Commit
·
f8ce820
1
Parent(s):
4ffbb47
update
Browse files- added_tokens.json +3 -0
- arguments.json +64 -0
- arguments.pkl +3 -0
- config.json +39 -0
- environ.txt +253 -0
- latest +1 -0
- pytorch_model.bin +3 -0
- script.sh +122 -0
- special_tokens_map.json +24 -0
- stderr.log +0 -0
- stdout.log +41 -0
- tokenizer.model +3 -0
- tokenizer_config.json +35 -0
- zero_to_fp32.py +587 -0
added_tokens.json
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"<pad>": 32000
|
| 3 |
+
}
|
arguments.json
ADDED
|
@@ -0,0 +1,64 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"model_name_or_path": "output/sft",
|
| 3 |
+
"max_length": 512,
|
| 4 |
+
"trust_remote_code": true,
|
| 5 |
+
"train_datasets": [
|
| 6 |
+
[
|
| 7 |
+
"PKU-SafeRLHF/train",
|
| 8 |
+
{
|
| 9 |
+
"proportion": 1.0,
|
| 10 |
+
"path": "PKU-SafeRLHF-harmless-only-30k"
|
| 11 |
+
}
|
| 12 |
+
]
|
| 13 |
+
],
|
| 14 |
+
"eval_datasets": [
|
| 15 |
+
[
|
| 16 |
+
"PKU-SafeRLHF/test",
|
| 17 |
+
{
|
| 18 |
+
"proportion": 1.0
|
| 19 |
+
}
|
| 20 |
+
]
|
| 21 |
+
],
|
| 22 |
+
"loss_type": "sequence-wise",
|
| 23 |
+
"epochs": 2,
|
| 24 |
+
"per_device_train_batch_size": 16,
|
| 25 |
+
"per_device_eval_batch_size": 16,
|
| 26 |
+
"gradient_accumulation_steps": 2,
|
| 27 |
+
"gradient_checkpointing": true,
|
| 28 |
+
"normalize_score_during_training": false,
|
| 29 |
+
"normalizer_type": "ExponentialMovingAverage",
|
| 30 |
+
"normalizer_momentum": 0.9,
|
| 31 |
+
"lr": 2e-05,
|
| 32 |
+
"lr_scheduler_type": "cosine",
|
| 33 |
+
"lr_warmup_ratio": 0.03,
|
| 34 |
+
"weight_decay": 0.1,
|
| 35 |
+
"seed": 42,
|
| 36 |
+
"fp16": false,
|
| 37 |
+
"bf16": true,
|
| 38 |
+
"tf32": true,
|
| 39 |
+
"eval_strategy": "epoch",
|
| 40 |
+
"eval_interval": 1000000,
|
| 41 |
+
"need_eval": false,
|
| 42 |
+
"eval_split_ratio": null,
|
| 43 |
+
"output_dir": "/data/jiongxiao_wang/rlhf_attack/safe-rlhf/output/rm_30k",
|
| 44 |
+
"log_type": "wandb",
|
| 45 |
+
"log_dir": "/data/jiongxiao_wang/rlhf_attack/safe-rlhf/output/rm_30k",
|
| 46 |
+
"log_project": "Safe-RLHF-RM",
|
| 47 |
+
"log_run_name": "reward-2024-01-05-20-03-25",
|
| 48 |
+
"save_16bit": false,
|
| 49 |
+
"save_interval": 1000000,
|
| 50 |
+
"local_rank": 0,
|
| 51 |
+
"zero_stage": 3,
|
| 52 |
+
"deepspeed": false,
|
| 53 |
+
"deepspeed_config": null,
|
| 54 |
+
"deepscale": false,
|
| 55 |
+
"deepscale_config": null,
|
| 56 |
+
"deepspeed_mpi": false,
|
| 57 |
+
"global_rank": 0,
|
| 58 |
+
"device": {
|
| 59 |
+
"type": "torch.device",
|
| 60 |
+
"repr": "device(type='cuda', index=0)"
|
| 61 |
+
},
|
| 62 |
+
"num_update_steps_per_epoch": 210,
|
| 63 |
+
"total_training_steps": 420
|
| 64 |
+
}
|
arguments.pkl
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:263070491ba24413e07d4f20f0a1b77192ec40be5d476255e6866ab417b1832c
|
| 3 |
+
size 1245
|
config.json
ADDED
|
@@ -0,0 +1,39 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"_name_or_path": "output/sft",
|
| 3 |
+
"architectures": [
|
| 4 |
+
"LlamaModelForScore"
|
| 5 |
+
],
|
| 6 |
+
"bias": false,
|
| 7 |
+
"bos_token_id": 1,
|
| 8 |
+
"do_normalize": false,
|
| 9 |
+
"eos_token_id": 2,
|
| 10 |
+
"hidden_act": "silu",
|
| 11 |
+
"hidden_size": 4096,
|
| 12 |
+
"initializer_range": 0.02,
|
| 13 |
+
"intermediate_size": 11008,
|
| 14 |
+
"max_position_embeddings": 2048,
|
| 15 |
+
"max_sequence_length": 2048,
|
| 16 |
+
"mean": [
|
| 17 |
+
7.4375
|
| 18 |
+
],
|
| 19 |
+
"model_type": "llama",
|
| 20 |
+
"momentum": 0.9,
|
| 21 |
+
"normalizer_type": "ExponentialMovingAverage",
|
| 22 |
+
"num_attention_heads": 32,
|
| 23 |
+
"num_hidden_layers": 32,
|
| 24 |
+
"num_key_value_heads": 32,
|
| 25 |
+
"pad_token_id": 32000,
|
| 26 |
+
"pretraining_tp": 1,
|
| 27 |
+
"rms_norm_eps": 1e-06,
|
| 28 |
+
"rope_scaling": null,
|
| 29 |
+
"score_dim": 1,
|
| 30 |
+
"score_type": "reward",
|
| 31 |
+
"tie_word_embeddings": false,
|
| 32 |
+
"torch_dtype": "float16",
|
| 33 |
+
"transformers_version": "4.31.0",
|
| 34 |
+
"use_cache": true,
|
| 35 |
+
"var": [
|
| 36 |
+
82.5
|
| 37 |
+
],
|
| 38 |
+
"vocab_size": 32001
|
| 39 |
+
}
|
environ.txt
ADDED
|
@@ -0,0 +1,253 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
ADDR2LINE=/data/jiongxiao_wang/anaconda3/envs/safe-rlhf/bin/x86_64-conda-linux-gnu-addr2line
|
| 2 |
+
AR=/data/jiongxiao_wang/anaconda3/envs/safe-rlhf/bin/x86_64-conda-linux-gnu-ar
|
| 3 |
+
AS=/data/jiongxiao_wang/anaconda3/envs/safe-rlhf/bin/x86_64-conda-linux-gnu-as
|
| 4 |
+
BASH_FUNC__spack_shell_wrapper()=() { for var in LD_LIBRARY_PATH DYLD_LIBRARY_PATH DYLD_FALLBACK_LIBRARY_PATH;
|
| 5 |
+
do
|
| 6 |
+
eval "if [ -n \"\${${var}-}\" ]; then export SPACK_$var=\${${var}}; fi";
|
| 7 |
+
done;
|
| 8 |
+
if [ -n "${ZSH_VERSION:-}" ]; then
|
| 9 |
+
emulate -L sh;
|
| 10 |
+
fi;
|
| 11 |
+
_sp_flags="";
|
| 12 |
+
while [ ! -z ${1+x} ] && [ "${1#-}" != "${1}" ]; do
|
| 13 |
+
_sp_flags="$_sp_flags $1";
|
| 14 |
+
shift;
|
| 15 |
+
done;
|
| 16 |
+
if [ -n "$_sp_flags" ] && [ "${_sp_flags#*h}" != "${_sp_flags}" ] || [ "${_sp_flags#*V}" != "${_sp_flags}" ]; then
|
| 17 |
+
command spack $_sp_flags "$@";
|
| 18 |
+
return;
|
| 19 |
+
fi;
|
| 20 |
+
_sp_subcommand="";
|
| 21 |
+
if [ ! -z ${1+x} ]; then
|
| 22 |
+
_sp_subcommand="$1";
|
| 23 |
+
shift;
|
| 24 |
+
fi;
|
| 25 |
+
case $_sp_subcommand in
|
| 26 |
+
"cd")
|
| 27 |
+
_sp_arg="";
|
| 28 |
+
if [ -n "$1" ]; then
|
| 29 |
+
_sp_arg="$1";
|
| 30 |
+
shift;
|
| 31 |
+
fi;
|
| 32 |
+
if [ "$_sp_arg" = "-h" ] || [ "$_sp_arg" = "--help" ]; then
|
| 33 |
+
command spack cd -h;
|
| 34 |
+
else
|
| 35 |
+
LOC="$(spack location $_sp_arg "$@")";
|
| 36 |
+
if [ -d "$LOC" ]; then
|
| 37 |
+
cd "$LOC";
|
| 38 |
+
else
|
| 39 |
+
return 1;
|
| 40 |
+
fi;
|
| 41 |
+
fi;
|
| 42 |
+
return
|
| 43 |
+
;;
|
| 44 |
+
"env")
|
| 45 |
+
_sp_arg="";
|
| 46 |
+
if [ -n "$1" ]; then
|
| 47 |
+
_sp_arg="$1";
|
| 48 |
+
shift;
|
| 49 |
+
fi;
|
| 50 |
+
if [ "$_sp_arg" = "-h" ] || [ "$_sp_arg" = "--help" ]; then
|
| 51 |
+
command spack env -h;
|
| 52 |
+
else
|
| 53 |
+
case $_sp_arg in
|
| 54 |
+
activate)
|
| 55 |
+
_a=" $@";
|
| 56 |
+
if [ -z ${1+x} ] || [ "${_a#* --sh}" != "$_a" ] || [ "${_a#* --csh}" != "$_a" ] || [ "${_a#* -h}" != "$_a" ] || [ "${_a#* --help}" != "$_a" ]; then
|
| 57 |
+
command spack env activate "$@";
|
| 58 |
+
else
|
| 59 |
+
stdout="$(command spack $_sp_flags env activate --sh "$@")" || return;
|
| 60 |
+
eval "$stdout";
|
| 61 |
+
fi
|
| 62 |
+
;;
|
| 63 |
+
deactivate)
|
| 64 |
+
_a=" $@";
|
| 65 |
+
if [ "${_a#* --sh}" != "$_a" ] || [ "${_a#* --csh}" != "$_a" ]; then
|
| 66 |
+
command spack env deactivate "$@";
|
| 67 |
+
else
|
| 68 |
+
if [ -n "$*" ]; then
|
| 69 |
+
command spack env deactivate -h;
|
| 70 |
+
else
|
| 71 |
+
stdout="$(command spack $_sp_flags env deactivate --sh)" || return;
|
| 72 |
+
eval "$stdout";
|
| 73 |
+
fi;
|
| 74 |
+
fi
|
| 75 |
+
;;
|
| 76 |
+
*)
|
| 77 |
+
command spack env $_sp_arg "$@"
|
| 78 |
+
;;
|
| 79 |
+
esac;
|
| 80 |
+
fi;
|
| 81 |
+
return
|
| 82 |
+
;;
|
| 83 |
+
"load" | "unload")
|
| 84 |
+
_a=" $@";
|
| 85 |
+
if [ "${_a#* --sh}" != "$_a" ] || [ "${_a#* --csh}" != "$_a" ] || [ "${_a#* -h}" != "$_a" ] || [ "${_a#* --list}" != "$_a" ] || [ "${_a#* --help}" != "$_a" ]; then
|
| 86 |
+
command spack $_sp_flags $_sp_subcommand "$@";
|
| 87 |
+
else
|
| 88 |
+
stdout="$(command spack $_sp_flags $_sp_subcommand --sh "$@")" || return;
|
| 89 |
+
eval "$stdout";
|
| 90 |
+
fi
|
| 91 |
+
;;
|
| 92 |
+
*)
|
| 93 |
+
command spack $_sp_flags $_sp_subcommand "$@"
|
| 94 |
+
;;
|
| 95 |
+
esac
|
| 96 |
+
}
|
| 97 |
+
BASH_FUNC_module()=() { eval `/usr/bin/modulecmd bash $*`
|
| 98 |
+
}
|
| 99 |
+
BASH_FUNC_spack()=() { : this is a shell function from: /nfs/cluster/spack/share/spack/setup-env.sh;
|
| 100 |
+
: the real spack script is here: /nfs/cluster/spack/bin/spack;
|
| 101 |
+
_spack_shell_wrapper "$@";
|
| 102 |
+
return $?
