Upload folder using huggingface_hub

135_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",
+    "q_proj",
+    "up_proj",
+    "v_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_rslora": false
+}

135_128_e3_3e-5/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:aa6f28805197b3fca2f4e7aa67504992f3f17741bba7d5cc756382db03a96b39
+size 791751704

135_128_e3_3e-5/latest

@@ -0,0 +1 @@
+global_step1167

135_128_e3_3e-5/rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c8c63f6f936492ebbbb93cce1f901d2436935004f0ede1c2e0569a0cc74d217a
+size 16389

135_128_e3_3e-5/rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d700f5bce8f096509b045f72252bbb04cced9d0dfa3ea4956d9a854761d2e0f2
+size 16389

135_128_e3_3e-5/rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f4416d7dc092e763d78945423301c7fd4afa04b28713843f18ceda8e818e10c9
+size 16389

135_128_e3_3e-5/rng_state_3.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:340418e06a179c2892b84ce28249b904a4c474d86312d94073ee4a4bfe137e3c
+size 16389

135_128_e3_3e-5/rng_state_4.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e371f53b80636e9c81cd9137ae7321de988c0ac68bd770ce2e5054721aa7c103
+size 16389

135_128_e3_3e-5/rng_state_5.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:806f06355fce66dc07cf318ced013113c7e37d9fec89ef8e774d9a48662448fa
+size 16389

135_128_e3_3e-5/rng_state_6.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:65ce7ed5dfdf59a4de9fbafe79b81df36c21652d55a2c35b4b74a583b9fad5d5
+size 16389

135_128_e3_3e-5/rng_state_7.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4178a14e696cb7af48b825b0176a47cf91809e81b2db67bb88223de01227b7d9
+size 16389

135_128_e3_3e-5/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:198e62dc2b2ef81e15cbd3cee044f71d5be43b2aa180a317c37b19e4837bfdb2
+size 1401

