Upload folder using huggingface_hub

129_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": [
+    "down_proj",
+    "k_proj",
+    "q_proj",
+    "o_proj",
+    "gate_proj",
+    "up_proj",
+    "v_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_rslora": false
+}

129_128_e3_3e-5/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5a5341ff6cf120a268ae7e510b955f1bdf3d84c27e983ffe4589339c05a37e2d
+size 791751704

129_128_e3_3e-5/latest

@@ -0,0 +1 @@
+global_step1035

129_128_e3_3e-5/rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7b1ee69b5e6941ec9d960860496745ec3c9df711350416598e9e00855c092a6a
+size 16389

129_128_e3_3e-5/rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:152eeaeb9f019cd9b3402f066b8f50993997c38f9ca7a870000c9b6967cd8e77
+size 16389

129_128_e3_3e-5/rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b3c63d8c627092e27a5a505c7e48a2aa3a5837847b99d1e64e84d6608f1262c7
+size 16389

129_128_e3_3e-5/rng_state_3.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:27b6b65ca906e6dca2c1fa0d5a31bf6c9b06fb0b8284b851c773df7da3fdb8e2
+size 16389

129_128_e3_3e-5/rng_state_4.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c447bb97e6ee9bd2f397b5eea567cda71ac6bd2b467b082c04cbbf77e6db65ff
+size 16389

129_128_e3_3e-5/rng_state_5.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b393ca7b0355c5063d036cd8f67b254ac23bd32b3c165d32b38839e5c43507ec
+size 16389

129_128_e3_3e-5/rng_state_6.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:150a22168929719880a1acde8c84d78003cf6651969e2045261fdfda24e3afaf
+size 16389

129_128_e3_3e-5/rng_state_7.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6aafeea5eb757e12d610be1b0ec028e4a31ec152590279f11f106253d4306b93
+size 16389

129_128_e3_3e-5/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ae3a2c931f69f69da991845d965fd769c51dc5a9c968d4fe0bbb9872018c10e8
+size 1401

