863403,
+      "learning_rate": 2.5240567412468816e-05,
+      "loss": 0.4523,
+      "step": 415
+    },
+    {
+      "epoch": 0.9071274298056156,
+      "grad_norm": 1.44025719165802,
+      "learning_rate": 2.5109316763929414e-05,
+      "loss": 0.4735,
+      "step": 420
+    },
+    {
+      "epoch": 0.91792656587473,
+      "grad_norm": 1.880956768989563,
+      "learning_rate": 2.497663238716495e-05,
+      "loss": 0.5053,
+      "step": 425
+    },
+    {
+      "epoch": 0.9287257019438445,
+      "grad_norm": 1.4297730922698975,
+      "learning_rate": 2.4842533099800822e-05,
+      "loss": 0.4486,
+      "step": 430
+    },
+    {
+      "epoch": 0.9395248380129589,
+      "grad_norm": 1.486853837966919,
+      "learning_rate": 2.4707037920128554e-05,
+      "loss": 0.4447,
+      "step": 435
+    },
+    {
+      "epoch": 0.9503239740820735,
+      "grad_norm": 1.5644670724868774,
+      "learning_rate": 2.4570166064408556e-05,
+      "loss": 0.4751,
+      "step": 440
+    },
+    {
+      "epoch": 0.9611231101511879,
+      "grad_norm": 1.4592996835708618,
+      "learning_rate": 2.4431936944144818e-05,
+      "loss": 0.4417,
+      "step": 445
+    },
+    {
+      "epoch": 0.9719222462203023,
+      "grad_norm": 1.404414176940918,
+      "learning_rate": 2.4292370163331936e-05,
+      "loss": 0.4525,
+      "step": 450
+    },
+    {
+      "epoch": 0.9827213822894169,
+      "grad_norm": 1.3904056549072266,
+      "learning_rate": 2.4151485515674837e-05,
+      "loss": 0.4497,
+      "step": 455
+    },
+    {
+      "epoch": 0.9935205183585313,
+      "grad_norm": 1.280365228652954,
+      "learning_rate": 2.400930298178155e-05,
+      "loss": 0.4004,
+      "step": 460
+    },
+    {
+      "epoch": 1.0043196544276458,
+      "grad_norm": 1.3646249771118164,
+      "learning_rate": 2.3865842726329556e-05,
+      "loss": 0.3918,
+      "step": 465
+    },
+    {
+      "epoch": 1.0151187904967602,
+      "grad_norm": 1.408378005027771,
+      "learning_rate": 2.3721125095205935e-05,
+      "loss": 0.3471,
+      "step": 470
+    },
+    {
+      "epoch": 1.0259179265658747,
+      "grad_norm": 1.5114684104919434,
+      "learning_rate": 2.357517061262189e-05,
+      "loss": 0.331,
+      "step": 475
+    },
+    {
+      "epoch": 1.0367170626349893,
+      "grad_norm": 1.246845006942749,
+      "learning_rate": 2.3427999978201956e-05,
+      "loss": 0.3334,
+      "step": 480
+    },
+    {
+      "epoch": 1.0475161987041037,
+      "grad_norm": 1.423335313796997,
+      "learning_rate": 2.3279634064048308e-05,
+      "loss": 0.3685,
+      "step": 485
+    },
+    {
+      "epoch": 1.0583153347732182,
+      "grad_norm": 1.4037672281265259,
+      "learning_rate": 2.3130093911780642e-05,
+      "loss": 0.331,
+      "step": 490
+    },
+    {
+      "epoch": 1.0691144708423326,
+      "grad_norm": 1.58006751537323,
+      "learning_rate": 2.2979400729552014e-05,
+      "loss": 0.3729,
+      "step": 495
+    },
+    {
+      "epoch": 1.079913606911447,
+      "grad_norm": 1.4677419662475586,
+      "learning_rate": 2.2827575889041007e-05,
+      "loss": 0.3692,
+      "step": 500
+    },
+    {
+      "epoch": 1.0907127429805616,
+      "grad_norm": 1.3571957349777222,
+      "learning_rate": 2.267464092242078e-05,
+      "loss": 0.3087,
+      "step": 505
+    },
+    {
+      "epoch": 1.101511879049676,
+      "grad_norm": 1.375675082206726,
+      "learning_rate": 2.2520617519305325e-05,
+      "loss": 0.3481,
+      "step": 510
+    },
+    {
+      "epoch": 1.1123110151187905,
+      "grad_norm": 1.7425004243850708,
+      "learning_rate": 2.2365527523673368e-05,
+      "loss": 0.3203,
+      "step": 515
+    },
+    {
+      "epoch": 1.123110151187905,
+      "grad_norm": 1.4006694555282593,
+      "learning_rate": 2.2209392930770424e-05,
+      "loss": 0.2964,
+      "step": 520
+    },
+    {
+      "epoch": 1.1339092872570196,
+      "grad_norm": 1.701220989227295,
+      "learning_rate": 2.2052235883989355e-05,
+      "loss": 0.3061,
+      "step": 525
+    },
+    {
+      "epoch": 1.144708423326134,
+      "grad_norm": 1.4398974180221558,
+      "learning_rate": 2.189407867172997e-05,
+      "loss": 0.3401,
+      "step": 530
+    },
+    {
+      "epoch": 1.1555075593952484,
+      "grad_norm": 1.5313234329223633,
+      "learning_rate": 2.1734943724237986e-05,
+      "loss": 0.3134,
+      "step": 535
+    },
+    {
