Upload 14 files

added_tokens.json

@@ -0,0 +1,3 @@
+{
+  "[PAD]": 50257
+}

config.json

@@ -0,0 +1,42 @@
+{
+  "_name_or_path": "Salesforce/codegen-350M-nl",
+  "activation_function": "gelu_new",
+  "architectures": [
+    "CodeGenForCausalLM"
+  ],
+  "attn_pdrop": 0.0,
+  "bos_token_id": 1,
+  "embd_pdrop": 0.0,
+  "eos_token_id": 50256,
+  "gradient_checkpointing": false,
+  "initializer_range": 0.02,
+  "layer_norm_epsilon": 1e-05,
+  "model_type": "codegen",
+  "n_ctx": 2048,
+  "n_embd": 1024,
+  "n_head": 16,
+  "n_inner": null,
+  "n_layer": 20,
+  "n_positions": 2048,
+  "resid_pdrop": 0.0,
+  "rotary_dim": 32,
+  "scale_attn_weights": true,
+  "summary_activation": null,
+  "summary_first_dropout": 0.1,
+  "summary_proj_to_labels": true,
+  "summary_type": "cls_index",
+  "summary_use_proj": true,
+  "task_specific_params": {
+    "text-generation": {
+      "do_sample": true,
+      "max_length": 50,
+      "temperature": 1.0
+    }
+  },
+  "tie_word_embeddings": false,
+  "tokenizer_class": "GPT2Tokenizer",
+  "torch_dtype": "float16",
+  "transformers_version": "4.29.2",
+  "use_cache": false,
+  "vocab_size": 50258
+}

generation_config.json

@@ -0,0 +1,6 @@
+{
+  "_from_model_config": true,
+  "bos_token_id": 1,
+  "eos_token_id": 50256,
+  "transformers_version": "4.29.2"
+}

latest

@@ -0,0 +1 @@
+global_step200

pytorch_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c3eed59218a9957c2c3158fb8fff8469556cd818f82f0e1cddad5609020eefde
+size 793507279

rng_state.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:674676e662eeb93778c2b153ffad13aa90b43355da1956ce0b1e01e72f48c8d7
+size 14503

special_tokens_map.json

@@ -0,0 +1,6 @@
+{
+  "bos_token": "<|endoftext|>",
+  "eos_token": "<|endoftext|>",
+  "pad_token": "[PAD]",
+  "unk_token": "<|endoftext|>"
+}

tokenizer_config.json

@@ -0,0 +1,10 @@
+{
+  "add_prefix_space": false,
+  "bos_token": "<|endoftext|>",
+  "clean_up_tokenization_spaces": true,
+  "eos_token": "<|endoftext|>",
+  "model_max_length": 514,
+  "padding_side": "right",
+  "tokenizer_class": "CodeGenTokenizer",
+  "unk_token": "<|endoftext|>"
+}