129_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
+  }
+}

129_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
+}

129_128_e3_3e-5/trainer_state.json

@@ -0,0 +1,1483 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.0,
+  "eval_steps": 500,
+  "global_step": 1035,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.014492753623188406,
+      "grad_norm": 0.9208874106407166,
+      "learning_rate": 2.307692307692308e-06,
+      "loss": 1.3214,
+      "step": 5
+    },
+    {
+      "epoch": 0.028985507246376812,
+      "grad_norm": 0.7672443985939026,
+      "learning_rate": 5.192307692307692e-06,
+      "loss": 1.3099,
+      "step": 10
+    },
+    {
+      "epoch": 0.043478260869565216,
+      "grad_norm": 0.6573767066001892,
+      "learning_rate": 8.076923076923077e-06,
+      "loss": 1.2703,
+      "step": 15
+    },
+    {
+      "epoch": 0.057971014492753624,
+      "grad_norm": 0.630136251449585,
+      "learning_rate": 1.0961538461538462e-05,
+      "loss": 1.2804,
+      "step": 20
+    },
+    {
+      "epoch": 0.07246376811594203,
+      "grad_norm": 0.5200535655021667,
+      "learning_rate": 1.3846153846153847e-05,
+      "loss": 1.2294,
+      "step": 25
+    },
+    {
+      "epoch": 0.08695652173913043,
+      "grad_norm": 0.5240834355354309,
+      "learning_rate": 1.673076923076923e-05,
+      "loss": 1.2896,
+      "step": 30
+    },
+    {
+      "epoch": 0.10144927536231885,
+      "grad_norm": 0.43760108947753906,
+      "learning_rate": 1.9615384615384617e-05,
+      "loss": 1.2526,
+      "step": 35
+    },
+    {
+      "epoch": 0.11594202898550725,
+      "grad_norm": 0.46089237928390503,
+      "learning_rate": 2.25e-05,
+      "loss": 1.2231,
+      "step": 40
+    },
+    {
+      "epoch": 0.13043478260869565,
+      "grad_norm": 0.6038466095924377,
+      "learning_rate": 2.5384615384615386e-05,
+      "loss": 1.1802,
+      "step": 45
+    },
+    {
+      "epoch": 0.14492753623188406,
+      "grad_norm": 0.46358007192611694,
+      "learning_rate": 2.8269230769230768e-05,
+      "loss": 1.1359,
+      "step": 50
+    },
+    {
+      "epoch": 0.15942028985507245,
+      "grad_norm": 0.5262047648429871,
+      "learning_rate": 2.9999693583261366e-05,
+      "loss": 1.1746,
+      "step": 55
+    },
+    {
+      "epoch": 0.17391304347826086,
+      "grad_norm": 0.5417841672897339,
+      "learning_rate": 2.999624653872102e-05,
+      "loss": 1.1513,
+      "step": 60
+    },
+    {
+      "epoch": 0.18840579710144928,
+      "grad_norm": 0.5249184966087341,
+      "learning_rate": 2.9988970311828866e-05,
+      "loss": 1.0661,
+      "step": 65
+    },
+    {
+      "epoch": 0.2028985507246377,
+      "grad_norm": 0.6231042146682739,
+      "learning_rate": 2.9977866760516453e-05,
+      "loss": 1.092,
+      "step": 70
+    },
+    {
+      "epoch": 0.21739130434782608,
+      "grad_norm": 0.6113649010658264,
+      "learning_rate": 2.9962938719994754e-05,
+      "loss": 1.0935,
+      "step": 75
+    },
+    {
+      "epoch": 0.2318840579710145,
+      "grad_norm": 0.6280838847160339,
+      "learning_rate": 2.9944190002030212e-05,
+      "loss": 1.0389,
+      "step": 80
+    },
+    {
+      "epoch": 0.2463768115942029,
+      "grad_norm": 0.6393106579780579,
+      "learning_rate": 2.9921625393971432e-05,
+      "loss": 1.0185,
+      "step": 85
+    },
+    {
+      "epoch": 0.2608695652173913,
+      "grad_norm": 0.6914243698120117,
+      "learning_rate": 2.9895250657526762e-05,
+      "loss": 0.9935,
+      "step": 90
+    },
+    {
+      "epoch": 0.2753623188405797,
+      "grad_norm": 0.7685122489929199,
+      "learning_rate": 2.9865072527293107e-05,
+      "loss": 0.9555,
+      "step": 95
+    },
+    {
+      "epoch": 0.2898550724637681,
+      "grad_norm": 0.7435173392295837,
+      "learning_rate": 2.983109870903625e-05,
+      "loss": 0.9597,
+      "step": 100
+    },
+    {
+      "epoch": 0.30434782608695654,
+      "grad_norm": 0.7294394969940186,
+      "learning_rate": 2.97933378777233e-05,
+      "loss": 0.9636,
+      "step": 105
+    },
+    {
+      "epoch": 0.3188405797101449,
+      "grad_norm": 0.7624271512031555,
+      "learning_rate": 2.9751799675307554e-05,
+      "loss": 0.8448,
+      "step": 110
+    },
+    {
+      "epoch": 0.3333333333333333,
+      "grad_norm": 0.7932360172271729,
+      "learning_rate": 2.9706494708266526e-05,
+      "loss": 0.9143,
+      "step": 115
+    },
+    {
+      "epoch": 0.34782608695652173,
+      "grad_norm": 0.8719151616096497,
+      "learning_rate": 2.965743454489363e-05,
+      "loss": 0.8836,
+      "step": 120
+    },
+    {
+      "epoch": 0.36231884057971014,
+      "grad_norm": 0.9567369222640991,
+      "learning_rate": 2.9604631712344344e-05,
+      "loss": 0.8437,
+      "step": 125
+    },
+    {
+      "epoch": 0.37681159420289856,
+      "grad_norm": 0.8909892439842224,
+      "learning_rate": 2.9548099693437444e-05,
+      "loss": 0.8801,
+      "step": 130
+    },
+    {
+      "epoch": 0.391304347826087,
+      "grad_norm": 0.9535673260688782,
+      "learning_rate": 2.9487852923212293e-05,
+      "loss": 0.7937,
+      "step": 135
+    },
+    {
+      "epoch": 0.4057971014492754,
+      "grad_norm": 0.9305369853973389,
+      "learning_rate": 2.942390678524296e-05,
+      "loss": 0.8069,
+      "step": 140
+    },
+    {
+      "epoch": 0.42028985507246375,
