first commit

all_results.json

@@ -0,0 +1,14 @@
+{
+    "epoch": 6000.0,
+    "eval_loss": 6.7578125,
+    "eval_runtime": 0.0933,
+    "eval_samples": 6,
+    "eval_samples_per_second": 64.277,
+    "eval_steps_per_second": 10.713,
+    "perplexity": 860.7572284206628,
+    "train_loss": 0.032272634650270145,
+    "train_runtime": 8757.5386,
+    "train_samples": 16,
+    "train_samples_per_second": 10.962,
+    "train_steps_per_second": 0.685
+}

config.json

@@ -0,0 +1,39 @@
+{
+  "_name_or_path": "gpt2",
+  "activation_function": "gelu_new",
+  "architectures": [
+    "GPT2LMHeadModel"
+  ],
+  "attn_pdrop": 0.1,
+  "bos_token_id": 50256,
+  "embd_pdrop": 0.1,
+  "eos_token_id": 50256,
+  "initializer_range": 0.02,
+  "layer_norm_epsilon": 1e-05,
+  "model_type": "gpt2",
+  "n_ctx": 1024,
+  "n_embd": 768,
+  "n_head": 12,
+  "n_inner": null,
+  "n_layer": 12,
+  "n_positions": 1024,
+  "reorder_and_upcast_attn": false,
+  "resid_pdrop": 0.1,
+  "scale_attn_by_inverse_layer_idx": false,
+  "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
+    }
+  },
+  "torch_dtype": "float16",
+  "transformers_version": "4.13.0",
+  "use_cache": false,
+  "vocab_size": 50257
+}

eval_results.json

@@ -0,0 +1,9 @@
+{
+    "epoch": 6000.0,
+    "eval_loss": 6.7578125,
+    "eval_runtime": 0.0933,
+    "eval_samples": 6,
+    "eval_samples_per_second": 64.277,
+    "eval_steps_per_second": 10.713,
+    "perplexity": 860.7572284206628
+}

pytorch_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:69cecef7afbe220a908f56c7db0d90e7f967f7493bfbb9e6c15ecb7505d3b831
+size 261495703

runs/Dec27_14-25-14_louis-gpu/1640611527.5667238/events.out.tfevents.1640611527.louis-gpu.524994.1

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7ec65cd63aa258e43e61cf3811726473e4d6ee955b65ddd4b2f57c4c9014c17f
+size 4685

runs/Dec27_14-25-14_louis-gpu/events.out.tfevents.1640611527.louis-gpu.524994.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:892349d299f41243d66cfa15a9576d1e554159343648f7f3c93bf6ad9f3826c6
+size 4121

runs/Dec27_14-25-14_louis-gpu/events.out.tfevents.1640611815.louis-gpu.524994.2

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0dc927c4a812500f3048e44100ebb2880fe0e294c57813a86c6de54e47bc0f5a
+size 311

runs/Dec27_14-31-31_louis-gpu/1640611904.130664/events.out.tfevents.1640611904.louis-gpu.554922.1

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6482c74293c5e965bc99557736c1cff5e2921970642b1aa3fb22a6769c99a3ba
+size 4685

runs/Dec27_14-31-31_louis-gpu/events.out.tfevents.1640611904.louis-gpu.554922.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bb1ba1633e5b652808f3a87ff26040f8c3b43d7372525983c5090bce1ef1f149
+size 13864

runs/Dec27_14-31-31_louis-gpu/events.out.tfevents.1640620664.louis-gpu.554922.2

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2458cde1503c48de144697eae42aea32b3dcf5568403eb911acf27da93c3dea0
+size 311

special_tokens_map.json

@@ -0,0 +1 @@
+{"bos_token": "<|endoftext|>", "eos_token": "<|endoftext|>", "unk_token": "<|endoftext|>"}

tokenizer_config.json

@@ -0,0 +1 @@
+{"unk_token": "<|endoftext|>", "bos_token": "<|endoftext|>", "eos_token": "<|endoftext|>", "add_prefix_space": false, "model_max_length": 1024, "special_tokens_map_file": null, "name_or_path": "gpt2", "tokenizer_class": "GPT2Tokenizer"}

