add model deps

README.md

@@ -0,0 +1,13 @@
+---
+license: llama2
+language:
+- th
+- en
+metrics:
+- accuracy
+pipeline_tag: text-generation
+tags:
+- openthai
+- pytorch
+library_name: transformers
+---

config.json

@@ -0,0 +1,29 @@
+{
+  "_name_or_path": "/project/lt900048-ai24tn/models/openthaigpt/openthai-llama-21000",
+  "architectures": [
+    "LlamaForCausalLM"
+  ],
+  "attention_bias": false,
+  "attention_dropout": 0.0,
+  "bos_token_id": 1,
+  "eos_token_id": 2,
+  "hidden_act": "silu",
+  "hidden_size": 4096,
+  "initializer_range": 0.02,
+  "intermediate_size": 11008,
+  "max_position_embeddings": 4096,
+  "model_type": "llama",
+  "num_attention_heads": 32,
+  "num_hidden_layers": 32,
+  "num_key_value_heads": 32,
+  "pad_token_id": 0,
+  "pretraining_tp": 1,
+  "rms_norm_eps": 1e-05,
+  "rope_scaling": null,
+  "rope_theta": 10000.0,
+  "tie_word_embeddings": false,
+  "torch_dtype": "float32",
+  "transformers_version": "4.40.1",
+  "use_cache": true,
+  "vocab_size": 56541
+}

generation_config.json

@@ -0,0 +1,7 @@
+{
+  "_from_model_config": true,
+  "bos_token_id": 1,
+  "eos_token_id": 2,
+  "pad_token_id": 0,
+  "transformers_version": "4.40.1"
+}

latest

@@ -0,0 +1 @@
+global_step75

special_tokens_map.json

@@ -0,0 +1,24 @@
+{
+  "bos_token": {
+    "content": "<s>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "eos_token": {
+    "content": "</s>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": "</s>",
+  "unk_token": {
+    "content": "<unk>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}

tokenizer.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:77901fd826845fd170d28f7c5eb4edf599ee7baa58e9e1c5bd4f527ca26f4893
+size 1167962

tokenizer_config.json

@@ -0,0 +1,44 @@
+{
+  "add_bos_token": true,
+  "add_eos_token": false,
+  "add_prefix_space": true,
+  "added_tokens_decoder": {
+    "0": {
+      "content": "<unk>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "1": {
+      "content": "<s>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "2": {
+      "content": "</s>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    }
+  },
+  "bos_token": "<s>",
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "</s>",
+  "legacy": true,
+  "model_max_length": 2048,
+  "pad_token": "</s>",
+  "padding_side": "right",
+  "sp_model_kwargs": {},
+  "spaces_between_special_tokens": false,
+  "tokenizer_class": "LlamaTokenizer",
+  "unk_token": "<unk>",
+  "use_default_system_prompt": true,
+  "use_fast": false
+}

