LLM-SpecDec/benchmark.py at main · joshualu777/LLM-SpecDec · GitHub

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
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172

import torch
import argparse
import contexttimer
from colorama import Fore, Style
from transformers import AutoTokenizer, AutoModelForCausalLM

from sampling import autoregressive_sampling, speculative_sampling, speculative_sampling_v2, speculative_sampling_merging
from globals import Decoder
import json
from  tqdm import tqdm
from alignment import TokenMapper
from datasets import load_dataset
from collections import defaultdict


import os
os.environ["TOKENIZERS_PARALLELISM"] = "true"

# my local models
MODELZOO = {
    # llama-1
    # https://huggingface.co/PY007/TinyLlama-1.1B-step-50K-105b
    "llama1b": "/share_nfs/fangjiarui/root/code/hf_models/TinyLlama-1.1B-step-50K-105b",
    "llama7b": "/share_nfs/tianzhi/code/llama-7b",
    "llama30b": "/share_nfs/fangjiarui/root/code/hf_models/llama-30b-hf",
    "llama2-7b" : "/share_nfs/fangjiarui/root/code/hf_models/llama-2-7b-hf",
    "llama2-70b" : "/share_nfs/fangjiarui/root/code/hf_models/llama-2-70b-hf",
    "bloom-560m": "/share_nfs/fangjiarui/root/code/hf_models/bloom-560m",
    "bloom7b": "/share_nfs/fangjiarui/root/code/hf_models/bloomz-7b1",
    "baichuan-7b": "/share_nfs/duanqiyuan/models/source_models/hf/baichuan-7B",
    "baichuan-13b": "/share_nfs/duanqiyuan/models/source_models/hf/Baichuan-13B-Base",
}

def parse_arguments():
    parser = argparse.ArgumentParser(description='args for main.py')

    parser.add_argument('--input', type=str, default="Suggest at least five related search terms to \"Mạng neural nhân tạo\".")
    parser.add_argument('--approx_model_name', type=str, default=MODELZOO["llama2-7b"])
    parser.add_argument('--target_model_name', type=str, default=MODELZOO["llama2-70b"])
    parser.add_argument('--verbose', '-v', action='store_true', default=False, help='enable verbose mode')
    parser.add_argument('--seed', '-s', type=int, default=None, help='set a random seed, which can makes the result reproducible')
    parser.add_argument('--benchmark', '-b', action='store_true', default=False, help='show benchmark results.')
    parser.add_argument('--profiling', '-p', action='store_true', default=False, help='collect torch profiler results.')
    parser.add_argument('--max_tokens', '-M', type=int, default=20, help='max token number generated.')
    parser.add_argument('--gamma', '-g', type=int, default=4, help='guess time.')
    args = parser.parse_args()
    return args


def benchmark(fn, info, *args, **kwargs):

    dataset = load_dataset("OpenAssistant/oasst1", split="train")

    threads = defaultdict(list)
    for message in dataset:
        if "message_tree_id" in message:
            threads[message["message_tree_id"]].append(message)

    test_sample_num = 5
    with contexttimer.Timer() as t:
        total_tokens = 0
        with tqdm(total=test_sample_num, desc=f"{info} benchmarking") as pbar:
            for tree_id, messages in list(threads.items())[:test_sample_num]:
                # Sort messages by hierarchy (if necessary)
                messages = sorted(messages, key=lambda x: x["parent_id"] or "")

                # Reconstruct the conversation
                content = "\n".join([msg["role"] + ": " + msg["text"] for msg in messages])
                # print("content", content)
                input_ids = Decoder().encode(content, return_tensors='pt').to('mps')
                if len(input_ids[0]) > 2048 :
                    continue
                output_ids = fn(input_ids, *args, **kwargs)
                generated_text = Decoder().decode(output_ids)
                # print("generated_text", generated_text)
                # print("end_generated_text")
                total_tokens += (len(output_ids[0]) - len(input_ids[0]))
                test_sample_num -= 1
                if test_sample_num < 0:
                    break

                pbar.update(1)

    print(f"\n [benchmark] {info} tokens/sec: {total_tokens / t.elapsed}, {t.elapsed} sec generates {total_tokens} tokens")

def benchmark_merge(fn, info, mapper, *args, **kwargs):

    dataset = load_dataset("OpenAssistant/oasst1", split="train")
    print(dataset[0].keys())

