File size: 4,583 Bytes
262d1c8
97d3016
e348d3e
97d3016
262d1c8
97d3016
954a751
 
97d3016
 
 
 
96c33c2
d0afb9d
97d3016
d0afb9d
97d3016
 
724f6b3
97d3016
 
 
 
262d1c8
97d3016
 
 
 
 
262d1c8
97d3016
 
 
 
 
954a751
 
 
 
97d3016
262d1c8
954a751
97d3016
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
88ef11a
23b0e90
97d3016
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
954a751
97d3016
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
954a751
97d3016
 
 
 
 
 
 
 
 
 
 
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
import spaces
import torch
from peft import PeftModel
import transformers
import gradio as gr
import os


assert (
    "LlamaTokenizer" in transformers._import_structure["models.llama"]
), "LLaMA is now in HuggingFace's main branch.\nPlease reinstall it: pip uninstall transformers && pip install git+https://github.com/huggingface/transformers.git"
from transformers import LlamaTokenizer, LlamaForCausalLM, GenerationConfig
access_token = os.environ.get('HF_TOKEN')

tokenizer = LlamaTokenizer.from_pretrained("meta-llama/Llama-2-7b-hf", token=access_token)

BASE_MODEL = "meta-llama/Llama-2-7b-hf"
LORA_WEIGHTS = "DSMI/LLaMA-E"

if torch.cuda.is_available():
    device = "cuda"
else:
    device = "cpu"

try:
    if torch.backends.mps.is_available():
        device = "mps"
except:
    pass

print("Device: " + str(device))

if device == "cuda":
    model = LlamaForCausalLM.from_pretrained(
        BASE_MODEL,
        load_in_8bit=False,
        torch_dtype=torch.float16,
        device_map="auto",
    )
    model = PeftModel.from_pretrained(
        model, LORA_WEIGHTS, torch_dtype=torch.float16, force_download=True
    )
elif device == "mps":
    model = LlamaForCausalLM.from_pretrained(
        BASE_MODEL,
        device_map={"": device},
        torch_dtype=torch.float16,
    )
    model = PeftModel.from_pretrained(
        model,
        LORA_WEIGHTS,
        device_map={"": device},
        torch_dtype=torch.float16,
    )
else:
    model = LlamaForCausalLM.from_pretrained(
        BASE_MODEL, 
        device_map={"": device}, 
        low_cpu_mem_usage=True,
        load_in_8bit=True,
    )
    model = PeftModel.from_pretrained(
        model,
        LORA_WEIGHTS,
        device_map={"": device},
    )

print("Model: " + str(model))

def generate_prompt(instruction, input=None):
    if input:
        return f"""Below is an instruction that describes a task, paired with an input that provides further context. Write a response that appropriately completes the request.
                ### Instruction:
                {instruction}
                ### Input:
                {input}
                ### Response:"""
    else:
        return f"""Below is an instruction that describes a task. Write a response that appropriately completes the request.
                ### Instruction:
                {instruction}
                ### Response:"""

if device != "cpu":
    model.half()
model.eval()
if torch.__version__ >= "2":
    model = torch.compile(model)

@spaces.GPU()
def evaluate(
    instruction,
    input=None,
    temperature=0.1,
    top_p=0.75,
    top_k=40,
    num_beams=2,
    max_new_tokens=64,
    **kwargs,
):
    prompt = generate_prompt(instruction, input)
    inputs = tokenizer(prompt, return_tensors="pt")
    input_ids = inputs["input_ids"].to(device)
    generation_config = GenerationConfig(
        temperature=temperature,
        top_p=top_p,
        top_k=top_k,
        num_beams=num_beams,
        **kwargs,
    )
    with torch.no_grad():
        generation_output = model.generate(
            input_ids=input_ids,
            generation_config=generation_config,
            return_dict_in_generate=True,
            output_scores=True,
            max_new_tokens=max_new_tokens,
        )
    s = generation_output.sequences[0]
    output = tokenizer.decode(s)
    return output.split("### Response:")[1].strip()


g = gr.Interface(
    fn=evaluate,
    inputs=[
        gr.components.Textbox(
            lines=2, label="Instruction", placeholder="Tell me about alpacas."
        ),
        gr.components.Textbox(lines=2, label="Input", placeholder="none"),
        gr.components.Slider(minimum=0, maximum=1, value=0.1, label="Temperature"),
        gr.components.Slider(minimum=0, maximum=1, value=0.75, label="Top p"),
        gr.components.Slider(minimum=0, maximum=100, step=1, value=40, label="Top k"),
        gr.components.Slider(minimum=1, maximum=4, step=1, value=4, label="Beams"),
        gr.components.Slider(
            minimum=1, maximum=512, step=1, value=128, label="Max tokens"
        ),
    ],
    outputs=[
        gr.Textbox(
            lines=5,
            label="Output",
        )
    ],
    title="πŸ¦™πŸ›οΈ LLaMA-E",
    description="LLaMA-E is a series of fine-tuned LLaMA model following the E-commerce instructions. It is developed by DSMI (http://dsmi.tech/) @ University of Technology Sydney, and trained on the 120k instruction set. This model is for academic research use only. For more details please contact: Kaize.Shi@uts.edu.au",
)
g.queue(concurrency_count=1)
g.launch()