+      "grad_norm": 1.0312312841415405,
+      "learning_rate": 2.935627760771012e-05,
+      "loss": 0.7559,
+      "step": 145
+    },
+    {
+      "epoch": 0.43478260869565216,
+      "grad_norm": 0.9863070249557495,
+      "learning_rate": 2.9284982659231786e-05,
+      "loss": 0.7702,
+      "step": 150
+    },
+    {
+      "epoch": 0.4492753623188406,
+      "grad_norm": 0.9744274020195007,
+      "learning_rate": 2.9210040144453882e-05,
+      "loss": 0.7468,
+      "step": 155
+    },
+    {
+      "epoch": 0.463768115942029,
+      "grad_norm": 1.0713469982147217,
+      "learning_rate": 2.9131469199401834e-05,
+      "loss": 0.7621,
+      "step": 160
+    },
+    {
+      "epoch": 0.4782608695652174,
+      "grad_norm": 1.2604933977127075,
+      "learning_rate": 2.904928988659431e-05,
+      "loss": 0.7128,
+      "step": 165
+    },
+    {
+      "epoch": 0.4927536231884058,
+      "grad_norm": 1.147759199142456,
+      "learning_rate": 2.896352318992041e-05,
+      "loss": 0.7373,
+      "step": 170
+    },
+    {
+      "epoch": 0.5072463768115942,
+      "grad_norm": 1.257650375366211,
+      "learning_rate": 2.887419100928158e-05,
+      "loss": 0.7357,
+      "step": 175
+    },
+    {
+      "epoch": 0.5217391304347826,
+      "grad_norm": 0.9941902756690979,
+      "learning_rate": 2.8781316154999635e-05,
+      "loss": 0.616,
+      "step": 180
+    },
+    {
+      "epoch": 0.5362318840579711,
+      "grad_norm": 1.1878472566604614,
+      "learning_rate": 2.8684922341992308e-05,
+      "loss": 0.6409,
+      "step": 185
+    },
+    {
+      "epoch": 0.5507246376811594,
+      "grad_norm": 1.050559163093567,
+      "learning_rate": 2.858503418371781e-05,
+      "loss": 0.6658,
+      "step": 190
+    },
+    {
+      "epoch": 0.5652173913043478,
+      "grad_norm": 1.074332594871521,
+      "learning_rate": 2.8481677185889985e-05,
+      "loss": 0.7082,
+      "step": 195
+    },
+    {
+      "epoch": 0.5797101449275363,
+      "grad_norm": 1.3266993761062622,
+      "learning_rate": 2.837487773996558e-05,
+      "loss": 0.6929,
+      "step": 200
+    },
+    {
+      "epoch": 0.5942028985507246,
+      "grad_norm": 1.1883240938186646,
+      "learning_rate": 2.8264663116405413e-05,
+      "loss": 0.6083,
+      "step": 205
+    },
+    {
+      "epoch": 0.6086956521739131,
+      "grad_norm": 1.1964757442474365,
+      "learning_rate": 2.8151061457711023e-05,
+      "loss": 0.6519,
+      "step": 210
+    },
+    {
+      "epoch": 0.6231884057971014,
+      "grad_norm": 1.4452741146087646,
+      "learning_rate": 2.803410177123872e-05,
+      "loss": 0.608,
+      "step": 215
+    },
+    {
+      "epoch": 0.6376811594202898,
+      "grad_norm": 1.1859878301620483,
+      "learning_rate": 2.7913813921792755e-05,
+      "loss": 0.6188,
+      "step": 220
+    },
+    {
+      "epoch": 0.6521739130434783,
+      "grad_norm": 1.0713025331497192,
+      "learning_rate": 2.7790228623999554e-05,
+      "loss": 0.5987,
+      "step": 225
+    },
+    {
+      "epoch": 0.6666666666666666,
+      "grad_norm": 1.199990153312683,
+      "learning_rate": 2.766337743446496e-05,
+      "loss": 0.5482,
+      "step": 230
+    },
+    {
+      "epoch": 0.6811594202898551,
+      "grad_norm": 1.252993106842041,
+      "learning_rate": 2.7533292743716508e-05,
+      "loss": 0.5561,
+      "step": 235
+    },
+    {
+      "epoch": 0.6956521739130435,
+      "grad_norm": 1.1550300121307373,
+      "learning_rate": 2.74000077679327e-05,
+      "loss": 0.5705,
+      "step": 240
+    },
+    {
+      "epoch": 0.7101449275362319,
+      "grad_norm": 1.130991816520691,
+      "learning_rate": 2.7263556540461533e-05,
+      "loss": 0.537,
+      "step": 245
+    },
+    {
+      "epoch": 0.7246376811594203,
+      "grad_norm": 1.2547308206558228,
+      "learning_rate": 2.7123973903130312e-05,
+      "loss": 0.5531,
+      "step": 250
+    },
+    {
+      "epoch": 0.7391304347826086,
+      "grad_norm": 1.2408415079116821,
+      "learning_rate": 2.6981295497349044e-05,
+      "loss": 0.5559,
+      "step": 255
+    },
+    {
+      "epoch": 0.7536231884057971,
+      "grad_norm": 1.2174203395843506,
+      "learning_rate": 2.6835557755009675e-05,
+      "loss": 0.4959,
+      "step": 260
+    },
+    {
+      "epoch": 0.7681159420289855,
+      "grad_norm": 1.1765241622924805,
+      "learning_rate": 2.668679788918345e-05,
+      "loss": 0.4753,
+      "step": 265
+    },
+    {
+      "epoch": 0.782608695652174,
+      "grad_norm": 1.2669579982757568,
+      "learning_rate": 2.653505388461885e-05,
+      "loss": 0.4868,
+      "step": 270
+    },
+    {
+      "epoch": 0.7971014492753623,
+      "grad_norm": 1.2592658996582031,
+      "learning_rate": 2.6380364488042448e-05,
+      "loss": 0.5029,
+      "step": 275
+    },
+    {
+      "epoch": 0.8115942028985508,
+      "grad_norm": 1.1942518949508667,
+      "learning_rate": 2.622276919826522e-05,
+      "loss": 0.5132,
+      "step": 280
+    },
+    {
+      "epoch": 0.8260869565217391,
+      "grad_norm": 1.058091640472412,
+      "learning_rate": 2.6062308256096808e-05,
+      "loss": 0.4747,
+      "step": 285
+    },
+    {
+      "epoch": 0.8405797101449275,
+      "grad_norm": 1.187317967414856,
+      "learning_rate": 2.589902263407034e-05,
+      "loss": 0.467,
+      "step": 290
+    },
+    {
+      "epoch": 0.855072463768116,
+      "grad_norm": 1.1849353313446045,
