File size: 4,188 Bytes
cd2fc0b |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 |
# Copyright (c) Meta Platforms, Inc. and affiliates.
# This software may be used and distributed according to the terms of the GNU General Public License version 3.
from typing import Tuple
import os
import sys
import torch
import fire
import time
import json
from tqdm import tqdm
from pathlib import Path
from fairscale.nn.model_parallel.initialize import initialize_model_parallel
from llama import ModelArgs, Transformer, Tokenizer, LLaMA
def setup_model_parallel() -> Tuple[int, int]:
local_rank = int(os.environ.get("LOCAL_RANK", -1))
world_size = int(os.environ.get("WORLD_SIZE", -1))
torch.distributed.init_process_group("nccl")
initialize_model_parallel(world_size)
torch.cuda.set_device(local_rank)
# seed must be the same in all processes
torch.manual_seed(1)
return local_rank, world_size
def load(ckpt_dir: str, tokenizer_path: str, local_rank: int, world_size: int) -> LLaMA:
start_time = time.time()
checkpoints = sorted(Path(ckpt_dir).glob("*.pth"))
assert (
world_size == len(checkpoints)
), f"Loading a checkpoint for MP={len(checkpoints)} but world size is {world_size}"
ckpt_path = checkpoints[local_rank]
print("Loading")
checkpoint = torch.load(ckpt_path, map_location="cpu")
with open(Path(ckpt_dir) / "params.json", "r") as f:
params = json.loads(f.read())
model_args: ModelArgs = ModelArgs(max_seq_len=1536, max_batch_size=32, **params)
tokenizer = Tokenizer(model_path=tokenizer_path)
model_args.vocab_size = tokenizer.n_words
torch.set_default_tensor_type(torch.cuda.HalfTensor)
model = Transformer(model_args)
torch.set_default_tensor_type(torch.FloatTensor)
model.load_state_dict(checkpoint, strict=False)
generator = LLaMA(model, tokenizer)
print(f"Loaded in {time.time() - start_time:.2f} seconds")
return generator
def read(lang):
score_map = dict()
f = open("/data/wyt/in-context/xglm/bleu_scores.txt", 'r')
for line in f:
idx = int(line.split("\t")[0])
score = float(line.split("\t")[1])
score_map[idx] = score
new_map = sorted(score_map.items(), key=lambda x:x[1], reverse=True)
top_k = [i[0] for i in new_map]
example_src_list, example_tgt_list = [], []
with open("/data/wyt/in-context/xglm/corpus/dev.%s" % (lang[:2]), 'r') as f:
src_list = [line.strip() for line in f]
with open("/data/wyt/in-context/xglm/corpus/dev.%s" % (lang[-2:]), 'r') as f:
tgt_list = [line.strip() for line in f]
f = open("/data/wyt/in-context/xglm/corpus/wmt19/test.%s.%s" % (lang, lang[:2]), 'r')
test_src_list = [line.strip() for line in f if line != "\n"]
demonstration = "Translate these sentences from German to English:\n" if lang == "de-en" else "Translate these sentences from English to German:\n"
for idx in top_k[:16]:
demonstration = demonstration + "%s\n%s\n" % (src_list[idx], tgt_list[idx])
return demonstration, test_src_list
def main(ckpt_dir: str, tokenizer_path: str, lang: str, temperature: float = 0.8, top_p: float = 0.95):
local_rank, world_size = setup_model_parallel()
if local_rank > 0:
sys.stdout = open(os.devnull, 'w')
print("temperature: ", temperature)
generator = load(ckpt_dir, tokenizer_path, local_rank, world_size)
demonstration, test_src_list = read(lang)
step_len = 8
for i in tqdm(range(0, len(test_src_list), step_len)):
prompt_list = [demonstration + test_src + "\n" for test_src in test_src_list[i:i+step_len]]
results = generator.generate(prompt_list, max_gen_len=1536, temperature=temperature, top_p=top_p)
if local_rank == 0:
f = open("%s.%s.temp%02d.raw" % (lang, lang[-2:], temperature * 10), 'a')
for res_idx, result in enumerate(results):
start = result.find(test_src_list[i + res_idx])
if start == -1:
tgt = ""
else:
lines = result[start:].split("\n")
tgt = lines[1] if len(lines) > 1 else ""
f.write(tgt + "\n")
if __name__ == "__main__":
fire.Fire(main)
|