+      "epoch": 1.1663066954643628,
+      "grad_norm": 1.3739843368530273,
+      "learning_rate": 2.157485361042398e-05,
+      "loss": 0.3141,
+      "step": 540
+    },
+    {
+      "epoch": 1.1771058315334773,
+      "grad_norm": 1.5831795930862427,
+      "learning_rate": 2.1413831034662536e-05,
+      "loss": 0.3209,
+      "step": 545
+    },
+    {
+      "epoch": 1.187904967602592,
+      "grad_norm": 1.38355553150177,
+      "learning_rate": 2.1251898833572303e-05,
+      "loss": 0.3108,
+      "step": 550
+    },
+    {
+      "epoch": 1.1987041036717063,
+      "grad_norm": 1.3868389129638672,
+      "learning_rate": 2.1089079972777238e-05,
+      "loss": 0.2764,
+      "step": 555
+    },
+    {
+      "epoch": 1.2095032397408207,
+      "grad_norm": 1.5070290565490723,
+      "learning_rate": 2.092539754364957e-05,
+      "loss": 0.2763,
+      "step": 560
+    },
+    {
+      "epoch": 1.2203023758099352,
+      "grad_norm": 1.3210084438323975,
+      "learning_rate": 2.07608747600349e-05,
+      "loss": 0.2935,
+      "step": 565
+    },
+    {
+      "epoch": 1.2311015118790496,
+      "grad_norm": 1.3744834661483765,
+      "learning_rate": 2.0595534954959984e-05,
+      "loss": 0.2675,
+      "step": 570
+    },
+    {
+      "epoch": 1.2419006479481642,
+      "grad_norm": 1.4556934833526611,
+      "learning_rate": 2.0429401577323576e-05,
+      "loss": 0.2875,
+      "step": 575
+    },
+    {
+      "epoch": 1.2526997840172787,
+      "grad_norm": 1.522120714187622,
+      "learning_rate": 2.0262498188570855e-05,
+      "loss": 0.2936,
+      "step": 580
+    },
+    {
+      "epoch": 1.263498920086393,
+      "grad_norm": 1.5570476055145264,
+      "learning_rate": 2.0094848459351853e-05,
+      "loss": 0.281,
+      "step": 585
+    },
+    {
+      "epoch": 1.2742980561555075,
+      "grad_norm": 1.2775851488113403,
+      "learning_rate": 1.992647616616447e-05,
+      "loss": 0.269,
+      "step": 590
+    },
+    {
+      "epoch": 1.285097192224622,
+      "grad_norm": 1.4673560857772827,
+      "learning_rate": 1.9757405187982397e-05,
+      "loss": 0.2801,
+      "step": 595
+    },
+    {
+      "epoch": 1.2958963282937366,
+      "grad_norm": 1.5904041528701782,
+      "learning_rate": 1.9587659502868546e-05,
+      "loss": 0.268,
+      "step": 600
+    },
+    {
+      "epoch": 1.306695464362851,
+      "grad_norm": 1.3792083263397217,
+      "learning_rate": 1.9417263184574453e-05,
+      "loss": 0.2863,
+      "step": 605
+    },
+    {
+      "epoch": 1.3174946004319654,
+      "grad_norm": 1.3500524759292603,
+      "learning_rate": 1.9246240399126036e-05,
+      "loss": 0.2656,
+      "step": 610
+    },
+    {
+      "epoch": 1.3282937365010798,
+      "grad_norm": 1.5083184242248535,
+      "learning_rate": 1.907461540139633e-05,
+      "loss": 0.2749,
+      "step": 615
+    },
+    {
+      "epoch": 1.3390928725701943,
+      "grad_norm": 1.4029912948608398,
+      "learning_rate": 1.8902412531665613e-05,
+      "loss": 0.2703,
+      "step": 620
+    },
+    {
+      "epoch": 1.349892008639309,
+      "grad_norm": 1.3081443309783936,
+      "learning_rate": 1.872965621216938e-05,
+      "loss": 0.2534,
+      "step": 625
+    },
+    {
+      "epoch": 1.3606911447084233,
+      "grad_norm": 1.5683013200759888,
+      "learning_rate": 1.855637094363474e-05,
+      "loss": 0.2366,
+      "step": 630
+    },
+    {
+      "epoch": 1.3714902807775378,
+      "grad_norm": 1.2933757305145264,
+      "learning_rate": 1.8382581301805676e-05,
+      "loss": 0.2598,
+      "step": 635
+    },
+    {
+      "epoch": 1.3822894168466522,
+      "grad_norm": 1.1082271337509155,
+      "learning_rate": 1.8208311933957606e-05,
+      "loss": 0.2596,
+      "step": 640
+    },
+    {
+      "epoch": 1.3930885529157666,
+      "grad_norm": 1.7207790613174438,
+      "learning_rate": 1.80335875554019e-05,
+      "loss": 0.2538,
+      "step": 645
+    },
+    {
+      "epoch": 1.4038876889848813,
+      "grad_norm": 1.2548213005065918,
+      "learning_rate": 1.7858432945980645e-05,
+      "loss": 0.2234,
+      "step": 650
+    },
+    {
+      "epoch": 1.4146868250539957,
+      "grad_norm": 1.3207660913467407,
+      "learning_rate": 1.7682872946552352e-05,
+      "loss": 0.2588,
+      "step": 655
+    },
+    {
+      "epoch": 1.42548596112311,
+      "grad_norm": 1.4590904712677002,
+      "learning_rate": 1.750693245546893e-05,
