|
- # Copyright 2021 Huawei Technologies Co., Ltd
- #
- # Licensed under the Apache License, Version 2.0 (the "License");
- # you may not use this file except in compliance with the License.
- # You may obtain a copy of the License at
- #
- # http://www.apache.org/licenses/LICENSE-2.0
- #
- # Unless required by applicable law or agreed to in writing, software
- # distributed under the License is distributed on an "AS IS" BASIS,
- # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
- # See the License for the specific language governing permissions and
- # limitations under the License.
- # ============================================================================
- """
- PanGu predict run
- """
- import os
-
- import numpy as np
-
- import mindspore.common.dtype as mstype
- import mindspore.communication.management as D
- from mindspore import context, Tensor
- from mindspore import export
- from mindspore.context import ParallelMode
- from mindspore.parallel import set_algo_parameters
- from mindspore.parallel._cost_model_context import _set_multi_subgraphs
- from mindspore.train.model import Model
- ## from mindspore.train.serialization import load_distributed_checkpoint
- from src.serialization import load_distributed_checkpoint
- from src.pangu_alpha import PanguAlpha, EvalNet
- from src.pangu_alpha_config import PANGUALPHAConfig, set_parse
- from src.utils_m53_exp4 import get_args
-
- import time
- import moxing as mox
- from src.utils_m53_exp4 import download_data, ckpt_copy_tar_new, get_ckpt_file_list
-
- def load_model(args_opt):
- r"""
- The main function for load model
- """
- # Set execution mode
- context.set_context(save_graphs=False,
- mode=context.GRAPH_MODE,
- device_target=args_opt.device_target)
- context.set_context(variable_memory_max_size="30GB")
- # Set parallel context
- if args_opt.distribute == "true":
- D.init()
- device_num = D.get_group_size()
- rank = D.get_rank()
- print("rank_id is {}, device_num is {}".format(rank, device_num))
-
- local_strategy_ckpt_path="/cache/ckpt_strategy.ckpt"
- if rank % 8 == 0:
- os.system('ulimit -s 102400')
- mox.file.copy(src_url=args_opt.strategy_load_ckpt_path, dst_url=local_strategy_ckpt_path)
- steps_name = args_opt.load_ckpt_obs_path.split('/')[-1].split('-')[-1]
- print("steps_name", steps_name)
- name_in = args_opt.load_obs_ckptname
- ckpt_copy_tar_new(args_opt.load_ckpt_obs_path+name_in, target_path=args_opt.load_ckpt_local_path)
- mox.file.copy(f'{args_opt.load_ckpt_obs_path}/{name_in}_word_embedding.npy', f'{args_opt.load_ckpt_local_path}/word_embedding.npy')
- mox.file.copy(f'{args_opt.load_ckpt_obs_path}/{name_in}_position_embedding.npy', f'{args_opt.load_ckpt_local_path}/position_embedding.npy')
- mox.file.copy(f'{args_opt.load_ckpt_obs_path}/{name_in}_top_query_embedding.npy', f'{args_opt.load_ckpt_local_path}/top_query_embedding.npy')
- print("setting env success.")