+      "learning_rate": 2.573295402598034e-05,
+      "loss": 0.491,
+      "step": 295
+    },
+    {
+      "epoch": 0.8695652173913043,
+      "grad_norm": 1.5354399681091309,
+      "learning_rate": 2.55641448362366e-05,
+      "loss": 0.5033,
+      "step": 300
+    },
+    {
+      "epoch": 0.8840579710144928,
+      "grad_norm": 1.2296185493469238,
+      "learning_rate": 2.5392638169036455e-05,
+      "loss": 0.4301,
+      "step": 305
+    },
+    {
+      "epoch": 0.8985507246376812,
+      "grad_norm": 1.1664255857467651,
+      "learning_rate": 2.5218477817358476e-05,
+      "loss": 0.4846,
+      "step": 310
+    },
+    {
+      "epoch": 0.9130434782608695,
+      "grad_norm": 1.411539912223816,
+      "learning_rate": 2.5041708251780228e-05,
+      "loss": 0.427,
+      "step": 315
+    },
+    {
+      "epoch": 0.927536231884058,
+      "grad_norm": 1.451964259147644,
+      "learning_rate": 2.4862374609123066e-05,
+      "loss": 0.4301,
+      "step": 320
+    },
+    {
+      "epoch": 0.9420289855072463,
+      "grad_norm": 1.4673717021942139,
+      "learning_rate": 2.468052268092674e-05,
+      "loss": 0.4055,
+      "step": 325
+    },
+    {
+      "epoch": 0.9565217391304348,
+      "grad_norm": 1.350848913192749,
+      "learning_rate": 2.449619890175688e-05,
+      "loss": 0.3974,
+      "step": 330
+    },
+    {
+      "epoch": 0.9710144927536232,
+      "grad_norm": 1.20427668094635,
+      "learning_rate": 2.430945033734826e-05,
+      "loss": 0.4052,
+      "step": 335
+    },
+    {
+      "epoch": 0.9855072463768116,
+      "grad_norm": 1.356813669204712,
+      "learning_rate": 2.412032467258691e-05,
+      "loss": 0.3487,
+      "step": 340
+    },
+    {
+      "epoch": 1.0,
+      "grad_norm": 1.3223190307617188,
+      "learning_rate": 2.3928870199334122e-05,
+      "loss": 0.434,
+      "step": 345
+    },
+    {
+      "epoch": 1.0144927536231885,
+      "grad_norm": 1.380501627922058,
+      "learning_rate": 2.373513580409548e-05,
+      "loss": 0.3635,
+      "step": 350
+    },
+    {
+      "epoch": 1.0289855072463767,
+      "grad_norm": 1.3034790754318237,
+      "learning_rate": 2.3539170955538052e-05,
+      "loss": 0.3201,
+      "step": 355
+    },
+    {
+      "epoch": 1.0434782608695652,
+      "grad_norm": 1.1440016031265259,
+      "learning_rate": 2.33410256918589e-05,
+      "loss": 0.3557,
+      "step": 360
+    },
+    {
+      "epoch": 1.0579710144927537,
+      "grad_norm": 1.2603141069412231,
+      "learning_rate": 2.3140750608008236e-05,
+      "loss": 0.3254,
+      "step": 365
+    },
+    {
+      "epoch": 1.0724637681159421,
+      "grad_norm": 1.3948999643325806,
+      "learning_rate": 2.2938396842770318e-05,
+      "loss": 0.3114,
+      "step": 370
+    },
+    {
+      "epoch": 1.0869565217391304,
+      "grad_norm": 1.4150797128677368,
+      "learning_rate": 2.2734016065705572e-05,
+      "loss": 0.3156,
+      "step": 375
+    },
+    {
+      "epoch": 1.1014492753623188,
+      "grad_norm": 1.3342996835708618,
+      "learning_rate": 2.252766046395713e-05,
+      "loss": 0.3186,
+      "step": 380
+    },
+    {
+      "epoch": 1.1159420289855073,
+      "grad_norm": 1.5770394802093506,
+      "learning_rate": 2.2319382728925234e-05,
+      "loss": 0.2912,
+      "step": 385
+    },
+    {
+      "epoch": 1.1304347826086956,
+      "grad_norm": 1.2255595922470093,
+      "learning_rate": 2.2109236042812854e-05,
+      "loss": 0.2855,
+      "step": 390
+    },
+    {
+      "epoch": 1.144927536231884,
+      "grad_norm": 1.399651288986206,
+      "learning_rate": 2.1897274065046008e-05,
+      "loss": 0.3021,
+      "step": 395
+    },
+    {
+      "epoch": 1.1594202898550725,
+      "grad_norm": 1.1826292276382446,
+      "learning_rate": 2.1683550918572232e-05,
+      "loss": 0.286,
+      "step": 400
+    },
+    {
+      "epoch": 1.1739130434782608,
+      "grad_norm": 1.3519924879074097,
+      "learning_rate": 2.1468121176040653e-05,
+      "loss": 0.2728,
+      "step": 405
+    },
+    {
+      "epoch": 1.1884057971014492,
+      "grad_norm": 1.4265503883361816,
+      "learning_rate": 2.1251039845867278e-05,
+      "loss": 0.2666,
+      "step": 410
+    },
+    {
+      "epoch": 1.2028985507246377,
+      "grad_norm": 1.5021287202835083,
+      "learning_rate": 2.1032362358188998e-05,
+      "loss": 0.2833,
+      "step": 415
+    },
+    {
+      "epoch": 1.2173913043478262,
+      "grad_norm": 1.220919132232666,
+      "learning_rate": 2.0812144550709882e-05,
+      "loss": 0.2625,
+      "step": 420
+    },
+    {
+      "epoch": 1.2318840579710144,
+      "grad_norm": 1.237191915512085,
+      "learning_rate": 2.0590442654443466e-05,
+      "loss": 0.2919,
+      "step": 425
+    },
+    {
+      "epoch": 1.2463768115942029,
+      "grad_norm": 1.7470474243164062,
+      "learning_rate": 2.036731327935454e-05,
+      "loss": 0.2468,
+      "step": 430
+    },
+    {
+      "epoch": 1.2608695652173914,
+      "grad_norm": 1.4158445596694946,
+      "learning_rate": 2.0142813399904212e-05,
+      "loss": 0.2625,
+      "step": 435
+    },
+    {
+      "epoch": 1.2753623188405796,
+      "grad_norm": 1.2676362991333008,
+      "learning_rate": 1.9917000340501904e-05,
+      "loss": 0.2425,
+      "step": 440
+    },
+    {
+      "epoch": 1.289855072463768,
+      "grad_norm": 1.094304084777832,
+      "learning_rate": 1.968993176086798e-05,
