Spaces:
Build error
Build error
swpa out chat logic
Browse files
app.py
CHANGED
@@ -7,17 +7,25 @@ import torch
|
|
7 |
import gradio as gr
|
8 |
|
9 |
|
10 |
-
def chat(message, history):
|
11 |
-
|
12 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
13 |
bot_input_ids = torch.cat([torch.LongTensor(history), new_user_input_ids], dim=-1)
|
14 |
history = model.generate(bot_input_ids, max_length=500, pad_token_id=tokenizer.eos_token_id).tolist()
|
15 |
-
response = tokenizer.decode(history[0]).replace("<|endoftext|>", "
|
16 |
-
|
17 |
-
# response = tokenizer.decode(chat_history_ids[:, bot_input_ids.shape[-1]:][0], skip_special_tokens=True)
|
18 |
-
print("The response is ", [response])
|
19 |
-
# history.append((message, response, new_user_input_ids, chat_history_ids))
|
20 |
-
return response, feedback(message)
|
21 |
|
22 |
|
23 |
def feedback(text):
|
|
|
7 |
import gradio as gr
|
8 |
|
9 |
|
10 |
+
# def chat(message, history):
|
11 |
+
# history = history if history is not None else []
|
12 |
+
# new_user_input_ids = tokenizer.encode(message+tokenizer.eos_token, return_tensors='pt')
|
13 |
+
# bot_input_ids = torch.cat([torch.LongTensor(history), new_user_input_ids], dim=-1)
|
14 |
+
# history = model.generate(bot_input_ids, max_length=500, pad_token_id=tokenizer.eos_token_id).tolist()
|
15 |
+
# # response = tokenizer.decode(history[0]).replace("<|endoftext|>", "\n")
|
16 |
+
# # pretty print last ouput tokens from bot
|
17 |
+
# response = tokenizer.decode(bot_input_ids.shape[-1][0], skip_special_tokens=True)
|
18 |
+
# print("The response is ", [response])
|
19 |
+
# # history.append((message, response, new_user_input_ids, chat_history_ids))
|
20 |
+
# return response, history, feedback(message)
|
21 |
+
|
22 |
+
|
23 |
+
def chat(message, history=[]):
|
24 |
+
new_user_input_ids = tokenizer.encode(input + tokenizer.eos_token, return_tensors='pt')
|
25 |
bot_input_ids = torch.cat([torch.LongTensor(history), new_user_input_ids], dim=-1)
|
26 |
history = model.generate(bot_input_ids, max_length=500, pad_token_id=tokenizer.eos_token_id).tolist()
|
27 |
+
response = tokenizer.decode(history[0]).replace("<|endoftext|>", "")
|
28 |
+
return response, history
|
|
|
|
|
|
|
|
|
29 |
|
30 |
|
31 |
def feedback(text):
|