Spaces:
Running
Running
Update app.py
Browse files
app.py
CHANGED
|
@@ -1,6 +1,7 @@
|
|
| 1 |
import gradio as gr
|
| 2 |
from transformers import AutoProcessor, Idefics3ForConditionalGeneration, TextIteratorStreamer
|
| 3 |
from threading import Thread
|
|
|
|
| 4 |
import re
|
| 5 |
import time
|
| 6 |
from PIL import Image
|
|
@@ -24,13 +25,13 @@ def model_inference(
|
|
| 24 |
input_dict, history, decoding_strategy, temperature, max_new_tokens,
|
| 25 |
repetition_penalty, top_p
|
| 26 |
):
|
| 27 |
-
|
| 28 |
text = input_dict["text"]
|
| 29 |
|
| 30 |
if len(input_dict["files"]) > 1:
|
| 31 |
-
images = [Image.open(image).convert("RGB") for image in input_dict["files"]]
|
| 32 |
elif len(input_dict["files"]) == 1:
|
| 33 |
-
images = [Image.open(input_dict["files"][0]).convert("RGB")]
|
| 34 |
else:
|
| 35 |
images = []
|
| 36 |
|
|
@@ -78,8 +79,9 @@ def model_inference(
|
|
| 78 |
# start token id = argmax last logit
|
| 79 |
start_token_id = int(np.argmax(prefill_out["logits"][:, -1, :], axis=-1)[0])
|
| 80 |
|
| 81 |
-
|
| 82 |
-
|
|
|
|
| 83 |
generation_args = {
|
| 84 |
"llm_session" : llm_session,
|
| 85 |
"embed_tokens_session": embed_tokens_session,
|
|
@@ -89,14 +91,32 @@ def model_inference(
|
|
| 89 |
"freqs_cos": freqs_cos,
|
| 90 |
"freqs_sin": freqs_sin,
|
| 91 |
"attention_mask": attention_mask.numpy(),
|
| 92 |
-
"max_new_tokens":
|
| 93 |
"eos_token_id": 2,
|
| 94 |
-
"start_pos": seqlen
|
|
|
|
| 95 |
}
|
| 96 |
-
|
| 97 |
|
|
|
|
| 98 |
thread = Thread(target=generate_autoregressive, kwargs=generation_args)
|
| 99 |
thread.start()
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 100 |
|
| 101 |
|
| 102 |
examples = [
|
|
|
|
| 1 |
import gradio as gr
|
| 2 |
from transformers import AutoProcessor, Idefics3ForConditionalGeneration, TextIteratorStreamer
|
| 3 |
from threading import Thread
|
| 4 |
+
from queue import Queue
|
| 5 |
import re
|
| 6 |
import time
|
| 7 |
from PIL import Image
|
|
|
|
| 25 |
input_dict, history, decoding_strategy, temperature, max_new_tokens,
|
| 26 |
repetition_penalty, top_p
|
| 27 |
):
|
| 28 |
+
print(input_dict)
|
| 29 |
text = input_dict["text"]
|
| 30 |
|
| 31 |
if len(input_dict["files"]) > 1:
|
| 32 |
+
images = [Image.open(image).convert("RGB") for image in input_dict["files"]["path"]]
|
| 33 |
elif len(input_dict["files"]) == 1:
|
| 34 |
+
images = [Image.open(input_dict["files"][0]["path"]).convert("RGB")]
|
| 35 |
else:
|
| 36 |
images = []
|
| 37 |
|
|
|
|
| 79 |
# start token id = argmax last logit
|
| 80 |
start_token_id = int(np.argmax(prefill_out["logits"][:, -1, :], axis=-1)[0])
|
| 81 |
|
| 82 |
+
# 创建输出队列用于线程间通信
|
| 83 |
+
output_queue = Queue()
|
| 84 |
+
|
| 85 |
generation_args = {
|
| 86 |
"llm_session" : llm_session,
|
| 87 |
"embed_tokens_session": embed_tokens_session,
|
|
|
|
| 91 |
"freqs_cos": freqs_cos,
|
| 92 |
"freqs_sin": freqs_sin,
|
| 93 |
"attention_mask": attention_mask.numpy(),
|
| 94 |
+
"max_new_tokens": max_new_tokens,
|
| 95 |
"eos_token_id": 2,
|
| 96 |
+
"start_pos": seqlen,
|
| 97 |
+
"output_queue": output_queue
|
| 98 |
}
|
|
|
|
| 99 |
|
| 100 |
+
# 在后台线程启动生成
|
| 101 |
thread = Thread(target=generate_autoregressive, kwargs=generation_args)
|
| 102 |
thread.start()
|
| 103 |
+
|
| 104 |
+
# 从队列中读取生成的文本并 yield
|
| 105 |
+
yield "..."
|
| 106 |
+
buffer = ""
|
| 107 |
+
|
| 108 |
+
while True:
|
| 109 |
+
text_chunk = output_queue.get() # 阻塞等待队列中的数据
|
| 110 |
+
|
| 111 |
+
if text_chunk is None: # 生成完成信号
|
| 112 |
+
break
|
| 113 |
+
|
| 114 |
+
buffer += text_chunk
|
| 115 |
+
time.sleep(0.01)
|
| 116 |
+
yield buffer
|
| 117 |
+
|
| 118 |
+
# 等待线程完成
|
| 119 |
+
thread.join()
|
| 120 |
|
| 121 |
|
| 122 |
examples = [
|