sloth/sftune.py at master · toranb/sloth · 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
import os
import json
import torch
import torch.nn as nn
import pandas as pd
from unsloth import FastLanguageModel
from transformers import TrainingArguments, Trainer
from trl import SFTTrainer
from datasets import load_dataset, Dataset
from unsloth.chat_templates import get_chat_template

max_seq_length = 2048

args = TrainingArguments(
        output_dir = "./workspace",
        per_device_train_batch_size = 16,
        per_device_eval_batch_size = 16,
        gradient_accumulation_steps = 1,
        max_grad_norm = 0.4,
        warmup_ratio = 0.05,
        num_train_epochs = 3,
        learning_rate = 2e-4,
        fp16 = not torch.cuda.is_bf16_supported(),
        bf16 = torch.cuda.is_bf16_supported(),
        logging_steps = 1,
        save_steps = 200,
        save_total_limit = 10,
        eval_delay = 1000,
        eval_accumulation_steps = 1,
        evaluation_strategy = "steps",
        eval_steps = 200,
        optim = "paged_adamw_8bit",
        weight_decay = 0.001,
        lr_scheduler_type = "cosine",
    )

model, tokenizer = FastLanguageModel.from_pretrained(
    model_name = "mistralai/Mistral-7B-Instruct-v0.3",
    max_seq_length = max_seq_length,
    dtype = torch.bfloat16,
    load_in_4bit = True,
    load_in_8bit = False,
    attn_implementation = "flash_attention_2",
)

model.config.use_cache = False

unsloth_eos_token = "eos_token"
unsloth_template = \
    "{{ bos_token + messages['instruction'] + '\n' }}"\
    "{{ messages['output'] + eos_token }}"

tokenizer = get_chat_template(
    tokenizer,
    chat_template = (unsloth_template, unsloth_eos_token),
    mapping = {"role" : "from", "content" : "value", "user" : "human", "assistant" : "gpt"},
    map_eos_token = True, # Maps <|im_end|> to </s> instead
)

def formatting_prompts_func(item):
    text = tokenizer.apply_chat_template(item, tokenize = False, add_generation_prompt = False)
    return { "text" : text }

full_dataset = load_dataset("json", data_files="./data.json")
dataset = full_dataset['train'].train_test_split(test_size=0.07)
train_dataset = dataset['train'].map(formatting_prompts_func)
eval_dataset = dataset['test'].map(formatting_prompts_func)

train_dataset = train_dataset.remove_columns(
    [col for col in dataset.column_names['train'] if col not in ['text']]
)

eval_dataset = eval_dataset.remove_columns(
    [col for col in dataset.column_names['train'] if col not in ['text']]
)

ft_model = FastLanguageModel.get_peft_model(
    model,
    r = 32,
    target_modules = ["q_proj", "k_proj", "v_proj", "o_proj", "gate_proj", "up_proj", "down_proj"],
    lora_alpha = 64,
    lora_dropout = 0,
    bias = "none",
    use_gradient_checkpointing = True,
    random_state = 3407,
)

ft_model.config.use_cache = False

trainer = SFTTrainer(
    model = ft_model,
    args = args,
    packing = False,
    dataset_num_proc = 2,
    train_dataset = train_dataset,
    eval_dataset = eval_dataset,
    dataset_text_field = "text",
    max_seq_length = max_seq_length,
)

trainer.train()

model.save_pretrained_merged("model", tokenizer, save_method = "merged_16bit")