硬件: 昇腾910B3
驱动版本:23.0.rc3.b101
镜像版本:swr.cn-central-221.ovaijisuan.com/mindformers/mindformers0.8.0_mindspore2.2.0:aarch_20231025
按照教程说明将skywork模型转为CKPT然后转为mindir。
ckpt是能正常推理的,但是使用MSLite推理MindIR时报错,提示cstdint找不到。
具体报错信息如下:
WARNING:root:mslite ascendc custom kernel path not found
WARNING:root:mslite ascendc custom kernel path not found
start compile Ascend C operator PromptFlashAttention. kernel name is prompt_flash_attention
run stack spill compile error: In file included from /usr/local/Ascend/ascend-toolkit/7.0.RC1/opp/built-in/op_impl/ai_core/tbe/impl/dynamic/../ascendc/prompt_flash_attention/prompt_flash_attention.cpp:4:
In file included from /usr/local/Ascend/ascend-toolkit/7.0.RC1/opp/built-in/op_impl/ai_core/tbe/impl/dynamic/../ascendc/prompt_flash_attention/prompt_flash_attention_base.h:10:
In file included from /usr/local/Ascend/ascend-toolkit/7.0.RC1/aarch64-linux/tikcpp/tikcfw/kernel_operator.h:7:
In file included from /usr/local/Ascend/ascend-toolkit/7.0.RC1/aarch64-linux/tikcpp/tikcfw/interface/kernel_tpipe.h:7:
In file included from /usr/local/Ascend/ascend-toolkit/7.0.RC1/aarch64-linux/tikcpp/tikcfw/interface/kernel_tensor.h:7:
In file included from /usr/local/Ascend/ascend-toolkit/7.0.RC1/aarch64-linux/tikcpp/tikcfw/impl/kernel_utils.h:17:
/usr/local/Ascend/ascend-toolkit/7.0.RC1/aarch64-linux/tikcpp/tikcfw/impl/kernel_macros.h:7:10: fatal error: 'cstdint' file not found
include
^~~~~~~~~
1 error generated.
[ERROR] ME(3618,fffc591f0160,python):2023-12-19-11:21:10.545.420 [mindspore/lite/src/extendrt/delegate/ascend_ge/ge_graph_executor.cc:1176] operator()] RunAsync failed.
[ERROR] ME(3618,ffffbe26b010,python):2023-12-19-11:21:10.559.002 [mindspore/lite/src/extendrt/delegate/ascend_ge/ge_graph_executor.cc:1459] RunGraph] Exec compute graph failed, graph id 1
[ERROR] ME(3618,ffffbe26b010,python):2023-12-19-11:21:10.559.165 [mindspore/lite/src/extendrt/session/delegate_session.cc:245] RunGraph] GraphSinkSession::RunGraph run graph failed
[ERROR] ME(3618,ffffbe26b010,python):2023-12-19-11:21:10.559.247 [mindspore/lite/src/extendrt/cxx_api/model/model_impl.cc:640] Predict] ModelImpl::Predict RunGraph failed with Common error code.
Traceback (most recent call last):
File "/home/gyy/projects/ASR/codes/mindformers-master/research/skywork/run_skywork_infer_lite.py", line 58, in
infer_main(args)
File "/home/gyy/projects/ASR/codes/mindformers-master/research/skywork/run_skywork_infer_lite.py", line 43, in infer_main
output = infer_model.infer(user_input, pad_token_id=0, input_seq_length=args.input_seq_length, eos_token_id=2)
File "/root/miniconda3/envs/mindspore2.2_py39/lib/python3.9/site-packages/mindformers/inference/infers/text_generator_infer.py", line 245, in infer
output_ids = self.generate(input_ids, do_sample, top_k, top_p, temperature,
File "/root/miniconda3/envs/mindspore2.2_py39/lib/python3.9/site-packages/mindformers/inference/infers/text_generator_infer.py", line 379, in generate
outputs = self._inc_infer(input_ids, current_index, valid_length, is_first_iteration, **kwargs)
File "/root/miniconda3/envs/mindspore2.2_py39/lib/python3.9/site-packages/mindformers/inference/infers/text_generator_infer.py", line 454, in _inc_infer
outputs = self.full_model.predict(lite_inputs)
File "/root/miniconda3/envs/mindspore2.2_py39/lib/python3.9/site-packages/mindspore_lite/model.py", line 375, in predict
return super(Model, self).predict(inputs_tensor, outputs)
File "/root/miniconda3/envs/mindspore2.2_py39/lib/python3.9/site-packages/mindspore_lite/base_model.py", line 80, in predict
raise RuntimeError(f"predict failed!")