|
| 103 |
+
}
|
| 104 |
+
BROWSER=/data/jiongxiao_wang/.vscode-server/bin/0ee08df0cf4527e40edc9aa28f4b5bd38bbff2b2/bin/helpers/browser.sh
|
| 105 |
+
BUILD=x86_64-conda-linux-gnu
|
| 106 |
+
CC=/data/jiongxiao_wang/anaconda3/envs/safe-rlhf/bin/x86_64-conda-linux-gnu-cc
|
| 107 |
+
CC_FOR_BUILD=/data/jiongxiao_wang/anaconda3/envs/safe-rlhf/bin/x86_64-conda-linux-gnu-cc
|
| 108 |
+
CFLAGS=-march=nocona -mtune=haswell -ftree-vectorize -fPIC -fstack-protector-strong -fno-plt -O2 -ffunction-sections -pipe -isystem /data/jiongxiao_wang/anaconda3/envs/safe-rlhf/include
|
| 109 |
+
CMAKE_ARGS=-DCMAKE_AR=/data/jiongxiao_wang/anaconda3/envs/safe-rlhf/bin/x86_64-conda-linux-gnu-ar -DCMAKE_CXX_COMPILER_AR=/data/jiongxiao_wang/anaconda3/envs/safe-rlhf/bin/x86_64-conda-linux-gnu-gcc-ar -DCMAKE_C_COMPILER_AR=/data/jiongxiao_wang/anaconda3/envs/safe-rlhf/bin/x86_64-conda-linux-gnu-gcc-ar -DCMAKE_RANLIB=/data/jiongxiao_wang/anaconda3/envs/safe-rlhf/bin/x86_64-conda-linux-gnu-ranlib -DCMAKE_CXX_COMPILER_RANLIB=/data/jiongxiao_wang/anaconda3/envs/safe-rlhf/bin/x86_64-conda-linux-gnu-gcc-ranlib -DCMAKE_C_COMPILER_RANLIB=/data/jiongxiao_wang/anaconda3/envs/safe-rlhf/bin/x86_64-conda-linux-gnu-gcc-ranlib -DCMAKE_LINKER=/data/jiongxiao_wang/anaconda3/envs/safe-rlhf/bin/x86_64-conda-linux-gnu-ld -DCMAKE_STRIP=/data/jiongxiao_wang/anaconda3/envs/safe-rlhf/bin/x86_64-conda-linux-gnu-strip -DCMAKE_BUILD_TYPE=Release
|
| 110 |
+
CMAKE_PREFIX_PATH=/data/jiongxiao_wang/anaconda3/envs/safe-rlhf:/data/jiongxiao_wang/anaconda3/envs/safe-rlhf/x86_64-conda-linux-gnu/sysroot/usr
|
| 111 |
+
COLORTERM=truecolor
|
| 112 |
+
CONDA_BUILD_SYSROOT=/data/jiongxiao_wang/anaconda3/envs/safe-rlhf/x86_64-conda-linux-gnu/sysroot
|
| 113 |
+
CONDA_DEFAULT_ENV=safe-rlhf
|
| 114 |
+
CONDA_EXE=/data/jiongxiao_wang/anaconda3/bin/conda
|
| 115 |
+
CONDA_PREFIX=/data/jiongxiao_wang/anaconda3/envs/safe-rlhf
|
| 116 |
+
CONDA_PREFIX_1=/data/jiongxiao_wang/anaconda3
|
| 117 |
+
CONDA_PROMPT_MODIFIER=(safe-rlhf)
|
| 118 |
+
CONDA_PYTHON_EXE=/data/jiongxiao_wang/anaconda3/bin/python
|
| 119 |
+
CONDA_ROOT=/data/jiongxiao_wang/anaconda3
|
| 120 |
+
CONDA_SHLVL=2
|
| 121 |
+
CONDA_TOOLCHAIN_BUILD=x86_64-conda-linux-gnu
|
| 122 |
+
CONDA_TOOLCHAIN_HOST=x86_64-conda-linux-gnu
|
| 123 |
+
CPP=/data/jiongxiao_wang/anaconda3/envs/safe-rlhf/bin/x86_64-conda-linux-gnu-cpp
|
| 124 |
+
CPPFLAGS=-DNDEBUG -D_FORTIFY_SOURCE=2 -O2 -isystem /data/jiongxiao_wang/anaconda3/envs/safe-rlhf/include
|
| 125 |
+
CROSS_RANK=0
|
| 126 |
+
CROSS_SIZE=1
|
| 127 |
+
CUDA_MODULE_LOADING=LAZY
|
| 128 |
+
CUDA_VISIBLE_DEVICES=0,1,2,3
|
| 129 |
+
CXX=/data/jiongxiao_wang/anaconda3/envs/safe-rlhf/bin/x86_64-conda-linux-gnu-c++
|
| 130 |
+
CXXFILT=/data/jiongxiao_wang/anaconda3/envs/safe-rlhf/bin/x86_64-conda-linux-gnu-c++filt
|
| 131 |
+
CXXFLAGS=-fvisibility-inlines-hidden -fmessage-length=0 -march=nocona -mtune=haswell -ftree-vectorize -fPIC -fstack-protector-strong -fno-plt -O2 -ffunction-sections -pipe -isystem /data/jiongxiao_wang/anaconda3/envs/safe-rlhf/include
|
| 132 |
+
CXX_FOR_BUILD=/data/jiongxiao_wang/anaconda3/envs/safe-rlhf/bin/x86_64-conda-linux-gnu-c++
|
| 133 |
+
DEBUG_CFLAGS=-march=nocona -mtune=haswell -ftree-vectorize -fPIC -fstack-protector-all -fno-plt -Og -g -Wall -Wextra -fvar-tracking-assignments -ffunction-sections -pipe -isystem /data/jiongxiao_wang/anaconda3/envs/safe-rlhf/include
|
| 134 |
+
DEBUG_CPPFLAGS=-D_DEBUG -D_FORTIFY_SOURCE=2 -Og -isystem /data/jiongxiao_wang/anaconda3/envs/safe-rlhf/include
|
| 135 |
+
DEBUG_CXXFLAGS=-fvisibility-inlines-hidden -fmessage-length=0 -march=nocona -mtune=haswell -ftree-vectorize -fPIC -fstack-protector-all -fno-plt -Og -g -Wall -Wextra -fvar-tracking-assignments -ffunction-sections -pipe -isystem /data/jiongxiao_wang/anaconda3/envs/safe-rlhf/include
|
| 136 |
+
ELFEDIT=/data/jiongxiao_wang/anaconda3/envs/safe-rlhf/bin/x86_64-conda-linux-gnu-elfedit
|
| 137 |
+
ENVIRONMENT=BATCH
|
| 138 |
+
GCC=/data/jiongxiao_wang/anaconda3/envs/safe-rlhf/bin/x86_64-conda-linux-gnu-gcc
|
| 139 |
+
GCC_AR=/data/jiongxiao_wang/anaconda3/envs/safe-rlhf/bin/x86_64-conda-linux-gnu-gcc-ar
|
| 140 |
+
GCC_NM=/data/jiongxiao_wang/anaconda3/envs/safe-rlhf/bin/x86_64-conda-linux-gnu-gcc-nm
|
| 141 |
+
GCC_RANLIB=/data/jiongxiao_wang/anaconda3/envs/safe-rlhf/bin/x86_64-conda-linux-gnu-gcc-ranlib
|
| 142 |
+
GIT_ASKPASS=/data/jiongxiao_wang/.vscode-server/bin/0ee08df0cf4527e40edc9aa28f4b5bd38bbff2b2/extensions/git/dist/askpass.sh
|
| 143 |
+
GPROF=/data/jiongxiao_wang/anaconda3/envs/safe-rlhf/bin/x86_64-conda-linux-gnu-gprof
|
| 144 |
+
GPU_DEVICE_ORDINAL=0,1,2,3
|
| 145 |
+
GXX=/data/jiongxiao_wang/anaconda3/envs/safe-rlhf/bin/x86_64-conda-linux-gnu-g++
|
| 146 |
+
HISTCONTROL=ignoredups
|
| 147 |
+
HISTSIZE=1000
|
| 148 |
+
HOME=/data/jiongxiao_wang
|
| 149 |
+
HOST=x86_64-conda-linux-gnu
|
| 150 |
+
HOSTNAME=compute-permanent-node-485
|
| 151 |
+
LANG=en_US.UTF-8
|
| 152 |
+
LD=/data/jiongxiao_wang/anaconda3/envs/safe-rlhf/bin/x86_64-conda-linux-gnu-ld
|
| 153 |
+
LDFLAGS=-Wl,-O2 -Wl,--sort-common -Wl,--as-needed -Wl,-z,relro -Wl,-z,now -Wl,--disable-new-dtags -Wl,--gc-sections -Wl,--allow-shlib-undefined -Wl,-rpath,/data/jiongxiao_wang/anaconda3/envs/safe-rlhf/lib -Wl,-rpath-link,/data/jiongxiao_wang/anaconda3/envs/safe-rlhf/lib -L/data/jiongxiao_wang/anaconda3/envs/safe-rlhf/lib
|
| 154 |
+
LD_GOLD=/data/jiongxiao_wang/anaconda3/envs/safe-rlhf/bin/x86_64-conda-linux-gnu-ld.gold
|
| 155 |
+
LD_LIBRARY_PATH=/data/jiongxiao_wang/anaconda3/envs/safe-rlhf/lib:/data/jiongxiao_wang/anaconda3/envs/safe-rlhf/lib:/data/jiongxiao_wang/anaconda3/envs/safe-rlhf/lib:
|
| 156 |
+
LESSOPEN=||/usr/bin/lesspipe.sh %s
|
| 157 |
+
LOADEDMODULES=
|
| 158 |
+
LOCAL_RANK=0
|
| 159 |
+
LOCAL_SIZE=4
|
| 160 |
+
LOGLEVEL=WARNING
|
| 161 |
+
LOGNAME=jiongxiao_wang
|
| 162 |
+