trainer_state.json

@@ -0,0 +1,616 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 2.7210884353741496,
+  "global_step": 200,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.03,
+      "learning_rate": 0.0,
+      "loss": 1.1564,
+      "step": 2
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 0.0,
+      "loss": 1.4501,
+      "step": 4
+    },
+    {
+      "epoch": 0.08,
+      "learning_rate": 0.0,
+      "loss": 0.9663,
+      "step": 6
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 0.0,
+      "loss": 1.1779,
+      "step": 8
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 0.0,
+      "loss": 1.0604,
+      "step": 10
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 0.0,
+      "loss": 0.9114,
+      "step": 12
+    },
+    {
+      "epoch": 0.19,
+      "learning_rate": 0.0,
+      "loss": 1.0516,
+      "step": 14
+    },
+    {
+      "epoch": 0.22,
+      "learning_rate": 0.0,
+      "loss": 1.2253,
+      "step": 16
+    },
+    {
+      "epoch": 0.24,
+      "learning_rate": 1.3333333333333334e-06,
+      "loss": 0.9658,
+      "step": 18
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 2.666666666666667e-06,
+      "loss": 0.9407,
+      "step": 20
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 4.000000000000001e-06,
+      "loss": 0.9526,
+      "step": 22
+    },
+    {
+      "epoch": 0.33,
+      "learning_rate": 5.333333333333334e-06,
+      "loss": 0.8532,
+      "step": 24
+    },
+    {
+      "epoch": 0.35,
+      "learning_rate": 6.666666666666667e-06,
+      "loss": 0.9836,
+      "step": 26
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 8.000000000000001e-06,
+      "loss": 0.6827,
+      "step": 28
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 9.333333333333334e-06,
+      "loss": 1.0519,
+      "step": 30
+    },
+    {
+      "epoch": 0.44,
+      "learning_rate": 1.0666666666666667e-05,
+      "loss": 0.7011,
+      "step": 32
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 1.2e-05,
+      "loss": 0.8513,
+      "step": 34
+    },
+    {
+      "epoch": 0.49,
+      "learning_rate": 1.3333333333333333e-05,
+      "loss": 0.781,
+      "step": 36
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 1.4666666666666666e-05,
+      "loss": 0.7113,
+      "step": 38
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 1.6000000000000003e-05,
+      "loss": 0.791,
+      "step": 40
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 1.7333333333333336e-05,
+      "loss": 0.7299,
+      "step": 42
+    },
+    {
+      "epoch": 0.6,
+      "learning_rate": 1.866666666666667e-05,
+      "loss": 0.7265,
+      "step": 44
+    },
+    {
+      "epoch": 0.63,
+      "learning_rate": 2e-05,
+      "loss": 0.7213,
+      "step": 46
+    },
+    {
+      "epoch": 0.65,
+      "learning_rate": 1.999447456932676e-05,
+      "loss": 0.7056,
+      "step": 48
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 1.997790438338385e-05,
+      "loss": 0.7687,
+      "step": 50
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 1.9950307753654016e-05,
+      "loss": 0.6776,
+      "step": 52
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 1.991171517679013e-05,
+      "loss": 0.6322,
+      "step": 54
+    },
+    {
+      "epoch": 0.76,
+      "learning_rate": 1.9862169300913784e-05,
+      "loss": 0.7253,
+      "step": 56
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 1.9801724878485438e-05,
+      "loss": 0.6503,
+      "step": 58
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 1.973044870579824e-05,
+      "loss": 0.7397,
+      "step": 60
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 1.964841954916235e-05,
+      "loss": 0.8342,
+      "step": 62
+    },
+    {
+      "epoch": 0.87,
+      "learning_rate": 1.955572805786141e-05,
+      "loss": 0.7061,
+      "step": 64
+    },
+    {
+      "epoch": 0.9,
+      "learning_rate": 1.945247666397725e-05,
+      "loss": 0.6767,
+      "step": 66
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 1.9338779469193638e-05,
+      "loss": 0.7581,
+      "step": 68
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 1.921476211870408e-05,
+      "loss": 0.6546,
+      "step": 70
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 1.908056166236305e-05,
+      "loss": 0.8286,
+      "step": 72
+    },
+    {
+      "epoch": 1.01,
+      "learning_rate": 1.8936326403234125e-05,
+      "loss": 0.8036,
+      "step": 74
+    },
+    {
+      "epoch": 1.03,
+      "learning_rate": 1.8782215733702286e-05,
+      "loss": 0.7079,
+      "step": 76
+    },
