模型超过2G怎么处理

[Worromots]

File "/workdir/common/anaconda3/envs/qpixart/lib/python3.9/runpy.py", line 87, in _run_code
exec(code, run_globals)
File "/mnt/dolphinfs/hdd_pool/docker/user/hadoop-platcv/tsl/q_mtpixart/Dipoorlet/dipoorlet/main.py", line 120, in
act_clip_val, weight_clip_val = tensor_calibration(onnx_graph, args)
File "/mnt/dolphinfs/hdd_pool/docker/user/hadoop-platcv/tsl/q_mtpixart/Dipoorlet/dipoorlet/tensor_cali/tensor_cali_base.py", line 6, in tensor_calibration
act_clip_val = tensor_cali_dispatcher(args.act_quant, onnx_graph, args)
File "/mnt/dolphinfs/hdd_pool/docker/user/hadoop-platcv/tsl/q_mtpixart/Dipoorlet/dipoorlet/utils.py", line 297, in wrapper
return dispatch(args[0])(*(args[1:]), **kw)
File "/mnt/dolphinfs/hdd_pool/docker/user/hadoop-platcv/tsl/q_mtpixart/Dipoorlet/dipoorlet/tensor_cali/basic_algorithm.py", line 35, in find_clip_val_hist
stats_min_max = forward_get_minmax(onnx_graph, args)
File "/mnt/dolphinfs/hdd_pool/docker/user/hadoop-platcv/tsl/q_mtpixart/Dipoorlet/dipoorlet/forward_net.py", line 200, in forward_get_minmax
ort_session = ort.InferenceSession(net.SerializeToString(), providers=providers)

[gushiqiao]

跑的是transformer类的模型嘛？

[Worromots]

是的 跑的DIT发自我的 iPhone在 2024年8月26日，14:51，gushiqiao ***@***.***> 写道： 跑的是transformer类的模型嘛？ —Reply to this email directly, view it on GitHub, or unsubscribe.You are receiving this because you authored the thread.Message ID: ***@***.***>

[gushiqiao]

超过2G的模型，Dipoorlet会依赖"onnxruntime.transformers.optimizer"进行优化， 具体代码在这里，https://github.com/ModelTC/Dipoorlet/blob/c89130744d45ae2e7cb77081b8799c2ff31ee08d/dipoorlet/__main__.py#L85。
至于DIT，其实要看"onnxruntime.transformers.optimizer"是否支持，目前我们在Dipoorlet里只支持了Unet的优化，可以在这里看到

Dipoorlet/dipoorlet/__main__.py

Line 53 in c891307

parser.add_argument("--model_type", help="Transformer model type", choices=["unet"], default=None)

[gushiqiao]

超过2G的模型，Dipoorlet会依赖"onnxruntime.transformers.optimizer"进行优化， 具体代码在这里，https://github.com/ModelTC/Dipoorlet/blob/c89130744d45ae2e7cb77081b8799c2ff31ee08d/dipoorlet/main.py#L85。 至于DIT，其实要看"onnxruntime.transformers.optimizer"是否支持，目前我们在Dipoorlet里只支持了Unet的优化，可以在这里看到

Dipoorlet/dipoorlet/__main__.py

Line 53 in c891307

parser.add_argument("--model_type", help="Transformer model type", choices=["unet"], default=None)

@Worromots