LS_COLORS=rs=0:di=38;5;27:ln=38;5;51:mh=44;38;5;15:pi=40;38;5;11:so=38;5;13:do=38;5;5:bd=48;5;232;38;5;11:cd=48;5;232;38;5;3:or=48;5;232;38;5;9:mi=05;48;5;232;38;5;15:su=48;5;196;38;5;15:sg=48;5;11;38;5;16:ca=48;5;196;38;5;226:tw=48;5;10;38;5;16:ow=48;5;10;38;5;21:st=48;5;21;38;5;15:ex=38;5;34:*.tar=38;5;9:*.tgz=38;5;9:*.arc=38;5;9:*.arj=38;5;9:*.taz=38;5;9:*.lha=38;5;9:*.lz4=38;5;9:*.lzh=38;5;9:*.lzma=38;5;9:*.tlz=38;5;9:*.txz=38;5;9:*.tzo=38;5;9:*.t7z=38;5;9:*.zip=38;5;9:*.z=38;5;9:*.Z=38;5;9:*.dz=38;5;9:*.gz=38;5;9:*.lrz=38;5;9:*.lz=38;5;9:*.lzo=38;5;9:*.xz=38;5;9:*.bz2=38;5;9:*.bz=38;5;9:*.tbz=38;5;9:*.tbz2=38;5;9:*.tz=38;5;9:*.deb=38;5;9:*.rpm=38;5;9:*.jar=38;5;9:*.war=38;5;9:*.ear=38;5;9:*.sar=38;5;9:*.rar=38;5;9:*.alz=38;5;9:*.ace=38;5;9:*.zoo=38;5;9:*.cpio=38;5;9:*.7z=38;5;9:*.rz=38;5;9:*.cab=38;5;9:*.jpg=38;5;13:*.jpeg=38;5;13:*.gif=38;5;13:*.bmp=38;5;13:*.pbm=38;5;13:*.pgm=38;5;13:*.ppm=38;5;13:*.tga=38;5;13:*.xbm=38;5;13:*.xpm=38;5;13:*.tif=38;5;13:*.tiff=38;5;13:*.png=38;5;13:*.svg=38;5;13:*.svgz=38;5;13:*.mng=38;5;13:*.pcx=38;5;13:*.mov=38;5;13:*.mpg=38;5;13:*.mpeg=38;5;13:*.m2v=38;5;13:*.mkv=38;5;13:*.webm=38;5;13:*.ogm=38;5;13:*.mp4=38;5;13:*.m4v=38;5;13:*.mp4v=38;5;13:*.vob=38;5;13:*.qt=38;5;13:*.nuv=38;5;13:*.wmv=38;5;13:*.asf=38;5;13:*.rm=38;5;13:*.rmvb=38;5;13:*.flc=38;5;13:*.avi=38;5;13:*.fli=38;5;13:*.flv=38;5;13:*.gl=38;5;13:*.dl=38;5;13:*.xcf=38;5;13:*.xwd=38;5;13:*.yuv=38;5;13:*.cgm=38;5;13:*.emf=38;5;13:*.axv=38;5;13:*.anx=38;5;13:*.ogv=38;5;13:*.ogx=38;5;13:*.aac=38;5;45:*.au=38;5;45:*.flac=38;5;45:*.mid=38;5;45:*.midi=38;5;45:*.mka=38;5;45:*.mp3=38;5;45:*.mpc=38;5;45:*.ogg=38;5;45:*.ra=38;5;45:*.wav=38;5;45:*.axa=38;5;45:*.oga=38;5;45:*.spx=38;5;45:*.xspf=38;5;45:
|
| 163 |
+
MAIL=/var/spool/mail/jiongxiao_wang
|
| 164 |
+
MASTER_ADDR=127.0.0.1
|
| 165 |
+
MASTER_PORT=47607
|
| 166 |
+
MESON_ARGS=--buildtype release
|
| 167 |
+
MODULEPATH=/usr/share/Modules/modulefiles:/etc/modulefiles
|
| 168 |
+
MODULESHOME=/usr/share/Modules
|
| 169 |
+
NIX_CONF_DIR=/nix
|
| 170 |
+
NM=/data/jiongxiao_wang/anaconda3/envs/safe-rlhf/bin/x86_64-conda-linux-gnu-nm
|
| 171 |
+
OBJCOPY=/data/jiongxiao_wang/anaconda3/envs/safe-rlhf/bin/x86_64-conda-linux-gnu-objcopy
|
| 172 |
+
OBJDUMP=/data/jiongxiao_wang/anaconda3/envs/safe-rlhf/bin/x86_64-conda-linux-gnu-objdump
|
| 173 |
+
PATH=/nix/var/nix/profiles/default/bin:/data/jiongxiao_wang/.nix-profile/bin:/data/jiongxiao_wang/.vscode-server/bin/0ee08df0cf4527e40edc9aa28f4b5bd38bbff2b2/bin/remote-cli:/nix/var/nix/profiles/default/bin:/data/jiongxiao_wang/.nix-profile/bin:/data/jiongxiao_wang/anaconda3/envs/safe-rlhf/bin:/data/jiongxiao_wang/anaconda3/condabin:/nix/var/nix/profiles/default/bin:/data/jiongxiao_wang/.nix-profile/bin:/nfs/cluster/spack/bin:/usr/local/bin:/usr/bin:/var/lib/snapd/snap/bin:/usr/local/sbin:/usr/sbin:/data/jiongxiao_wang/.local/bin:/data/jiongxiao_wang/bin
|
| 174 |
+
PWD=/data/jiongxiao_wang/rlhf_attack/safe-rlhf
|
| 175 |
+
PYTHONHASHSEED=42
|
| 176 |
+
PYTHONPATH=/data/jiongxiao_wang/rlhf_attack/safe-rlhf
|
| 177 |
+
RANK=0
|
| 178 |
+
RANLIB=/data/jiongxiao_wang/anaconda3/envs/safe-rlhf/bin/x86_64-conda-linux-gnu-ranlib
|
| 179 |
+
READELF=/data/jiongxiao_wang/anaconda3/envs/safe-rlhf/bin/x86_64-conda-linux-gnu-readelf
|
| 180 |
+
ROCR_VISIBLE_DEVICES=0,1,2,3
|
| 181 |
+
SHELL=/bin/bash
|
| 182 |
+
SHLVL=7
|
| 183 |
+
SIZE=/data/jiongxiao_wang/anaconda3/envs/safe-rlhf/bin/x86_64-conda-linux-gnu-size
|
| 184 |
+
SLURMD_NODENAME=compute-permanent-node-485
|
| 185 |
+
SLURM_CLUSTER_NAME=cluster
|
| 186 |
+
SLURM_CONF=/var/spool/slurmd/conf-cache/slurm.conf
|
| 187 |
+
SLURM_CPUS_ON_NODE=8
|
| 188 |
+
SLURM_GPUS_ON_NODE=4
|
| 189 |
+
SLURM_GTIDS=0
|
| 190 |
+
SLURM_JOBID=1047466
|
| 191 |
+
SLURM_JOB_ACCOUNT=chaowei_xiao
|
| 192 |
+
SLURM_JOB_CPUS_PER_NODE=8
|
| 193 |
+
SLURM_JOB_END_TIME=1704657746
|
| 194 |
+
SLURM_JOB_GID=10043
|
| 195 |
+
SLURM_JOB_GPUS=0,1,2,3
|
| 196 |
+
SLURM_JOB_ID=1047466
|
| 197 |
+
SLURM_JOB_NAME=rlhf
|
| 198 |
+
SLURM_JOB_NODELIST=compute-permanent-node-485
|
| 199 |
+
SLURM_JOB_NUM_NODES=1
|
| 200 |
+
SLURM_JOB_PARTITION=compute
|
| 201 |
+
SLURM_JOB_QOS=default_qos
|
| 202 |
+
SLURM_JOB_START_TIME=1704484946
|
| 203 |
+
SLURM_JOB_UID=10193
|
| 204 |
+
SLURM_JOB_USER=jiongxiao_wang
|
| 205 |
+
SLURM_LOCALID=0
|
| 206 |
+
SLURM_MEM_PER_NODE=40960
|
| 207 |
+
SLURM_NNODES=1
|
| 208 |
+
SLURM_NODEID=0
|
| 209 |
+
SLURM_NODELIST=compute-permanent-node-485
|
| 210 |
+
SLURM_NODE_ALIASES=(null)
|
| 211 |
+
SLURM_NPROCS=1
|
| 212 |
+
SLURM_NTASKS=1
|
| 213 |
+
SLURM_PRIO_PROCESS=0
|
| 214 |
+
SLURM_PROCID=0
|
| 215 |
+
SLURM_SUBMIT_DIR=/data/jiongxiao_wang/rlhf_attack/safe-rlhf
|
| 216 |
+
SLURM_SUBMIT_HOST=watch-tower-login
|
| 217 |
+
SLURM_TASKS_PER_NODE=1
|
| 218 |
+
SLURM_TASK_PID=249941
|
| 219 |
+
SLURM_TOPOLOGY_ADDR=watch-tower.watch-tower:2094be69a339ca402351893d.compute-permanent-node-485
|
| 220 |
+
SLURM_TOPOLOGY_ADDR_PATTERN=switch.switch.node
|
| 221 |
+
SLURM_WORKING_CLUSTER=cluster:watch-tower-bastion:6817:9984:109
|
| 222 |
+
SPACK_PYTHON=/usr/bin/python3
|
| 223 |
+
SPACK_ROOT=/nfs/cluster/spack
|
| 224 |
+
SSH_CLIENT=131.93.180.124 52843 22
|
| 225 |
+
SSH_CONNECTION=131.93.180.124 52843 172.16.0.238 22
|
| 226 |
+
STRINGS=/data/jiongxiao_wang/anaconda3/envs/safe-rlhf/bin/x86_64-conda-linux-gnu-strings
|
| 227 |
+
STRIP=/data/jiongxiao_wang/anaconda3/envs/safe-rlhf/bin/x86_64-conda-linux-gnu-strip
|
| 228 |
+
TERM=xterm-256color
|
| 229 |
+
TERM_PROGRAM=vscode
|
| 230 |
+
TERM_PROGRAM_VERSION=1.85.1
|
| 231 |
+
TF2_BEHAVIOR=1
|
| 232 |
+
TF_CPP_MIN_LOG_LEVEL=1
|
| 233 |
+
TMPDIR=/tmp
|
| 234 |
+
TPU_ML_PLATFORM=Tensorflow
|
| 235 |
+
USER=jiongxiao_wang
|
| 236 |
+
VSCODE_GIT_ASKPASS_EXTRA_ARGS=
|
| 237 |
+
VSCODE_GIT_ASKPASS_MAIN=/data/jiongxiao_wang/.vscode-server/bin/0ee08df0cf4527e40edc9aa28f4b5bd38bbff2b2/extensions/git/dist/askpass-main.js
|
| 238 |
+
VSCODE_GIT_ASKPASS_NODE=/data/jiongxiao_wang/.vscode-server/bin/0ee08df0cf4527e40edc9aa28f4b5bd38bbff2b2/node
|
| 239 |
+
VSCODE_GIT_IPC_HANDLE=/run/user/10193/vscode-git-0b983ec108.sock
|
| 240 |
+
VSCODE_IPC_HOOK_CLI=/run/user/10193/vscode-ipc-dff0bf4e-b4ff-48e4-bd2a-e4dcda6129b0.sock
|
| 241 |
+
WANDB_API_KEY=f6021dca133c93e80a7dae4620bd335d4d08cac6
|
| 242 |
+
WANDB_SERVICE=2-250380-tcp-localhost-36113
|
| 243 |
+
WORLD_SIZE=4
|
| 244 |
+
XDG_DATA_DIRS=/usr/local/share:/usr/share:/var/lib/snapd/desktop
|
| 245 |
+
XDG_RUNTIME_DIR=/run/user/10193
|
| 246 |
+
XDG_SESSION_ID=629456
|
| 247 |
+
ZE_AFFINITY_MASK=0,1,2,3