+      "loss": 0.2743,
+      "step": 445
+    },
+    {
+      "epoch": 1.3043478260869565,
+      "grad_norm": 1.3659477233886719,
+      "learning_rate": 1.946166564131075e-05,
+      "loss": 0.2513,
+      "step": 450
+    },
+    {
+      "epoch": 1.318840579710145,
+      "grad_norm": 1.4266231060028076,
+      "learning_rate": 1.9232260267921616e-05,
+      "loss": 0.2428,
+      "step": 455
+    },
+    {
+      "epoch": 1.3333333333333333,
+      "grad_norm": 1.3393499851226807,
+      "learning_rate": 1.9001774217692157e-05,
+      "loss": 0.2564,
+      "step": 460
+    },
+    {
+      "epoch": 1.3478260869565217,
+      "grad_norm": 1.4260315895080566,
+      "learning_rate": 1.877026634355691e-05,
+      "loss": 0.2264,
+      "step": 465
+    },
+    {
+      "epoch": 1.3623188405797102,
+      "grad_norm": 1.239377737045288,
+      "learning_rate": 1.8537795759365712e-05,
+      "loss": 0.2185,
+      "step": 470
+    },
+    {
+      "epoch": 1.3768115942028984,
+      "grad_norm": 3.305755615234375,
+      "learning_rate": 1.83044218247894e-05,
+      "loss": 0.2643,
+      "step": 475
+    },
+    {
+      "epoch": 1.391304347826087,
+      "grad_norm": 1.292801856994629,
+      "learning_rate": 1.8070204130162743e-05,
+      "loss": 0.245,
+      "step": 480
+    },
+    {
+      "epoch": 1.4057971014492754,
+      "grad_norm": 1.4896929264068604,
+      "learning_rate": 1.7835202481268512e-05,
+      "loss": 0.2349,
+      "step": 485
+    },
+    {
+      "epoch": 1.4202898550724639,
+      "grad_norm": 1.3014402389526367,
+      "learning_rate": 1.759947688406648e-05,
+      "loss": 0.2651,
+      "step": 490
+    },
+    {
+      "epoch": 1.434782608695652,
+      "grad_norm": 1.429904818534851,
+      "learning_rate": 1.7363087529371348e-05,
+      "loss": 0.2082,
+      "step": 495
+    },
+    {
+      "epoch": 1.4492753623188406,
+      "grad_norm": 2.28893780708313,
+      "learning_rate": 1.7126094777483467e-05,
+      "loss": 0.2128,
+      "step": 500
+    },
+    {
+      "epoch": 1.463768115942029,
+      "grad_norm": 1.4327991008758545,
+      "learning_rate": 1.6888559142776264e-05,
+      "loss": 0.2216,
+      "step": 505
+    },
+    {
+      "epoch": 1.4782608695652173,
+      "grad_norm": 1.2945911884307861,
+      "learning_rate": 1.6650541278244334e-05,
+      "loss": 0.2427,
+      "step": 510
+    },
+    {
+      "epoch": 1.4927536231884058,
+      "grad_norm": 1.4880881309509277,
+      "learning_rate": 1.6412101960016163e-05,
+      "loss": 0.2098,
+      "step": 515
+    },
+    {
+      "epoch": 1.5072463768115942,
+      "grad_norm": 1.5571190118789673,
+      "learning_rate": 1.6173302071835345e-05,
+      "loss": 0.2056,
+      "step": 520
+    },
+    {
+      "epoch": 1.5217391304347827,
+      "grad_norm": 1.2675591707229614,
+      "learning_rate": 1.593420258951441e-05,
+      "loss": 0.219,
+      "step": 525
+    },
+    {
+      "epoch": 1.5362318840579712,
+      "grad_norm": 1.32280433177948,
+      "learning_rate": 1.5694864565365052e-05,
+      "loss": 0.1952,
+      "step": 530
+    },
+    {
+      "epoch": 1.5507246376811594,
+      "grad_norm": 1.2481799125671387,
+      "learning_rate": 1.54553491126089e-05,
+      "loss": 0.1726,
+      "step": 535
+    },
+    {
+      "epoch": 1.5652173913043477,
+      "grad_norm": 1.3095006942749023,
+      "learning_rate": 1.5215717389772667e-05,
+      "loss": 0.1808,
+      "step": 540
+    },
+    {
+      "epoch": 1.5797101449275361,
+      "grad_norm": 1.3961431980133057,
+      "learning_rate": 1.4976030585071798e-05,
+      "loss": 0.2074,
+      "step": 545
+    },
+    {
+      "epoch": 1.5942028985507246,
+      "grad_norm": 1.3327604532241821,
+      "learning_rate": 1.4736349900786485e-05,
+      "loss": 0.1984,
+      "step": 550
+    },
+    {
+      "epoch": 1.608695652173913,
+      "grad_norm": 1.2703512907028198,
+      "learning_rate": 1.4496736537634117e-05,
+      "loss": 0.2078,
+      "step": 555
+    },
+    {
+      "epoch": 1.6231884057971016,
+      "grad_norm": 1.1081833839416504,
+      "learning_rate": 1.4257251679142138e-05,
+      "loss": 0.1873,
+      "step": 560
+    },
+    {
+      "epoch": 1.6376811594202898,
+      "grad_norm": 1.0600528717041016,
+      "learning_rate": 1.401795647602523e-05,
+      "loss": 0.1736,
+      "step": 565
+    },
+    {
+      "epoch": 1.6521739130434783,
+      "grad_norm": 1.1113861799240112,
+      "learning_rate": 1.377891203057098e-05,
+      "loss": 0.1844,
+      "step": 570
+    },
+    {
+      "epoch": 1.6666666666666665,
+      "grad_norm": 1.1133737564086914,
+      "learning_rate": 1.3540179381037828e-05,
+      "loss": 0.1623,
+      "step": 575
+    },
+    {
+      "epoch": 1.681159420289855,
+      "grad_norm": 0.9907549619674683,
+      "learning_rate": 1.3301819486069398e-05,
+      "loss": 0.1637,
+      "step": 580
+    },
+    {
+      "epoch": 1.6956521739130435,
+      "grad_norm": 1.2914941310882568,
+      "learning_rate": 1.3063893209129138e-05,
+      "loss": 0.1916,
+      "step": 585
+    },
+    {
+      "epoch": 1.710144927536232,
+      "grad_norm": 1.346624493598938,
+      "learning_rate": 1.2826461302959302e-05,
+      "loss": 0.1694,
+      "step": 590
+    },
+    {
+      "epoch": 1.7246376811594204,
+      "grad_norm": 1.4284813404083252,
+      "learning_rate": 1.2589584394068174e-05,
