Spaces:
Sleeping
Sleeping
phamngoctukts
commited on
Commit
•
ff4edf1
1
Parent(s):
deacc32
Update app.py
Browse files
app.py
CHANGED
@@ -18,12 +18,10 @@ import os
|
|
18 |
tk = token = os.environ.get("HF_TOKEN")
|
19 |
login(tk)
|
20 |
model_id = "meta-llama/Llama-3.2-1B"
|
21 |
-
|
22 |
-
|
23 |
-
|
24 |
-
|
25 |
-
device_map="auto"
|
26 |
-
)
|
27 |
r = sr.Recognizer()
|
28 |
|
29 |
@dataclass
|
@@ -85,7 +83,7 @@ def process_audio(audio:tuple, state:AppState):
|
|
85 |
return gr.Audio(recording=False), state
|
86 |
return None, state
|
87 |
|
88 |
-
def response(state:AppState):
|
89 |
if not state.pause_detected and not state.started_talking:
|
90 |
return None, AppState()
|
91 |
audio_buffer = BytesIO()
|
@@ -107,7 +105,54 @@ def response(state:AppState):
|
|
107 |
if textin != "":
|
108 |
print("Đang nghĩ...")
|
109 |
textout=str(text2text(textin))
|
110 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
111 |
state.conversation.append({"role": "user", "content": "Trợ lý: " + textout})
|
112 |
if textout != "":
|
113 |
print("Đang đọc...")
|
|
|
18 |
tk = token = os.environ.get("HF_TOKEN")
|
19 |
login(tk)
|
20 |
model_id = "meta-llama/Llama-3.2-1B"
|
21 |
+
ckpt = "meta-llama/Llama-3.2-11B-Vision-Instruct"
|
22 |
+
model = MllamaForConditionalGeneration.from_pretrained(ckpt,
|
23 |
+
torch_dtype=torch.bfloat16).to("cpu")
|
24 |
+
processor = AutoProcessor.from_pretrained(ckpt)
|
|
|
|
|
25 |
r = sr.Recognizer()
|
26 |
|
27 |
@dataclass
|
|
|
83 |
return gr.Audio(recording=False), state
|
84 |
return None, state
|
85 |
|
86 |
+
def response(state:AppState, message, history, max_new_tokens=250):
|
87 |
if not state.pause_detected and not state.started_talking:
|
88 |
return None, AppState()
|
89 |
audio_buffer = BytesIO()
|
|
|
105 |
if textin != "":
|
106 |
print("Đang nghĩ...")
|
107 |
textout=str(text2text(textin))
|
108 |
+
|
109 |
+
|
110 |
+
for i, msg in enumerate(history):
|
111 |
+
if isinstance(msg[0], tuple):
|
112 |
+
messages.append({"role": "user", "content": [{"type": "text", "text": history[i+1][0]}, {"type": "image"}]})
|
113 |
+
messages.append({"role": "assistant", "content": [{"type": "text", "text": history[i+1][1]}]})
|
114 |
+
images.append(Image.open(msg[0][0]).convert("RGB"))
|
115 |
+
elif isinstance(history[i-1], tuple) and isinstance(msg[0], str):
|
116 |
+
# messages are already handled
|
117 |
+
pass
|
118 |
+
elif isinstance(history[i-1][0], str) and isinstance(msg[0], str): # text only turn
|
119 |
+
messages.append({"role": "user", "content": [{"type": "text", "text": msg[0]}]})
|
120 |
+
messages.append({"role": "assistant", "content": [{"type": "text", "text": msg[1]}]})
|
121 |
+
|
122 |
+
# add current message
|
123 |
+
if len(message["files"]) == 1:
|
124 |
+
if isinstance(message["files"][0], str): # examples
|
125 |
+
image = Image.open(message["files"][0]).convert("RGB")
|
126 |
+
else: # regular input
|
127 |
+
image = Image.open(message["files"][0]["path"]).convert("RGB")
|
128 |
+
images.append(image)
|
129 |
+
messages.append({"role": "user", "content": [{"type": "text", "text": txt}, {"type": "image"}]})
|
130 |
+
else:
|
131 |
+
messages.append({"role": "user", "content": [{"type": "text", "text": txt}]})
|
132 |
+
|
133 |
+
|
134 |
+
texts = processor.apply_chat_template(messages, add_generation_prompt=True)
|
135 |
+
|
136 |
+
if images == []:
|
137 |
+
inputs = processor(text=texts, return_tensors="pt").to("cpu")
|
138 |
+
else:
|
139 |
+
inputs = processor(text=texts, images=images, return_tensors="pt").to("cpu")
|
140 |
+
streamer = TextIteratorStreamer(processor, skip_special_tokens=True, skip_prompt=True)
|
141 |
+
generation_kwargs = dict(inputs, streamer=streamer, max_new_tokens=max_new_tokens)
|
142 |
+
generated_text = streamer
|
143 |
+
|
144 |
+
thread = Thread(target=model.generate, kwargs=generation_kwargs)
|
145 |
+
thread.start()
|
146 |
+
buffer = ""
|
147 |
+
|
148 |
+
for new_text in streamer:
|
149 |
+
buffer += new_text
|
150 |
+
generated_text_without_prompt = buffer
|
151 |
+
time.sleep(0.01)
|
152 |
+
yield buffer
|
153 |
+
|
154 |
+
|
155 |
+
textout = generated_text.replace('*','')
|
156 |
state.conversation.append({"role": "user", "content": "Trợ lý: " + textout})
|
157 |
if textout != "":
|
158 |
print("Đang đọc...")
|