|
| 248 |
+
_=/data/jiongxiao_wang/anaconda3/envs/safe-rlhf/bin/deepspeed
|
| 249 |
+
_CE_CONDA=
|
| 250 |
+
_CE_M=
|
| 251 |
+
_CONDA_PYTHON_SYSCONFIGDATA_NAME=_sysconfigdata_x86_64_conda_cos6_linux_gnu
|
| 252 |
+
build_alias=x86_64-conda-linux-gnu
|
| 253 |
+
host_alias=x86_64-conda-linux-gnu
|
latest
ADDED
|
@@ -0,0 +1 @@
|
|
|
|
|
|
|
| 1 |
+
global_step420
|
pytorch_model.bin
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:63ad93ceaa8fea8843727c34dc01cbef154a0dce8fe6fbc496226572a9fe4944
|
| 3 |
+
size 26429523941
|
script.sh
ADDED
|
@@ -0,0 +1,122 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
#!/usr/bin/env bash
|
| 2 |
+
#
|
| 3 |
+
# Copyright 2023 PKU-Alignment Team. All Rights Reserved.
|
| 4 |
+
#
|
| 5 |
+
# Licensed under the Apache License, Version 2.0 (the "License");
|
| 6 |
+
# you may not use this file except in compliance with the License.
|
| 7 |
+
# You may obtain a copy of the License at
|
| 8 |
+
#
|
| 9 |
+
# http://www.apache.org/licenses/LICENSE-2.0
|
| 10 |
+
#
|
| 11 |
+
# Unless required by applicable law or agreed to in writing, software
|
| 12 |
+
# distributed under the License is distributed on an "AS IS" BASIS,
|
| 13 |
+
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
| 14 |
+
# See the License for the specific language governing permissions and
|
| 15 |
+
# limitations under the License.
|
| 16 |
+
# ==============================================================================
|
| 17 |
+
export WANDB_API_KEY="f6021dca133c93e80a7dae4620bd335d4d08cac6"
|
| 18 |
+
|
| 19 |
+
if [ -z "${BASH_VERSION}" ]; then
|
| 20 |
+
echo "Please use bash to run this script." >&2
|
| 21 |
+
exit 1
|
| 22 |
+
fi
|
| 23 |
+
|
| 24 |
+
set -x
|
| 25 |
+
|
| 26 |
+
SCRIPT_DIR="$(cd "$(dirname "$0")" &>/dev/null && pwd)"
|
| 27 |
+
ROOT_DIR="$(dirname "${SCRIPT_DIR}")"
|
| 28 |
+
export PYTHONPATH="${ROOT_DIR}${PYTHONPATH:+:${PYTHONPATH}}"
|
| 29 |
+
export LOGLEVEL="${LOGLEVEL:-WARNING}"
|
| 30 |
+
|
| 31 |
+
MODEL_NAME_OR_PATH="output/llama-7b/sft"
|
| 32 |
+
OUTPUT_DIR="${ROOT_DIR}/output/harmless/rm"
|
| 33 |
+
DATASET_PATH="PKU-SafeRLHF-harmless-only-30k"
|
| 34 |
+
ZERO_STAGE=3
|
| 35 |
+
while [[ "$#" -gt 0 ]]; do
|
| 36 |
+
arg="$1"
|
| 37 |
+
shift
|
| 38 |
+
case "${arg}" in
|
| 39 |
+
--model_name_or_path)
|
| 40 |
+
MODEL_NAME_OR_PATH="$1"
|
| 41 |
+
shift
|
| 42 |
+
;;
|
| 43 |
+
--model_name_or_path=*)
|
| 44 |
+
MODEL_NAME_OR_PATH="${arg#*=}"
|
| 45 |
+
;;
|
| 46 |
+
--output_dir)
|
| 47 |
+
OUTPUT_DIR="$1"
|
| 48 |
+
shift
|
| 49 |
+
;;
|
| 50 |
+
--output_dir=*)
|
| 51 |
+
OUTPUT_DIR="${arg#*=}"
|
| 52 |
+
;;
|
| 53 |
+
--zero_stage)
|
| 54 |
+
ZERO_STAGE="$1"
|
| 55 |
+
shift
|
| 56 |
+
;;
|
| 57 |
+
--zero_stage=*)
|
| 58 |
+
ZERO_STAGE="${arg#*=}"
|
| 59 |
+
;;
|
| 60 |
+
--dataset_path)
|
| 61 |
+
DATASET_PATH="$1"
|
| 62 |
+
shift
|
| 63 |
+
;;
|
| 64 |
+
*)
|
| 65 |
+
echo "Unknown parameter passed: '${arg}'" >&2
|
| 66 |
+
exit 1
|
| 67 |
+
;;
|
| 68 |
+
esac
|
| 69 |
+
done
|
| 70 |
+
|
| 71 |
+
mkdir -p "${OUTPUT_DIR}"
|
| 72 |
+
OUTPUT_DIR="$(cd "${OUTPUT_DIR}" &>/dev/null && pwd)"
|
| 73 |
+
if [[ ! -f "${OUTPUT_DIR}/.gitignore" ]]; then
|
| 74 |
+
echo '*' >"${OUTPUT_DIR}/.gitignore"
|
| 75 |
+
fi
|
| 76 |
+
|
| 77 |
+
cp -f "$0" "${OUTPUT_DIR}/script.sh"
|
| 78 |
+
|
| 79 |
+
if [[ -z "${WANDB_API_KEY}" ]]; then
|
| 80 |
+
export WANDB_MODE="offline"
|
| 81 |
+
fi
|
| 82 |
+
|
| 83 |
+
MASTER_PORT_START=10000
|
| 84 |
+
MASTER_PORT_END=65535
|
| 85 |
+
MASTER_PORT="$(
|
| 86 |
+
comm -23 \
|
| 87 |
+
<(seq "${MASTER_PORT_START}" "${MASTER_PORT_END}" | sort) \
|
| 88 |
+
<(ss -Htan | awk '{ print $4 }' | awk -F ':' '{ print $NF }' | sort -u) |
|
| 89 |
+
shuf | head -n 1
|
| 90 |
+
)"
|
| 91 |
+
|
| 92 |
+
exec 1> >(tee "${OUTPUT_DIR}/stdout.log" >&1) 2> >(tee "${OUTPUT_DIR}/stderr.log" >&2)
|
| 93 |
+
|
| 94 |
+
deepspeed --num_nodes=1 --num_gpus=4 \
|
| 95 |
+
--master_port "${MASTER_PORT}" \
|
| 96 |
+
--module safe_rlhf.values.reward \
|
| 97 |
+
--train_datasets PKU-SafeRLHF/train:1.0:"${DATASET_PATH}" \
|
| 98 |
+
--eval_datasets PKU-SafeRLHF/test \
|
| 99 |
+
--model_name_or_path "${MODEL_NAME_OR_PATH}" \
|
| 100 |
+
--max_length 512 \
|
| 101 |
+
--trust_remote_code True \
|
| 102 |
+
--loss_type sequence-wise \
|
| 103 |
+
--epochs 2 \
|
| 104 |
+
--per_device_train_batch_size 16 \
|
| 105 |
+
--per_device_eval_batch_size 16 \
|
| 106 |
+
--gradient_accumulation_steps 2 \
|
| 107 |
+
--gradient_checkpointing \
|
| 108 |
+
--normalize_score_during_training False \
|
| 109 |
+
--normalizer_type ExponentialMovingAverage \
|
| 110 |
+
--normalizer_momentum 0.9 \
|
| 111 |
+
--learning_rate 2e-5 \
|
| 112 |
+
--lr_scheduler_type cosine \
|
| 113 |
+
--lr_warmup_ratio 0.03 \
|
| 114 |
+
--weight_decay 0.1 \
|
| 115 |
+
--seed 42 \
|
| 116 |
+
--eval_strategy epoch \
|
| 117 |
+
--output_dir "${OUTPUT_DIR}" \
|
| 118 |
+
--log_type wandb \
|
| 119 |
+
--log_project Safe-RLHF-RM \
|
| 120 |
+
--zero_stage "${ZERO_STAGE}" \
|
| 121 |
+
--bf16 True \
|
| 122 |
+
--tf32 True
|
special_tokens_map.json
ADDED
|
@@ -0,0 +1,24 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"bos_token": {
|
| 3 |
+
"content": "<s>",
|
| 4 |
+
"lstrip": false,
|
| 5 |
+
"normalized": true,
|
| 6 |
+
"rstrip": false,
|
| 7 |
+
"single_word": false
|
| 8 |
+
},
|
| 9 |
+
"eos_token": {
|
| 10 |
+
"content": "</s>",
|
| 11 |
+
"lstrip": false,
|
| 12 |
+
"normalized": true,
|
| 13 |
+
"rstrip": false,
|
| 14 |
+
"single_word": false
|
| 15 |
+
},
|
| 16 |
+
"pad_token": "<pad>",
|
| 17 |
+
"unk_token": {
|
| 18 |
+
"content": "<unk>",
|
| 19 |
+
"lstrip": false,
|
| 20 |
+
"normalized": true,
|
| 21 |
+
"rstrip": false,
|
| 22 |
+
"single_word": false
|
| 23 |
+
}
|
| 24 |
+
}
|
stderr.log
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
stdout.log
ADDED
|
@@ -0,0 +1,41 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
[2024-01-05 20:02:29,789] [INFO] [real_accelerator.py:161:get_accelerator] Setting ds_accelerator to cuda (auto detect)
|
| 2 |
+
[2024-01-05 20:02:34,703] [WARNING] [runner.py:202:fetch_hostfile] Unable to find hostfile, will proceed with training with local resources only.