+      "loss": 0.1714,
+      "step": 595
+    },
+    {
+      "epoch": 1.7391304347826086,
+      "grad_norm": 1.0757873058319092,
+      "learning_rate": 1.2353322967249495e-05,
+      "loss": 0.1619,
+      "step": 600
+    },
+    {
+      "epoch": 1.7536231884057971,
+      "grad_norm": 1.236033320426941,
+      "learning_rate": 1.2117737350138172e-05,
+      "loss": 0.1596,
+      "step": 605
+    },
+    {
+      "epoch": 1.7681159420289854,
+      "grad_norm": 1.3429259061813354,
+      "learning_rate": 1.188288769780601e-05,
+      "loss": 0.1565,
+      "step": 610
+    },
+    {
+      "epoch": 1.7826086956521738,
+      "grad_norm": 1.3207345008850098,
+      "learning_rate": 1.1648833977401575e-05,
+      "loss": 0.1542,
+      "step": 615
+    },
+    {
+      "epoch": 1.7971014492753623,
+      "grad_norm": 1.280550241470337,
+      "learning_rate": 1.1415635952838e-05,
+      "loss": 0.1492,
+      "step": 620
+    },
+    {
+      "epoch": 1.8115942028985508,
+      "grad_norm": 1.3305364847183228,
+      "learning_rate": 1.1183353169532721e-05,
+      "loss": 0.1534,
+      "step": 625
+    },
+    {
+      "epoch": 1.8260869565217392,
+      "grad_norm": 1.0829124450683594,
+      "learning_rate": 1.0952044939202995e-05,
+      "loss": 0.1184,
+      "step": 630
+    },
+    {
+      "epoch": 1.8405797101449275,
+      "grad_norm": 1.1254221200942993,
+      "learning_rate": 1.0721770324721043e-05,
+      "loss": 0.1677,
+      "step": 635
+    },
+    {
+      "epoch": 1.855072463768116,
+      "grad_norm": 1.0419032573699951,
+      "learning_rate": 1.0492588125032843e-05,
+      "loss": 0.1561,
+      "step": 640
+    },
+    {
+      "epoch": 1.8695652173913042,
+      "grad_norm": 1.1556445360183716,
+      "learning_rate": 1.0264556860144187e-05,
+      "loss": 0.1425,
+      "step": 645
+    },
+    {
+      "epoch": 1.8840579710144927,
+      "grad_norm": 1.2914941310882568,
+      "learning_rate": 1.0037734756178059e-05,
+      "loss": 0.1459,
+      "step": 650
+    },
+    {
+      "epoch": 1.8985507246376812,
+      "grad_norm": 1.3069145679473877,
+      "learning_rate": 9.812179730506989e-06,
+      "loss": 0.1235,
+      "step": 655
+    },
+    {
+      "epoch": 1.9130434782608696,
+      "grad_norm": 1.1081457138061523,
+      "learning_rate": 9.58794937696431e-06,
+      "loss": 0.1406,
+      "step": 660
+    },
+    {
+      "epoch": 1.927536231884058,
+      "grad_norm": 1.1222493648529053,
+      "learning_rate": 9.365100951137967e-06,
+      "loss": 0.1673,
+      "step": 665
+    },
+    {
+      "epoch": 1.9420289855072463,
+      "grad_norm": 1.182753324508667,
+      "learning_rate": 9.143691355750718e-06,
+      "loss": 0.1355,
+      "step": 670
+    },
+    {
+      "epoch": 1.9565217391304348,
+      "grad_norm": 1.212599515914917,
+      "learning_rate": 8.923777126130456e-06,
+      "loss": 0.122,
+      "step": 675
+    },
+    {
+      "epoch": 1.971014492753623,
+      "grad_norm": 1.2320023775100708,
+      "learning_rate": 8.705414415774314e-06,
+      "loss": 0.1311,
+      "step": 680
+    },
+    {
+      "epoch": 1.9855072463768115,
+      "grad_norm": 1.1107462644577026,
+      "learning_rate": 8.488658982010277e-06,
+      "loss": 0.1346,
+      "step": 685
+    },
+    {
+      "epoch": 2.0,
+      "grad_norm": 1.215120792388916,
+      "learning_rate": 8.273566171759952e-06,
+      "loss": 0.1351,
+      "step": 690
+    },
+    {
+      "epoch": 2.0144927536231885,
+      "grad_norm": 1.147032618522644,
+      "learning_rate": 8.060190907406136e-06,
+      "loss": 0.1167,
+      "step": 695
+    },
+    {
+      "epoch": 2.028985507246377,
+      "grad_norm": 0.9225956797599792,
+      "learning_rate": 7.848587672768811e-06,
+      "loss": 0.0985,
+      "step": 700
+    },
+    {
+      "epoch": 2.0434782608695654,
+      "grad_norm": 0.9207383394241333,
+      "learning_rate": 7.63881049919304e-06,
+      "loss": 0.0998,
+      "step": 705
+    },
+    {
+      "epoch": 2.0579710144927534,
+      "grad_norm": 0.869804859161377,
+      "learning_rate": 7.430912951752516e-06,
+      "loss": 0.0972,
+      "step": 710
+    },
+    {
+      "epoch": 2.072463768115942,
+      "grad_norm": 1.1430824995040894,
+      "learning_rate": 7.224948115572088e-06,
+      "loss": 0.1069,
+      "step": 715
+    },
+    {
+      "epoch": 2.0869565217391304,
+      "grad_norm": 0.9880947470664978,
+      "learning_rate": 7.020968582272843e-06,
+      "loss": 0.094,
+      "step": 720
+    },
+    {
+      "epoch": 2.101449275362319,
+      "grad_norm": 1.1354153156280518,
+      "learning_rate": 6.819026436543241e-06,
+      "loss": 0.1113,
+      "step": 725
+    },
+    {
+      "epoch": 2.1159420289855073,
+      "grad_norm": 0.8495379686355591,
+      "learning_rate": 6.619173242839658e-06,
+      "loss": 0.1103,
+      "step": 730
+    },
+    {
+      "epoch": 2.130434782608696,
+      "grad_norm": 0.9505207538604736,
+      "learning_rate": 6.421460032219795e-06,
+      "loss": 0.0932,
+      "step": 735
+    },
+    {
+      "epoch": 2.1449275362318843,
+      "grad_norm": 0.9441226124763489,
+      "learning_rate": 6.225937289312244e-06,
+      "loss": 0.0783,
+      "step": 740
+    },
+    {
+      "epoch": 2.1594202898550723,
+      "grad_norm": 0.9335423111915588,
+      "learning_rate": 6.032654939425613e-06,
+      "loss": 0.0928,
+      "step": 745