135_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
+  }
+}

135_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
+}

135_128_e3_3e-5/trainer_state.json

@@ -0,0 +1,1665 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.0,
+  "eval_steps": 500,
+  "global_step": 1167,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.012853470437017995,
+      "grad_norm": 0.8533712029457092,
+      "learning_rate": 2.033898305084746e-06,
+      "loss": 1.4375,
+      "step": 5
+    },
+    {
+      "epoch": 0.02570694087403599,
+      "grad_norm": 0.6216090321540833,
+      "learning_rate": 4.576271186440678e-06,
+      "loss": 1.3952,
+      "step": 10
+    },
+    {
+      "epoch": 0.038560411311053984,
+      "grad_norm": 0.6457392573356628,
+      "learning_rate": 7.1186440677966106e-06,
+      "loss": 1.413,
+      "step": 15
+    },
+    {
+      "epoch": 0.05141388174807198,
+      "grad_norm": 0.5996050834655762,
+      "learning_rate": 9.661016949152542e-06,
+      "loss": 1.3787,
+      "step": 20
+    },
+    {
+      "epoch": 0.06426735218508997,
+      "grad_norm": 0.6159489154815674,
+      "learning_rate": 1.2203389830508475e-05,
+      "loss": 1.3436,
+      "step": 25
+    },
+    {
+      "epoch": 0.07712082262210797,
+      "grad_norm": 0.5571305155754089,
+      "learning_rate": 1.4745762711864408e-05,
+      "loss": 1.3388,
+      "step": 30
+    },
+    {
+      "epoch": 0.08997429305912596,
+      "grad_norm": 0.5749513506889343,
+      "learning_rate": 1.728813559322034e-05,
+      "loss": 1.3769,
+      "step": 35
+    },
+    {
+      "epoch": 0.10282776349614396,
+      "grad_norm": 0.5699848532676697,
+      "learning_rate": 1.983050847457627e-05,
+      "loss": 1.3602,
+      "step": 40
+    },
+    {
+      "epoch": 0.11568123393316196,
+      "grad_norm": 0.469750314950943,
+      "learning_rate": 2.2372881355932205e-05,
+      "loss": 1.2902,
+      "step": 45
+    },
+    {
+      "epoch": 0.12853470437017994,
+      "grad_norm": 0.5320536494255066,
+      "learning_rate": 2.4915254237288138e-05,
+      "loss": 1.3016,
+      "step": 50
+    },
+    {
+      "epoch": 0.14138817480719795,
+      "grad_norm": 0.5024668574333191,
+      "learning_rate": 2.7457627118644068e-05,
+      "loss": 1.2321,
+      "step": 55
+    },
+    {
+      "epoch": 0.15424164524421594,
+      "grad_norm": 0.5402373671531677,
+      "learning_rate": 3e-05,
+      "loss": 1.236,
+      "step": 60
+    },
+    {
+      "epoch": 0.16709511568123395,
+      "grad_norm": 0.6514788269996643,
+      "learning_rate": 2.9998492649592036e-05,
+      "loss": 1.2305,
+      "step": 65
+    },
+    {
+      "epoch": 0.17994858611825193,
+      "grad_norm": 0.7798457145690918,
+      "learning_rate": 2.9993970901315514e-05,
+      "loss": 1.2032,
+      "step": 70
+    },
+    {
+      "epoch": 0.1928020565552699,
+      "grad_norm": 0.579349160194397,
+      "learning_rate": 2.9986435663951645e-05,
+      "loss": 1.194,
+      "step": 75
+    },
+    {
+      "epoch": 0.20565552699228792,
+      "grad_norm": 0.7016724348068237,
+      "learning_rate": 2.997588845193284e-05,
+      "loss": 1.1708,
+      "step": 80
+    },
+    {
+      "epoch": 0.2185089974293059,
+      "grad_norm": 4.64923620223999,
+      "learning_rate": 2.9962331385038356e-05,
+      "loss": 1.1424,
+      "step": 85
+    },
+    {
+      "epoch": 0.23136246786632392,
+      "grad_norm": 0.7824217081069946,
+      "learning_rate": 2.9945767187968226e-05,
+      "loss": 1.0931,
+      "step": 90
+    },
+    {
+      "epoch": 0.2442159383033419,
+      "grad_norm": 0.7213613390922546,
+      "learning_rate": 2.9926199189795684e-05,
+      "loss": 1.1182,
+      "step": 95
+    },
+    {
+      "epoch": 0.2570694087403599,
+      "grad_norm": 0.7232889533042908,
+      "learning_rate": 2.990363132329806e-05,
+      "loss": 1.1185,
+      "step": 100
+    },
+    {
+      "epoch": 0.2699228791773779,
+      "grad_norm": 0.7367932796478271,
+      "learning_rate": 2.9878068124166394e-05,
+      "loss": 1.0805,
+      "step": 105
+    },
+    {
+      "epoch": 0.2827763496143959,
+      "grad_norm": 0.8606106042861938,
+      "learning_rate": 2.9849514730093843e-05,
+      "loss": 1.0088,
+      "step": 110
+    },
+    {
+      "epoch": 0.29562982005141386,
+      "grad_norm": 0.8659786581993103,
+      "learning_rate": 2.9817976879743092e-05,
+      "loss": 1.0113,
+      "step": 115
+    },
+    {
+      "epoch": 0.30848329048843187,
+      "grad_norm": 0.9298407435417175,
+      "learning_rate": 2.9783460911593024e-05,
+      "loss": 1.0187,
+      "step": 120
+    },
+    {
+      "epoch": 0.3213367609254499,
+      "grad_norm": 0.9039083123207092,
+      "learning_rate": 2.97459737626648e-05,
+      "loss": 0.9848,
+      "step": 125
+    },
+    {
+      "epoch": 0.3341902313624679,
+      "grad_norm": 0.8701486587524414,
+      "learning_rate": 2.970552296712764e-05,
+      "loss": 0.9496,
+      "step": 130
+    },
+    {
+      "epoch": 0.34704370179948585,
+      "grad_norm": 1.0542534589767456,
+      "learning_rate": 2.966211665478464e-05,
+      "loss": 0.9648,
+      "step": 135
+    },
+    {
+      "epoch": 0.35989717223650386,
+      "grad_norm": 0.8997522592544556,
+      "learning_rate": 2.961576354943881e-05,
+      "loss": 0.9306,
+      "step": 140
+    },
+    {
+      "epoch": 0.37275064267352187,
+      "grad_norm": 0.9883838891983032,
+      "learning_rate": 2.9566472967139796e-05,
+      "loss": 0.9157,
+      "step": 145
+    },
+    {
+      "epoch": 0.3856041131105398,
+      "grad_norm": 0.9717779755592346,
+      "learning_rate": 2.9514254814311492e-05,
+      "loss": 0.8654,
+      "step": 150
+    },
+    {
+      "epoch": 0.39845758354755784,
+      "grad_norm": 1.089228630065918,
+      "learning_rate": 2.9459119585761107e-05,
+      "loss": 0.8875,
+      "step": 155
+    },
+    {
+      "epoch": 0.41131105398457585,
+      "grad_norm": 1.1141870021820068,
+      "learning_rate": 2.9401078362569866e-05,
+      "loss": 0.8371,
+      "step": 160
+    },
+    {
+      "epoch": 0.4241645244215938,