|
| 3 |
+
Detected CUDA_VISIBLE_DEVICES=0,1,2,3 but ignoring it because one or several of --include/--exclude/--num_gpus/--num_nodes cl args were used. If you want to use CUDA_VISIBLE_DEVICES don't pass any of these arguments to deepspeed.
|
| 4 |
+
[2024-01-05 20:02:34,703] [INFO] [runner.py:571:main] cmd = /data/jiongxiao_wang/anaconda3/envs/safe-rlhf/bin/python -u -m deepspeed.launcher.launch --world_info=eyJsb2NhbGhvc3QiOiBbMCwgMSwgMiwgM119 --master_addr=127.0.0.1 --master_port=47607 --module --enable_each_rank_log=None safe_rlhf.values.reward --train_datasets PKU-SafeRLHF/train:1.0:PKU-SafeRLHF-harmless-only-30k --eval_datasets PKU-SafeRLHF/test --model_name_or_path output/sft --max_length 512 --trust_remote_code True --loss_type sequence-wise --epochs 2 --per_device_train_batch_size 16 --per_device_eval_batch_size 16 --gradient_accumulation_steps 2 --gradient_checkpointing --normalize_score_during_training False --normalizer_type ExponentialMovingAverage --normalizer_momentum 0.9 --learning_rate 2e-5 --lr_scheduler_type cosine --lr_warmup_ratio 0.03 --weight_decay 0.1 --seed 42 --eval_strategy epoch --output_dir /data/jiongxiao_wang/rlhf_attack/safe-rlhf/output/rm_30k --log_type wandb --log_project Safe-RLHF-RM --zero_stage 3 --bf16 True --tf32 True
|
| 5 |
+
[2024-01-05 20:02:36,745] [INFO] [real_accelerator.py:161:get_accelerator] Setting ds_accelerator to cuda (auto detect)
|
| 6 |
+
[2024-01-05 20:02:39,776] [INFO] [launch.py:145:main] WORLD INFO DICT: {'localhost': [0, 1, 2, 3]}
|
| 7 |
+
[2024-01-05 20:02:39,776] [INFO] [launch.py:151:main] nnodes=1, num_local_procs=4, node_rank=0
|
| 8 |
+
[2024-01-05 20:02:39,776] [INFO] [launch.py:162:main] global_rank_mapping=defaultdict(<class 'list'>, {'localhost': [0, 1, 2, 3]})
|
| 9 |
+
[2024-01-05 20:02:39,776] [INFO] [launch.py:163:main] dist_world_size=4
|
| 10 |
+
[2024-01-05 20:02:39,776] [INFO] [launch.py:165:main] Setting CUDA_VISIBLE_DEVICES=0,1,2,3
|
| 11 |
+
[2024-01-05 20:02:42,171] [INFO] [real_accelerator.py:161:get_accelerator] Setting ds_accelerator to cuda (auto detect)
|
| 12 |
+
[2024-01-05 20:02:42,172] [INFO] [real_accelerator.py:161:get_accelerator] Setting ds_accelerator to cuda (auto detect)
|
| 13 |
+
[2024-01-05 20:02:42,174] [INFO] [real_accelerator.py:161:get_accelerator] Setting ds_accelerator to cuda (auto detect)
|
| 14 |
+
[2024-01-05 20:02:42,175] [INFO] [real_accelerator.py:161:get_accelerator] Setting ds_accelerator to cuda (auto detect)
|
| 15 |
+
[2024-01-05 20:02:52,431] [INFO] [comm.py:637:init_distributed] cdb=None
|
| 16 |
+
[2024-01-05 20:02:52,431] [INFO] [comm.py:668:init_distributed] Initializing TorchBackend in DeepSpeed with backend nccl
|
| 17 |
+
[2024-01-05 20:02:52,437] [INFO] [comm.py:637:init_distributed] cdb=None
|
| 18 |
+
[2024-01-05 20:02:52,442] [INFO] [comm.py:637:init_distributed] cdb=None
|
| 19 |
+
[2024-01-05 20:02:52,443] [INFO] [comm.py:637:init_distributed] cdb=None
|
| 20 |
+
Set logger level to WARNING.
|
| 21 |
+
ninja: no work to do.
|
| 22 |
+
Time to load fused_adam op: 0.13281846046447754 seconds
|
| 23 |
+
Time to load fused_adam op: 0.20154452323913574 seconds
|
| 24 |
+
Time to load fused_adam op: 0.2013847827911377 seconds
|
| 25 |
+
Time to load fused_adam op: 0.2015242576599121 seconds
|
| 26 |
+
Parameter Offload: Total persistent parameters: 270336 in 66 params
|
| 27 |
+
***** Running training *****
|
| 28 |
+
Saving model to "/data/jiongxiao_wang/rlhf_attack/safe-rlhf/output/rm_30k" ...
|
| 29 |
+
Saving DeepSpeed Checkpoints...
|
| 30 |
+
Converting DeepSpeed Checkpoints to Hugging Face format...
|
| 31 |
+
[2024-01-05 20:56:50,901] [INFO] [real_accelerator.py:161:get_accelerator] Setting ds_accelerator to cuda (auto detect)
|
| 32 |
+
Processing zero checkpoint './global_step420'
|
| 33 |
+
Detected checkpoint of type zero stage 3, world_size: 4
|
| 34 |
+
Parsing checkpoint created by deepspeed==0.12.6
|
| 35 |
+
Reconstructed Trainable fp32 state dict with 291 params 6607351808 elements
|
| 36 |
+
Saving fp32 state dict to pytorch_model.bin
|
| 37 |
+
Model saved!
|
| 38 |
+
[2024-01-05 20:58:03,159] [INFO] [launch.py:347:main] Process 250383 exits successfully.
|
| 39 |
+
[2024-01-05 20:58:03,159] [INFO] [launch.py:347:main] Process 250381 exits successfully.
|
| 40 |
+
[2024-01-05 20:58:04,160] [INFO] [launch.py:347:main] Process 250382 exits successfully.
|
| 41 |
+
[2024-01-05 20:58:11,167] [INFO] [launch.py:347:main] Process 250380 exits successfully.
|
tokenizer.model
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9e556afd44213b6bd1be2b850ebbbd98f5481437a8021afaf58ee7fb1818d347
|
| 3 |
+
size 499723
|
tokenizer_config.json
ADDED
|
@@ -0,0 +1,35 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"add_bos_token": true,