- # 下载模型文件结束后,写一个文件来表示下载成功
- f = open("/tmp/download_ckpt.txt", 'w')
- f.close()
- # 此处用于阻塞其他进程,直到刷包以及下载数据集完成为止
- while not os.path.exists("/tmp/download_ckpt.txt"):
- time.sleep(1)
- print("\n\n************Checkpoint download succeed!*************\n\n", flush=True)
- if rank % 8 == 0:
- print(os.listdir(args_opt.load_ckpt_local_path), flush=True)
-
- context.reset_auto_parallel_context()
- context.set_auto_parallel_context(
- parallel_mode=ParallelMode.SEMI_AUTO_PARALLEL,
- gradients_mean=False,
- full_batch=True,
- loss_repeated_mean=True,
- enable_parallel_optimizer=False,
- strategy_ckpt_load_file=local_strategy_ckpt_path,
- pipeline_stages=args_opt.stage_num)
- set_algo_parameters(elementwise_op_strategy_follow=True)
- _set_multi_subgraphs()
-
- else:
- rank = 0
- device_num = 1
- context.reset_auto_parallel_context()
- context.set_auto_parallel_context(
- strategy_ckpt_load_file=local_strategy_ckpt_path)
-
- use_past = (args_opt.use_past == "true")
- print('local_rank:{}, start to run...'.format(rank), flush=True)
- if args_opt.export:
- use_past = True
- # Set model property
- model_parallel_num = args_opt.op_level_model_parallel_num
- data_parallel_num = int(device_num / model_parallel_num)
- per_batch_size = args_opt.per_batch_size
- batch_size = per_batch_size * data_parallel_num
- # Now only support single batch_size for predict
- if args_opt.run_type == "predict":
- batch_size = 1
- config = PANGUALPHAConfig(
- data_parallel_num=data_parallel_num,
- model_parallel_num=model_parallel_num,
- batch_size=batch_size,
- seq_length=args_opt.seq_length,
- vocab_size=args_opt.vocab_size,
- embedding_size=args_opt.embedding_size,
- num_layers=args_opt.num_layers,
- num_heads=args_opt.num_heads,
- expand_ratio=4,
- post_layernorm_residual=False,
- dropout_rate=0.0,
- compute_dtype=mstype.float16,
- use_past=use_past,
- stage_num=args_opt.stage_num,
- micro_size=args_opt.micro_size,
- eod_reset=False,
- word_emb_dp=True,
- load_ckpt_path=args_opt.load_ckpt_local_path,
- param_init_type=mstype.float32 if args_opt.param_init_type == 'fp32' else mstype.float16)
- print("===config is: ", config, flush=True)
- print("=====args_opt is: ", args_opt, flush=True)
-
- ckpt_name = args_opt.load_ckpt_name
- # Define network
- pangu_alpha = PanguAlpha(config)
- eval_net = EvalNet(pangu_alpha)
- eval_net.set_train(False)
- model_predict = Model(eval_net)
- # Compile network and obtain tensor layout for loading ckpt
- inputs_np = Tensor(np.ones(shape=(config.batch_size, config.seq_length)), mstype.int32)
- current_index = Tensor(np.array([0]), mstype.int32)
-
- if args_opt.distribute == "false":
- predict_layout = None
- elif config.use_past:
- batch_valid_length = Tensor(np.array([0]), mstype.int32)
- init_true = Tensor([True], mstype.bool_)
- inputs_np_1 = Tensor(np.ones(shape=(config.batch_size, 1)), mstype.int32)
- model_predict.predict_network.add_flags_recursive(is_first_iteration=True)
- predict_layout = model_predict.infer_predict_layout(inputs_np, current_index, init_true, batch_valid_length)
- model_predict.predict_network.add_flags_recursive(is_first_iteration=False)
- _ = model_predict.infer_predict_layout(inputs_np_1, current_index, init_true, batch_valid_length)
- else:
- predict_layout = model_predict.infer_predict_layout(inputs_np, current_index)
- ##------------------------------------------------------------------------------------------------------
- print("======start load_distributed checkpoint", flush=True)
- ckpt_file_list = get_ckpt_file_list(args_opt.load_ckpt_local_path, device_num=128)
- # For 2.6B and 13B models, the number of ckpt files is 512.
- ## ckpt_name = 'filerted'
- ## ckpt_file_list = [os.path.join(args_opt.load_ckpt_path, f"{ckpt_name}_{ckpt_rank}.ckpt") for ckpt_rank in
- ## range(0, 512)]
- print(ckpt_file_list)
- print(f"Loading from path {ckpt_file_list[0]}", flush=True)
- # Load checkpoint files
- print(ckpt_file_list)
- print(predict_layout)
- load_distributed_checkpoint(eval_net, ckpt_file_list, predict_strategy=predict_layout)
- print("================load param ok=================", flush=True)
- ##-------------------------------------------------------------------------------------------------------
- return model_predict, config
-
- def export_mindir(model_predict, config):
- """Export mindir model"""
- inputs_np = Tensor(np.ones(shape=(config.batch_size, config.seq_length)), mstype.int32)
- current_index = Tensor(np.array([0]), mstype.int32)
-
- batch_valid_length = Tensor(np.array([0]), mstype.int32)
- init_true = Tensor([True], mstype.bool_)
- inputs_np_1 = Tensor(np.ones(shape=(config.batch_size, 1)), mstype.int32)
-
- model_predict.predict_network.add_flags_recursive(is_first_iteration=True)
- export(model_predict.predict_network, inputs_np, current_index,
- init_true, batch_valid_length, file_name='pangu_alpha_1024', file_format='MINDIR')
- model_predict.predict_network.add_flags_recursive(is_first_iteration=False)
- export(model_predict.predict_network, inputs_np_1, current_index,
- init_true, batch_valid_length, file_name='pangu_alpha_1', file_format='MINDIR')
- print("Export finished and now exit.")
