Spaces:
Runtime error
Runtime error
import gradio as gr | |
import gc, copy, re | |
from huggingface_hub import hf_hub_download | |
from rwkv.model import RWKV | |
from rwkv.utils import PIPELINE, PIPELINE_ARGS | |
ctx_limit = 2048 | |
title = "ru_rwkv5_extract_qa_04B_65536_ctx8192_L24_D1024.pth" | |
model_path = hf_hub_download(repo_id="Sigma-AI/ru_rwkv5_extract_qa", filename=f"{title}") | |
model = RWKV(model=model_path, strategy='cpu bf16') | |
pipeline = PIPELINE(model, "rwkv_vocab_v20230424") | |
def generate_prompt(context, question): | |
context = context.strip().replace('\r\n','\n').replace('\n\n','\n').replace('\n\n','\n') | |
question = question.strip().replace('\r\n','\n').replace('\n\n','\n').replace('\n\n','\n') | |
return f"""CONTEXT:{context} | |
QUESTION:{question} | |
ANSWER:""" | |
examples = [ | |
["Вторая мировая война (началась 01.09.1939 и закончилась 02.09.1945) — война двух мировых военно-политических коалиций, ставшая крупнейшим вооружённым конфликтом в истории человечества.В ней участвовали 62 государства из 74 существовавших на тот момент (80 % населения Земного шара).\nБоевые действия велись на территории Европы, Азии и Африки и в водах всех океанов. Это единственный конфликт, в котором было применено ядерное оружие. В результате войны погибло более 70 миллионов человек, из которых большинство — мирные жители.Число участвовавших стран менялось в течение войны. Некоторые из них вели активные военные действия, другие помогали Союзникам поставками продовольствия, а многие участвовали в войне только номинально.", "Было ли применено ядерное оружие?", 300, 1, 0.5, 0.4, 0.4], | |
] | |
def evaluate( | |
instruction, | |
input=None, | |
token_count=200, | |
temperature=1.0, | |
top_p=0.5, | |
presencePenalty = 0.4, | |
countPenalty = 0.4, | |
): | |
args = PIPELINE_ARGS(temperature = max(0.2, float(temperature)), top_p = float(top_p), | |
alpha_frequency = countPenalty, | |
alpha_presence = presencePenalty, | |
token_ban = [], # ban the generation of some tokens | |
token_stop = [0]) # stop generation whenever you see any token here | |
instruction = re.sub(r'\n{2,}', '\n', instruction).strip().replace('\r\n','\n') | |
input = re.sub(r'\n{2,}', '\n', input).strip().replace('\r\n','\n') | |
ctx = generate_prompt(instruction, input) | |
all_tokens = [] | |
out_last = 0 | |
out_str = '' | |
occurrence = {} | |
state = None | |
for i in range(int(token_count)): | |
out, state = model.forward(pipeline.encode(ctx)[-ctx_limit:] if i == 0 else [token], state) | |
for n in occurrence: | |
out[n] -= (args.alpha_presence + occurrence[n] * args.alpha_frequency) | |
token = pipeline.sample_logits(out, temperature=args.temperature, top_p=args.top_p) | |
if token in args.token_stop: | |
break | |
all_tokens += [token] | |
for xxx in occurrence: | |
occurrence[xxx] *= 0.996 | |
if token not in occurrence: | |
occurrence[token] = 1 | |
else: | |
occurrence[token] += 1 | |
tmp = pipeline.decode(all_tokens[out_last:]) | |
if '\ufffd' not in tmp: | |
out_str += tmp | |
yield out_str.strip() | |
out_last = i + 1 | |
if '\n\n' in out_str: | |
break | |
del out | |
del state | |
gc.collect() | |
yield out_str.strip() | |
def user(message, chatbot): | |
chatbot = chatbot or [] | |
return "", chatbot + [[message, None]] | |
def alternative(chatbot, history): | |
if not chatbot or not history: | |
return chatbot, history | |
chatbot[-1][1] = None | |
history[0] = copy.deepcopy(history[1]) | |
return chatbot, history | |
with gr.Blocks(title=title) as demo: | |
gr.HTML(f"<div style=\"text-align: center;\">\n<h1>🌍World - {title}</h1>\n</div>") | |
with gr.Tab("Extract QA"): | |
gr.Markdown(f"100% RNN RWKV-LM **trained on 100+ world languages**. Demo limited to ctxlen {ctx_limit}") | |
with gr.Row(): | |
with gr.Column(): | |
context = gr.Textbox(lines=2, label="Context", value='') | |
question = gr.Textbox(lines=2, label="Question", placeholder="") | |
token_count = gr.Slider(10, 300, label="Max Tokens", step=10, value=300) | |
temperature = gr.Slider(0.2, 2.0, label="Temperature", step=0.1, value=1.2) | |
top_p = gr.Slider(0.0, 1.0, label="Top P", step=0.05, value=0.5) | |
presence_penalty = gr.Slider(0.0, 1.0, label="Presence Penalty", step=0.1, value=0.4) | |
count_penalty = gr.Slider(0.0, 1.0, label="Count Penalty", step=0.1, value=0.4) | |
with gr.Column(): | |
with gr.Row(): | |
submit = gr.Button("Submit", variant="primary") | |
clear = gr.Button("Clear", variant="secondary") | |
output = gr.Textbox(label="Output", lines=5) | |
data = gr.Dataset(components=[context, question, token_count, temperature, top_p, presence_penalty, count_penalty], samples=examples, label="Example", headers=["Context", "Question", "Max Tokens", "Temperature", "Top P", "Presence Penalty", "Count Penalty"]) | |
submit.click(evaluate, [context, question, token_count, temperature, top_p, presence_penalty, count_penalty], [output]) | |
clear.click(lambda: None, [], [output]) | |
data.click(lambda x: x, [data], [context, question, token_count, temperature, top_p, presence_penalty, count_penalty]) | |
demo.queue(max_size=10) | |
demo.launch(share=False) | |