trainer_state.json

@@ -0,0 +1,555 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.0,
+  "eval_steps": 500,
+  "global_step": 75,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.04,
+      "grad_norm": 20.00289492370743,
+      "learning_rate": 0.0,
+      "loss": 1.5756,
+      "step": 1
+    },
+    {
+      "epoch": 0.08,
+      "grad_norm": 169.81482789708753,
+      "learning_rate": 5e-05,
+      "loss": 14.3496,
+      "step": 2
+    },
+    {
+      "epoch": 0.12,
+      "grad_norm": 170.94221266882175,
+      "learning_rate": 7.924812503605781e-05,
+      "loss": 14.3495,
+      "step": 3
+    },
+    {
+      "epoch": 0.16,
+      "grad_norm": 96.07266456886373,
+      "learning_rate": 0.0001,
+      "loss": 5.728,
+      "step": 4
+    },
+    {
+      "epoch": 0.2,
+      "grad_norm": 74.89495923839695,
+      "learning_rate": 0.0001,
+      "loss": 3.198,
+      "step": 5
+    },
+    {
+      "epoch": 0.24,
+      "grad_norm": 16.06625851910231,
+      "learning_rate": 9.859154929577466e-05,
+      "loss": 1.711,
+      "step": 6
+    },
+    {
+      "epoch": 0.28,
+      "grad_norm": 11.432322037094815,
+      "learning_rate": 9.718309859154931e-05,
+      "loss": 1.1097,
+      "step": 7
+    },
+    {
+      "epoch": 0.32,
+      "grad_norm": 10.90988949596415,
+      "learning_rate": 9.577464788732394e-05,
+      "loss": 1.1647,
+      "step": 8
+    },
+    {
+      "epoch": 0.36,
+      "grad_norm": 5.166602608719267,
+      "learning_rate": 9.43661971830986e-05,
+      "loss": 0.6874,
+      "step": 9
+    },
+    {
+      "epoch": 0.4,
+      "grad_norm": 5.063308573446352,
+      "learning_rate": 9.295774647887325e-05,
+      "loss": 0.6268,
+      "step": 10
+    },
+    {
+      "epoch": 0.44,
+      "grad_norm": 7.227736963113414,
+      "learning_rate": 9.15492957746479e-05,
+      "loss": 0.5151,
+      "step": 11
+    },
+    {
+      "epoch": 0.48,
+      "grad_norm": 4.259327811156309,
+      "learning_rate": 9.014084507042254e-05,
+      "loss": 0.4389,
+      "step": 12
+    },
+    {
+      "epoch": 0.52,
+      "grad_norm": 6.156983867098677,
+      "learning_rate": 8.873239436619719e-05,
+      "loss": 0.5211,
+      "step": 13
+    },
+    {
+      "epoch": 0.56,
+      "grad_norm": 5.241850109217824,
+      "learning_rate": 8.732394366197182e-05,
+      "loss": 0.3401,
+      "step": 14
+    },
+    {
+      "epoch": 0.6,
+      "grad_norm": 45.75950192245342,
+      "learning_rate": 8.591549295774647e-05,
+      "loss": 0.6114,
+      "step": 15
+    },
+    {
+      "epoch": 0.64,
+      "grad_norm": 12.747985157681715,
+      "learning_rate": 8.450704225352113e-05,
+      "loss": 0.4109,
+      "step": 16
+    },
+    {
+      "epoch": 0.68,
+      "grad_norm": 6.204387028637272,
+      "learning_rate": 8.309859154929578e-05,
+      "loss": 0.3238,
+      "step": 17
+    },
+    {
+      "epoch": 0.72,
+      "grad_norm": 1.784294966675183,
+      "learning_rate": 8.169014084507043e-05,
+      "loss": 0.2746,
+      "step": 18
+    },
+    {
+      "epoch": 0.76,
+      "grad_norm": 1.0640451461171732,
+      "learning_rate": 8.028169014084508e-05,
+      "loss": 0.2216,
+      "step": 19
+    },
+    {
+      "epoch": 0.8,
+      "grad_norm": 2.247533777438539,
+      "learning_rate": 7.887323943661972e-05,