+      "grad_norm": 1.1549341678619385,
+      "learning_rate": 2.9340142809865967e-05,
+      "loss": 0.8254,
+      "step": 165
+    },
+    {
+      "epoch": 0.4370179948586118,
+      "grad_norm": 1.1419929265975952,
+      "learning_rate": 2.9276325174480108e-05,
+      "loss": 0.8228,
+      "step": 170
+    },
+    {
+      "epoch": 0.4498714652956298,
+      "grad_norm": 1.3094439506530762,
+      "learning_rate": 2.9209638282484122e-05,
+      "loss": 0.8233,
+      "step": 175
+    },
+    {
+      "epoch": 0.46272493573264784,
+      "grad_norm": 1.0400540828704834,
+      "learning_rate": 2.9140095536613182e-05,
+      "loss": 0.7837,
+      "step": 180
+    },
+    {
+      "epoch": 0.4755784061696658,
+      "grad_norm": 1.3066167831420898,
+      "learning_rate": 2.9067710913572146e-05,
+      "loss": 0.7731,
+      "step": 185
+    },
+    {
+      "epoch": 0.4884318766066838,
+      "grad_norm": 1.0634539127349854,
+      "learning_rate": 2.8992498961226482e-05,
+      "loss": 0.7418,
+      "step": 190
+    },
+    {
+      "epoch": 0.5012853470437018,
+      "grad_norm": 1.0969005823135376,
+      "learning_rate": 2.8914474795678462e-05,
+      "loss": 0.7657,
+      "step": 195
+    },
+    {
+      "epoch": 0.5141388174807198,
+      "grad_norm": 1.1349319219589233,
+      "learning_rate": 2.8833654098229132e-05,
+      "loss": 0.7219,
+      "step": 200
+    },
+    {
+      "epoch": 0.5269922879177378,
+      "grad_norm": 1.2453835010528564,
+      "learning_rate": 2.8750053112226654e-05,
+      "loss": 0.7274,
+      "step": 205
+    },
+    {
+      "epoch": 0.5398457583547558,
+      "grad_norm": 1.1184139251708984,
+      "learning_rate": 2.8663688639801748e-05,
+      "loss": 0.7502,
+      "step": 210
+    },
+    {
+      "epoch": 0.5526992287917738,
+      "grad_norm": 1.2294700145721436,
+      "learning_rate": 2.8574578038490772e-05,
+      "loss": 0.7119,
+      "step": 215
+    },
+    {
+      "epoch": 0.5655526992287918,
+      "grad_norm": 1.284010410308838,
+      "learning_rate": 2.8482739217747228e-05,
+      "loss": 0.7198,
+      "step": 220
+    },
+    {
+      "epoch": 0.5784061696658098,
+      "grad_norm": 1.288525938987732,
+      "learning_rate": 2.838819063534231e-05,
+      "loss": 0.6813,
+      "step": 225
+    },
+    {
+      "epoch": 0.5912596401028277,
+      "grad_norm": 1.1404322385787964,
+      "learning_rate": 2.8290951293655244e-05,
+      "loss": 0.6412,
+      "step": 230
+    },
+    {
+      "epoch": 0.6041131105398457,
+      "grad_norm": 1.4699451923370361,
+      "learning_rate": 2.819104073585422e-05,
+      "loss": 0.6683,
+      "step": 235
+    },
+    {
+      "epoch": 0.6169665809768637,
+      "grad_norm": 1.3198682069778442,
+      "learning_rate": 2.808847904196857e-05,
+      "loss": 0.6331,
+      "step": 240
+    },
+    {
+      "epoch": 0.6298200514138818,
+      "grad_norm": 1.3318793773651123,
+      "learning_rate": 2.798328682485312e-05,
+      "loss": 0.6054,
+      "step": 245
+    },
+    {
+      "epoch": 0.6426735218508998,
+      "grad_norm": 1.2329274415969849,
+      "learning_rate": 2.7875485226045385e-05,
+      "loss": 0.5831,
+      "step": 250
+    },
+    {
+      "epoch": 0.6555269922879178,
+      "grad_norm": 1.466707468032837,
+      "learning_rate": 2.7765095911516548e-05,
+      "loss": 0.6048,
+      "step": 255
+    },
+    {
+      "epoch": 0.6683804627249358,
+      "grad_norm": 1.8981208801269531,
+      "learning_rate": 2.765214106731706e-05,
+      "loss": 0.5972,
+      "step": 260
+    },
+    {
+      "epoch": 0.6812339331619537,
+      "grad_norm": 1.2886542081832886,
+      "learning_rate": 2.7536643395117645e-05,
+      "loss": 0.5754,
+      "step": 265
+    },
+    {
+      "epoch": 0.6940874035989717,
+      "grad_norm": 1.4969333410263062,
+      "learning_rate": 2.7418626107646747e-05,
+      "loss": 0.5626,
+      "step": 270
+    },
+    {
+      "epoch": 0.7069408740359897,
+      "grad_norm": 1.3916356563568115,
+      "learning_rate": 2.729811292402522e-05,
+      "loss": 0.569,
+      "step": 275
+    },
+    {
+      "epoch": 0.7197943444730077,
+      "grad_norm": 1.2441447973251343,
+      "learning_rate": 2.7175128064999272e-05,
+      "loss": 0.5667,
+      "step": 280
+    },
+    {
+      "epoch": 0.7326478149100257,
+      "grad_norm": 1.3603755235671997,
+      "learning_rate": 2.704969624807254e-05,
+      "loss": 0.592,
+      "step": 285
+    },
+    {
+      "epoch": 0.7455012853470437,
+      "grad_norm": 1.2956894636154175,
+      "learning_rate": 2.6921842682538436e-05,
+      "loss": 0.5895,
+      "step": 290
+    },
+    {
+      "epoch": 0.7583547557840618,
+      "grad_norm": 1.487562894821167,
+      "learning_rate": 2.6791593064413498e-05,
+      "loss": 0.5443,
+      "step": 295
+    },
+    {
+      "epoch": 0.7712082262210797,
+      "grad_norm": 1.4298415184020996,
+      "learning_rate": 2.6658973571273077e-05,
+      "loss": 0.5424,
+      "step": 300
+    },
+    {
+      "epoch": 0.7840616966580977,
+      "grad_norm": 1.221649408340454,
+      "learning_rate": 2.65240108569901e-05,
+      "loss": 0.5068,
+      "step": 305
+    },
+    {
+      "epoch": 0.7969151670951157,
+      "grad_norm": 1.3684046268463135,
+      "learning_rate": 2.6386732046378242e-05,
+      "loss": 0.4996,
+      "step": 310
+    },
+    {
+      "epoch": 0.8097686375321337,
+      "grad_norm": 1.3003875017166138,
+      "learning_rate": 2.6247164729740317e-05,
+      "loss": 0.5085,
+      "step": 315
+    },
+    {
+      "epoch": 0.8226221079691517,
+      "grad_norm": 1.4060603380203247,
+      "learning_rate": 2.6105336957323214e-05,
+      "loss": 0.4606,
+      "step": 320
+    },
+    {
+      "epoch": 0.8354755784061697,
+      "grad_norm": 1.23789644241333,
+      "learning_rate": 2.5961277233680356e-05,
+      "loss": 0.5189,
+      "step": 325
+    },
+    {
+      "epoch": 0.8483290488431876,
+      "grad_norm": 1.3305301666259766,
+      "learning_rate": 2.5815014511942828e-05,
+      "loss": 0.4949,
+      "step": 330