|
| 3 |
+
"add_eos_token": false,
|
| 4 |
+
"bos_token": {
|
| 5 |
+
"__type": "AddedToken",
|
| 6 |
+
"content": "<s>",
|
| 7 |
+
"lstrip": false,
|
| 8 |
+
"normalized": true,
|
| 9 |
+
"rstrip": false,
|
| 10 |
+
"single_word": false
|
| 11 |
+
},
|
| 12 |
+
"clean_up_tokenization_spaces": false,
|
| 13 |
+
"eos_token": {
|
| 14 |
+
"__type": "AddedToken",
|
| 15 |
+
"content": "</s>",
|
| 16 |
+
"lstrip": false,
|
| 17 |
+
"normalized": true,
|
| 18 |
+
"rstrip": false,
|
| 19 |
+
"single_word": false
|
| 20 |
+
},
|
| 21 |
+
"legacy": true,
|
| 22 |
+
"model_max_length": 512,
|
| 23 |
+
"pad_token": null,
|
| 24 |
+
"padding_side": "right",
|
| 25 |
+
"sp_model_kwargs": {},
|
| 26 |
+
"tokenizer_class": "LlamaTokenizer",
|
| 27 |
+
"unk_token": {
|
| 28 |
+
"__type": "AddedToken",
|
| 29 |
+
"content": "<unk>",
|
| 30 |
+
"lstrip": false,
|
| 31 |
+
"normalized": true,
|
| 32 |
+
"rstrip": false,
|
| 33 |
+
"single_word": false
|
| 34 |
+
}
|
| 35 |
+
}
|
zero_to_fp32.py
ADDED
|
@@ -0,0 +1,587 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
#!/usr/bin/env python
|
| 2 |
+
|
| 3 |
+
# Copyright (c) Microsoft Corporation.
|
| 4 |
+
# SPDX-License-Identifier: Apache-2.0
|
| 5 |
+
|
| 6 |
+
# DeepSpeed Team
|
| 7 |
+
|
| 8 |
+
# This script extracts fp32 consolidated weights from a zero 1, 2 and 3 DeepSpeed checkpoints. It gets
|
| 9 |
+
# copied into the top level checkpoint dir, so the user can easily do the conversion at any point in
|
| 10 |
+
# the future. Once extracted, the weights don't require DeepSpeed and can be used in any
|
| 11 |
+
# application.
|
| 12 |
+
#
|
| 13 |
+
# example: python zero_to_fp32.py . pytorch_model.bin
|
| 14 |
+
|
| 15 |
+
import argparse
|
| 16 |
+
import torch
|
| 17 |
+
import glob
|
| 18 |
+
import math
|
| 19 |
+
import os
|
| 20 |
+
import re
|
| 21 |
+
from collections import OrderedDict
|
| 22 |
+
from dataclasses import dataclass
|
| 23 |
+
|
| 24 |
+
# while this script doesn't use deepspeed to recover data, since the checkpoints are pickled with
|
| 25 |
+
# DeepSpeed data structures it has to be available in the current python environment.
|
| 26 |
+
from deepspeed.utils import logger
|
| 27 |
+
from deepspeed.checkpoint.constants import (DS_VERSION, OPTIMIZER_STATE_DICT, SINGLE_PARTITION_OF_FP32_GROUPS,
|
| 28 |
+
FP32_FLAT_GROUPS, ZERO_STAGE, PARTITION_COUNT, PARAM_SHAPES, BUFFER_NAMES,
|
| 29 |
+
FROZEN_PARAM_SHAPES, FROZEN_PARAM_FRAGMENTS)
|
| 30 |
+
|
| 31 |
+
|
| 32 |
+
@dataclass
|
| 33 |
+
class zero_model_state:
|
| 34 |
+
buffers: dict()
|
| 35 |
+
param_shapes: dict()
|
| 36 |
+
shared_params: list
|
| 37 |
+
ds_version: int
|
| 38 |
+
frozen_param_shapes: dict()
|
| 39 |
+
frozen_param_fragments: dict()
|
| 40 |
+
|
| 41 |
+
|
| 42 |
+
debug = 0
|
| 43 |
+
|
| 44 |
+
# load to cpu
|
| 45 |
+
device = torch.device('cpu')
|
| 46 |
+
|
| 47 |
+
|
| 48 |
+
def atoi(text):
|
| 49 |
+
return int(text) if text.isdigit() else text
|
| 50 |
+
|
| 51 |
+
|
| 52 |
+
def natural_keys(text):
|
| 53 |
+
'''
|
| 54 |
+
alist.sort(key=natural_keys) sorts in human order
|
| 55 |
+
http://nedbatchelder.com/blog/200712/human_sorting.html
|
| 56 |
+
(See Toothy's implementation in the comments)
|
| 57 |
+
'''
|
| 58 |
+
return [atoi(c) for c in re.split(r'(\d+)', text)]
|
| 59 |
+
|
| 60 |
+
|
| 61 |
+
def get_model_state_file(checkpoint_dir, zero_stage):
|
| 62 |
+
if not os.path.isdir(checkpoint_dir):
|
| 63 |
+
raise FileNotFoundError(f"Directory '{checkpoint_dir}' doesn't exist")
|
| 64 |
+
|
| 65 |
+
# there should be only one file
|
| 66 |
+
if zero_stage <= 2:
|
| 67 |
+
file = os.path.join(checkpoint_dir, "mp_rank_00_model_states.pt")
|
| 68 |
+
elif zero_stage == 3:
|
| 69 |
+
file = os.path.join(checkpoint_dir, "zero_pp_rank_0_mp_rank_00_model_states.pt")
|
| 70 |
+
|
| 71 |
+
if not os.path.exists(file):
|
| 72 |
+
raise FileNotFoundError(f"can't find model states file at '{file}'")
|
| 73 |
+
|
| 74 |
+
return file
|
| 75 |
+
|
| 76 |
+
|
| 77 |
+
def get_checkpoint_files(checkpoint_dir, glob_pattern):
|
| 78 |
+
# XXX: need to test that this simple glob rule works for multi-node setup too
|
| 79 |
+
ckpt_files = sorted(glob.glob(os.path.join(checkpoint_dir, glob_pattern)), key=natural_keys)
|
| 80 |
+
|
| 81 |
+
if len(ckpt_files) == 0:
|
| 82 |
+
raise FileNotFoundError(f"can't find {glob_pattern} files in directory '{checkpoint_dir}'")
|
| 83 |
+
|
| 84 |
+
return ckpt_files
|
| 85 |
+
|
| 86 |
+
|
| 87 |
+
def get_optim_files(checkpoint_dir):
|
| 88 |
+
return get_checkpoint_files(checkpoint_dir, "*_optim_states.pt")
|
| 89 |
+
|
| 90 |
+
|
| 91 |
+
def get_model_state_files(checkpoint_dir):
|
| 92 |
+
return get_checkpoint_files(checkpoint_dir, "*_model_states.pt")
|
| 93 |
+
|
| 94 |
+
|
| 95 |
+
def parse_model_states(files):
|
| 96 |
+
zero_model_states = []
|
| 97 |
+
for file in files:
|
| 98 |
+
state_dict = torch.load(file, map_location=device)
|
| 99 |
+
|
| 100 |
+
if BUFFER_NAMES not in state_dict:
|
| 101 |
+
raise ValueError(f"{file} is not a model state checkpoint")
|
| 102 |
+
buffer_names = state_dict[BUFFER_NAMES]
|
| 103 |
+
if debug:
|
| 104 |
+
print("Found buffers:", buffer_names)
|
| 105 |
+
|
| 106 |
+
# recover just the buffers while restoring them to fp32 if they were saved in fp16
|
| 107 |
+
buffers = {k: v.float() for k, v in state_dict["module"].items() if k in buffer_names}
|
| 108 |
+
param_shapes = state_dict[PARAM_SHAPES]
|
| 109 |
+
|
| 110 |
+
# collect parameters that are included in param_shapes
|
| 111 |
+
param_names = []
|
| 112 |
+
for s in param_shapes:
|
| 113 |
+
for name in s.keys():
|
| 114 |
+
param_names.append(name)
|
| 115 |
+
|
| 116 |
+
# update with frozen parameters
|
| 117 |
+
frozen_param_shapes = state_dict.get(FROZEN_PARAM_SHAPES, None)
|
| 118 |
+
if frozen_param_shapes is not None:
|
| 119 |
+
if debug:
|
| 120 |
+
print(f"Found frozen_param_shapes: {frozen_param_shapes}")
|
| 121 |
+
param_names += list(frozen_param_shapes.keys())
|
| 122 |
+
|
| 123 |
+
# handle shared params
|
| 124 |
+
shared_params = [[k, v] for k, v in state_dict["shared_params"].items()]
|
| 125 |
+
|
| 126 |
+
ds_version = state_dict.get(DS_VERSION, None)
|
| 127 |
+
|
| 128 |
+
frozen_param_fragments = state_dict.get(FROZEN_PARAM_FRAGMENTS, None)
|
| 129 |
+
|
| 130 |
+
z_model_state = zero_model_state(buffers=buffers,
|
| 131 |
+
param_shapes=param_shapes,
|
| 132 |
+
shared_params=shared_params,
|
| 133 |
+
ds_version=ds_version,
|
| 134 |
+
frozen_param_shapes=frozen_param_shapes,
|
| 135 |
+
frozen_param_fragments=frozen_param_fragments)
|
| 136 |
+
zero_model_states.append(z_model_state)
|
| 137 |
+
|
| 138 |
+
return zero_model_states
|
| 139 |
+
|
| 140 |
+
|
| 141 |
+
def parse_optim_states(files, ds_checkpoint_dir):
|
| 142 |
+
|
| 143 |
+
total_files = len(files)
|
| 144 |
+
state_dicts = []
|
| 145 |
+
for f in files:
|
| 146 |
+
state_dict = torch.load(f, map_location=device)
|
| 147 |
+
# immediately discard the potentially huge 2 optimizer states as we only care for fp32 master weights
|
| 148 |
+
# and also handle the case where it was already removed by another helper script
|
| 149 |
+
state_dict["optimizer_state_dict"].pop("optimizer_state_dict", None)
|
| 150 |
+
state_dicts.append(state_dict)
|
| 151 |
+
|
| 152 |
+
if not ZERO_STAGE in state_dicts[0][OPTIMIZER_STATE_DICT]:
|
| 153 |
+
raise ValueError(f"{files[0]} is not a zero checkpoint")
|
| 154 |
+
zero_stage = state_dicts[0][OPTIMIZER_STATE_DICT][ZERO_STAGE]
|
| 155 |
+
world_size = state_dicts[0][OPTIMIZER_STATE_DICT][PARTITION_COUNT]
|
| 156 |
+
|
| 157 |
+
# For ZeRO-2 each param group can have different partition_count as data parallelism for expert
|
| 158 |
+
# parameters can be different from data parallelism for non-expert parameters. So we can just
|
| 159 |
+
# use the max of the partition_count to get the dp world_size.
|
| 160 |
+
|
| 161 |
+
if type(world_size) is list:
|
| 162 |
+
world_size = max(world_size)
|
| 163 |
+
|
| 164 |
+
if world_size != total_files:
|
| 165 |
+
raise ValueError(
|
| 166 |
+
f"Expected {world_size} of '*_optim_states.pt' under '{ds_checkpoint_dir}' but found {total_files} files. "
|
| 167 |
+
"Possibly due to an overwrite of an old checkpoint, or a checkpoint didn't get saved by one or more processes."