train_results.json

@@ -0,0 +1,8 @@
+{
+    "epoch": 6000.0,
+    "train_loss": 0.032272634650270145,
+    "train_runtime": 8757.5386,
+    "train_samples": 16,
+    "train_samples_per_second": 10.962,
+    "train_steps_per_second": 0.685
+}

trainer_state.json

@@ -0,0 +1,337 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 6000.0,
+  "global_step": 6000,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 200.0,
+      "eval_loss": 4.5546875,
+      "eval_runtime": 0.0952,
+      "eval_samples_per_second": 63.02,
+      "eval_steps_per_second": 10.503,
+      "step": 200
+    },
+    {
+      "epoch": 400.0,
+      "eval_loss": 5.13671875,
+      "eval_runtime": 0.0952,
+      "eval_samples_per_second": 62.996,
+      "eval_steps_per_second": 10.499,
+      "step": 400
+    },
+    {
+      "epoch": 500.0,
+      "learning_rate": 5e-05,
+      "loss": 0.3672,
+      "step": 500
+    },
+    {
+      "epoch": 600.0,
+      "eval_loss": 5.375,
+      "eval_runtime": 0.0955,
+      "eval_samples_per_second": 62.813,
+      "eval_steps_per_second": 10.469,
+      "step": 600
+    },
+    {
+      "epoch": 800.0,
+      "eval_loss": 5.5390625,
+      "eval_runtime": 0.0955,
+      "eval_samples_per_second": 62.807,
+      "eval_steps_per_second": 10.468,
+      "step": 800
+    },
+    {
+      "epoch": 1000.0,
+      "learning_rate": 5e-05,
+      "loss": 0.0064,
+      "step": 1000
+    },
+    {
+      "epoch": 1000.0,
+      "eval_loss": 5.6484375,
+      "eval_runtime": 0.0939,
+      "eval_samples_per_second": 63.886,
+      "eval_steps_per_second": 10.648,
+      "step": 1000
+    },
+    {
+      "epoch": 1200.0,
+      "eval_loss": 5.76171875,
+      "eval_runtime": 0.0951,
+      "eval_samples_per_second": 63.073,
+      "eval_steps_per_second": 10.512,
+      "step": 1200
+    },
+    {
+      "epoch": 1400.0,
+      "eval_loss": 5.83203125,
+      "eval_runtime": 0.0957,
+      "eval_samples_per_second": 62.694,
+      "eval_steps_per_second": 10.449,
+      "step": 1400
+    },
+    {
+      "epoch": 1500.0,
+      "learning_rate": 5e-05,
+      "loss": 0.0032,
+      "step": 1500
+    },
+    {
+      "epoch": 1600.0,
+      "eval_loss": 5.93359375,
+      "eval_runtime": 0.0954,
+      "eval_samples_per_second": 62.898,
+      "eval_steps_per_second": 10.483,
+      "step": 1600
+    },
+    {
+      "epoch": 1800.0,
+      "eval_loss": 5.953125,
+      "eval_runtime": 0.0955,
+      "eval_samples_per_second": 62.854,
+      "eval_steps_per_second": 10.476,
+      "step": 1800
+    },
+    {
+      "epoch": 2000.0,
+      "learning_rate": 5e-05,
+      "loss": 0.0024,
+      "step": 2000
+    },
+    {
+      "epoch": 2000.0,
+      "eval_loss": 6.03515625,
+      "eval_runtime": 0.0935,
+      "eval_samples_per_second": 64.153,
+      "eval_steps_per_second": 10.692,
+      "step": 2000
+    },
+    {
+      "epoch": 2200.0,
+      "eval_loss": 6.1015625,
+      "eval_runtime": 0.0954,
+      "eval_samples_per_second": 62.861,
+      "eval_steps_per_second": 10.477,
+      "step": 2200
+    },
+    {
+      "epoch": 2400.0,
+      "eval_loss": 6.15625,
+      "eval_runtime": 0.0954,
+      "eval_samples_per_second": 62.877,
+      "eval_steps_per_second": 10.479,
+      "step": 2400
+    },
+    {
+      "epoch": 2500.0,
+      "learning_rate": 5e-05,
+      "loss": 0.0018,
+      "step": 2500
+    },
+    {
+      "epoch": 2600.0,
+      "eval_loss": 6.26171875,
+      "eval_runtime": 0.0955,
+      "eval_samples_per_second": 62.805,
+      "eval_steps_per_second": 10.468,
+      "step": 2600
+    },
+    {
+      "epoch": 2800.0,
+      "eval_loss": 6.25390625,
+      "eval_runtime": 0.0952,
+      "eval_samples_per_second": 63.053,
+      "eval_steps_per_second": 10.509,
+      "step": 2800
+    },
+    {
+      "epoch": 3000.0,
+      "learning_rate": 5e-05,
+      "loss": 0.0014,
+      "step": 3000
+    },
+    {
+      "epoch": 3000.0,
+      "eval_loss": 6.32421875,
+      "eval_runtime": 0.0938,
+      "eval_samples_per_second": 63.961,
+      "eval_steps_per_second": 10.66,
+      "step": 3000
+    },
+    {