+    },
+    {
+      "epoch": 0.8611825192802056,
+      "grad_norm": 1.5520987510681152,
+      "learning_rate": 2.566657818800041e-05,
+      "loss": 0.4667,
+      "step": 335
+    },
+    {
+      "epoch": 0.8740359897172236,
+      "grad_norm": 1.3534374237060547,
+      "learning_rate": 2.5515998094593555e-05,
+      "loss": 0.4906,
+      "step": 340
+    },
+    {
+      "epoch": 0.8868894601542416,
+      "grad_norm": 1.250394582748413,
+      "learning_rate": 2.5363304495317633e-05,
+      "loss": 0.4763,
+      "step": 345
+    },
+    {
+      "epoch": 0.8997429305912596,
+      "grad_norm": 1.4147182703018188,
+      "learning_rate": 2.5208528078540524e-05,
+      "loss": 0.4791,
+      "step": 350
+    },
+    {
+      "epoch": 0.9125964010282777,
+      "grad_norm": 1.3888635635375977,
+      "learning_rate": 2.50516999512349e-05,
+      "loss": 0.4879,
+      "step": 355
+    },
+    {
+      "epoch": 0.9254498714652957,
+      "grad_norm": 1.3531410694122314,
+      "learning_rate": 2.4892851632726306e-05,
+      "loss": 0.4186,
+      "step": 360
+    },
+    {
+      "epoch": 0.9383033419023136,
+      "grad_norm": 1.4474471807479858,
+      "learning_rate": 2.4732015048358447e-05,
+      "loss": 0.4037,
+      "step": 365
+    },
+    {
+      "epoch": 0.9511568123393316,
+      "grad_norm": 1.4082612991333008,
+      "learning_rate": 2.4569222523076796e-05,
+      "loss": 0.4194,
+      "step": 370
+    },
+    {
+      "epoch": 0.9640102827763496,
+      "grad_norm": 1.2963148355484009,
+      "learning_rate": 2.440450677493194e-05,
+      "loss": 0.4144,
+      "step": 375
+    },
+    {
+      "epoch": 0.9768637532133676,
+      "grad_norm": 1.3042285442352295,
+      "learning_rate": 2.4237900908503894e-05,
+      "loss": 0.4393,
+      "step": 380
+    },
+    {
+      "epoch": 0.9897172236503856,
+      "grad_norm": 1.5556186437606812,
+      "learning_rate": 2.406943840824876e-05,
+      "loss": 0.3883,
+      "step": 385
+    },
+    {
+      "epoch": 1.0025706940874035,
+      "grad_norm": 1.4160274267196655,
+      "learning_rate": 2.3899153131769004e-05,
+      "loss": 0.4251,
+      "step": 390
+    },
+    {
+      "epoch": 1.0154241645244215,
+      "grad_norm": 1.8217828273773193,
+      "learning_rate": 2.3727079303008752e-05,
+      "loss": 0.3359,
+      "step": 395
+    },
+    {
+      "epoch": 1.0282776349614395,
+      "grad_norm": 1.5839332342147827,
+      "learning_rate": 2.355325150537548e-05,
+      "loss": 0.3086,
+      "step": 400
+    },
+    {
+      "epoch": 1.0411311053984575,
+      "grad_norm": 1.3588227033615112,
+      "learning_rate": 2.337770467478939e-05,
+      "loss": 0.3136,
+      "step": 405
+    },
+    {
+      "epoch": 1.0539845758354756,
+      "grad_norm": 1.3605998754501343,
+      "learning_rate": 2.320047409266206e-05,
+      "loss": 0.3145,
+      "step": 410
+    },
+    {
+      "epoch": 1.0668380462724936,
+      "grad_norm": 1.2298566102981567,
+      "learning_rate": 2.3021595378805517e-05,
+      "loss": 0.3139,
+      "step": 415
+    },
+    {
+      "epoch": 1.0796915167095116,
+      "grad_norm": 1.417944073677063,
+      "learning_rate": 2.284110448427341e-05,
+      "loss": 0.3339,
+      "step": 420
+    },
+    {
+      "epoch": 1.0925449871465296,
+      "grad_norm": 1.3479814529418945,
+      "learning_rate": 2.2659037684135527e-05,
+      "loss": 0.3204,
+      "step": 425
+    },
+    {
+      "epoch": 1.1053984575835476,
+      "grad_norm": 1.4097672700881958,
+      "learning_rate": 2.2475431570187282e-05,
+      "loss": 0.2964,
+      "step": 430
+    },
+    {
+      "epoch": 1.1182519280205656,
+      "grad_norm": 1.4279357194900513,
+      "learning_rate": 2.2290323043595433e-05,
+      "loss": 0.301,
+      "step": 435
+    },
+    {
+      "epoch": 1.1311053984575836,
+      "grad_norm": 1.399916410446167,
+      "learning_rate": 2.210374930748172e-05,
+      "loss": 0.2832,
+      "step": 440
+    },
+    {
+      "epoch": 1.1439588688946016,
+      "grad_norm": 1.6761506795883179,
+      "learning_rate": 2.1915747859445784e-05,
+      "loss": 0.3444,
+      "step": 445
+    },
+    {
+      "epoch": 1.1568123393316196,
+      "grad_norm": 1.565943956375122,
+      "learning_rate": 2.1726356484028865e-05,
+      "loss": 0.3129,
+      "step": 450
+    },
+    {
+      "epoch": 1.1696658097686377,
+      "grad_norm": 1.512202262878418,
+      "learning_rate": 2.1535613245119913e-05,
+      "loss": 0.269,
+      "step": 455
+    },
+    {
+      "epoch": 1.1825192802056554,
+      "grad_norm": 1.3803540468215942,
+      "learning_rate": 2.1343556478305444e-05,
+      "loss": 0.3009,
+      "step": 460
+    },
+    {
+      "epoch": 1.1953727506426735,
+      "grad_norm": 1.3840749263763428,
+      "learning_rate": 2.1150224783164913e-05,
+      "loss": 0.2894,
+      "step": 465
+    },
+    {
+      "epoch": 1.2082262210796915,
+      "grad_norm": 1.244518518447876,
+      "learning_rate": 2.0955657015512916e-05,
+      "loss": 0.3098,
+      "step": 470
+    },
+    {
+      "epoch": 1.2210796915167095,
+      "grad_norm": 1.4942463636398315,
+      "learning_rate": 2.0759892279589983e-05,
+      "loss": 0.2729,
+      "step": 475
+    },
+    {
+      "epoch": 1.2339331619537275,
+      "grad_norm": 1.4378920793533325,
+      "learning_rate": 2.056296992020339e-05,
+      "loss": 0.2914,
+      "step": 480
+    },
+    {
+      "epoch": 1.2467866323907455,
+      "grad_norm": 1.4088345766067505,
+      "learning_rate": 2.036492951481963e-05,
+      "loss": 0.2792,
+      "step": 485
+    },
+    {
+      "epoch": 1.2596401028277635,
+      "grad_norm": 1.4317700862884521,
+      "learning_rate": 2.016581086561016e-05,
+      "loss": 0.2655,
+      "step": 490
+    },
+    {
+      "epoch": 1.2724935732647815,
+      "grad_norm": 1.3015400171279907,
+      "learning_rate": 1.996565399145192e-05,
+      "loss": 0.2799,
+      "step": 495
+    },
+    {
+      "epoch": 1.2853470437017995,
+      "grad_norm": 1.526664137840271,
+      "learning_rate": 1.976449911988438e-05,