+      "loss": 0.2191,
+      "step": 660
+    },
+    {
+      "epoch": 1.4362850971922247,
+      "grad_norm": 1.2964551448822021,
+      "learning_rate": 1.7330636425044553e-05,
+      "loss": 0.2564,
+      "step": 665
+    },
+    {
+      "epoch": 1.4470842332613392,
+      "grad_norm": 1.47450590133667,
+      "learning_rate": 1.7154009858016867e-05,
+      "loss": 0.2269,
+      "step": 670
+    },
+    {
+      "epoch": 1.4578833693304536,
+      "grad_norm": 1.328837275505066,
+      "learning_rate": 1.697707780400102e-05,
+      "loss": 0.2227,
+      "step": 675
+    },
+    {
+      "epoch": 1.468682505399568,
+      "grad_norm": 1.7263257503509521,
+      "learning_rate": 1.67998653559371e-05,
+      "loss": 0.2424,
+      "step": 680
+    },
+    {
+      "epoch": 1.4794816414686824,
+      "grad_norm": 1.1844679117202759,
+      "learning_rate": 1.6622397646531352e-05,
+      "loss": 0.2334,
+      "step": 685
+    },
+    {
+      "epoch": 1.490280777537797,
+      "grad_norm": 1.612632393836975,
+      "learning_rate": 1.644469984469182e-05,
+      "loss": 0.2282,
+      "step": 690
+    },
+    {
+      "epoch": 1.5010799136069113,
+      "grad_norm": 1.284920573234558,
+      "learning_rate": 1.6266797151958815e-05,
+      "loss": 0.2237,
+      "step": 695
+    },
+    {
+      "epoch": 1.511879049676026,
+      "grad_norm": 1.3669025897979736,
+      "learning_rate": 1.6088714798930806e-05,
+      "loss": 0.2606,
+      "step": 700
+    },
+    {
+      "epoch": 1.5226781857451404,
+      "grad_norm": 1.5955615043640137,
+      "learning_rate": 1.59104780416861e-05,
+      "loss": 0.2024,
+      "step": 705
+    },
+    {
+      "epoch": 1.5334773218142548,
+      "grad_norm": 1.303550362586975,
+      "learning_rate": 1.5732112158200995e-05,
+      "loss": 0.1914,
+      "step": 710
+    },
+    {
+      "epoch": 1.5442764578833694,
+      "grad_norm": 1.3817285299301147,
+      "learning_rate": 1.5553642444764808e-05,
+      "loss": 0.2429,
+      "step": 715
+    },
+    {
+      "epoch": 1.5550755939524838,
+      "grad_norm": 1.3321316242218018,
+      "learning_rate": 1.5375094212392253e-05,
+      "loss": 0.2218,
+      "step": 720
+    },
+    {
+      "epoch": 1.5658747300215983,
+      "grad_norm": 1.338300108909607,
+      "learning_rate": 1.5196492783233815e-05,
+      "loss": 0.2002,
+      "step": 725
+    },
+    {
+      "epoch": 1.5766738660907127,
+      "grad_norm": 1.3739572763442993,
+      "learning_rate": 1.501786348698446e-05,
+      "loss": 0.2005,
+      "step": 730
+    },
+    {
+      "epoch": 1.5874730021598271,
+      "grad_norm": 1.3973687887191772,
+      "learning_rate": 1.4839231657291343e-05,
+      "loss": 0.2254,
+      "step": 735
+    },
+    {
+      "epoch": 1.5982721382289418,
+      "grad_norm": 1.4655439853668213,
+      "learning_rate": 1.4660622628160921e-05,
+      "loss": 0.2028,
+      "step": 740
+    },
+    {
+      "epoch": 1.6090712742980562,
+      "grad_norm": 1.309818148612976,
+      "learning_rate": 1.4482061730366009e-05,
+      "loss": 0.1926,
+      "step": 745
+    },
+    {
+      "epoch": 1.6198704103671706,
+      "grad_norm": 1.3012702465057373,
+      "learning_rate": 1.4303574287853324e-05,
+      "loss": 0.1891,
+      "step": 750
+    },
+    {
+      "epoch": 1.6306695464362853,
+      "grad_norm": 1.3752199411392212,
+      "learning_rate": 1.4125185614151967e-05,
+      "loss": 0.1696,
+      "step": 755
+    },
+    {
+      "epoch": 1.6414686825053995,
+      "grad_norm": 1.3779516220092773,
+      "learning_rate": 1.3946921008783418e-05,
+      "loss": 0.1949,
+      "step": 760
+    },
+    {
+      "epoch": 1.652267818574514,
+      "grad_norm": 1.3249905109405518,
+      "learning_rate": 1.3768805753673465e-05,
+      "loss": 0.1843,
+      "step": 765
+    },
+    {
+      "epoch": 1.6630669546436285,
+      "grad_norm": 1.2693579196929932,
+      "learning_rate": 1.359086510956668e-05,
+      "loss": 0.1843,
+      "step": 770
+    },
+    {
+      "epoch": 1.673866090712743,
+      "grad_norm": 1.1808803081512451,
+      "learning_rate": 1.3413124312443874e-05,
+      "loss": 0.1762,
+      "step": 775
+    },
+    {
+      "epoch": 1.6846652267818576,
+      "grad_norm": 1.4530106782913208,
+      "learning_rate": 1.3235608569943059e-05,
+      "loss": 0.1694,
+      "step": 780
+    },
+    {
+      "epoch": 1.6954643628509718,