+      "loss": 0.2169,
+      "step": 20
+    },
+    {
+      "epoch": 0.84,
+      "grad_norm": 0.5843446520161867,
+      "learning_rate": 7.746478873239437e-05,
+      "loss": 0.2063,
+      "step": 21
+    },
+    {
+      "epoch": 0.88,
+      "grad_norm": 0.6140022871361251,
+      "learning_rate": 7.605633802816902e-05,
+      "loss": 0.1924,
+      "step": 22
+    },
+    {
+      "epoch": 0.92,
+      "grad_norm": 0.6328264593401415,
+      "learning_rate": 7.464788732394367e-05,
+      "loss": 0.1687,
+      "step": 23
+    },
+    {
+      "epoch": 0.96,
+      "grad_norm": 0.47188590510134204,
+      "learning_rate": 7.323943661971832e-05,
+      "loss": 0.1502,
+      "step": 24
+    },
+    {
+      "epoch": 1.0,
+      "grad_norm": 0.4569187747764233,
+      "learning_rate": 7.183098591549297e-05,
+      "loss": 0.1322,
+      "step": 25
+    },
+    {
+      "epoch": 1.04,
+      "grad_norm": 0.5022295511205257,
+      "learning_rate": 7.042253521126761e-05,
+      "loss": 0.1283,
+      "step": 26
+    },
+    {
+      "epoch": 1.08,
+      "grad_norm": 0.32747245962381083,
+      "learning_rate": 6.901408450704226e-05,
+      "loss": 0.105,
+      "step": 27
+    },
+    {
+      "epoch": 1.12,
+      "grad_norm": 0.4071591561018996,
+      "learning_rate": 6.76056338028169e-05,
+      "loss": 0.1127,
+      "step": 28
+    },
+    {
+      "epoch": 1.16,
+      "grad_norm": 0.4886479527811691,
+      "learning_rate": 6.619718309859155e-05,
+      "loss": 0.1097,
+      "step": 29
+    },
+    {
+      "epoch": 1.2,
+      "grad_norm": 0.34862267892547183,
+      "learning_rate": 6.47887323943662e-05,
+      "loss": 0.1007,
+      "step": 30
+    },
+    {
+      "epoch": 1.24,
+      "grad_norm": 0.33123650236209157,
+      "learning_rate": 6.338028169014085e-05,
+      "loss": 0.0945,
+      "step": 31
+    },
+    {
+      "epoch": 1.28,
+      "grad_norm": 0.3246463353672584,
+      "learning_rate": 6.197183098591549e-05,
+      "loss": 0.0957,
+      "step": 32
+    },
+    {
+      "epoch": 1.32,
+      "grad_norm": 0.3503498413159293,
+      "learning_rate": 6.056338028169014e-05,
+      "loss": 0.1007,
+      "step": 33
+    },
+    {
+      "epoch": 1.3599999999999999,
+      "grad_norm": 0.2994313557882548,
+      "learning_rate": 5.915492957746479e-05,
+      "loss": 0.0961,
+      "step": 34
+    },
+    {
+      "epoch": 1.4,
+      "grad_norm": 0.3098559906263102,
+      "learning_rate": 5.774647887323944e-05,
+      "loss": 0.0891,
+      "step": 35
+    },
+    {
+      "epoch": 1.44,
+      "grad_norm": 0.3320370344139516,
+      "learning_rate": 5.633802816901409e-05,
+      "loss": 0.093,
+      "step": 36
+    },
+    {
+      "epoch": 1.48,
+      "grad_norm": 0.2975761075734135,
+      "learning_rate": 5.492957746478874e-05,
+      "loss": 0.0824,
+      "step": 37
+    },
+    {
+      "epoch": 1.52,
+      "grad_norm": 0.2609696051747955,
+      "learning_rate": 5.352112676056338e-05,
+      "loss": 0.0849,
+      "step": 38
+    },
+    {
+      "epoch": 1.56,
+      "grad_norm": 0.2746218271104525,
+      "learning_rate": 5.2112676056338026e-05,
+      "loss": 0.0812,
+      "step": 39
+    },
+    {
+      "epoch": 1.6,
+      "grad_norm": 0.27280756684282476,
+      "learning_rate": 5.070422535211268e-05,