+      "loss": 0.2585,
+      "step": 500
+    },
+    {
+      "epoch": 1.2982005141388175,
+      "grad_norm": 1.4342806339263916,
+      "learning_rate": 1.956238667902455e-05,
+      "loss": 0.2529,
+      "step": 505
+    },
+    {
+      "epoch": 1.3110539845758356,
+      "grad_norm": 1.2547333240509033,
+      "learning_rate": 1.93593572894418e-05,
+      "loss": 0.2631,
+      "step": 510
+    },
+    {
+      "epoch": 1.3239074550128533,
+      "grad_norm": 1.3297747373580933,
+      "learning_rate": 1.9155451755993896e-05,
+      "loss": 0.2461,
+      "step": 515
+    },
+    {
+      "epoch": 1.3367609254498714,
+      "grad_norm": 1.9114813804626465,
+      "learning_rate": 1.8950711059626032e-05,
+      "loss": 0.2488,
+      "step": 520
+    },
+    {
+      "epoch": 1.3496143958868894,
+      "grad_norm": 1.2873486280441284,
+      "learning_rate": 1.8745176349134498e-05,
+      "loss": 0.2249,
+      "step": 525
+    },
+    {
+      "epoch": 1.3624678663239074,
+      "grad_norm": 1.1826640367507935,
+      "learning_rate": 1.853888893289661e-05,
+      "loss": 0.2525,
+      "step": 530
+    },
+    {
+      "epoch": 1.3753213367609254,
+      "grad_norm": 1.2898495197296143,
+      "learning_rate": 1.833189027056848e-05,
+      "loss": 0.2436,
+      "step": 535
+    },
+    {
+      "epoch": 1.3881748071979434,
+      "grad_norm": 1.6029397249221802,
+      "learning_rate": 1.8124221964752535e-05,
+      "loss": 0.2506,
+      "step": 540
+    },
+    {
+      "epoch": 1.4010282776349614,
+      "grad_norm": 1.3479888439178467,
+      "learning_rate": 1.7915925752636164e-05,
+      "loss": 0.2413,
+      "step": 545
+    },
+    {
+      "epoch": 1.4138817480719794,
+      "grad_norm": 1.4753426313400269,
+      "learning_rate": 1.7707043497603426e-05,
+      "loss": 0.2179,
+      "step": 550
+    },
+    {
+      "epoch": 1.4267352185089974,
+      "grad_norm": 1.2107239961624146,
+      "learning_rate": 1.7497617180821287e-05,
+      "loss": 0.2244,
+      "step": 555
+    },
+    {
+      "epoch": 1.4395886889460154,
+      "grad_norm": 1.2898181676864624,
+      "learning_rate": 1.728768889280229e-05,
+      "loss": 0.2232,
+      "step": 560
+    },
+    {
+      "epoch": 1.4524421593830334,
+      "grad_norm": 1.6774007081985474,
+      "learning_rate": 1.7077300824945174e-05,
+      "loss": 0.2123,
+      "step": 565
+    },
+    {
+      "epoch": 1.4652956298200515,
+      "grad_norm": 1.2464574575424194,
+      "learning_rate": 1.6866495261055265e-05,
+      "loss": 0.2027,
+      "step": 570
+    },
+    {
+      "epoch": 1.4781491002570695,
+      "grad_norm": 1.2632747888565063,
+      "learning_rate": 1.6655314568846265e-05,
+      "loss": 0.2226,
+      "step": 575
+    },
+    {
+      "epoch": 1.4910025706940875,
+      "grad_norm": 1.2137638330459595,
+      "learning_rate": 1.6443801191425176e-05,
+      "loss": 0.1948,
+      "step": 580
+    },
+    {
+      "epoch": 1.5038560411311055,
+      "grad_norm": 1.2996569871902466,
+      "learning_rate": 1.6231997638762096e-05,
+      "loss": 0.2121,
+      "step": 585
+    },
+    {
+      "epoch": 1.5167095115681235,
+      "grad_norm": 1.252046823501587,
+      "learning_rate": 1.6019946479146567e-05,
+      "loss": 0.1898,
+      "step": 590
+    },
+    {
+      "epoch": 1.5295629820051415,
+      "grad_norm": 1.3210327625274658,
+      "learning_rate": 1.5807690330632176e-05,
+      "loss": 0.2255,
+      "step": 595
+    },
+    {
+      "epoch": 1.5424164524421595,
+      "grad_norm": 1.248801827430725,
+      "learning_rate": 1.5595271852471204e-05,
+      "loss": 0.1808,
+      "step": 600
+    },
+    {
+      "epoch": 1.5552699228791775,
+      "grad_norm": 1.257836103439331,
+      "learning_rate": 1.5382733736540936e-05,
+      "loss": 0.1956,
+      "step": 605
+    },
+    {
+      "epoch": 1.5681233933161953,
+      "grad_norm": 1.1201304197311401,
+      "learning_rate": 1.517011869876348e-05,
+      "loss": 0.1915,
+      "step": 610
+    },
+    {
+      "epoch": 1.5809768637532133,
+      "grad_norm": 1.5343998670578003,
+      "learning_rate": 1.4957469470520697e-05,
+      "loss": 0.2092,
+      "step": 615
+    },
+    {
+      "epoch": 1.5938303341902313,
+      "grad_norm": 1.2799941301345825,
+      "learning_rate": 1.4744828790066041e-05,
+      "loss": 0.1776,
+      "step": 620
+    },
+    {
+      "epoch": 1.6066838046272494,
+      "grad_norm": 1.2523736953735352,
+      "learning_rate": 1.4532239393935035e-05,
+      "loss": 0.1991,
+      "step": 625
+    },
+    {
+      "epoch": 1.6195372750642674,
+      "grad_norm": 1.3685052394866943,
+      "learning_rate": 1.4319744008356085e-05,
+      "loss": 0.1748,
+      "step": 630
+    },
+    {
+      "epoch": 1.6323907455012854,
+      "grad_norm": 1.2179163694381714,
+      "learning_rate": 1.4107385340663331e-05,
+      "loss": 0.1858,
+      "step": 635
+    },
+    {
+      "epoch": 1.6452442159383034,
+      "grad_norm": 1.6119420528411865,
+      "learning_rate": 1.3895206070713373e-05,
+      "loss": 0.1791,
+      "step": 640
+    },
+    {
+      "epoch": 1.6580976863753212,
+      "grad_norm": 1.2677887678146362,
+      "learning_rate": 1.3683248842307416e-05,
+      "loss": 0.1554,
+      "step": 645
+    },
+    {
+      "epoch": 1.6709511568123392,
+      "grad_norm": 1.2102363109588623,
+      "learning_rate": 1.3471556254620751e-05,
+      "loss": 0.1841,
+      "step": 650
+    },
+    {
+      "epoch": 1.6838046272493572,
+      "grad_norm": 1.3049168586730957,
+      "learning_rate": 1.3260170853641178e-05,
+      "loss": 0.1816,
+      "step": 655
+    },
+    {
+      "epoch": 1.6966580976863752,
+      "grad_norm": 1.2198703289031982,
+      "learning_rate": 1.3049135123618073e-05,
+      "loss": 0.1648,
+      "step": 660
+    },
+    {
+      "epoch": 1.7095115681233932,
+      "grad_norm": 1.3759807348251343,
+      "learning_rate": 1.283849147852395e-05,
+      "loss": 0.1859,
+      "step": 665
+    },
+    {
+      "epoch": 1.7223650385604112,
+      "grad_norm": 1.135671615600586,
