File size: 6,341 Bytes
6a07688
 
 
 
65be4e0
6a07688
 
 
65be4e0
6a07688
65be4e0
 
 
6a07688
 
 
 
 
 
 
 
 
65be4e0
6a07688
291e9e3
6a07688
 
65be4e0
6a07688
 
 
 
13565ec
 
 
 
 
 
6a07688
 
 
 
 
 
 
 
 
 
 
 
 
 
 
65be4e0
6a07688
65be4e0
 
 
6a07688
 
 
65be4e0
 
 
6a07688
65be4e0
 
 
 
 
 
 
 
 
 
 
 
6a07688
65be4e0
 
 
 
 
 
 
 
 
6a07688
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
65be4e0
 
 
 
 
 
 
 
 
 
 
6a07688
65be4e0
 
 
6a07688
65be4e0
 
 
 
6a07688
65be4e0
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
6a07688
 
 
65be4e0
6a07688
65be4e0
 
6a07688
65be4e0
6a07688
65be4e0
6a07688
 
 
 
 
 
 
 
 
65be4e0
 
 
 
 
 
6a07688
 
 
 
65be4e0
 
 
 
 
 
 
 
 
 
 
6a07688
65be4e0
 
 
 
 
6a07688
65be4e0
 
 
6a07688
 
 
 
 
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
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
import sys
import logging

import datasets
from datasets import load_dataset
from peft import LoraConfig
import torch
import transformers
from trl import SFTTrainer
from transformers import AutoModelForCausalLM, AutoTokenizer, TrainingArguments, BitsAndBytesConfig

"""
A simple example on using SFTTrainer and Accelerate to finetune Phi-3 models. For
a more advanced example, please follow HF alignment-handbook/scripts/run_sft.py.
This example has utilized DeepSpeed ZeRO3 offload to reduce the memory usage. The
script can be run on V100 or later generation GPUs. Here are some suggestions on 
futher reducing memory consumption:
    - reduce batch size
    - decrease lora dimension
    - restrict lora target modules
Please follow these steps to run the script:
1. Install dependencies: 
    conda install -c conda-forge accelerate
    pip3 install -i https://pypi.org/simple/ bitsandbytes
    pip3 install peft transformers trl datasets
    pip3 install deepspeed
2. Setup accelerate and deepspeed config based on the machine used:
    accelerate config
Here is a sample config for deepspeed zero3:
    compute_environment: LOCAL_MACHINE
    debug: false
    deepspeed_config:
      gradient_accumulation_steps: 1
      offload_optimizer_device: none
      offload_param_device: none
      zero3_init_flag: true
      zero3_save_16bit_model: true
      zero_stage: 3
    distributed_type: DEEPSPEED
    downcast_bf16: 'no'
    enable_cpu_affinity: false
    machine_rank: 0
    main_training_function: main
    mixed_precision: bf16
    num_machines: 1
    num_processes: 4
    rdzv_backend: static
    same_network: true
    tpu_env: []
    tpu_use_cluster: false
    tpu_use_sudo: false
    use_cpu: false
3. check accelerate config:
    accelerate env
4. Run the code:
    accelerate launch sample_finetune.py
"""

logger = logging.getLogger(__name__)


###################
# Hyper-parameters
###################
training_config = {
    "bf16": True,
    "do_eval": False,
    "learning_rate": 5.0e-06,
    "log_level": "info",
    "logging_steps": 20,
    "logging_strategy": "steps",
    "lr_scheduler_type": "cosine",
    "num_train_epochs": 1,
    "max_steps": -1,
    "output_dir": "./checkpoint_dir",
    "overwrite_output_dir": True,
    "per_device_eval_batch_size": 4,
    "per_device_train_batch_size": 4,
    "remove_unused_columns": True,
    "save_steps": 100,
    "save_total_limit": 1,
    "seed": 0,
    "gradient_checkpointing": True,
    "gradient_checkpointing_kwargs":{"use_reentrant": False},
    "gradient_accumulation_steps": 1,
    "warmup_ratio": 0.2,
    }

peft_config = {
    "r": 16,
    "lora_alpha": 32,
    "lora_dropout": 0.05,
    "bias": "none",
    "task_type": "CAUSAL_LM",
    "target_modules": "all-linear",
    "modules_to_save": None,
}
train_conf = TrainingArguments(**training_config)
peft_conf = LoraConfig(**peft_config)