-
-
- def run_predict(model_predict, config, args_opt):
- """run predict"""
- from src.tokenization_jieba import JIEBATokenizer
- from src.generate import generate, generate_increment
- # Define tokenizer
- tokenizer = JIEBATokenizer(os.path.join(args_opt.tokenizer_path, 'vocab10.vocab'),
- os.path.join(args_opt.tokenizer_path, 'vocab10.model'))
-
- # Tokenize input sentence to ids
- sample = "今天是一个好天气"
- tokenized_token = tokenizer.tokenize(sample)
- start_sentence = tokenizer.convert_tokens_to_ids(tokenized_token)
- input_ids = np.array(start_sentence).reshape(1, -1)
- # Call inference
- generate_func = generate_increment if config.use_past else generate
- output_ids = generate_func(model_predict, input_ids, args_opt)
- # Decode output ids to sentence
- output_samples = tokenizer.convert_ids_to_tokens(output_ids.tolist())
- print('Output is:', output_samples, flush=True)
-
- def run_predict_langs21(model_predict, config, args_opt):
- """run predict"""
- from tokenizer.tokenizer_spm import SpmTokenizer
- from src.generate import generate, generate_increment, generate_increment2
- from tokenizer.tokenizer_spm import langs_ID, translate_ID
- import jieba
-
- D.init()
- rank = D.get_rank()
-
- work_dir = '/home/work/user-job-dir/pangu_alpha-r1.3'
- # Define tokenizer
- vocab_file = work_dir + '/tokenizer/spm.128k.model.1'
- tokenizer = SpmTokenizer(vocab_file)
- EOT = tokenizer.eot_id
- # inference mode
- generate_func = generate_increment if config.use_past else generate
-
- #------------------------------------------------------------------
- # Tokenize input sentence to ids, example
- sample = "你 今天 中午 吃的 什么 ?"
- tokenized_token = tokenizer.tokenize(sample)
- start_sentence = tokenizer.convert_tokens_to_ids(tokenized_token)
- input_ids = np.array(start_sentence).reshape(1, -1)
- # Call inference
- print('000000000000'*20)
- print(input_ids)
- output_ids = generate_func(model_predict, input_ids, args_opt, dynamic_generate_length=20)
- # Decode output ids to sentence
- output_samples = tokenizer.convert_ids_to_tokens(output_ids.tolist())
- print('\nExample output is:', output_samples, flush=True)
- #------------------------------------------------------------------
-
- result = []
- times_stat = []
- obs_sub_dir = args_opt.load_obs_ckptname.split('_')[0]
- obs_save_dir = f"obs://research-my2/taoht-13b/multi_langs_translate/mPanGu_langs53/{obs_sub_dir}/wmt-10/"
- local_output_save_path = f"/cache/output.txt"
-
- years = 2020
- if args_opt.language == 'hi':
- years = 2014
- elif args_opt.language == 'fr':
- years = 2015
- elif args_opt.language == 'ro':
- years = 2016
- elif args_opt.language == 'lv':
- years = 2017
- else:
- years = 2018
-
- translate_file_path = work_dir + f'/tokenizer/wmt/wmt-10/newstest{years}-en{args_opt.language}-wmt.txt'
- if args_opt.language_idx_wmt == 0:
- src_langs = 'en'
- tag_langs = args_opt.language
- obs_upload_path = f"{obs_save_dir}/output_{src_langs}_2_{tag_langs}--wmt-newstest{years}_87000_remove_deplicate.txt"
- else:
- src_langs = args_opt.language
- tag_langs = 'en'