+      "learning_rate": 1.2628282253530037e-05,
+      "loss": 0.1418,
+      "step": 670
+    },
+    {
+      "epoch": 1.7352185089974292,
+      "grad_norm": 1.3504388332366943,
+      "learning_rate": 1.2418549696497832e-05,
+      "loss": 0.1717,
+      "step": 675
+    },
+    {
+      "epoch": 1.7480719794344473,
+      "grad_norm": 1.2006444931030273,
+      "learning_rate": 1.2209335959488037e-05,
+      "loss": 0.1474,
+      "step": 680
+    },
+    {
+      "epoch": 1.7609254498714653,
+      "grad_norm": 1.6840342283248901,
+      "learning_rate": 1.2000683090288901e-05,
+      "loss": 0.1669,
+      "step": 685
+    },
+    {
+      "epoch": 1.7737789203084833,
+      "grad_norm": 1.1025441884994507,
+      "learning_rate": 1.1792633023965433e-05,
+      "loss": 0.1554,
+      "step": 690
+    },
+    {
+      "epoch": 1.7866323907455013,
+      "grad_norm": 1.1083474159240723,
+      "learning_rate": 1.1585227574431263e-05,
+      "loss": 0.1642,
+      "step": 695
+    },
+    {
+      "epoch": 1.7994858611825193,
+      "grad_norm": 1.1549644470214844,
+      "learning_rate": 1.1378508426044943e-05,
+      "loss": 0.1566,
+      "step": 700
+    },
+    {
+      "epoch": 1.8123393316195373,
+      "grad_norm": 1.1245815753936768,
+      "learning_rate": 1.1172517125232142e-05,
+      "loss": 0.134,
+      "step": 705
+    },
+    {
+      "epoch": 1.8251928020565553,
+      "grad_norm": 1.2178139686584473,
+      "learning_rate": 1.0967295072135706e-05,
+      "loss": 0.1473,
+      "step": 710
+    },
+    {
+      "epoch": 1.8380462724935733,
+      "grad_norm": 1.1886050701141357,
+      "learning_rate": 1.0762883512295037e-05,
+      "loss": 0.1359,
+      "step": 715
+    },
+    {
+      "epoch": 1.8508997429305913,
+      "grad_norm": 1.1713837385177612,
+      "learning_rate": 1.0559323528356542e-05,
+      "loss": 0.1354,
+      "step": 720
+    },
+    {
+      "epoch": 1.8637532133676094,
+      "grad_norm": 1.1516737937927246,
+      "learning_rate": 1.0356656031816873e-05,
+      "loss": 0.1415,
+      "step": 725
+    },
+    {
+      "epoch": 1.8766066838046274,
+      "grad_norm": 1.1565130949020386,
+      "learning_rate": 1.0154921754800501e-05,
+      "loss": 0.1401,
+      "step": 730
+    },
+    {
+      "epoch": 1.8894601542416454,
+      "grad_norm": 1.1375868320465088,
+      "learning_rate": 9.9541612418734e-06,
+      "loss": 0.1418,
+      "step": 735
+    },
+    {
+      "epoch": 1.9023136246786634,
+      "grad_norm": 1.3106184005737305,
+      "learning_rate": 9.754414841894377e-06,
+      "loss": 0.131,
+      "step": 740
+    },
+    {
+      "epoch": 1.9151670951156814,
+      "grad_norm": 1.120499849319458,
+      "learning_rate": 9.55572269990576e-06,
+      "loss": 0.1404,
+      "step": 745
+    },
+    {
+      "epoch": 1.9280205655526992,
+      "grad_norm": 1.2646766901016235,
+      "learning_rate": 9.358124749065068e-06,
+      "loss": 0.14,
+      "step": 750
+    },
+    {
+      "epoch": 1.9408740359897172,
+      "grad_norm": 1.2874771356582642,
+      "learning_rate": 9.161660702619186e-06,
+      "loss": 0.1332,
+      "step": 755
+    },
+    {
+      "epoch": 1.9537275064267352,
+      "grad_norm": 1.1476320028305054,
+      "learning_rate": 8.966370045922881e-06,
+      "loss": 0.1162,
+      "step": 760
+    },
+    {
+      "epoch": 1.9665809768637532,
+      "grad_norm": 1.2806767225265503,
+      "learning_rate": 8.77229202850294e-06,
+      "loss": 0.1383,
+      "step": 765
+    },
+    {
+      "epoch": 1.9794344473007712,
+      "grad_norm": 1.2029035091400146,
+      "learning_rate": 8.57946565616987e-06,
+      "loss": 0.1215,
+      "step": 770
+    },
+    {
+      "epoch": 1.9922879177377892,
+      "grad_norm": 1.167834758758545,
+      "learning_rate": 8.38792968317847e-06,
+      "loss": 0.1281,
+      "step": 775
+    },
+    {
+      "epoch": 2.005141388174807,
+      "grad_norm": 0.8894156217575073,
+      "learning_rate": 8.19772260443901e-06,
+      "loss": 0.0983,
+      "step": 780
+    },
+    {
+      "epoch": 2.017994858611825,
+      "grad_norm": 1.0540406703948975,
+      "learning_rate": 8.008882647780518e-06,
+      "loss": 0.1097,
+      "step": 785
+    },
+    {
+      "epoch": 2.030848329048843,
+      "grad_norm": 1.2565796375274658,
+      "learning_rate": 7.821447766267752e-06,
+      "loss": 0.0955,
+      "step": 790
+    },
+    {
+      "epoch": 2.043701799485861,
+      "grad_norm": 0.9756515026092529,
+      "learning_rate": 7.635455630573408e-06,
+      "loss": 0.1034,
+      "step": 795
+    },
+    {
+      "epoch": 2.056555269922879,
+      "grad_norm": 0.9432031512260437,
+      "learning_rate": 7.450943621407026e-06,
+      "loss": 0.0978,
+      "step": 800
+    },
+    {
+      "epoch": 2.069408740359897,
+      "grad_norm": 1.0281354188919067,
+      "learning_rate": 7.267948822002248e-06,
+      "loss": 0.0853,
+      "step": 805
+    },
+    {
+      "epoch": 2.082262210796915,
+      "grad_norm": 1.0572271347045898,
+      "learning_rate": 7.086508010663808e-06,
+      "loss": 0.1039,
+      "step": 810
+    },
+    {
+      "epoch": 2.095115681233933,
+      "grad_norm": 1.2560112476348877,
+      "learning_rate": 6.906657653375839e-06,
+      "loss": 0.1033,
+      "step": 815
+    },
+    {
+      "epoch": 2.107969151670951,
+      "grad_norm": 1.0965826511383057,
+      "learning_rate": 6.7284338964729366e-06,
+      "loss": 0.0935,
+      "step": 820
+    },
+    {
+      "epoch": 2.120822622107969,
+      "grad_norm": 0.7812572717666626,
+      "learning_rate": 6.5518725593754465e-06,
+      "loss": 0.0858,
+      "step": 825
+    },
+    {
+      "epoch": 2.133676092544987,
+      "grad_norm": 1.017379879951477,
+      "learning_rate": 6.377009127390523e-06,
+      "loss": 0.0987,
+      "step": 830
+    },
+    {
+      "epoch": 2.146529562982005,
+      "grad_norm": 1.188006043434143,
+      "learning_rate": 6.203878744580218e-06,
+      "loss": 0.096,
+      "step": 835
+    },
+    {
+      "epoch": 2.159383033419023,
+      "grad_norm": 0.9553003311157227,