+    {
+      "epoch": 1.06,
+      "learning_rate": 1.8618399959331642e-05,
+      "loss": 0.6375,
+      "step": 78
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 1.844506011066308e-05,
+      "loss": 0.4525,
+      "step": 80
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 1.826238774315995e-05,
+      "loss": 0.5356,
+      "step": 82
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 1.8070584725522763e-05,
+      "loss": 0.536,
+      "step": 84
+    },
+    {
+      "epoch": 1.17,
+      "learning_rate": 1.7869863016606893e-05,
+      "loss": 0.6215,
+      "step": 86
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 1.766044443118978e-05,
+      "loss": 0.4808,
+      "step": 88
+    },
+    {
+      "epoch": 1.22,
+      "learning_rate": 1.7442560394846518e-05,
+      "loss": 0.5008,
+      "step": 90
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 1.7216451688204623e-05,
+      "loss": 0.6458,
+      "step": 92
+    },
+    {
+      "epoch": 1.28,
+      "learning_rate": 1.698236818086073e-05,
+      "loss": 0.5105,
+      "step": 94
+    },
+    {
+      "epoch": 1.31,
+      "learning_rate": 1.6740568555253153e-05,
+      "loss": 0.4312,
+      "step": 96
+    },
+    {
+      "epoch": 1.33,
+      "learning_rate": 1.649132002079552e-05,
+      "loss": 0.5643,
+      "step": 98
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 1.6234898018587336e-05,
+      "loss": 0.6065,
+      "step": 100
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 1.5971585917027864e-05,
+      "loss": 0.4018,
+      "step": 102
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 1.570167469866962e-05,
+      "loss": 0.496,
+      "step": 104
+    },
+    {
+      "epoch": 1.44,
+      "learning_rate": 1.5425462638657597e-05,
+      "loss": 0.429,
+      "step": 106
+    },
+    {
+      "epoch": 1.47,
+      "learning_rate": 1.5143254975109538e-05,
+      "loss": 0.6738,
+      "step": 108
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 1.4855363571801523e-05,
+      "loss": 0.5518,
+      "step": 110
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 1.4562106573531632e-05,
+      "loss": 0.5336,
+      "step": 112
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 1.4263808054542541e-05,
+      "loss": 0.6872,
+      "step": 114
+    },
+    {
+      "epoch": 1.58,
+      "learning_rate": 1.396079766039157e-05,
+      "loss": 0.5226,
+      "step": 116
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 1.3653410243663953e-05,
+      "loss": 0.6211,
+      "step": 118
+    },
+    {
+      "epoch": 1.63,
+      "learning_rate": 1.3341985493931877e-05,
+      "loss": 0.5429,
+      "step": 120
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 1.3026867562368262e-05,
+      "loss": 0.6971,
+      "step": 122
+    },
+    {
+      "epoch": 1.69,
+      "learning_rate": 1.2708404681430054e-05,
+      "loss": 0.5812,
+      "step": 124
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 1.238694878003138e-05,
+      "loss": 0.4746,
+      "step": 126
+    },
+    {
+      "epoch": 1.74,
+      "learning_rate": 1.2062855094631777e-05,
+      "loss": 0.4972,
+      "step": 128
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 1.1736481776669307e-05,
+      "loss": 0.4216,
+      "step": 130
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 1.1408189496772369e-05,
+      "loss": 0.5295,
+      "step": 132
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 1.1078341046187588e-05,
+      "loss": 0.4123,
+      "step": 134
+    },
+    {
+      "epoch": 1.85,
+      "learning_rate": 1.0747300935864245e-05,
+      "loss": 0.5449,
+      "step": 136
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 1.0415434993638269e-05,
+      "loss": 0.5406,
+      "step": 138
+    },
+    {
+      "epoch": 1.9,
+      "learning_rate": 1.0083109959960974e-05,
+      "loss": 0.553,
+      "step": 140
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 9.750693082619274e-06,
+      "loss": 0.4892,
+      "step": 142
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 9.418551710895243e-06,
+      "loss": 0.5435,
+      "step": 144
+    },
+    {
+      "epoch": 1.99,
+      "learning_rate": 9.087052889613519e-06,
+      "loss": 0.4862,
+      "step": 146
+    },
+    {
+      "epoch": 2.01,
+      "learning_rate": 8.756562953525151e-06,
+      "loss": 0.4767,
+      "step": 148
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 8.427447122476148e-06,
+      "loss": 0.3959,
+      "step": 150
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 8.100069097808103e-06,