+      "loss": 0.0857,
+      "step": 40
+    },
+    {
+      "epoch": 1.6400000000000001,
+      "grad_norm": 0.3333732319784279,
+      "learning_rate": 4.929577464788733e-05,
+      "loss": 0.083,
+      "step": 41
+    },
+    {
+      "epoch": 1.6800000000000002,
+      "grad_norm": 0.28187480716143415,
+      "learning_rate": 4.788732394366197e-05,
+      "loss": 0.0805,
+      "step": 42
+    },
+    {
+      "epoch": 1.72,
+      "grad_norm": 0.25874624247743283,
+      "learning_rate": 4.647887323943662e-05,
+      "loss": 0.0806,
+      "step": 43
+    },
+    {
+      "epoch": 1.76,
+      "grad_norm": 0.28417848530718226,
+      "learning_rate": 4.507042253521127e-05,
+      "loss": 0.0776,
+      "step": 44
+    },
+    {
+      "epoch": 1.8,
+      "grad_norm": 0.24698649624452082,
+      "learning_rate": 4.366197183098591e-05,
+      "loss": 0.0765,
+      "step": 45
+    },
+    {
+      "epoch": 1.8399999999999999,
+      "grad_norm": 0.23537664520973256,
+      "learning_rate": 4.225352112676056e-05,
+      "loss": 0.0758,
+      "step": 46
+    },
+    {
+      "epoch": 1.88,
+      "grad_norm": 0.2534413438607467,
+      "learning_rate": 4.0845070422535214e-05,
+      "loss": 0.0804,
+      "step": 47
+    },
+    {
+      "epoch": 1.92,
+      "grad_norm": 0.20702488497520458,
+      "learning_rate": 3.943661971830986e-05,
+      "loss": 0.0715,
+      "step": 48
+    },
+    {
+      "epoch": 1.96,
+      "grad_norm": 0.2613262372736133,
+      "learning_rate": 3.802816901408451e-05,
+      "loss": 0.084,
+      "step": 49
+    },
+    {
+      "epoch": 2.0,
+      "grad_norm": 0.22492435603929029,
+      "learning_rate": 3.661971830985916e-05,
+      "loss": 0.0556,
+      "step": 50
+    },
+    {
+      "epoch": 2.04,
+      "grad_norm": 0.2611962185259755,
+      "learning_rate": 3.5211267605633805e-05,
+      "loss": 0.0481,
+      "step": 51
+    },
+    {
+      "epoch": 2.08,
+      "grad_norm": 0.15962947759230994,
+      "learning_rate": 3.380281690140845e-05,
+      "loss": 0.0449,
+      "step": 52
+    },
+    {
+      "epoch": 2.12,
+      "grad_norm": 0.1694677846392973,
+      "learning_rate": 3.23943661971831e-05,
+      "loss": 0.0482,
+      "step": 53
+    },
+    {
+      "epoch": 2.16,
+      "grad_norm": 0.1438746814861677,
+      "learning_rate": 3.0985915492957744e-05,
+      "loss": 0.0425,
+      "step": 54
+    },
+    {
+      "epoch": 2.2,
+      "grad_norm": 0.17430208968882352,
+      "learning_rate": 2.9577464788732395e-05,
+      "loss": 0.043,
+      "step": 55
+    },
+    {
+      "epoch": 2.24,
+      "grad_norm": 0.17207620385513733,
+      "learning_rate": 2.8169014084507046e-05,
+      "loss": 0.0449,
+      "step": 56
+    },
+    {
+      "epoch": 2.2800000000000002,
+      "grad_norm": 0.18001842523178951,
+      "learning_rate": 2.676056338028169e-05,
+      "loss": 0.0407,
+      "step": 57
+    },
+    {
+      "epoch": 2.32,
+      "grad_norm": 0.20877963876068653,
+      "learning_rate": 2.535211267605634e-05,
+      "loss": 0.0418,
+      "step": 58
+    },
+    {
+      "epoch": 2.36,
+      "grad_norm": 0.2086206476860082,
+      "learning_rate": 2.3943661971830986e-05,
+      "loss": 0.0432,
+      "step": 59
+    },
+    {
+      "epoch": 2.4,
