Spaces:
Sleeping
Sleeping
File size: 4,991 Bytes
1c0e302 075341e 1c0e302 4671f72 1c0e302 4671f72 1c0e302 075341e 1c0e302 148548f 5f6f93d 1c0e302 fbf1718 1c0e302 66bd4f7 1c0e302 1578b0f 1c0e302 3817432 1c0e302 4679d7b 3f33715 1c0e302 e9c890c 40ac173 8acee1d 4679d7b f8fd9ee 0df9548 1bbbd28 8212aed 4679d7b 0555380 4679d7b |
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 |
import pathlib
import gradio as gr
import transformers
from transformers import AutoTokenizer
from transformers import AutoModelForCausalLM
from transformers import GenerationConfig
from typing import List, Dict, Union
from typing import Any, TypeVar
Pathable = Union[str, pathlib.Path]
def load_model(name: str) -> Any:
return AutoModelForCausalLM.from_pretrained(name)
def load_tokenizer(name: str) -> Any:
return AutoTokenizer.from_pretrained(name)
def create_generator():
return GenerationConfig(
temperature=1.0,
top_p=0.75,
num_beams=4,
)
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:"""
# model= load_model(name = 's3nh/pythia-410m-70k-steps-self-instruct-polish')
# tokenizer = load_tokenizer(name = 's3nh/pythia-410m-70k-steps-self-instruct-polish')
generation_config = create_generator()
def evaluate(instruction, input, model, tokenizer):
prompt = generate_prompt(instruction, input)
inputs = tokenizer(prompt, return_tensors="pt")
input_ids = inputs["input_ids"]
generation_output = model.generate(
input_ids=input_ids,
generation_config=generation_config,
return_dict_in_generate=True,
output_scores=True,
max_new_tokens=256
)
result = []
for s in generation_output.sequences:
output = tokenizer.decode(s)
result.append( output.split("### Response:")[1].strip())
return ' '.join(el for el in result)
def inference(model_name, text, input):
model = load_model(model_name)
tokenizer = load_tokenizer(model_name)
output = evaluate(instruction = text, input = input, model = model, tokenizer = tokenizer)
return output
def choose_model(name):
return load_model(name), load_tokenizer(name)
with gr.Accordion(label="Parameters", open=False, elem_id="parameters-accordion"):
temperature = gr.Slider(
label="Temperature",
value=0.7,
minimum=0.0,
maximum=1.0,
step=0.1,
interactive=True,
info="Higher values produce more diverse outputs",
)
top_p = gr.Slider(
label="Top-p (nucleus sampling)",
value=0.9,
minimum=0.0,
maximum=1,
step=0.05,
interactive=True,
info="Higher values sample more low-probability tokens",
)
max_new_tokens = gr.Slider(
label="Max new tokens",
value=1024,
minimum=0,
maximum=2048,
step=4,
interactive=True,
info="The maximum numbers of new tokens",
)
repetition_penalty = gr.Slider(
label="Repetition Penalty",
value=1.2,
minimum=0.0,
maximum=10,
step=0.1,
interactive=True,
info="The parameter for repetition penalty. 1.0 means no penalty.",
)
io = gr.Interface(
inference,
inputs = [
gr.Dropdown(["s3nh/pythia-1.4b-deduped-16k-steps-self-instruct-polish", "s3nh/pythia-410m-91k-steps-self-instruct-polish", "s3nh/tiny-gpt2-instruct-polish",
"s3nh/pythia-410m-103k-steps-self-instruct-polish", "https://huggingface.co/s3nh/DialoGPT-large-instruct-polish-3000-steps",
"https://huggingface.co/s3nh/pythia-410m-70k-steps-self-instruct-polish", "https://huggingface.co/s3nh/tiny-gpt2-instruct-polish",
"s3nh/Cerebras-GPT-590M-3000steps-polish", "s3nh/gpt-j-6b-3500steps-polish", "s3nh/DialoGPT-medium-4000steps-polish",
"s3nh/DialoGPT-small-5000steps-polish",
"Lajonbot/pythia-160m-53500-self-instruct-polish",
"Lajonbot/gpt-neo-125m-self-instruct-polish-66k-steps",
"Lajonbot/pythia-160m-33k-steps-self-instruct-polish",
"Lajonbot/pythia-410m-21k-steps-self-instruct-polish",
"Lajonbot/llama-30b-hf-pl-lora",
#"Amazon-LightGPT-pl-qlora",
#"wizard-mega-13b-pl-lora",
#"stablelm-base-alpha-3b-Lora-polish",
#"dolly-v2-3b-Lora-polish",
#"LaMini-GPT-1.5B-Lora-polish"],
]),
gr.Textbox(
lines = 3,
max_lines = 10,
placeholder = "Add question here",
interactive = True,
show_label = False
),
gr.Textbox(
lines = 3,
max_lines = 10,
placeholder = "Add context here",
interactive = True,
show_label = False
)],
outputs = [gr.Textbox(lines = 1, label = 'Pythia410m', interactive = False)],
cache_examples = False,
)
io.launch(debug = True) |