    # Group messages by thread_id
    threads = defaultdict(list)
    for message in dataset:
        if "message_tree_id" in message:
            threads[message["message_tree_id"]].append(message)

    test_sample_num = 5
    with contexttimer.Timer() as t:
        total_tokens = 0
        with tqdm(total=test_sample_num, desc=f"{info} benchmarking") as pbar:
            for tree_id, messages in list(threads.items())[:test_sample_num]:
                # Sort messages by hierarchy (if necessary)
                messages = sorted(messages, key=lambda x: x["parent_id"] or "")

                # Reconstruct the conversation
                content = "\n".join([msg["role"] + ": " + msg["text"] for msg in messages])
                # print("content: ", content)
                input_ids_draft = Decoder().encode_draft(content, return_tensors='pt').to('mps')
                input_ids_target = Decoder().encode_target(content, return_tensors='pt').to('mps')
                if len(input_ids_draft[0]) > 2048 or len(input_ids_target[0]) > 2048 :
                    print("too long")
                    continue
                output_ids = fn(mapper, input_ids_draft, input_ids_target, *args, **kwargs)
                # print(len(output_ids))
                generated_text = Decoder().decode_target(output_ids)
                # print("generated_text", generated_text)
                # print("end_generated_text")
                total_tokens += (len(output_ids[0]) - len(input_ids_target[0]))
                test_sample_num -= 1
                if test_sample_num < 0:
                    break

                pbar.update(1)

    print(f"\n [benchmark] {info} tokens/sec: {total_tokens / t.elapsed}, {t.elapsed} sec generates {total_tokens} tokens")

def generate(input_text, approx_model_name, target_model_name, num_tokens=100, gamma = 4,
             random_seed = None):
    # NOTE() approx_model_name and target_model_name should use the same tokenizer!

    torch_device = 'mps' if torch.backends.mps.is_available() else 'cpu'

    small_tokenizer = AutoTokenizer.from_pretrained(approx_model_name, trust_remote_code=True, use_auth_token=True)
    large_tokenizer = AutoTokenizer.from_pretrained(target_model_name, trust_remote_code=True, use_auth_token=True)

    Decoder().set_tokenizer(small_tokenizer, large_tokenizer)

    # print(f"begin loading models: \n {approx_model_name} \n {target_model_name}")
    small_model = AutoModelForCausalLM.from_pretrained(approx_model_name,
                                                       torch_dtype=torch.float16,
                                                       device_map="auto",
                                                       trust_remote_code=True)
    large_model = AutoModelForCausalLM.from_pretrained(target_model_name,
                                                       torch_dtype=torch.float16,
                                                       device_map="auto",
                                                       trust_remote_code=True)
    # print("finish loading models")

    input_ids = small_tokenizer.encode(input_text, return_tensors='pt').to(torch_device)

    top_k = 5000
    top_p = 0

    token_map = TokenMapper(approx_model_name, target_model_name)

    torch.manual_seed(123)
    benchmark(autoregressive_sampling, "AS_large", large_model, num_tokens, top_k = top_k, top_p=top_p)

    torch.manual_seed(123)
    benchmark(autoregressive_sampling, "AS_small", small_model, num_tokens, top_k = top_k, top_p=top_p)

    #torch.manual_seed(123)
    #benchmark(speculative_sampling, "SP", small_model, large_model, max_len = num_tokens, gamma = gamma, top_k = top_k, top_p=top_p, random_seed = random_seed)

    torch.manual_seed(123)
    benchmark_merge(speculative_sampling_merging, "SP", token_map, small_model, large_model, max_len = num_tokens, gamma = gamma, top_k = top_k, top_p=top_p, random_seed = random_seed)

if __name__ == "__main__":
    args = parse_arguments()

    generate(args.input, args.approx_model_name, args.target_model_name, num_tokens=args.max_tokens, gamma=args.gamma)