+      "grad_norm": 1.5051404237747192,
+      "learning_rate": 1.3058343057784458e-05,
+      "loss": 0.187,
+      "step": 785
+    },
+    {
+      "epoch": 1.7062634989200864,
+      "grad_norm": 1.2471327781677246,
+      "learning_rate": 1.2881352916199988e-05,
+      "loss": 0.1605,
+      "step": 790
+    },
+    {
+      "epoch": 1.7170626349892009,
+      "grad_norm": 1.6013789176940918,
+      "learning_rate": 1.270466324636786e-05,
+      "loss": 0.1847,
+      "step": 795
+    },
+    {
+      "epoch": 1.7278617710583153,
+      "grad_norm": 1.384282112121582,
+      "learning_rate": 1.252829910685263e-05,
+      "loss": 0.1791,
+      "step": 800
+    },
+    {
+      "epoch": 1.73866090712743,
+      "grad_norm": 1.2057247161865234,
+      "learning_rate": 1.2352285510051324e-05,
+      "loss": 0.1714,
+      "step": 805
+    },
+    {
+      "epoch": 1.7494600431965441,
+      "grad_norm": 1.1892156600952148,
+      "learning_rate": 1.2176647418646156e-05,
+      "loss": 0.1624,
+      "step": 810
+    },
+    {
+      "epoch": 1.7602591792656588,
+      "grad_norm": 1.2634958028793335,
+      "learning_rate": 1.2001409742064245e-05,
+      "loss": 0.1557,
+      "step": 815
+    },
+    {
+      "epoch": 1.7710583153347732,
+      "grad_norm": 1.4009325504302979,
+      "learning_rate": 1.1826597332944873e-05,
+      "loss": 0.1613,
+      "step": 820
+    },
+    {
+      "epoch": 1.7818574514038876,
+      "grad_norm": 1.1138142347335815,
+      "learning_rate": 1.1652234983614848e-05,
+      "loss": 0.1532,
+      "step": 825
+    },
+    {
+      "epoch": 1.7926565874730023,
+      "grad_norm": 1.3123793601989746,
+      "learning_rate": 1.1478347422572396e-05,
+      "loss": 0.1467,
+      "step": 830
+    },
+    {
+      "epoch": 1.8034557235421165,
+      "grad_norm": 1.1863819360733032,
+      "learning_rate": 1.1304959310980097e-05,
+      "loss": 0.1482,
+      "step": 835
+    },
+    {
+      "epoch": 1.8142548596112311,
+      "grad_norm": 1.1802626848220825,
+      "learning_rate": 1.113209523916737e-05,
+      "loss": 0.1355,
+      "step": 840
+    },
+    {
+      "epoch": 1.8250539956803455,
+      "grad_norm": 1.225582242012024,
+      "learning_rate": 1.0959779723143024e-05,
+      "loss": 0.1551,
+      "step": 845
+    },
+    {
+      "epoch": 1.83585313174946,
+      "grad_norm": 1.0997288227081299,
+      "learning_rate": 1.0788037201118346e-05,
+      "loss": 0.1584,
+      "step": 850
+    },
+    {
+      "epoch": 1.8466522678185746,
+      "grad_norm": 1.3584377765655518,
+      "learning_rate": 1.0616892030041184e-05,
+      "loss": 0.1498,
+      "step": 855
+    },
+    {
+      "epoch": 1.857451403887689,
+      "grad_norm": 1.1767578125,
+      "learning_rate": 1.0446368482141612e-05,
+      "loss": 0.1305,
+      "step": 860
+    },
+    {
+      "epoch": 1.8682505399568035,
+      "grad_norm": 1.2314656972885132,
+      "learning_rate": 1.027649074148956e-05,
+      "loss": 0.1493,
+      "step": 865
+    },
+    {
+      "epoch": 1.8790496760259179,
+      "grad_norm": 1.1895897388458252,
+      "learning_rate": 1.0107282900565009e-05,
+      "loss": 0.1517,
+      "step": 870
+    },
+    {
+      "epoch": 1.8898488120950323,
+      "grad_norm": 1.0842256546020508,
+      "learning_rate": 9.938768956841077e-06,
+      "loss": 0.152,
+      "step": 875
+    },
+    {
+      "epoch": 1.900647948164147,
+      "grad_norm": 1.24657142162323,
+      "learning_rate": 9.770972809380696e-06,
+      "loss": 0.1313,
+      "step": 880
+    },
+    {
+      "epoch": 1.9114470842332614,
+      "grad_norm": 0.9886346459388733,
+      "learning_rate": 9.603918255447141e-06,
+      "loss": 0.1278,
+      "step": 885
+    },
+    {
+      "epoch": 1.9222462203023758,
+      "grad_norm": 1.1505898237228394,
+      "learning_rate": 9.437628987129084e-06,
+      "loss": 0.145,
+      "step": 890
+    },
+    {
+      "epoch": 1.9330453563714904,
+      "grad_norm": 1.2165902853012085,
+      "learning_rate": 9.272128587980498e-06,
+      "loss": 0.1464,
+      "step": 895
+    },
+    {
+      "epoch": 1.9438444924406046,
+      "grad_norm": 1.154052972793579,
+      "learning_rate": 9.107440529675971e-06,
+      "loss": 0.1231,
+      "step": 900
+    },
+    {
+      "epoch": 1.9546436285097193,
+      "grad_norm": 1.1554679870605469,
+      "learning_rate": 8.943588168681937e-06,
+      "loss": 0.147,
+      "step": 905