+      "learning_rate": 6.032516206698289e-06,
+      "loss": 0.0896,
+      "step": 840
+    },
+    {
+      "epoch": 2.172236503856041,
+      "grad_norm": 0.9677857756614685,
+      "learning_rate": 5.862955954196931e-06,
+      "loss": 0.0922,
+      "step": 845
+    },
+    {
+      "epoch": 2.185089974293059,
+      "grad_norm": 0.9444218277931213,
+      "learning_rate": 5.69523206530491e-06,
+      "loss": 0.0895,
+      "step": 850
+    },
+    {
+      "epoch": 2.197943444730077,
+      "grad_norm": 0.7982380986213684,
+      "learning_rate": 5.529378249178538e-06,
+      "loss": 0.0922,
+      "step": 855
+    },
+    {
+      "epoch": 2.210796915167095,
+      "grad_norm": 1.0864180326461792,
+      "learning_rate": 5.365427839126795e-06,
+      "loss": 0.0913,
+      "step": 860
+    },
+    {
+      "epoch": 2.223650385604113,
+      "grad_norm": 0.9678763747215271,
+      "learning_rate": 5.203413785912031e-06,
+      "loss": 0.0988,
+      "step": 865
+    },
+    {
+      "epoch": 2.236503856041131,
+      "grad_norm": 0.9454684257507324,
+      "learning_rate": 5.043368651127455e-06,
+      "loss": 0.0931,
+      "step": 870
+    },
+    {
+      "epoch": 2.2493573264781492,
+      "grad_norm": 1.0195993185043335,
+      "learning_rate": 4.885324600652963e-06,
+      "loss": 0.0905,
+      "step": 875
+    },
+    {
+      "epoch": 2.2622107969151672,
+      "grad_norm": 0.9516244530677795,
+      "learning_rate": 4.729313398190428e-06,
+      "loss": 0.0799,
+      "step": 880
+    },
+    {
+      "epoch": 2.2750642673521853,
+      "grad_norm": 0.8987143635749817,
+      "learning_rate": 4.575366398879799e-06,
+      "loss": 0.083,
+      "step": 885
+    },
+    {
+      "epoch": 2.2879177377892033,
+      "grad_norm": 0.9396174550056458,
+      "learning_rate": 4.4235145429973725e-06,
+      "loss": 0.093,
+      "step": 890
+    },
+    {
+      "epoch": 2.3007712082262213,
+      "grad_norm": 0.8572186231613159,
+      "learning_rate": 4.273788349737403e-06,
+      "loss": 0.0745,
+      "step": 895
+    },
+    {
+      "epoch": 2.3136246786632393,
+      "grad_norm": 0.8053613901138306,
+      "learning_rate": 4.126217911078359e-06,
+      "loss": 0.0787,
+      "step": 900
+    },
+    {
+      "epoch": 2.3264781491002573,
+      "grad_norm": 0.9791682958602905,
+      "learning_rate": 3.980832885735027e-06,
+      "loss": 0.0907,
+      "step": 905
+    },
+    {
+      "epoch": 2.3393316195372753,
+      "grad_norm": 0.8168100118637085,
+      "learning_rate": 3.837662493197704e-06,
+      "loss": 0.0769,
+      "step": 910
+    },
+    {
+      "epoch": 2.352185089974293,
+      "grad_norm": 0.8076280951499939,
+      "learning_rate": 3.696735507859685e-06,
+      "loss": 0.088,
+      "step": 915
+    },
+    {
+      "epoch": 2.365038560411311,
+      "grad_norm": 0.8363491296768188,
+      "learning_rate": 3.5580802532341293e-06,
+      "loss": 0.0783,
+      "step": 920
+    },
+    {
+      "epoch": 2.377892030848329,
+      "grad_norm": 0.9027379751205444,
+      "learning_rate": 3.42172459626167e-06,
+      "loss": 0.0832,
+      "step": 925
+    },
+    {
+      "epoch": 2.390745501285347,
+      "grad_norm": 0.9106459021568298,
+      "learning_rate": 3.287695941709655e-06,
+      "loss": 0.0809,
+      "step": 930
+    },
+    {
+      "epoch": 2.403598971722365,
+      "grad_norm": 0.8326634168624878,
+      "learning_rate": 3.156021226664367e-06,
+      "loss": 0.0814,
+      "step": 935
+    },
+    {
+      "epoch": 2.416452442159383,
+      "grad_norm": 0.8519615530967712,
+      "learning_rate": 3.026726915117201e-06,
+      "loss": 0.081,
+      "step": 940
+    },
+    {
+      "epoch": 2.429305912596401,
+      "grad_norm": 0.7807072997093201,
+      "learning_rate": 2.899838992645918e-06,
+      "loss": 0.0714,
+      "step": 945
+    },
+    {
+      "epoch": 2.442159383033419,
+      "grad_norm": 0.8008826375007629,
+      "learning_rate": 2.7753829611920906e-06,
+      "loss": 0.0793,
+      "step": 950
+    },
+    {
+      "epoch": 2.455012853470437,
+      "grad_norm": 0.8866891860961914,
+      "learning_rate": 2.653383833935675e-06,
+      "loss": 0.0795,
+      "step": 955
+    },
+    {
+      "epoch": 2.467866323907455,
+      "grad_norm": 0.6699994802474976,
+      "learning_rate": 2.5338661302679146e-06,
+      "loss": 0.076,
+      "step": 960
+    },
+    {
+      "epoch": 2.480719794344473,
+      "grad_norm": 0.7317408919334412,
+      "learning_rate": 2.4168538708633924e-06,
+      "loss": 0.071,
+      "step": 965
+    },
+    {
+      "epoch": 2.493573264781491,
+      "grad_norm": 0.8876296877861023,
+      "learning_rate": 2.3023705728523635e-06,
+      "loss": 0.0751,
+      "step": 970
+    },
+    {
+      "epoch": 2.506426735218509,
+      "grad_norm": 0.9609207510948181,
+      "learning_rate": 2.190439245094294e-06,
+      "loss": 0.0837,
+      "step": 975
+    },
+    {
+      "epoch": 2.519280205655527,
+      "grad_norm": 0.8199177980422974,
+      "learning_rate": 2.081082383553523e-06,
+      "loss": 0.075,
+      "step": 980
+    },
+    {
+      "epoch": 2.532133676092545,
+      "grad_norm": 0.784706711769104,
+      "learning_rate": 1.9743219667780325e-06,
+      "loss": 0.0693,
+      "step": 985
+    },
+    {
+      "epoch": 2.544987146529563,
+      "grad_norm": 0.8273710608482361,
+      "learning_rate": 1.8701794514821896e-06,
+      "loss": 0.0758,
+      "step": 990
+    },
+    {
+      "epoch": 2.557840616966581,
+      "grad_norm": 1.0023750066757202,
+      "learning_rate": 1.7686757682343891e-06,
+      "loss": 0.0748,
+      "step": 995
+    },
+    {
+      "epoch": 2.570694087403599,
+      "grad_norm": 0.8674479722976685,
+      "learning_rate": 1.6698313172504127e-06,
+      "loss": 0.0766,
+      "step": 1000
+    },
+    {
+      "epoch": 2.583547557840617,
+      "grad_norm": 0.8136624693870544,
+      "learning_rate": 1.5736659642933847e-06,
+      "loss": 0.0819,
+      "step": 1005
+    },
+    {
+      "epoch": 2.596401028277635,
+      "grad_norm": 0.8119713068008423,