|
| 168 |
+
)
|
| 169 |
+
|
| 170 |
+
# the groups are named differently in each stage
|
| 171 |
+
if zero_stage <= 2:
|
| 172 |
+
fp32_groups_key = SINGLE_PARTITION_OF_FP32_GROUPS
|
| 173 |
+
elif zero_stage == 3:
|
| 174 |
+
fp32_groups_key = FP32_FLAT_GROUPS
|
| 175 |
+
else:
|
| 176 |
+
raise ValueError(f"unknown zero stage {zero_stage}")
|
| 177 |
+
|
| 178 |
+
if zero_stage <= 2:
|
| 179 |
+
fp32_flat_groups = [state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key] for i in range(len(state_dicts))]
|
| 180 |
+
elif zero_stage == 3:
|
| 181 |
+
# if there is more than one param group, there will be multiple flattened tensors - one
|
| 182 |
+
# flattened tensor per group - for simplicity merge them into a single tensor
|
| 183 |
+
#
|
| 184 |
+
# XXX: could make the script more memory efficient for when there are multiple groups - it
|
| 185 |
+
# will require matching the sub-lists of param_shapes for each param group flattened tensor
|
| 186 |
+
|
| 187 |
+
fp32_flat_groups = [
|
| 188 |
+
torch.cat(state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key], 0) for i in range(len(state_dicts))
|
| 189 |
+
]
|
| 190 |
+
|
| 191 |
+
return zero_stage, world_size, fp32_flat_groups
|
| 192 |
+
|
| 193 |
+
|
| 194 |
+
def _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir):
|
| 195 |
+
"""
|
| 196 |
+
Returns fp32 state_dict reconstructed from ds checkpoint
|
| 197 |
+
|
| 198 |
+
Args:
|
| 199 |
+
- ``ds_checkpoint_dir``: path to the deepspeed checkpoint folder (where the optimizer files are)
|
| 200 |
+
|
| 201 |
+
"""
|
| 202 |
+
print(f"Processing zero checkpoint '{ds_checkpoint_dir}'")
|
| 203 |
+
|
| 204 |
+
optim_files = get_optim_files(ds_checkpoint_dir)
|
| 205 |
+
zero_stage, world_size, fp32_flat_groups = parse_optim_states(optim_files, ds_checkpoint_dir)
|
| 206 |
+
print(f"Detected checkpoint of type zero stage {zero_stage}, world_size: {world_size}")
|
| 207 |
+
|
| 208 |
+
model_files = get_model_state_files(ds_checkpoint_dir)
|
| 209 |
+
|
| 210 |
+
zero_model_states = parse_model_states(model_files)
|
| 211 |
+
print(f'Parsing checkpoint created by deepspeed=={zero_model_states[0].ds_version}')
|
| 212 |
+
|
| 213 |
+
if zero_stage <= 2:
|
| 214 |
+
return _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states)
|
| 215 |
+
elif zero_stage == 3:
|
| 216 |
+
return _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states)
|
| 217 |
+
|
| 218 |
+
|
| 219 |
+
def _zero2_merge_frozen_params(state_dict, zero_model_states):
|
| 220 |
+
if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
|
| 221 |
+
return
|
| 222 |
+
|
| 223 |
+
frozen_param_shapes = zero_model_states[0].frozen_param_shapes
|
| 224 |
+
frozen_param_fragments = zero_model_states[0].frozen_param_fragments
|
| 225 |
+
|
| 226 |
+
if debug:
|
| 227 |
+
num_elem = sum(s.numel() for s in frozen_param_shapes.values())
|
| 228 |
+
print(f'rank 0: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
|
| 229 |
+
|
| 230 |
+
wanted_params = len(frozen_param_shapes)
|
| 231 |
+
wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
|
| 232 |
+
avail_numel = sum([p.numel() for p in frozen_param_fragments.values()])
|
| 233 |
+
print(f'Frozen params: Have {avail_numel} numels to process.')
|
| 234 |
+
print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
|
| 235 |
+
|
| 236 |
+
total_params = 0
|
| 237 |
+
total_numel = 0
|
| 238 |
+
for name, shape in frozen_param_shapes.items():
|
| 239 |
+
total_params += 1
|
| 240 |
+
unpartitioned_numel = shape.numel()
|
| 241 |
+
total_numel += unpartitioned_numel
|
| 242 |
+
|
| 243 |
+
state_dict[name] = frozen_param_fragments[name]
|
| 244 |
+
|
| 245 |
+
if debug:
|
| 246 |
+
print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
|
| 247 |
+
|
| 248 |
+
print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
|
| 249 |
+
|
| 250 |
+
|
| 251 |
+
def _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
|
| 252 |
+
param_shapes = zero_model_states[0].param_shapes
|
| 253 |
+
|
| 254 |
+
# Reconstruction protocol:
|
| 255 |
+
#
|
| 256 |
+
# XXX: document this
|
| 257 |
+
|
| 258 |
+
if debug:
|
| 259 |
+
for i in range(world_size):
|
| 260 |
+
for j in range(len(fp32_flat_groups[0])):
|
| 261 |
+
print(f"{FP32_FLAT_GROUPS}[{i}][{j}].shape={fp32_flat_groups[i][j].shape}")
|
| 262 |
+
|
| 263 |
+
# XXX: memory usage doubles here (zero2)
|
| 264 |
+
num_param_groups = len(fp32_flat_groups[0])
|
| 265 |
+
merged_single_partition_of_fp32_groups = []
|
| 266 |
+
for i in range(num_param_groups):
|
| 267 |
+
merged_partitions = [sd[i] for sd in fp32_flat_groups]
|
| 268 |
+
full_single_fp32_vector = torch.cat(merged_partitions, 0)
|
| 269 |
+
merged_single_partition_of_fp32_groups.append(full_single_fp32_vector)
|
| 270 |
+
avail_numel = sum(
|
| 271 |
+
[full_single_fp32_vector.numel() for full_single_fp32_vector in merged_single_partition_of_fp32_groups])
|
| 272 |
+
|
| 273 |
+
if debug:
|
| 274 |
+
wanted_params = sum([len(shapes) for shapes in param_shapes])
|
| 275 |
+
wanted_numel = sum([sum(shape.numel() for shape in shapes.values()) for shapes in param_shapes])
|
| 276 |
+
# not asserting if there is a mismatch due to possible padding
|
| 277 |
+
print(f"Have {avail_numel} numels to process.")
|
| 278 |
+
print(f"Need {wanted_numel} numels in {wanted_params} params.")
|
| 279 |
+
|
| 280 |
+
# params
|
| 281 |
+
# XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
|
| 282 |
+
# out-of-core computing solution
|
| 283 |
+
total_numel = 0
|
| 284 |
+
total_params = 0
|
| 285 |
+
for shapes, full_single_fp32_vector in zip(param_shapes, merged_single_partition_of_fp32_groups):
|
| 286 |
+
offset = 0
|
| 287 |
+
avail_numel = full_single_fp32_vector.numel()
|
| 288 |
+
for name, shape in shapes.items():
|
| 289 |
+
|
| 290 |
+
unpartitioned_numel = shape.numel()
|
| 291 |
+
total_numel += unpartitioned_numel
|
| 292 |
+
total_params += 1
|
| 293 |
+
|
| 294 |
+
if debug:
|
| 295 |
+
print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
|
| 296 |
+
state_dict[name] = full_single_fp32_vector.narrow(0, offset, unpartitioned_numel).view(shape)
|
| 297 |
+
offset += unpartitioned_numel
|
| 298 |
+
|
| 299 |
+
# Z2 started to align to 2*world_size to improve nccl performance. Therefore both offset and
|
| 300 |
+
# avail_numel can differ by anywhere between 0..2*world_size. Due to two unrelated complex
|
| 301 |
+
# paddings performed in the code it's almost impossible to predict the exact numbers w/o the
|
| 302 |
+
# live optimizer object, so we are checking that the numbers are within the right range
|
| 303 |
+
align_to = 2 * world_size
|
| 304 |
+
|
| 305 |
+
def zero2_align(x):
|
| 306 |
+
return align_to * math.ceil(x / align_to)
|
| 307 |
+
|
| 308 |
+
if debug:
|
| 309 |
+
print(f"original offset={offset}, avail_numel={avail_numel}")
|
| 310 |
+
|
| 311 |
+
offset = zero2_align(offset)
|
| 312 |
+
avail_numel = zero2_align(avail_numel)
|
| 313 |
+
|
| 314 |
+
if debug:
|
| 315 |
+
print(f"aligned offset={offset}, avail_numel={avail_numel}")
|
| 316 |
+
|
| 317 |
+
# Sanity check
|
| 318 |
+
if offset != avail_numel:
|
| 319 |
+
raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
|
| 320 |
+
|
| 321 |
+
print(f"Reconstructed fp32 state dict with {total_params} params {total_numel} elements")
|
| 322 |
+
|
| 323 |
+
|
| 324 |
+
def _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states):
|
| 325 |
+
state_dict = OrderedDict()
|
| 326 |
+
|
| 327 |
+
# buffers
|
| 328 |
+
buffers = zero_model_states[0].buffers
|
| 329 |
+
state_dict.update(buffers)
|
| 330 |
+
if debug:
|
| 331 |
+
print(f"added {len(buffers)} buffers")
|
| 332 |
+
|
| 333 |
+
_zero2_merge_frozen_params(state_dict, zero_model_states)
|
| 334 |
+
|
| 335 |
+
_zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
|
| 336 |
+
|
| 337 |
+
# recover shared parameters
|
| 338 |
+
for pair in zero_model_states[0].shared_params:
|
| 339 |
+
if pair[1] in state_dict:
|
| 340 |
+
state_dict[pair[0]] = state_dict[pair[1]]
|
| 341 |
+
|
| 342 |
+
return state_dict
|
| 343 |
+
|
| 344 |
+
|
| 345 |
+
def zero3_partitioned_param_info(unpartitioned_numel, world_size):
|
| 346 |
+
remainder = unpartitioned_numel % world_size
|
| 347 |
+
padding_numel = (world_size - remainder) if remainder else 0
|
| 348 |
+
partitioned_numel = math.ceil(unpartitioned_numel / world_size)
|
| 349 |
+
return partitioned_numel, padding_numel
|
| 350 |
+
|
| 351 |
+
|
| 352 |
+
def _zero3_merge_frozen_params(state_dict, world_size, zero_model_states):
|
| 353 |
+
if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
|
| 354 |
+
return
|
| 355 |
+
|
| 356 |
+
if debug:
|
| 357 |
+
for i in range(world_size):
|
| 358 |
+
num_elem = sum(s.numel() for s in zero_model_states[i].frozen_param_fragments.values())
|
| 359 |
+
print(f'rank {i}: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
|
| 360 |
+
|
| 361 |
+
frozen_param_shapes = zero_model_states[0].frozen_param_shapes
|
| 362 |
+
wanted_params = len(frozen_param_shapes)
|
| 363 |
+
wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
|
| 364 |
+
avail_numel = sum([p.numel() for p in zero_model_states[0].frozen_param_fragments.values()]) * world_size
|
| 365 |
+
print(f'Frozen params: Have {avail_numel} numels to process.')
|
| 366 |
+
print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
|
| 367 |
+
|
| 368 |
+
total_params = 0
|
| 369 |
+
total_numel = 0
|
| 370 |
+
for name, shape in zero_model_states[0].frozen_param_shapes.items():
|
| 371 |
+
total_params += 1
|
| 372 |
+
unpartitioned_numel = shape.numel()
|
| 373 |
+
total_numel += unpartitioned_numel
|
| 374 |
+
|
| 375 |
+
param_frags = tuple(model_state.frozen_param_fragments[name] for model_state in zero_model_states)
|
| 376 |
+
state_dict[name] = torch.cat(param_frags, 0).narrow(0, 0, unpartitioned_numel).view(shape)
|
| 377 |
+
|
| 378 |
+
partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
|
| 379 |
+
|
| 380 |
+
if debug:
|
| 381 |
+
print(
|
| 382 |
+
f"Frozen params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
|
| 383 |
+
)
|
| 384 |
+
|
| 385 |
+
print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
|
| 386 |
+
|
| 387 |
+
|
| 388 |
+
def _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
|
| 389 |
+
param_shapes = zero_model_states[0].param_shapes
|
| 390 |
+
avail_numel = fp32_flat_groups[0].numel() * world_size
|
| 391 |
+
# Reconstruction protocol: For zero3 we need to zip the partitions together at boundary of each
|
| 392 |
+
# param, re-consolidating each param, while dealing with padding if any
|
| 393 |
+
|
| 394 |
+
# merge list of dicts, preserving order
|
| 395 |
+
param_shapes = {k: v for d in param_shapes for k, v in d.items()}
|
| 396 |
+
|
| 397 |
+
if debug:
|
| 398 |
+
for i in range(world_size):
|
| 399 |
+
print(f"{FP32_FLAT_GROUPS}[{i}].shape={fp32_flat_groups[i].shape}")
|
| 400 |
+
|
| 401 |
+
wanted_params = len(param_shapes)
|
| 402 |
+
wanted_numel = sum(shape.numel() for shape in param_shapes.values())
|
| 403 |
+
# not asserting if there is a mismatch due to possible padding
|
| 404 |
+
avail_numel = fp32_flat_groups[0].numel() * world_size
|
| 405 |
+
print(f"Trainable params: Have {avail_numel} numels to process.")