+    },
+    {
+      "epoch": 1.9654427645788337,
+      "grad_norm": 1.1286513805389404,
+      "learning_rate": 8.780594742944159e-06,
+      "loss": 0.1224,
+      "step": 910
+    },
+    {
+      "epoch": 1.9762419006479481,
+      "grad_norm": 1.0984737873077393,
+      "learning_rate": 8.618483368592088e-06,
+      "loss": 0.1183,
+      "step": 915
+    },
+    {
+      "epoch": 1.9870410367170628,
+      "grad_norm": 1.2652884721755981,
+      "learning_rate": 8.457277036660463e-06,
+      "loss": 0.134,
+      "step": 920
+    },
+    {
+      "epoch": 1.997840172786177,
+      "grad_norm": 0.9271209239959717,
+      "learning_rate": 8.296998609828704e-06,
+      "loss": 0.1336,
+      "step": 925
+    },
+    {
+      "epoch": 2.0086393088552916,
+      "grad_norm": 1.0432496070861816,
+      "learning_rate": 8.13767081917841e-06,
+      "loss": 0.1097,
+      "step": 930
+    },
+    {
+      "epoch": 2.019438444924406,
+      "grad_norm": 1.0845433473587036,
+      "learning_rate": 7.97931626096964e-06,
+      "loss": 0.0992,
+      "step": 935
+    },
+    {
+      "epoch": 2.0302375809935205,
+      "grad_norm": 1.0280168056488037,
+      "learning_rate": 7.821957393436202e-06,
+      "loss": 0.1189,
+      "step": 940
+    },
+    {
+      "epoch": 2.041036717062635,
+      "grad_norm": 1.010212779045105,
+      "learning_rate": 7.665616533600628e-06,
+      "loss": 0.1008,
+      "step": 945
+    },
+    {
+      "epoch": 2.0518358531317493,
+      "grad_norm": 1.1928298473358154,
+      "learning_rate": 7.5103158541090665e-06,
+      "loss": 0.107,
+      "step": 950
+    },
+    {
+      "epoch": 2.062634989200864,
+      "grad_norm": 0.9925423264503479,
+      "learning_rate": 7.356077380086726e-06,
+      "loss": 0.0902,
+      "step": 955
+    },
+    {
+      "epoch": 2.0734341252699786,
+      "grad_norm": 1.107944369316101,
+      "learning_rate": 7.202922986014228e-06,
+      "loss": 0.105,
+      "step": 960
+    },
+    {
+      "epoch": 2.084233261339093,
+      "grad_norm": 1.0484154224395752,
+      "learning_rate": 7.050874392625302e-06,
+      "loss": 0.1129,
+      "step": 965
+    },
+    {
+      "epoch": 2.0950323974082075,
+      "grad_norm": 1.0642160177230835,
+      "learning_rate": 6.899953163826292e-06,
+      "loss": 0.1069,
+      "step": 970
+    },
+    {
+      "epoch": 2.1058315334773217,
+      "grad_norm": 1.1226906776428223,
+      "learning_rate": 6.7501807036379125e-06,
+      "loss": 0.1007,
+      "step": 975
+    },
+    {
+      "epoch": 2.1166306695464363,
+      "grad_norm": 1.067341685295105,
+      "learning_rate": 6.601578253159698e-06,
+      "loss": 0.0871,
+      "step": 980
+    },
+    {
+      "epoch": 2.127429805615551,
+      "grad_norm": 0.9834439754486084,
+      "learning_rate": 6.454166887557508e-06,
+      "loss": 0.1012,
+      "step": 985
+    },
+    {
+      "epoch": 2.138228941684665,
+      "grad_norm": 1.065042495727539,
+      "learning_rate": 6.307967513074605e-06,
+      "loss": 0.106,
+      "step": 990
+    },
+    {
+      "epoch": 2.14902807775378,
+      "grad_norm": 0.9570000171661377,
+      "learning_rate": 6.163000864066698e-06,
+      "loss": 0.0944,
+      "step": 995
+    },
+    {
+      "epoch": 2.159827213822894,
+      "grad_norm": 1.0731011629104614,
+      "learning_rate": 6.019287500061326e-06,
+      "loss": 0.087,
+      "step": 1000
+    },
+    {
+      "epoch": 2.1706263498920086,
+      "grad_norm": 0.9897378087043762,
+      "learning_rate": 5.876847802842052e-06,
+      "loss": 0.1039,
+      "step": 1005
+    },
+    {
+      "epoch": 2.1814254859611233,
+      "grad_norm": 0.9978522658348083,
+      "learning_rate": 5.73570197355788e-06,
+      "loss": 0.0947,
+      "step": 1010
+    },
+    {
+      "epoch": 2.1922246220302375,
+      "grad_norm": 1.2583086490631104,
+      "learning_rate": 5.595870029858268e-06,
+      "loss": 0.1082,
+      "step": 1015
+    },
+    {
+      "epoch": 2.203023758099352,
+      "grad_norm": 0.9107490181922913,
+      "learning_rate": 5.4573718030541965e-06,
+      "loss": 0.0982,
+      "step": 1020
+    },
+    {
+      "epoch": 2.2138228941684663,
+      "grad_norm": 0.9486368894577026,
+      "learning_rate": 5.320226935305609e-06,
+      "loss": 0.0906,
+      "step": 1025
+    },
+    {
+      "epoch": 2.224622030237581,
+      "grad_norm": 1.0325815677642822,
+      "learning_rate": 5.184454876835746e-06,