+      "learning_rate": 1.4801990366811836e-06,
+      "loss": 0.0723,
+      "step": 1010
+    },
+    {
+      "epoch": 2.609254498714653,
+      "grad_norm": 0.8546728491783142,
+      "learning_rate": 1.389449319402002e-06,
+      "loss": 0.0692,
+      "step": 1015
+    },
+    {
+      "epoch": 2.622107969151671,
+      "grad_norm": 0.7734676003456116,
+      "learning_rate": 1.3014350513389556e-06,
+      "loss": 0.0752,
+      "step": 1020
+    },
+    {
+      "epoch": 2.6349614395886887,
+      "grad_norm": 0.7785642743110657,
+      "learning_rate": 1.2161739216044243e-06,
+      "loss": 0.0746,
+      "step": 1025
+    },
+    {
+      "epoch": 2.6478149100257067,
+      "grad_norm": 0.8515306711196899,
+      "learning_rate": 1.1336830659849046e-06,
+      "loss": 0.0794,
+      "step": 1030
+    },
+    {
+      "epoch": 2.6606683804627247,
+      "grad_norm": 0.87235027551651,
+      "learning_rate": 1.0539790634970459e-06,
+      "loss": 0.0665,
+      "step": 1035
+    },
+    {
+      "epoch": 2.6735218508997427,
+      "grad_norm": 0.8246081471443176,
+      "learning_rate": 9.770779330555978e-07,
+      "loss": 0.0782,
+      "step": 1040
+    },
+    {
+      "epoch": 2.6863753213367607,
+      "grad_norm": 0.7857927083969116,
+      "learning_rate": 9.029951302539458e-07,
+      "loss": 0.063,
+      "step": 1045
+    },
+    {
+      "epoch": 2.6992287917737787,
+      "grad_norm": 0.7189505696296692,
+      "learning_rate": 8.317455442578231e-07,
+      "loss": 0.0672,
+      "step": 1050
+    },
+    {
+      "epoch": 2.7120822622107967,
+      "grad_norm": 0.7249512672424316,
+      "learning_rate": 7.633434948128992e-07,
+      "loss": 0.0788,
+      "step": 1055
+    },
+    {
+      "epoch": 2.7249357326478147,
+      "grad_norm": 0.6243793368339539,
+      "learning_rate": 6.978027293667943e-07,
+      "loss": 0.0661,
+      "step": 1060
+    },
+    {
+      "epoch": 2.7377892030848328,
+      "grad_norm": 0.8508261442184448,
+      "learning_rate": 6.351364203061116e-07,
+      "loss": 0.0729,
+      "step": 1065
+    },
+    {
+      "epoch": 2.7506426735218508,
+      "grad_norm": 0.7222792506217957,
+      "learning_rate": 5.753571623090531e-07,
+      "loss": 0.0708,
+      "step": 1070
+    },
+    {
+      "epoch": 2.763496143958869,
+      "grad_norm": 0.769237220287323,
+      "learning_rate": 5.184769698141445e-07,
+      "loss": 0.0705,
+      "step": 1075
+    },
+    {
+      "epoch": 2.776349614395887,
+      "grad_norm": 0.6945733428001404,
+      "learning_rate": 4.645072746055684e-07,
+      "loss": 0.0733,
+      "step": 1080
+    },
+    {
+      "epoch": 2.789203084832905,
+      "grad_norm": 0.7368326187133789,
+      "learning_rate": 4.134589235156022e-07,
+      "loss": 0.0783,
+      "step": 1085
+    },
+    {
+      "epoch": 2.802056555269923,
+      "grad_norm": 0.6354882717132568,
+      "learning_rate": 3.653421762446252e-07,
+      "loss": 0.0723,
+      "step": 1090
+    },
+    {
+      "epoch": 2.814910025706941,
+      "grad_norm": 0.7055100798606873,
+      "learning_rate": 3.201667032991212e-07,
+      "loss": 0.0768,
+      "step": 1095
+    },
+    {
+      "epoch": 2.827763496143959,
+      "grad_norm": 0.7156299948692322,
+      "learning_rate": 2.7794158404810034e-07,
+      "loss": 0.0713,
+      "step": 1100
+    },
+    {
+      "epoch": 2.840616966580977,
+      "grad_norm": 0.7245969176292419,
+      "learning_rate": 2.3867530489832643e-07,
+      "loss": 0.0818,
+      "step": 1105
+    },
+    {
+      "epoch": 2.853470437017995,
+      "grad_norm": 0.6645637154579163,
+      "learning_rate": 2.0237575758871697e-07,
+      "loss": 0.065,
+      "step": 1110
+    },
+    {
+      "epoch": 2.866323907455013,
+      "grad_norm": 0.8119460344314575,
+      "learning_rate": 1.6905023760426797e-07,
+      "loss": 0.0682,
+      "step": 1115
+    },
+    {
+      "epoch": 2.879177377892031,
+      "grad_norm": 0.6375195384025574,
+      "learning_rate": 1.3870544270979724e-07,
+      "loss": 0.07,
+      "step": 1120
+    },
+    {
+      "epoch": 2.892030848329049,
+      "grad_norm": 0.8233482241630554,
+      "learning_rate": 1.1134747160383263e-07,
+      "loss": 0.0689,
+      "step": 1125
+    },
+    {
+      "epoch": 2.904884318766067,
+      "grad_norm": 0.7489317655563354,
+      "learning_rate": 8.698182269289423e-08,
+      "loss": 0.0674,
+      "step": 1130
+    },
+    {
+      "epoch": 2.917737789203085,
+      "grad_norm": 0.6949130296707153,
+      "learning_rate": 6.561339298642999e-08,
+      "loss": 0.0704,
+      "step": 1135
+    },
+    {
+      "epoch": 2.930591259640103,
+      "grad_norm": 0.6076891422271729,
+      "learning_rate": 4.7246477112599665e-08,
+      "loss": 0.0667,
+      "step": 1140
+    },
+    {
+      "epoch": 2.943444730077121,
+      "grad_norm": 0.7674615383148193,
+      "learning_rate": 3.188476645515692e-08,
+      "loss": 0.0783,
+      "step": 1145
+    },
+    {
+      "epoch": 2.956298200514139,
+      "grad_norm": 0.7354020476341248,
+      "learning_rate": 1.953134841154447e-08,
+      "loss": 0.0668,
+      "step": 1150
+    },
+    {
+      "epoch": 2.969151670951157,
+      "grad_norm": 0.6061502695083618,
+      "learning_rate": 1.0188705772393725e-08,
+      "loss": 0.068,
+      "step": 1155
+    },
+    {
+      "epoch": 2.982005141388175,
+      "grad_norm": 0.7489121556282043,
+      "learning_rate": 3.858716222530645e-09,
+      "loss": 0.0706,
+      "step": 1160
+    },
+    {
+      "epoch": 2.994858611825193,
+      "grad_norm": 0.6796852946281433,
+      "learning_rate": 5.426519635975958e-10,
+      "loss": 0.0679,
+      "step": 1165
+    }
+  ],
+  "logging_steps": 5,
+  "max_steps": 1167,
+  "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": 1.4602586872883446e+18,
+  "train_batch_size": 2,
+  "trial_name": null,
+  "trial_params": null
+}

135_128_e3_3e-5/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:335ca983baa5044859a69642576ffac61bef0df5adec32f663ca9c38cdc8a275
+size 8209

135_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)