Closed ooe1123 closed 2 months ago
llm -> large_language_modelにrename。
ailia SDKで下記のエラー。
Traceback (most recent call last):
File "/usr/local/lib/python3.11/site-packages/ailia/wrapper.py", line 657, in set_input_blob_data
buf_shape = self.get_blob_shape(idx)
^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.11/site-packages/ailia/wrapper.py", line 491, in get_blob_shape
core.check_error(code, self.__net)
File "/usr/local/lib/python3.11/site-packages/ailia/core.py", line 931, in check_error
raise e(detail)
ailia.core.AiliaUnsettledShapeException: code: -18 (The shape of the blob is not defined.)
+ error detail : Unsettled blob shape.
During handling of the above exception, another exception occurred:
Traceback (most recent call last):
File "/Users/kyakuno/Desktop/ailia/ailia-models-ax/large_language_model/llava/llava.py", line 541, in <module>
main()
File "/Users/kyakuno/Desktop/ailia/ailia-models-ax/large_language_model/llava/llava.py", line 537, in main
recognize_from_image(models)
File "/Users/kyakuno/Desktop/ailia/ailia-models-ax/large_language_model/llava/llava.py", line 492, in recognize_from_image
output_ids = predict(models, prompt, img)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/Users/kyakuno/Desktop/ailia/ailia-models-ax/large_language_model/llava/llava.py", line 453, in predict
generated_ids = greedy_search(net, inputs_embeds, attention_mask)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/Users/kyakuno/Desktop/ailia/ailia-models-ax/large_language_model/llava/llava.py", line 392, in greedy_search
logits, past_key_values = forward(
^^^^^^^^
File "/Users/kyakuno/Desktop/ailia/ailia-models-ax/large_language_model/llava/llava.py", line 147, in forward
output = net.predict([input_ids, inputs_embeds, position_ids, past_key_values])
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/usr/local/lib/python3.11/site-packages/ailia/wrapper.py", line 405, in predict
self.set_input_blob_data(input[i], idx)
File "/usr/local/lib/python3.11/site-packages/ailia/wrapper.py", line 661, in set_input_blob_data
buf_shape = input.shape
^^^^^^^^^^^
AttributeError: 'list' object has no attribute 'shape'
下記に変更する必要がある。
input_list = [input_ids, inputs_embeds, position_ids]
input_list.extend(past_key_values)
output = net.predict(input_list)
CPU推論でもモデルをメモリに乗せるだけで28.51GB使う。 datatype対応してFP16のままメモリに乗せられるようにした方が良さそう。
1分以上かかるが推論自体はできた。
kyakuno@mbakk llava % python3 llava.py -e 1
INFO arg_utils.py (13) : Start!
INFO arg_utils.py (163) : env_id: 1
INFO arg_utils.py (166) : CPU-AppleAccelerate
INFO model_utils.py (89) : ONNX file and Prototxt file are prepared!
INFO model_utils.py (89) : ONNX file and Prototxt file are prepared!
INFO model_utils.py (89) : ONNX file and Prototxt file are prepared!
prompt: What are the things I should be cautious about when I visit here?
INFO llava.py (468) : view.jpg
INFO llava.py (475) : Start inference...
When visiting this location, which features a pier extending over a large body of water, there are a few things to be cautious about. First, be mindful of the weather conditions, as the pier may be affected by strong winds or storms, which could make it unsafe to walk on. Second, be aware of the water depth and any potential hazards, such as submerged rocks or debris, that could pose a risk to your safety. Additionally, be cautious of the presence of wildlife in the area, as there might be birds or other animals that could pose a threat or disturbance. Finally, be respectful of the environment and other visitors, and follow any posted rules or guidelines to ensure a safe and enjoyable experience for everyone.
INFO llava.py (503) : Script finished successfully.
https://github.com/axinc-ai/ailia-models/issues/1411 のPRです。