+      "loss": 0.0978,
+      "step": 1030
+    },
+    {
+      "epoch": 2.2354211663066956,
+      "grad_norm": 0.9269861578941345,
+      "learning_rate": 5.05007488317265e-06,
+      "loss": 0.0924,
+      "step": 1035
+    },
+    {
+      "epoch": 2.24622030237581,
+      "grad_norm": 0.7919679284095764,
+      "learning_rate": 4.917106012418294e-06,
+      "loss": 0.0824,
+      "step": 1040
+    },
+    {
+      "epoch": 2.2570194384449245,
+      "grad_norm": 1.0633175373077393,
+      "learning_rate": 4.7855671225457225e-06,
+      "loss": 0.0874,
+      "step": 1045
+    },
+    {
+      "epoch": 2.267818574514039,
+      "grad_norm": 0.7829254269599915,
+      "learning_rate": 4.655476868724566e-06,
+      "loss": 0.0904,
+      "step": 1050
+    },
+    {
+      "epoch": 2.2786177105831533,
+      "grad_norm": 1.023898720741272,
+      "learning_rate": 4.526853700675325e-06,
+      "loss": 0.0958,
+      "step": 1055
+    },
+    {
+      "epoch": 2.289416846652268,
+      "grad_norm": 0.7806884050369263,
+      "learning_rate": 4.399715860052781e-06,
+      "loss": 0.0928,
+      "step": 1060
+    },
+    {
+      "epoch": 2.300215982721382,
+      "grad_norm": 0.849102258682251,
+      "learning_rate": 4.274081377858909e-06,
+      "loss": 0.0928,
+      "step": 1065
+    },
+    {
+      "epoch": 2.311015118790497,
+      "grad_norm": 0.8940497040748596,
+      "learning_rate": 4.149968071885682e-06,
+      "loss": 0.1011,
+      "step": 1070
+    },
+    {
+      "epoch": 2.3218142548596115,
+      "grad_norm": 1.3051635026931763,
+      "learning_rate": 4.027393544188129e-06,
+      "loss": 0.0784,
+      "step": 1075
+    },
+    {
+      "epoch": 2.3326133909287257,
+      "grad_norm": 1.0761061906814575,
+      "learning_rate": 3.9063751785879365e-06,
+      "loss": 0.0827,
+      "step": 1080
+    },
+    {
+      "epoch": 2.3434125269978403,
+      "grad_norm": 0.9575039744377136,
+      "learning_rate": 3.786930138208046e-06,
+      "loss": 0.0935,
+      "step": 1085
+    },
+    {
+      "epoch": 2.3542116630669545,
+      "grad_norm": 0.7994723320007324,
+      "learning_rate": 3.6690753630385436e-06,
+      "loss": 0.0865,
+      "step": 1090
+    },
+    {
+      "epoch": 2.365010799136069,
+      "grad_norm": 0.8735499382019043,
+      "learning_rate": 3.552827567534186e-06,
+      "loss": 0.0866,
+      "step": 1095
+    },
+    {
+      "epoch": 2.375809935205184,
+      "grad_norm": 1.0202031135559082,
+      "learning_rate": 3.4382032382439036e-06,
+      "loss": 0.0827,
+      "step": 1100
+    },
+    {
+      "epoch": 2.386609071274298,
+      "grad_norm": 0.8190648555755615,
+      "learning_rate": 3.3252186314726447e-06,
+      "loss": 0.0826,
+      "step": 1105
+    },
+    {
+      "epoch": 2.3974082073434126,
+      "grad_norm": 0.8609358668327332,
+      "learning_rate": 3.213889770975881e-06,
+      "loss": 0.0813,
+      "step": 1110
+    },
+    {
+      "epoch": 2.408207343412527,
+      "grad_norm": 0.8619616627693176,
+      "learning_rate": 3.1042324456870514e-06,
+      "loss": 0.0702,
+      "step": 1115
+    },
+    {
+      "epoch": 2.4190064794816415,
+      "grad_norm": 0.7532960176467896,
+      "learning_rate": 2.996262207478356e-06,
+      "loss": 0.0742,
+      "step": 1120
+    },
+    {
+      "epoch": 2.429805615550756,
+      "grad_norm": 0.9083415269851685,
+      "learning_rate": 2.8899943689551493e-06,
+      "loss": 0.0894,
+      "step": 1125
+    },
+    {
+      "epoch": 2.4406047516198703,
+      "grad_norm": 0.8385311961174011,
+      "learning_rate": 2.7854440012842684e-06,
+      "loss": 0.0818,
+      "step": 1130
+    },
+    {
+      "epoch": 2.451403887688985,
+      "grad_norm": 0.9755520820617676,
+      "learning_rate": 2.6826259320565903e-06,
+      "loss": 0.0895,
+      "step": 1135
+    },
+    {
+      "epoch": 2.462203023758099,
+      "grad_norm": 0.9728091955184937,
+      "learning_rate": 2.581554743184158e-06,
+      "loss": 0.0754,
+      "step": 1140
+    },
+    {
+      "epoch": 2.473002159827214,
+      "grad_norm": 0.7588487863540649,
+      "learning_rate": 2.482244768832146e-06,
+      "loss": 0.0767,
+      "step": 1145
+    },
+    {
+      "epoch": 2.4838012958963285,
+      "grad_norm": 0.822422981262207,
+      "learning_rate": 2.384710093385929e-06,
+      "loss": 0.0725,
+      "step": 1150
+    },
+    {
+      "epoch": 2.4946004319654427,