+      "grad_norm": 0.18489125316064906,
+      "learning_rate": 2.2535211267605634e-05,
+      "loss": 0.0443,
+      "step": 60
+    },
+    {
+      "epoch": 2.44,
+      "grad_norm": 0.18836568321121092,
+      "learning_rate": 2.112676056338028e-05,
+      "loss": 0.0416,
+      "step": 61
+    },
+    {
+      "epoch": 2.48,
+      "grad_norm": 0.18671273407693323,
+      "learning_rate": 1.971830985915493e-05,
+      "loss": 0.0392,
+      "step": 62
+    },
+    {
+      "epoch": 2.52,
+      "grad_norm": 0.183141175658178,
+      "learning_rate": 1.830985915492958e-05,
+      "loss": 0.0406,
+      "step": 63
+    },
+    {
+      "epoch": 2.56,
+      "grad_norm": 0.19023565118928507,
+      "learning_rate": 1.6901408450704224e-05,
+      "loss": 0.0392,
+      "step": 64
+    },
+    {
+      "epoch": 2.6,
+      "grad_norm": 0.1734836339026565,
+      "learning_rate": 1.5492957746478872e-05,
+      "loss": 0.0408,
+      "step": 65
+    },
+    {
+      "epoch": 2.64,
+      "grad_norm": 0.2124529691628785,
+      "learning_rate": 1.4084507042253523e-05,
+      "loss": 0.0402,
+      "step": 66
+    },
+    {
+      "epoch": 2.68,
+      "grad_norm": 0.16929391483607104,
+      "learning_rate": 1.267605633802817e-05,
+      "loss": 0.0422,
+      "step": 67
+    },
+    {
+      "epoch": 2.7199999999999998,
+      "grad_norm": 0.1865969813928253,
+      "learning_rate": 1.1267605633802817e-05,
+      "loss": 0.0423,
+      "step": 68
+    },
+    {
+      "epoch": 2.76,
+      "grad_norm": 0.19199511508921327,
+      "learning_rate": 9.859154929577465e-06,
+      "loss": 0.0428,
+      "step": 69
+    },
+    {
+      "epoch": 2.8,
+      "grad_norm": 0.15471843235506497,
+      "learning_rate": 8.450704225352112e-06,
+      "loss": 0.0383,
+      "step": 70
+    },
+    {
+      "epoch": 2.84,
+      "grad_norm": 0.17211726924185472,
+      "learning_rate": 7.042253521126762e-06,
+      "loss": 0.0422,
+      "step": 71
+    },
+    {
+      "epoch": 2.88,
+      "grad_norm": 0.16597891624795819,
+      "learning_rate": 5.6338028169014084e-06,
+      "loss": 0.0358,
+      "step": 72
+    },
+    {
+      "epoch": 2.92,
+      "grad_norm": 0.16696499746864643,
+      "learning_rate": 4.225352112676056e-06,
+      "loss": 0.0384,
+      "step": 73
+    },
+    {
+      "epoch": 2.96,
+      "grad_norm": 0.16841918706948866,
+      "learning_rate": 2.8169014084507042e-06,
+      "loss": 0.0399,
+      "step": 74
+    },
+    {
+      "epoch": 3.0,
+      "grad_norm": 0.12380367991746002,
+      "learning_rate": 1.4084507042253521e-06,
+      "loss": 0.0275,
+      "step": 75
+    },
+    {
+      "epoch": 3.0,
+      "step": 75,
+      "total_flos": 175054227243008.0,
+      "train_loss": 0.6997585766762495,
+      "train_runtime": 937.3587,
+      "train_samples_per_second": 103.388,
+      "train_steps_per_second": 0.08
+    }
+  ],
+  "logging_steps": 1.0,
+  "max_steps": 75,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 3,
+  "save_steps": 500,
+  "total_flos": 175054227243008.0,
+  "train_batch_size": 12,
+  "trial_name": null,
+  "trial_params": null
+}

training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d7fc176810fd9b3cd41a111aa8142543838dc193aad79a2d257a1ff5b70e73dc
+size 7032

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)