+      "loss": 0.3562,
+      "step": 152
+    },
+    {
+      "epoch": 2.1,
+      "learning_rate": 7.774790660436857e-06,
+      "loss": 0.3327,
+      "step": 154
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 7.451971271053455e-06,
+      "loss": 0.352,
+      "step": 156
+    },
+    {
+      "epoch": 2.15,
+      "learning_rate": 7.131967672889101e-06,
+      "loss": 0.36,
+      "step": 158
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 6.815133497483157e-06,
+      "loss": 0.391,
+      "step": 160
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 6.501818873889856e-06,
+      "loss": 0.3324,
+      "step": 162
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 6.192370041755505e-06,
+      "loss": 0.2525,
+      "step": 164
+    },
+    {
+      "epoch": 2.26,
+      "learning_rate": 5.887128968693887e-06,
+      "loss": 0.2986,
+      "step": 166
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 5.586432972382561e-06,
+      "loss": 0.3165,
+      "step": 168
+    },
+    {
+      "epoch": 2.31,
+      "learning_rate": 5.290614347797802e-06,
+      "loss": 0.3816,
+      "step": 170
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 5.000000000000003e-06,
+      "loss": 0.3999,
+      "step": 172
+    },
+    {
+      "epoch": 2.37,
+      "learning_rate": 4.714911082875446e-06,
+      "loss": 0.3504,
+      "step": 174
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 4.435662644233594e-06,
+      "loss": 0.3623,
+      "step": 176
+    },
+    {
+      "epoch": 2.42,
+      "learning_rate": 4.162563277652104e-06,
+      "loss": 0.4049,
+      "step": 178
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 3.89591478145437e-06,
+      "loss": 0.3554,
+      "step": 180
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 3.636011825196365e-06,
+      "loss": 0.4636,
+      "step": 182
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 3.3831416240314085e-06,
+      "loss": 0.2884,
+      "step": 184
+    },
+    {
+      "epoch": 2.53,
+      "learning_rate": 3.1375836213126653e-06,
+      "loss": 0.2063,
+      "step": 186
+    },
+    {
+      "epoch": 2.56,
+      "learning_rate": 2.8996091797841976e-06,
+      "loss": 0.4578,
+      "step": 188
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.669481281701739e-06,
+      "loss": 0.2891,
+      "step": 190
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.447454238214654e-06,
+      "loss": 0.3871,
+      "step": 192
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.2337734083302164e-06,
+      "loss": 0.3701,
+      "step": 194
+    },
+    {
+      "epoch": 2.67,
+      "learning_rate": 2.0286749277707783e-06,
+      "loss": 0.236,
+      "step": 196
+    },
+    {
+      "epoch": 2.69,
+      "learning_rate": 1.8323854480234348e-06,
+      "loss": 0.2797,
+      "step": 198
+    },
+    {
+      "epoch": 2.72,
+      "learning_rate": 1.6451218858706374e-06,
+      "loss": 0.2397,
+      "step": 200
+    }
+  ],
+  "max_steps": 219,
+  "num_train_epochs": 3,
+  "total_flos": 1906858253312.0,
+  "trial_name": null,
+  "trial_params": null
+}

training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3605f28a8901ab7bfced047095e893870a362227b8573e1ca4b3bc0bac1dc4c4
+size 4975

zero_to_fp32.py

@@ -0,0 +1,578 @@
+#!/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 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_dicts.append(torch.load(f, map_location=device))
+
+    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):
+    """
+    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)
+    elif zero_stage == 3:
+        return _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states)
+
+
+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 _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()
+            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):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    _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):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    _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):
+    """
+    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``
+
+    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)
+
+
+def convert_zero_checkpoint_to_fp32_state_dict(checkpoint_dir, output_file, tag=None):
+    """
+    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``
+    """
+
+    state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag)
+    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("-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)