+      "grad_norm": 1.2029268741607666,
+      "learning_rate": 2.288964549453633e-06,
+      "loss": 0.0824,
+      "step": 1155
+    },
+    {
+      "epoch": 2.5053995680345573,
+      "grad_norm": 0.8733217716217041,
+      "learning_rate": 2.1950217159043197e-06,
+      "loss": 0.076,
+      "step": 1160
+    },
+    {
+      "epoch": 2.5161987041036715,
+      "grad_norm": 0.9950538873672485,
+      "learning_rate": 2.102894915942243e-06,
+      "loss": 0.0938,
+      "step": 1165
+    },
+    {
+      "epoch": 2.526997840172786,
+      "grad_norm": 0.7292869687080383,
+      "learning_rate": 2.0125972152172777e-06,
+      "loss": 0.0823,
+      "step": 1170
+    },
+    {
+      "epoch": 2.537796976241901,
+      "grad_norm": 0.7431409358978271,
+      "learning_rate": 1.924141419971949e-06,
+      "loss": 0.0821,
+      "step": 1175
+    },
+    {
+      "epoch": 2.548596112311015,
+      "grad_norm": 0.8313660025596619,
+      "learning_rate": 1.8375400752251943e-06,
+      "loss": 0.0772,
+      "step": 1180
+    },
+    {
+      "epoch": 2.5593952483801297,
+      "grad_norm": 0.9040440320968628,
+      "learning_rate": 1.75280546299322e-06,
+      "loss": 0.0902,
+      "step": 1185
+    },
+    {
+      "epoch": 2.570194384449244,
+      "grad_norm": 0.6281514763832092,
+      "learning_rate": 1.6699496005476244e-06,
+      "loss": 0.0722,
+      "step": 1190
+    },
+    {
+      "epoch": 2.5809935205183585,
+      "grad_norm": 0.8100899457931519,
+      "learning_rate": 1.5889842387110742e-06,
+      "loss": 0.0781,
+      "step": 1195
+    },
+    {
+      "epoch": 2.591792656587473,
+      "grad_norm": 0.9115588068962097,
+      "learning_rate": 1.5099208601907777e-06,
+      "loss": 0.0854,
+      "step": 1200
+    },
+    {
+      "epoch": 2.6025917926565874,
+      "grad_norm": 0.7780690789222717,
+      "learning_rate": 1.4327706779499927e-06,
+      "loss": 0.0753,
+      "step": 1205
+    },
+    {
+      "epoch": 2.613390928725702,
+      "grad_norm": 0.7972307801246643,
+      "learning_rate": 1.3575446336177506e-06,
+      "loss": 0.0746,
+      "step": 1210
+    },
+    {
+      "epoch": 2.624190064794816,
+      "grad_norm": 0.772161066532135,
+      "learning_rate": 1.2842533959371066e-06,
+      "loss": 0.0769,
+      "step": 1215
+    },
+    {
+      "epoch": 2.634989200863931,
+      "grad_norm": 0.8315443992614746,
+      "learning_rate": 1.2129073592520685e-06,
+      "loss": 0.0786,
+      "step": 1220
+    },
+    {
+      "epoch": 2.6457883369330455,
+      "grad_norm": 0.6547254323959351,
+      "learning_rate": 1.1435166420334436e-06,
+      "loss": 0.0767,
+      "step": 1225
+    },
+    {
+      "epoch": 2.6565874730021597,
+      "grad_norm": 1.0073431730270386,
+      "learning_rate": 1.0760910854438104e-06,
+      "loss": 0.0864,
+      "step": 1230
+    },
+    {
+      "epoch": 2.6673866090712743,
+      "grad_norm": 0.7978554964065552,
+      "learning_rate": 1.0106402519418173e-06,
+      "loss": 0.0848,
+      "step": 1235
+    },
+    {
+      "epoch": 2.6781857451403885,
+      "grad_norm": 0.6678413152694702,
+      "learning_rate": 9.471734239260288e-07,
+      "loss": 0.0688,
+      "step": 1240
+    },
+    {
+      "epoch": 2.688984881209503,
+      "grad_norm": 0.7504372000694275,
+      "learning_rate": 8.856996024184477e-07,
+      "loss": 0.0818,
+      "step": 1245
+    },
+    {
+      "epoch": 2.699784017278618,
+      "grad_norm": 0.7164337038993835,
+      "learning_rate": 8.262275057879926e-07,
+      "loss": 0.0774,
+      "step": 1250
+    },
+    {
+      "epoch": 2.710583153347732,
+      "grad_norm": 0.9779824018478394,
+      "learning_rate": 7.6876556851401e-07,
+      "loss": 0.0752,
+      "step": 1255
+    },
+    {
+      "epoch": 2.7213822894168467,
+      "grad_norm": 0.8393222093582153,
+      "learning_rate": 7.133219399901097e-07,
+      "loss": 0.0844,
+      "step": 1260
+    },
+    {
+      "epoch": 2.732181425485961,
+      "grad_norm": 0.6940068602561951,
+      "learning_rate": 6.599044833683632e-07,
+      "loss": 0.0693,
+      "step": 1265
+    },
+    {
+      "epoch": 2.7429805615550755,
+      "grad_norm": 0.8789917230606079,
+      "learning_rate": 6.085207744441529e-07,
+      "loss": 0.0764,
+      "step": 1270
+    },
+    {
+      "epoch": 2.75377969762419,
+      "grad_norm": 0.8124670386314392,
+      "learning_rate": 5.591781005817542e-07,