+      "epoch": 3200.0,
+      "eval_loss": 6.28125,
+      "eval_runtime": 0.0934,
+      "eval_samples_per_second": 64.273,
+      "eval_steps_per_second": 10.712,
+      "step": 3200
+    },
+    {
+      "epoch": 3400.0,
+      "eval_loss": 6.34765625,
+      "eval_runtime": 0.0935,
+      "eval_samples_per_second": 64.183,
+      "eval_steps_per_second": 10.697,
+      "step": 3400
+    },
+    {
+      "epoch": 3500.0,
+      "learning_rate": 5e-05,
+      "loss": 0.0014,
+      "step": 3500
+    },
+    {
+      "epoch": 3600.0,
+      "eval_loss": 6.36328125,
+      "eval_runtime": 0.0935,
+      "eval_samples_per_second": 64.157,
+      "eval_steps_per_second": 10.693,
+      "step": 3600
+    },
+    {
+      "epoch": 3800.0,
+      "eval_loss": 6.5078125,
+      "eval_runtime": 0.0954,
+      "eval_samples_per_second": 62.909,
+      "eval_steps_per_second": 10.485,
+      "step": 3800
+    },
+    {
+      "epoch": 4000.0,
+      "learning_rate": 5e-05,
+      "loss": 0.0009,
+      "step": 4000
+    },
+    {
+      "epoch": 4000.0,
+      "eval_loss": 6.5546875,
+      "eval_runtime": 0.0937,
+      "eval_samples_per_second": 64.049,
+      "eval_steps_per_second": 10.675,
+      "step": 4000
+    },
+    {
+      "epoch": 4200.0,
+      "eval_loss": 6.46484375,
+      "eval_runtime": 0.0954,
+      "eval_samples_per_second": 62.864,
+      "eval_steps_per_second": 10.477,
+      "step": 4200
+    },
+    {
+      "epoch": 4400.0,
+      "eval_loss": 6.59765625,
+      "eval_runtime": 0.0937,
+      "eval_samples_per_second": 64.005,
+      "eval_steps_per_second": 10.667,
+      "step": 4400
+    },
+    {
+      "epoch": 4500.0,
+      "learning_rate": 5e-05,
+      "loss": 0.0008,
+      "step": 4500
+    },
+    {
+      "epoch": 4600.0,
+      "eval_loss": 6.640625,
+      "eval_runtime": 0.0955,
+      "eval_samples_per_second": 62.799,
+      "eval_steps_per_second": 10.467,
+      "step": 4600
+    },
+    {
+      "epoch": 4800.0,
+      "eval_loss": 6.625,
+      "eval_runtime": 0.0954,
+      "eval_samples_per_second": 62.869,
+      "eval_steps_per_second": 10.478,
+      "step": 4800
+    },
+    {
+      "epoch": 5000.0,
+      "learning_rate": 5e-05,
+      "loss": 0.0005,
+      "step": 5000
+    },
+    {
+      "epoch": 5000.0,
+      "eval_loss": 6.66796875,
+      "eval_runtime": 0.0937,
+      "eval_samples_per_second": 64.004,
+      "eval_steps_per_second": 10.667,
+      "step": 5000
+    },
+    {
+      "epoch": 5200.0,
+      "eval_loss": 6.671875,
+      "eval_runtime": 0.0955,
+      "eval_samples_per_second": 62.801,
+      "eval_steps_per_second": 10.467,
+      "step": 5200
+    },
+    {
+      "epoch": 5400.0,
+      "eval_loss": 6.78125,
+      "eval_runtime": 0.0955,
+      "eval_samples_per_second": 62.857,
+      "eval_steps_per_second": 10.476,
+      "step": 5400
+    },
+    {
+      "epoch": 5500.0,
+      "learning_rate": 5e-05,
+      "loss": 0.0005,
+      "step": 5500
+    },
+    {
+      "epoch": 5600.0,
+      "eval_loss": 6.83984375,
+      "eval_runtime": 0.0958,
+      "eval_samples_per_second": 62.636,
+      "eval_steps_per_second": 10.439,
+      "step": 5600
+    },
+    {
+      "epoch": 5800.0,
+      "eval_loss": 6.64453125,
+      "eval_runtime": 0.0954,
+      "eval_samples_per_second": 62.879,
+      "eval_steps_per_second": 10.48,
+      "step": 5800
+    },
+    {
+      "epoch": 6000.0,
+      "learning_rate": 5e-05,
+      "loss": 0.0007,
+      "step": 6000
+    },
+    {
+      "epoch": 6000.0,
+      "eval_loss": 6.7578125,
+      "eval_runtime": 0.0935,
+      "eval_samples_per_second": 64.204,
+      "eval_steps_per_second": 10.701,
+      "step": 6000
+    },
+    {
+      "epoch": 6000.0,
+      "step": 6000,
+      "total_flos": 5.016807094930637e+16,
+      "train_loss": 0.032272634650270145,
+      "train_runtime": 8757.5386,
+      "train_samples_per_second": 10.962,
+      "train_steps_per_second": 0.685
+    }
+  ],
+  "max_steps": 6000,
+  "num_train_epochs": 6000,
+  "total_flos": 5.016807094930637e+16,
+  "trial_name": null,
+  "trial_params": null
+}

