Spaces:
Runtime error
Runtime error
Update app.py
Browse files
app.py
CHANGED
|
@@ -30,23 +30,32 @@ def predict(text,
|
|
| 30 |
temperature,
|
| 31 |
max_length_tokens,
|
| 32 |
max_context_length_tokens,):
|
|
|
|
|
|
|
|
|
|
| 33 |
if text=="":
|
| 34 |
-
yield
|
| 35 |
return
|
|
|
|
|
|
|
| 36 |
try:
|
| 37 |
model
|
| 38 |
except:
|
| 39 |
-
yield [
|
| 40 |
return
|
| 41 |
|
|
|
|
| 42 |
inputs = generate_prompt_with_history(text,history,tokenizer,max_length=max_context_length_tokens)
|
| 43 |
if inputs is None:
|
| 44 |
-
yield
|
| 45 |
return
|
| 46 |
else:
|
| 47 |
prompt,inputs=inputs
|
| 48 |
begin_length = len(prompt)
|
| 49 |
-
|
|
|
|
|
|
|
|
|
|
| 50 |
input_ids = inputs["input_ids"][:,-max_context_length_tokens:].to(device)
|
| 51 |
torch.cuda.empty_cache()
|
| 52 |
|
|
@@ -78,8 +87,18 @@ def predict(text,
|
|
| 78 |
yield a,b,"Generate: Success"
|
| 79 |
except:
|
| 80 |
pass
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 81 |
|
|
|
|
|
|
|
|
|
|
| 82 |
|
|
|
|
|
|
|
| 83 |
def reset_chat():
|
| 84 |
#id_new = chatbot.new_conversation()
|
| 85 |
#chatbot.change_conversation(id_new)
|
|
@@ -162,15 +181,14 @@ with gr.Blocks(theme=small_and_beautiful_theme) as demo:
|
|
| 162 |
predict_args = dict(
|
| 163 |
fn=predict,
|
| 164 |
inputs=[
|
| 165 |
-
|
| 166 |
-
chatbotGr,
|
| 167 |
history,
|
| 168 |
top_p,
|
| 169 |
temperature,
|
| 170 |
max_length_tokens,
|
| 171 |
max_context_length_tokens,
|
| 172 |
],
|
| 173 |
-
outputs=[chatbotGr,
|
| 174 |
show_progress=True,
|
| 175 |
)
|
| 176 |
|
|
|
|
| 30 |
temperature,
|
| 31 |
max_length_tokens,
|
| 32 |
max_context_length_tokens,):
|
| 33 |
+
global model, tokenizer, device
|
| 34 |
+
|
| 35 |
+
#wenn eingabe leer - nix tun
|
| 36 |
if text=="":
|
| 37 |
+
yield history,"Empty context."
|
| 38 |
return
|
| 39 |
+
|
| 40 |
+
#wenn Model nicht gefunden -> Fehler
|
| 41 |
try:
|
| 42 |
model
|
| 43 |
except:
|
| 44 |
+
yield [],"No Model Found"
|
| 45 |
return
|
| 46 |
|
| 47 |
+
#Prompt generieren -> mit Kontext bezogen auch auf vorhergehende Eingaben in dem chat
|
| 48 |
inputs = generate_prompt_with_history(text,history,tokenizer,max_length=max_context_length_tokens)
|
| 49 |
if inputs is None:
|
| 50 |
+
yield history,"Input too long."
|
| 51 |
return
|
| 52 |
else:
|
| 53 |
prompt,inputs=inputs
|
| 54 |
begin_length = len(prompt)
|
| 55 |
+
|
| 56 |
+
#####################################################################################################
|
| 57 |
+
#ist glaube ich unnötig, da ich mit Pipeline arbeiten -> mal schauen, ich behalte es noch...
|
| 58 |
+
"""
|
| 59 |
input_ids = inputs["input_ids"][:,-max_context_length_tokens:].to(device)
|
| 60 |
torch.cuda.empty_cache()
|
| 61 |
|
|
|
|
| 87 |
yield a,b,"Generate: Success"
|
| 88 |
except:
|
| 89 |
pass
|
| 90 |
+
"""
|
| 91 |
+
##########################################################################
|
| 92 |
+
#Prompt ist erzeugt, nun mit pipeline eine Antwort von der KI bekommen!
|
| 93 |
+
pipe = pipeline("text-generation", model=model, tokenizer=tokenizer, device=0)
|
| 94 |
+
bot_message = pipe(prompt)
|
| 95 |
|
| 96 |
+
#chatbot - history erweitern und an chatbotGr zurückschicken
|
| 97 |
+
history = history.append((text, bot_message))
|
| 98 |
+
return history, "Erfolg!"
|
| 99 |
|
| 100 |
+
|
| 101 |
+
#neuen Chat beginnen
|
| 102 |
def reset_chat():
|
| 103 |
#id_new = chatbot.new_conversation()
|
| 104 |
#chatbot.change_conversation(id_new)
|
|
|
|
| 181 |
predict_args = dict(
|
| 182 |
fn=predict,
|
| 183 |
inputs=[
|
| 184 |
+
user_input,
|
|
|
|
| 185 |
history,
|
| 186 |
top_p,
|
| 187 |
temperature,
|
| 188 |
max_length_tokens,
|
| 189 |
max_context_length_tokens,
|
| 190 |
],
|
| 191 |
+
outputs=[chatbotGr, status_display],
|
| 192 |
show_progress=True,
|
| 193 |
)
|
| 194 |
|