+    },
+    {
+      "epoch": 2.1739130434782608,
+      "grad_norm": 0.8520256280899048,
+      "learning_rate": 5.8416623358004865e-06,
+      "loss": 0.0983,
+      "step": 750
+    },
+    {
+      "epoch": 2.1884057971014492,
+      "grad_norm": 0.9851803779602051,
+      "learning_rate": 5.653008247007395e-06,
+      "loss": 0.0842,
+      "step": 755
+    },
+    {
+      "epoch": 2.2028985507246377,
+      "grad_norm": 0.9417587518692017,
+      "learning_rate": 5.4667408444941455e-06,
+      "loss": 0.0912,
+      "step": 760
+    },
+    {
+      "epoch": 2.217391304347826,
+      "grad_norm": 1.1524457931518555,
+      "learning_rate": 5.2829076902855725e-06,
+      "loss": 0.0898,
+      "step": 765
+    },
+    {
+      "epoch": 2.2318840579710146,
+      "grad_norm": 0.9601754546165466,
+      "learning_rate": 5.1015557248389526e-06,
+      "loss": 0.1049,
+      "step": 770
+    },
+    {
+      "epoch": 2.246376811594203,
+      "grad_norm": 0.9812232851982117,
+      "learning_rate": 4.922731255058062e-06,
+      "loss": 0.0869,
+      "step": 775
+    },
+    {
+      "epoch": 2.260869565217391,
+      "grad_norm": 0.9177423119544983,
+      "learning_rate": 4.7464799424690645e-06,
+      "loss": 0.084,
+      "step": 780
+    },
+    {
+      "epoch": 2.2753623188405796,
+      "grad_norm": 1.089078664779663,
+      "learning_rate": 4.572846791561185e-06,
+      "loss": 0.0991,
+      "step": 785
+    },
+    {
+      "epoch": 2.289855072463768,
+      "grad_norm": 0.837441623210907,
+      "learning_rate": 4.401876138295117e-06,
+      "loss": 0.0789,
+      "step": 790
+    },
+    {
+      "epoch": 2.3043478260869565,
+      "grad_norm": 0.8127453327178955,
+      "learning_rate": 4.23361163878217e-06,
+      "loss": 0.0809,
+      "step": 795
+    },
+    {
+      "epoch": 2.318840579710145,
+      "grad_norm": 0.7465061545372009,
+      "learning_rate": 4.06809625813701e-06,
+      "loss": 0.0852,
+      "step": 800
+    },
+    {
+      "epoch": 2.3333333333333335,
+      "grad_norm": 1.0076066255569458,
+      "learning_rate": 3.9053722595067995e-06,
+      "loss": 0.0879,
+      "step": 805
+    },
+    {
+      "epoch": 2.3478260869565215,
+      "grad_norm": 0.8961737155914307,
+      "learning_rate": 3.7454811932796562e-06,
+      "loss": 0.0848,
+      "step": 810
+    },
+    {
+      "epoch": 2.36231884057971,
+      "grad_norm": 0.8941988348960876,
+      "learning_rate": 3.5884638864750073e-06,
+      "loss": 0.0796,
+      "step": 815
+    },
+    {
+      "epoch": 2.3768115942028984,
+      "grad_norm": 0.9690755605697632,
+      "learning_rate": 3.4343604323187645e-06,
+      "loss": 0.0922,
+      "step": 820
+    },
+    {
+      "epoch": 2.391304347826087,
+      "grad_norm": 0.9332724809646606,
+      "learning_rate": 3.2832101800057694e-06,
+      "loss": 0.0859,
+      "step": 825
+    },
+    {
+      "epoch": 2.4057971014492754,
+      "grad_norm": 0.9866376519203186,
+      "learning_rate": 3.1350517246522807e-06,
+      "loss": 0.0802,
+      "step": 830
+    },
+    {
+      "epoch": 2.420289855072464,
+      "grad_norm": 0.9095158576965332,
+      "learning_rate": 2.9899228974409966e-06,
+      "loss": 0.0707,
+      "step": 835
+    },
+    {
+      "epoch": 2.4347826086956523,
+      "grad_norm": 0.8134745955467224,
+      "learning_rate": 2.8478607559611185e-06,
+      "loss": 0.0841,
+      "step": 840
+    },
+    {
+      "epoch": 2.449275362318841,
+      "grad_norm": 0.8455703258514404,
+      "learning_rate": 2.7089015747459946e-06,
+      "loss": 0.0794,
+      "step": 845
+    },
+    {
+      "epoch": 2.463768115942029,
+      "grad_norm": 0.8130114078521729,
+      "learning_rate": 2.5730808360106416e-06,
+      "loss": 0.0741,
+      "step": 850
+    },
+    {
+      "epoch": 2.4782608695652173,
+      "grad_norm": 0.6896876096725464,
+      "learning_rate": 2.4404332205916663e-06,
+      "loss": 0.0774,
+      "step": 855
+    },
+    {
+      "epoch": 2.4927536231884058,
+      "grad_norm": 0.848038911819458,
+      "learning_rate": 2.3109925990917306e-06,
+      "loss": 0.0631,
+      "step": 860
+    },
+    {
+      "epoch": 2.5072463768115942,
+      "grad_norm": 0.7451518177986145,
+      "learning_rate": 2.184792023230951e-06,
+      "loss": 0.0784,
+      "step": 865
+    },
+    {
+      "epoch": 2.5217391304347827,
+      "grad_norm": 0.8208417296409607,
+      "learning_rate": 2.061863717407395e-06,
+      "loss": 0.0748,
+      "step": 870
+    },
+    {
+      "epoch": 2.536231884057971,
+      "grad_norm": 0.8441823124885559,
+      "learning_rate": 1.9422390704687894e-06,
+      "loss": 0.0822,
+      "step": 875
+    },
+    {
+      "epoch": 2.550724637681159,
+      "grad_norm": 0.7636905312538147,
+      "learning_rate": 1.8259486276976367e-06,
+      "loss": 0.068,
+      "step": 880
+    },
+    {
+      "epoch": 2.5652173913043477,
+      "grad_norm": 0.7833094596862793,
+      "learning_rate": 1.713022083011677e-06,
+      "loss": 0.0785,
+      "step": 885
+    },
+    {
+      "epoch": 2.579710144927536,
+      "grad_norm": 0.7898631691932678,
+      "learning_rate": 1.603488271381765e-06,
+      "loss": 0.072,
+      "step": 890
+    },
+    {
+      "epoch": 2.5942028985507246,
+      "grad_norm": 0.9022684693336487,
+      "learning_rate": 1.4973751614690905e-06,
+      "loss": 0.0775,
+      "step": 895
+    },
+    {