+      "loss": 0.076,
+      "step": 1275
+    },
+    {
+      "epoch": 2.7645788336933044,
+      "grad_norm": 0.8622475266456604,
+      "learning_rate": 5.11883459680812e-07,
+      "loss": 0.0819,
+      "step": 1280
+    },
+    {
+      "epoch": 2.775377969762419,
+      "grad_norm": 0.7363976240158081,
+      "learning_rate": 4.6664355918389244e-07,
+      "loss": 0.0778,
+      "step": 1285
+    },
+    {
+      "epoch": 2.786177105831533,
+      "grad_norm": 0.7811018824577332,
+      "learning_rate": 4.234648151252063e-07,
+      "loss": 0.0718,
+      "step": 1290
+    },
+    {
+      "epoch": 2.796976241900648,
+      "grad_norm": 0.7207102179527283,
+      "learning_rate": 3.823533512206845e-07,
+      "loss": 0.0735,
+      "step": 1295
+    },
+    {
+      "epoch": 2.8077753779697625,
+      "grad_norm": 0.8438379764556885,
+      "learning_rate": 3.4331499799948484e-07,
+      "loss": 0.0727,
+      "step": 1300
+    },
+    {
+      "epoch": 2.818574514038877,
+      "grad_norm": 0.8162416219711304,
+      "learning_rate": 3.063552919770984e-07,
+      "loss": 0.0726,
+      "step": 1305
+    },
+    {
+      "epoch": 2.8293736501079914,
+      "grad_norm": 0.6508810520172119,
+      "learning_rate": 2.7147947487014434e-07,
+      "loss": 0.0709,
+      "step": 1310
+    },
+    {
+      "epoch": 2.8401727861771056,
+      "grad_norm": 0.7725307941436768,
+      "learning_rate": 2.38692492852986e-07,
+      "loss": 0.0762,
+      "step": 1315
+    },
+    {
+      "epoch": 2.85097192224622,
+      "grad_norm": 0.7084739804267883,
+      "learning_rate": 2.0799899585623894e-07,
+      "loss": 0.0705,
+      "step": 1320
+    },
+    {
+      "epoch": 2.861771058315335,
+      "grad_norm": 0.825351357460022,
+      "learning_rate": 1.7940333690731004e-07,
+      "loss": 0.0769,
+      "step": 1325
+    },
+    {
+      "epoch": 2.8725701943844495,
+      "grad_norm": 0.6752949357032776,
+      "learning_rate": 1.5290957151304795e-07,
+      "loss": 0.075,
+      "step": 1330
+    },
+    {
+      "epoch": 2.8833693304535637,
+      "grad_norm": 0.9011345505714417,
+      "learning_rate": 1.2852145708457498e-07,
+      "loss": 0.0717,
+      "step": 1335
+    },
+    {
+      "epoch": 2.8941684665226783,
+      "grad_norm": 0.6693414449691772,
+      "learning_rate": 1.0624245240439811e-07,
+      "loss": 0.071,
+      "step": 1340
+    },
+    {
+      "epoch": 2.9049676025917925,
+      "grad_norm": 0.6685000658035278,
+      "learning_rate": 8.607571713588502e-08,
+      "loss": 0.0797,
+      "step": 1345
+    },
+    {
+      "epoch": 2.915766738660907,
+      "grad_norm": 0.8162562251091003,
+      "learning_rate": 6.802411137514897e-08,
+      "loss": 0.077,
+      "step": 1350
+    },
+    {
+      "epoch": 2.926565874730022,
+      "grad_norm": 0.6770502328872681,
+      "learning_rate": 5.20901952454167e-08,
+      "loss": 0.0718,
+      "step": 1355
+    },
+    {
+      "epoch": 2.937365010799136,
+      "grad_norm": 0.6245712637901306,
+      "learning_rate": 3.827622853395551e-08,
+      "loss": 0.072,
+      "step": 1360
+    },
+    {
+      "epoch": 2.9481641468682507,
+      "grad_norm": 0.693945586681366,
+      "learning_rate": 2.6584170371578008e-08,
+      "loss": 0.0689,
+      "step": 1365
+    },
+    {
+      "epoch": 2.958963282937365,
+      "grad_norm": 0.6926335096359253,
+      "learning_rate": 1.701567895479761e-08,
+      "loss": 0.0698,
+      "step": 1370
+    },
+    {
+      "epoch": 2.9697624190064795,
+      "grad_norm": 0.8302611112594604,
+      "learning_rate": 9.572111310653387e-09,
+      "loss": 0.0742,
+      "step": 1375
+    },
+    {
+      "epoch": 2.980561555075594,
+      "grad_norm": 0.7730148434638977,
+      "learning_rate": 4.254523104260666e-09,
+      "loss": 0.0758,
+      "step": 1380
+    },
+    {
+      "epoch": 2.9913606911447084,
+      "grad_norm": 0.8508152365684509,
+      "learning_rate": 1.0636684890874548e-09,
+      "loss": 0.0694,
+      "step": 1385
+    }
+  ],
+  "logging_steps": 5,
+  "max_steps": 1389,
+  "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": 1.7559409401465405e+18,
+  "train_batch_size": 2,
+  "trial_name": null,
+  "trial_params": null
+}

145_128_e3_3e-5/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dc17d81d75a769efb05eb561c226f1908c442e84ace6f82dcb6f751ff6455e8f
+size 8209

145_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)