###############
# Setup logging
###############
logging.basicConfig(
    format="%(asctime)s - %(levelname)s - %(name)s - %(message)s",
    datefmt="%Y-%m-%d %H:%M:%S",
    handlers=[logging.StreamHandler(sys.stdout)],
)
log_level = train_conf.get_process_log_level()
logger.setLevel(log_level)
datasets.utils.logging.set_verbosity(log_level)
transformers.utils.logging.set_verbosity(log_level)
transformers.utils.logging.enable_default_handler()
transformers.utils.logging.enable_explicit_format()

# Log on each process a small summary
logger.warning(
    f"Process rank: {train_conf.local_rank}, device: {train_conf.device}, n_gpu: {train_conf.n_gpu}"
    + f" distributed training: {bool(train_conf.local_rank != -1)}, 16-bits training: {train_conf.fp16}"
)
logger.info(f"Training/evaluation parameters {train_conf}")
logger.info(f"PEFT parameters {peft_conf}")


################
# Modle Loading
################
checkpoint_path = "microsoft/Phi-3-mini-4k-instruct"
# checkpoint_path = "microsoft/Phi-3-mini-128k-instruct"
model_kwargs = dict(
    use_cache=False,
    trust_remote_code=True,
    attn_implementation="flash_attention_2",  # loading the model with flash-attenstion support
    torch_dtype=torch.bfloat16,
    device_map=None
)
model = AutoModelForCausalLM.from_pretrained(checkpoint_path, **model_kwargs)
tokenizer = AutoTokenizer.from_pretrained(checkpoint_path)
tokenizer.model_max_length = 2048
tokenizer.pad_token = tokenizer.unk_token  # use unk rather than eos token to prevent endless generation
tokenizer.pad_token_id = tokenizer.convert_tokens_to_ids(tokenizer.pad_token)
tokenizer.padding_side = 'right'


##################
# Data Processing
##################
def apply_chat_template(
    example,
    tokenizer,
):
    messages = example["messages"]
    # Add an empty system message if there is none
    if messages[0]["role"] != "system":
        messages.insert(0, {"role": "system", "content": ""})
    example["text"] = tokenizer.apply_chat_template(
        messages, tokenize=False, add_generation_prompt=False)
    return example

raw_dataset = load_dataset("HuggingFaceH4/ultrachat_200k")
train_dataset = raw_dataset["train_sft"]
test_dataset = raw_dataset["test_sft"]
column_names = list(train_dataset.features)

processed_train_dataset = train_dataset.map(
    apply_chat_template,
    fn_kwargs={"tokenizer": tokenizer},
    num_proc=10,
    remove_columns=column_names,
    desc="Applying chat template to train_sft",
)

processed_test_dataset = test_dataset.map(
    apply_chat_template,
    fn_kwargs={"tokenizer": tokenizer},
    num_proc=10,
    remove_columns=column_names,
    desc="Applying chat template to test_sft",
)


###########
# Training
###########
trainer = SFTTrainer(
    model=model,
    args=train_conf,
    peft_config=peft_conf,
    train_dataset=processed_train_dataset,
    eval_dataset=processed_test_dataset,
    max_seq_length=2048,
    dataset_text_field="text",
    tokenizer=tokenizer,
    packing=True
)
train_result = trainer.train()
metrics = train_result.metrics
trainer.log_metrics("train", metrics)
trainer.save_metrics("train", metrics)
trainer.save_state()


#############
# Evaluation
#############
tokenizer.padding_side = 'left'
metrics = trainer.evaluate()
metrics["eval_samples"] = len(processed_test_dataset)
trainer.log_metrics("eval", metrics)
trainer.save_metrics("eval", metrics)


# ############
# # Save model
# ############
trainer.save_model(train_conf.output_dir)