RuntimeError: predict failed!
硬件: 昇腾910B3 驱动版本:23.0.rc3.b101 镜像版本:swr.cn-central-221.ovaijisuan.com/mindformers/mindformers0.8.0_mindspore2.2.0:aarch_20231025 按照教程说明将skywork模型转为CKPT然后转为mindir。 ckpt是能正常推理的,但是使用MSLite推理MindIR时报错,提示cstdint找不到。 具体报错信息如下: WARNING:root:mslite ascendc custom kernel path not found WARNING:root:mslite ascendc custom kernel path not found start compile Ascend C operator PromptFlashAttention. kernel name is prompt_flash_attention run stack spill compile error: In file included from /usr/local/Ascend/ascend-toolkit/7.0.RC1/opp/built-in/op_impl/ai_core/tbe/impl/dynamic/../ascendc/prompt_flash_attention/prompt_flash_attention.cpp:4: In file included from /usr/local/Ascend/ascend-toolkit/7.0.RC1/opp/built-in/op_impl/ai_core/tbe/impl/dynamic/../ascendc/prompt_flash_attention/prompt_flash_attention_base.h:10: In file included from /usr/local/Ascend/ascend-toolkit/7.0.RC1/aarch64-linux/tikcpp/tikcfw/kernel_operator.h:7: In file included from /usr/local/Ascend/ascend-toolkit/7.0.RC1/aarch64-linux/tikcpp/tikcfw/interface/kernel_tpipe.h:7: In file included from /usr/local/Ascend/ascend-toolkit/7.0.RC1/aarch64-linux/tikcpp/tikcfw/interface/kernel_tensor.h:7: In file included from /usr/local/Ascend/ascend-toolkit/7.0.RC1/aarch64-linux/tikcpp/tikcfw/impl/kernel_utils.h:17: /usr/local/Ascend/ascend-toolkit/7.0.RC1/aarch64-linux/tikcpp/tikcfw/impl/kernel_macros.h:7:10: fatal error: 'cstdint' file not found
include
1 error generated.
[ERROR] ME(3618,fffc591f0160,python):2023-12-19-11:21:10.545.420 [mindspore/lite/src/extendrt/delegate/ascend_ge/ge_graph_executor.cc:1176] operator()] RunAsync failed. [ERROR] ME(3618,ffffbe26b010,python):2023-12-19-11:21:10.559.002 [mindspore/lite/src/extendrt/delegate/ascend_ge/ge_graph_executor.cc:1459] RunGraph] Exec compute graph failed, graph id 1 [ERROR] ME(3618,ffffbe26b010,python):2023-12-19-11:21:10.559.165 [mindspore/lite/src/extendrt/session/delegate_session.cc:245] RunGraph] GraphSinkSession::RunGraph run graph failed [ERROR] ME(3618,ffffbe26b010,python):2023-12-19-11:21:10.559.247 [mindspore/lite/src/extendrt/cxx_api/model/model_impl.cc:640] Predict] ModelImpl::Predict RunGraph failed with Common error code. Traceback (most recent call last): File "/home/gyy/projects/ASR/codes/mindformers-master/research/skywork/run_skywork_infer_lite.py", line 58, in
infer_main(args)
File "/home/gyy/projects/ASR/codes/mindformers-master/research/skywork/run_skywork_infer_lite.py", line 43, in infer_main
output = infer_model.infer(user_input, pad_token_id=0, input_seq_length=args.input_seq_length, eos_token_id=2)
File "/root/miniconda3/envs/mindspore2.2_py39/lib/python3.9/site-packages/mindformers/inference/infers/text_generator_infer.py", line 245, in infer
output_ids = self.generate(input_ids, do_sample, top_k, top_p, temperature,
File "/root/miniconda3/envs/mindspore2.2_py39/lib/python3.9/site-packages/mindformers/inference/infers/text_generator_infer.py", line 379, in generate
outputs = self._inc_infer(input_ids, current_index, valid_length, is_first_iteration, **kwargs)
File "/root/miniconda3/envs/mindspore2.2_py39/lib/python3.9/site-packages/mindformers/inference/infers/text_generator_infer.py", line 454, in _inc_infer
outputs = self.full_model.predict(lite_inputs)
File "/root/miniconda3/envs/mindspore2.2_py39/lib/python3.9/site-packages/mindspore_lite/model.py", line 375, in predict
return super(Model, self).predict(inputs_tensor, outputs)
File "/root/miniconda3/envs/mindspore2.2_py39/lib/python3.9/site-packages/mindspore_lite/base_model.py", line 80, in predict
raise RuntimeError(f"predict failed!")
RuntimeError: predict failed!