- obs_upload_path = f"{obs_save_dir}/output_{src_langs}_2_{tag_langs}--wmt-newstest{years}_87000_remove_deplicate.txt"
-
- if not mox.file.exists(obs_save_dir):
- print("Creating translate output bueckt dir {}".format(obs_save_dir))
- mox.file.make_dirs(obs_save_dir)
-
- with open(translate_file_path, 'r', encoding='utf-8') as f:
- if 'newstest' in translate_file_path:
- length_all = 2000
- for idx, data in enumerate(f.readlines()):
- if data:
- if src_langs == 'en':
- data_txt = data.split("\t")[0]
- tokenized_en = tokenizer.tokenize(''+data_txt)
- en_id = tokenizer.convert_tokens_to_ids(tokenized_en)
-
- langs_input = [langs_ID[src_langs], langs_ID[src_langs], langs_ID[src_langs]] +\
- en_id + \
- [translate_ID, translate_ID, translate_ID] + \
- [langs_ID[tag_langs], langs_ID[tag_langs], langs_ID[tag_langs]]
- out_max_len = min(len(en_id)*3+20, 512)
- else:
- data_txt = data.split("\t")[1].replace("\n", "")
- tokenized_langs = tokenizer.tokenize(''+data_txt)
- data_id = tokenizer.convert_tokens_to_ids(tokenized_langs)
-
- langs_input = [langs_ID[src_langs], langs_ID[src_langs], langs_ID[src_langs]] + \
- data_id + \
- [translate_ID, translate_ID, translate_ID] + \
- [langs_ID[tag_langs], langs_ID[tag_langs], langs_ID[tag_langs]]
- out_max_len = min(len(data_id)*3+20, 512)
-
- # Call inference
- time_start = time.time()
- output_ids = generate_func(model_predict, np.array([langs_input]), args_opt, dynamic_generate_length=out_max_len).tolist()
- output_ids = output_ids[len(langs_input):]
- time_1 = time.time()
- if len(output_ids) >0:
- times_stat.append((time_1-time_start)/len(output_ids))
- # Decode output ids to sentence
- langs_output = tokenizer.convert_ids_to_tokens(output_ids)
- result.append(langs_output)
- if rank == 0:
- print(f"------------------------{idx}--------------------------------")
- print(" INPUT is : ", data_txt, "\n")
- print(" OUTPUT is : " + langs_output)
-
- if rank == 0 and idx%20 == 0:
- with open(local_output_save_path, 'w')as f_output:
- for i, i_txt in enumerate(result):
- f_output.writelines(str(i) + '\t' + i_txt +"\n")
- try:
- mox.file.copy(local_output_save_path, obs_upload_path)
- except:
- print("Copy to obs Error...")
- print(tag_langs, "translate time: ", np.average(times_stat), " s/tokens"+'\n\n')
- if rank == 0 and idx == (length_all-1):
- with open(local_output_save_path, 'w')as f_output:
- for i, i_txt in enumerate(result):
- f_output.writelines(str(i) + '\t' + i_txt +"\n")
- mox.file.copy(local_output_save_path, obs_upload_path)
- time.sleep(3)
- if rank == 0:
- with open(local_output_save_path, 'w')as f_output:
- for i, i_txt in enumerate(result):
- f_output.writelines(str(i) + '\t' + i_txt +"\n")
- print("Copy the output file {} to the obs:{}".format(local_output_save_path, obs_upload_path))
- mox.file.copy(local_output_save_path, obs_upload_path)
- time.sleep(2)
-
- def main():
- """Main process for predict or export model"""
- opt = get_args(True)
- set_parse(opt)
- model_predict, config = load_model(opt)
- if opt.export:
- export_mindir(model_predict, config)
- else:
- run_predict_langs21(model_predict, config, opt)
-
-
- if __name__ == "__main__":
- main()
|