|
| 406 |
+
print(f"Trainable params: Need {wanted_numel} numels in {wanted_params} params.")
|
| 407 |
+
|
| 408 |
+
# params
|
| 409 |
+
# XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
|
| 410 |
+
# out-of-core computing solution
|
| 411 |
+
offset = 0
|
| 412 |
+
total_numel = 0
|
| 413 |
+
total_params = 0
|
| 414 |
+
for name, shape in param_shapes.items():
|
| 415 |
+
|
| 416 |
+
unpartitioned_numel = shape.numel()
|
| 417 |
+
total_numel += unpartitioned_numel
|
| 418 |
+
total_params += 1
|
| 419 |
+
|
| 420 |
+
partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
|
| 421 |
+
|
| 422 |
+
if debug:
|
| 423 |
+
print(
|
| 424 |
+
f"Trainable params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
|
| 425 |
+
)
|
| 426 |
+
|
| 427 |
+
# XXX: memory usage doubles here
|
| 428 |
+
state_dict[name] = torch.cat(
|
| 429 |
+
tuple(fp32_flat_groups[i].narrow(0, offset, partitioned_numel) for i in range(world_size)),
|
| 430 |
+
0).narrow(0, 0, unpartitioned_numel).view(shape)
|
| 431 |
+
offset += partitioned_numel
|
| 432 |
+
|
| 433 |
+
offset *= world_size
|
| 434 |
+
|
| 435 |
+
# Sanity check
|
| 436 |
+
if offset != avail_numel:
|
| 437 |
+
raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
|
| 438 |
+
|
| 439 |
+
print(f"Reconstructed Trainable fp32 state dict with {total_params} params {total_numel} elements")
|
| 440 |
+
|
| 441 |
+
|
| 442 |
+
def _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states):
|
| 443 |
+
state_dict = OrderedDict()
|
| 444 |
+
|
| 445 |
+
# buffers
|
| 446 |
+
buffers = zero_model_states[0].buffers
|
| 447 |
+
state_dict.update(buffers)
|
| 448 |
+
if debug:
|
| 449 |
+
print(f"added {len(buffers)} buffers")
|
| 450 |
+
|
| 451 |
+
_zero3_merge_frozen_params(state_dict, world_size, zero_model_states)
|
| 452 |
+
|
| 453 |
+
_zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
|
| 454 |
+
|
| 455 |
+
# recover shared parameters
|
| 456 |
+
for pair in zero_model_states[0].shared_params:
|
| 457 |
+
if pair[1] in state_dict:
|
| 458 |
+
state_dict[pair[0]] = state_dict[pair[1]]
|
| 459 |
+
|
| 460 |
+
return state_dict
|
| 461 |
+
|
| 462 |
+
|
| 463 |
+
def get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag=None):
|
| 464 |
+
"""
|
| 465 |
+
Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated state_dict that can be loaded with
|
| 466 |
+
``load_state_dict()`` and used for training without DeepSpeed or shared with others, for example
|
| 467 |
+
via a model hub.
|
| 468 |
+
|
| 469 |
+
Args:
|
| 470 |
+
- ``checkpoint_dir``: path to the desired checkpoint folder
|
| 471 |
+
- ``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``
|
| 472 |
+
|
| 473 |
+
Returns:
|
| 474 |
+
- pytorch ``state_dict``
|
| 475 |
+
|
| 476 |
+
Note: this approach may not work if your application doesn't have sufficient free CPU memory and
|
| 477 |
+
you may need to use the offline approach using the ``zero_to_fp32.py`` script that is saved with
|
| 478 |
+
the checkpoint.
|
| 479 |
+
|
| 480 |
+
A typical usage might be ::
|
| 481 |
+
|
| 482 |
+
from deepspeed.utils.zero_to_fp32 import get_fp32_state_dict_from_zero_checkpoint
|
| 483 |
+
# do the training and checkpoint saving
|
| 484 |
+
state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir) # already on cpu
|
| 485 |
+
model = model.cpu() # move to cpu
|
| 486 |
+
model.load_state_dict(state_dict)
|
| 487 |
+
# submit to model hub or save the model to share with others
|
| 488 |
+
|
| 489 |
+
In this example the ``model`` will no longer be usable in the deepspeed context of the same
|
| 490 |
+
application. i.e. you will need to re-initialize the deepspeed engine, since
|
| 491 |
+
``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
|
| 492 |
+
|
| 493 |
+
If you want it all done for you, use ``load_state_dict_from_zero_checkpoint`` instead.
|
| 494 |
+
|
| 495 |
+
"""
|
| 496 |
+
if tag is None:
|
| 497 |
+
latest_path = os.path.join(checkpoint_dir, 'latest')
|
| 498 |
+
if os.path.isfile(latest_path):
|
| 499 |
+
with open(latest_path, 'r') as fd:
|
| 500 |
+
tag = fd.read().strip()
|
| 501 |
+
else:
|
| 502 |
+
raise ValueError(f"Unable to find 'latest' file at {latest_path}")
|
| 503 |
+
|
| 504 |
+
ds_checkpoint_dir = os.path.join(checkpoint_dir, tag)
|
| 505 |
+
|
| 506 |
+
if not os.path.isdir(ds_checkpoint_dir):
|
| 507 |
+
raise FileNotFoundError(f"Directory '{ds_checkpoint_dir}' doesn't exist")
|
| 508 |
+
|
| 509 |
+
return _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir)
|
| 510 |
+
|
| 511 |
+
|
| 512 |
+
def convert_zero_checkpoint_to_fp32_state_dict(checkpoint_dir, output_file, tag=None):
|
| 513 |
+
"""
|
| 514 |
+
Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict`` file that can be
|
| 515 |
+
loaded with ``torch.load(file)`` + ``load_state_dict()`` and used for training without DeepSpeed.
|
| 516 |
+
|
| 517 |
+
Args:
|
| 518 |
+
- ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
|
| 519 |
+
- ``output_file``: path to the pytorch fp32 state_dict output file (e.g. path/pytorch_model.bin)
|
| 520 |
+
- ``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``
|
| 521 |
+
"""
|
| 522 |
+
|
| 523 |
+
state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag)
|
| 524 |
+
print(f"Saving fp32 state dict to {output_file}")
|
| 525 |
+
torch.save(state_dict, output_file)
|
| 526 |
+
|
| 527 |
+
|
| 528 |
+
def load_state_dict_from_zero_checkpoint(model, checkpoint_dir, tag=None):
|
| 529 |
+
"""
|
| 530 |
+
1. Put the provided model to cpu
|
| 531 |
+
2. Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict``
|
| 532 |
+
3. Load it into the provided model
|
| 533 |
+
|
| 534 |
+
Args:
|
| 535 |
+
- ``model``: the model object to update
|
| 536 |
+
- ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
|
| 537 |
+
- ``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``
|
| 538 |
+
|
| 539 |
+
Returns:
|
| 540 |
+
- ``model`: modified model
|
| 541 |
+
|
| 542 |
+
Make sure you have plenty of CPU memory available before you call this function. If you don't
|
| 543 |
+
have enough use the ``zero_to_fp32.py`` utility to do the conversion. You will find it
|
| 544 |
+
conveniently placed for you in the checkpoint folder.
|
| 545 |
+
|
| 546 |
+
A typical usage might be ::
|
| 547 |
+
|
| 548 |
+
from deepspeed.utils.zero_to_fp32 import load_state_dict_from_zero_checkpoint
|
| 549 |
+
model = load_state_dict_from_zero_checkpoint(trainer.model, checkpoint_dir)
|
| 550 |
+
# submit to model hub or save the model to share with others
|
| 551 |
+
|
| 552 |
+
Note, that once this was run, the ``model`` will no longer be usable in the deepspeed context
|
| 553 |
+
of the same application. i.e. you will need to re-initialize the deepspeed engine, since
|
| 554 |
+
``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
|
| 555 |
+
|
| 556 |
+
"""
|
| 557 |
+
logger.info(f"Extracting fp32 weights")
|
| 558 |
+
state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag)
|
| 559 |
+
|
| 560 |
+
logger.info(f"Overwriting model with fp32 weights")
|
| 561 |
+
model = model.cpu()
|
| 562 |
+
model.load_state_dict(state_dict, strict=False)
|
| 563 |
+
|
| 564 |
+
return model
|
| 565 |
+
|
| 566 |
+
|
| 567 |
+
if __name__ == "__main__":
|
| 568 |
+
|
| 569 |
+
parser = argparse.ArgumentParser()
|
| 570 |
+
parser.add_argument("checkpoint_dir",
|
| 571 |
+
type=str,
|
| 572 |
+
help="path to the desired checkpoint folder, e.g., path/checkpoint-12")
|
| 573 |
+
parser.add_argument(
|
| 574 |
+
"output_file",
|
| 575 |
+
type=str,
|
| 576 |
+
help="path to the pytorch fp32 state_dict output file (e.g. path/checkpoint-12/pytorch_model.bin)")
|
| 577 |
+
parser.add_argument("-t",
|
| 578 |
+
"--tag",
|
| 579 |
+
type=str,
|
| 580 |
+
default=None,
|
| 581 |
+
help="checkpoint tag used as a unique identifier for checkpoint. e.g., global_step1")
|
| 582 |
+
parser.add_argument("-d", "--debug", action='store_true', help="enable debug")
|
| 583 |
+
args = parser.parse_args()
|
| 584 |
+
|
| 585 |
+
debug = args.debug
|
| 586 |
+
|
| 587 |
+
convert_zero_checkpoint_to_fp32_state_dict(args.checkpoint_dir, args.output_file, tag=args.tag)
|