training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e1849b6eaf414c68539e989e27c22fac5ac73af887a82d599007c57f86465dee
+size 3887

zero_to_fp32.py

@@ -0,0 +1,453 @@
+#!/usr/bin/env python
+
+# 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
+from collections import OrderedDict
+
+# 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.
+import deepspeed
+from deepspeed.utils import logger
+
+debug = 0
+
+# load to cpu
+device = torch.device('cpu')
+
+
+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_optim_files(checkpoint_dir):
+    # XXX: need to test that this simple glob rule works for multi-node setup too
+    optim_files = sorted(glob.glob(os.path.join(checkpoint_dir, "*_optim_states.pt")))
+
+    if len(optim_files) == 0:
+        raise FileNotFoundError(
+            f"can't find '*_optim_states.pt' files in directory '{checkpoint_dir}'")
+
+    return optim_files
+
+
+def parse_model_state(file):
+    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
+    }
+    return buffers
+
+
+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"]
+    param_shapes = state_dicts[0]["param_shapes"]
+    # 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, param_shapes, 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, param_shapes, 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_file = get_model_state_file(ds_checkpoint_dir, zero_stage)
+    buffers = parse_model_state(model_file)
+
+    if zero_stage == 2:
+        return _get_fp32_state_dict_from_zero2_checkpoint(world_size,
+                                                          param_shapes,
+                                                          fp32_flat_groups,
+                                                          buffers)
+    elif zero_stage == 3:
+        return _get_fp32_state_dict_from_zero3_checkpoint(world_size,
+                                                          param_shapes,
+                                                          fp32_flat_groups,
+                                                          buffers)
+
+
+def _get_fp32_state_dict_from_zero2_checkpoint(world_size,
+                                               param_shapes,
+                                               fp32_flat_groups,
+                                               buffers):
+
+    # 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.")
+
+    state_dict = OrderedDict()
+
+    # buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    # 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"
+    )
+
+    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 _get_fp32_state_dict_from_zero3_checkpoint(world_size,
+                                               param_shapes,
+                                               fp32_flat_groups,
+                                               buffers):
+
+    # 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
+
+    avail_numel = fp32_flat_groups[0].numel() * world_size
+    # 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
+        print(f"Have {avail_numel} numels to process.")
+        print(f"Need {wanted_numel} numels in {wanted_params} params.")
+
+    state_dict = OrderedDict()
+
+    # buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    # 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"{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 fp32 state dict with {total_params} params {total_numel} elements"
+    )
+
+    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)