def test_from_pipeline_conversation(self):
model_id = "facebook/blenderbot_small-90M"
# from model id
conversation_agent_from_model_id = pipeline("conversational",
model=model_id,
tokenizer=model_id)
# from model object
model = BlenderbotSmallForConditionalGeneration.from_pretrained(
model_id)
tokenizer = BlenderbotSmallTokenizer.from_pretrained(model_id)
conversation_agent_from_model = pipeline("conversational",
model=model,
tokenizer=tokenizer)
conversation = Conversation("My name is Sarah and I live in London")
conversation_copy = Conversation(
"My name is Sarah and I live in London")
result_model_id = conversation_agent_from_model_id([conversation])
result_model = conversation_agent_from_model([conversation_copy])
# check for equality
self.assertEqual(
result_model_id.generated_responses[0],
"hi sarah, i live in london as well. do you have any plans for the weekend?",
)
self.assertEqual(
result_model_id.generated_responses[0],
result_model.generated_responses[0],
)
def Loader():
f = open("step.txt", "w")
f.write("0")
f.close()
mname = 'facebook/blenderbot-90M'
model = BlenderbotSmallForConditionalGeneration.from_pretrained(mname)
tokenizer = BlenderbotSmallTokenizer.from_pretrained(mname)
return tokenizer, model
def __init__(self, size, env, device, max_context_length):
"""
The Blender chatbot model was proposed in Recipes for building an open-domain chatbot Stephen Roller,
Emily Dinan, Naman Goyal, Da Ju, Mary Williamson, Yinhan Liu, Jing Xu, Myle Ott, Kurt Shuster, Eric M. Smith,
Y-Lan Boureau, Jason Weston on 30 Apr 2020.
Args:
size (str): model size
env (BaseEnv): dialogue environment
device (str): device (one of ['CPU', 'CUDA', 'CUDA:N']
max_context_length (int): max history context length
(it means that length of input context tokens)
"""
assert size in ['small', 'medium', 'large', 'xlarge'], \
"model size must be one of ['small', 'medium', 'large', 'xlarge']"
if size == "small":
super().__init__("facebook/blenderbot_small-90M", env)
self.model = BlenderbotSmallForConditionalGeneration.from_pretrained(
self.name).to(device)
self.tokenizer = BlenderbotSmallTokenizer.from_pretrained(
self.name)
else:
if size == "medium":
super().__init__("facebook/blenderbot-400M-distill", env)
elif size == "large":
super().__init__("facebook/blenderbot-1B-distill", env)
elif size == "xlarge":
super().__init__("facebook/blenderbot-3B", env)
self.model = BlenderbotForConditionalGeneration.from_pretrained(
self.name).to(device)
self.tokenizer = BlenderbotTokenizer.from_pretrained(self.name)
self.size = size
self.device = device.lower()
self.max_context_length = max_context_length
self.eos = "</s> <s>"
print('Done!')
def tokenizer(self):
return BlenderbotSmallTokenizer.from_pretrained(self.ckpt)
def tokenizer(self):
# use "old" tokenizer here because of bug when downloading new tokenizer
return BlenderbotSmallTokenizer.from_pretrained("facebook/blenderbot-90M")
def tokenizer(self):
return BlenderbotSmallTokenizer.from_pretrained(self.model_name)
from transformers import TFBlenderbotSmallModel, BlenderbotSmallTokenizer, BlenderbotSmallConfig
import tensorflow as tf
from tensorflow import keras
from config_ import cfg
import numpy as np
mname = 'facebook/blenderbot_small-90M'
blen_model = TFBlenderbotSmallModel.from_pretrained(mname)
tokenizer = BlenderbotSmallTokenizer.from_pretrained(mname)
config = BlenderbotSmallConfig.from_pretrained(mname)
que = []
ans = []
start_token = 1
end_token = 2
with open(cfg.data_file) as f:
data = f.read().split("\n")
state = None
q = None
a = None
for i in range(len(data)):
if data[i] != " " and data[i] != "":
if data[i + 1] != "" and data[i + 1] != " ":
if q is None:
q = data[i]
a = data[i + 1]
que.append(q)
ans.append(a)
tokenizer = GPT2Tokenizer.from_pretrained("microsoft/DialoGPT-small")
#------dialogpt medium------#
model = GPT2LMHeadModel.from_pretrained("microsoft/DialoGPT-medium")
tokenizer = GPT2Tokenizer.from_pretrained("microsoft/DialoGPT-medium")
print("dialogpt is done!")
elif download_type == 'gptneo':
#------gptneo small------#
model = AutoModelForCausalLM.from_pretrained("EleutherAI/gpt-neo-125M")
tokenizer = AutoTokenizer.from_pretrained("EleutherAI/gpt-neo-125M")
#------gptneo large------#
#model = AutoModelForCausalLM.from_pretrained("EleutherAI/gpt-neo-1.3B")
#tokenizer = AutoTokenizer.from_pretrained("EleutherAI/gpt-neo-1.3B")
print("gptneo is done!")
elif download_type == 'blender':
#------blender small------#
model = BlenderbotSmallForConditionalGeneration.from_pretrained(
"facebook/blenderbot_small-90M")
tokenizer = BlenderbotSmallTokenizer.from_pretrained(
"facebook/blenderbot_small-90M")
#------blender medium------#
model = BlenderbotForConditionalGeneration.from_pretrained(
"facebook/blenderbot-400M-distill")
tokenizer = BlenderbotTokenizer.from_pretrained(
"facebook/blenderbot-400M-distill")
print("blender is done!")
def get_model():
mname = 'facebook/blenderbot-90M'
model = BlenderbotForConditionalGeneration.from_pretrained(mname)
tokenizer = BlenderbotSmallTokenizer.from_pretrained(mname)
return model, tokenizer