+      "epoch": 2.608695652173913,
+      "grad_norm": 0.8136042356491089,
+      "learning_rate": 1.3947098484835496e-06,
+      "loss": 0.0752,
+      "step": 900
+    },
+    {
+      "epoch": 2.6231884057971016,
+      "grad_norm": 0.8685534596443176,
+      "learning_rate": 1.2955185472652176e-06,
+      "loss": 0.0785,
+      "step": 905
+    },
+    {
+      "epoch": 2.63768115942029,
+      "grad_norm": 0.8825992941856384,
+      "learning_rate": 1.1998265855905487e-06,
+      "loss": 0.0784,
+      "step": 910
+    },
+    {
+      "epoch": 2.6521739130434785,
+      "grad_norm": 0.8254004716873169,
+      "learning_rate": 1.1076583977051386e-06,
+      "loss": 0.0715,
+      "step": 915
+    },
+    {
+      "epoch": 2.6666666666666665,
+      "grad_norm": 0.8967143893241882,
+      "learning_rate": 1.0190375180845985e-06,
+      "loss": 0.0714,
+      "step": 920
+    },
+    {
+      "epoch": 2.681159420289855,
+      "grad_norm": 0.7813670039176941,
+      "learning_rate": 9.339865754252036e-07,
+      "loss": 0.0901,
+      "step": 925
+    },
+    {
+      "epoch": 2.6956521739130435,
+      "grad_norm": 1.0012155771255493,
+      "learning_rate": 8.525272868658346e-07,
+      "loss": 0.0781,
+      "step": 930
+    },
+    {
+      "epoch": 2.710144927536232,
+      "grad_norm": 0.6551988124847412,
+      "learning_rate": 7.746804524426299e-07,
+      "loss": 0.0663,
+      "step": 935
+    },
+    {
+      "epoch": 2.7246376811594204,
+      "grad_norm": 0.8370825052261353,
+      "learning_rate": 7.004659497778942e-07,
+      "loss": 0.0753,
+      "step": 940
+    },
+    {
+      "epoch": 2.7391304347826084,
+      "grad_norm": 0.763660728931427,
+      "learning_rate": 6.29902729004449e-07,
+      "loss": 0.0741,
+      "step": 945
+    },
+    {
+      "epoch": 2.753623188405797,
+      "grad_norm": 0.808006763458252,
+      "learning_rate": 5.630088079268874e-07,
+      "loss": 0.0795,
+      "step": 950
+    },
+    {
+      "epoch": 2.7681159420289854,
+      "grad_norm": 0.74526047706604,
+      "learning_rate": 4.998012674208369e-07,
+      "loss": 0.0815,
+      "step": 955
+    },
+    {
+      "epoch": 2.782608695652174,
+      "grad_norm": 0.627281129360199,
+      "learning_rate": 4.402962470714889e-07,
+      "loss": 0.0661,
+      "step": 960
+    },
+    {
+      "epoch": 2.7971014492753623,
+      "grad_norm": 0.7469335198402405,
+      "learning_rate": 3.84508941052481e-07,
+      "loss": 0.0766,
+      "step": 965
+    },
+    {
+      "epoch": 2.8115942028985508,
+      "grad_norm": 0.5805714726448059,
+      "learning_rate": 3.324535942461637e-07,
+      "loss": 0.071,
+      "step": 970
+    },
+    {
+      "epoch": 2.8260869565217392,
+      "grad_norm": 0.7372128367424011,
+      "learning_rate": 2.8414349860627965e-07,
+      "loss": 0.0698,
+      "step": 975
+    },
+    {
+      "epoch": 2.8405797101449277,
+      "grad_norm": 0.7484320402145386,
+      "learning_rate": 2.3959098976395863e-07,
+      "loss": 0.0785,
+      "step": 980
+    },
+    {
+      "epoch": 2.855072463768116,
+      "grad_norm": 0.8040164709091187,
+      "learning_rate": 1.9880744387790585e-07,
+      "loss": 0.064,
+      "step": 985
+    },
+    {
+      "epoch": 2.869565217391304,
+      "grad_norm": 0.6598532795906067,
+      "learning_rate": 1.6180327472957835e-07,
+      "loss": 0.0799,
+      "step": 990
+    },
+    {
+      "epoch": 2.8840579710144927,
+      "grad_norm": 0.7085109353065491,
+      "learning_rate": 1.2858793106410584e-07,
+      "loss": 0.07,
+      "step": 995
+    },
+    {
+      "epoch": 2.898550724637681,
+      "grad_norm": 0.676478385925293,
+      "learning_rate": 9.916989417762201e-08,
+      "loss": 0.07,
+      "step": 1000
+    },
+    {
+      "epoch": 2.9130434782608696,
+      "grad_norm": 0.734004557132721,
+      "learning_rate": 7.355667575162373e-08,
+      "loss": 0.0699,
+      "step": 1005
+    },
+    {
+      "epoch": 2.927536231884058,
+      "grad_norm": 0.7295432090759277,
+      "learning_rate": 5.175481593492348e-08,
+      "loss": 0.0739,
+      "step": 1010
+    },
+    {
+      "epoch": 2.942028985507246,
+      "grad_norm": 0.7546738386154175,
+      "learning_rate": 3.376988167366535e-08,
+      "loss": 0.0686,
+      "step": 1015
+    },
+    {
+      "epoch": 2.9565217391304346,
+      "grad_norm": 0.7214581370353699,
+      "learning_rate": 1.9606465289848662e-08,
+      "loss": 0.073,
+      "step": 1020
+    },
+    {
+      "epoch": 2.971014492753623,
+      "grad_norm": 0.7831776142120361,
+      "learning_rate": 9.268183308706046e-09,
+      "loss": 0.0805,
+      "step": 1025
+    },
+    {
+      "epoch": 2.9855072463768115,
+      "grad_norm": 0.7942865490913391,
+      "learning_rate": 2.757675535262072e-09,
+      "loss": 0.0669,
+      "step": 1030
+    },
+    {
+      "epoch": 3.0,
+      "grad_norm": 0.8652397394180298,
+      "learning_rate": 7.66043802674643e-11,
+      "loss": 0.0753,
+      "step": 1035
+    }
+  ],
+  "logging_steps": 5,
+  "max_steps": 1035,
+  "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": 1.4426652672461373e+18,
+  "train_batch_size": 2,
+  "trial_name": null,
+  "trial_params": null
+}

129_128_e3_3e-5/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:705687a2cde173e0e7684a4a0c5192f0e169323dcdcb71d0734078a9d633388b
+size 8209

129_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)