SNLI_VE Inference Error

[Pefect96]

When I modify the model path and run the evaluate_snli_ve_base.sh or evaluate_snli_ve.sh to test the performance, I meet the following error:

Traceback (most recent call last):
File "/home/dev/OFA-main/run_scripts/snli_ve/../../evaluate.py", line 203, in
cli_main()
File "/home/dev/OFA-main/run_scripts/snli_ve/../../evaluate.py", line 193, in cli_main
distributed_utils.call_main(
File "/home/dev/OFA-main/fairseq/fairseq/distributed/utils.py", line 374, in call_main
distributed_main(cfg.distributed_training.device_id, main, cfg, kwargs)
File "/home/dev/OFA-main/fairseq/fairseq/distributed/utils.py", line 348, in distributed_main
main(cfg, **kwargs)
File "/home/dev/OFA-main/run_scripts/snli_ve/../../evaluate.py", line 168, in main
if torch.any(scores) and isinstance(scores[0], tuple):
TypeError: any(): argument 'input' (position 1) must be Tensor, not list

[Pefect96]

I have solved this problem, it seems that I modified the code earlier due to version problems.

[Yeemkt]

e

When I modify the model path and run the evaluate_snli_ve_base.sh or evaluate_snli_ve.sh to test the performance, I meet the following error:

Traceback (most recent call last): File "/home/dev/OFA-main/run_scripts/snli_ve/../../evaluate.py", line 203, in cli_main() File "/home/dev/OFA-main/run_scripts/snli_ve/../../evaluate.py", line 193, in cli_main distributed_utils.call_main( File "/home/dev/OFA-main/fairseq/fairseq/distributed/utils.py", line 374, in call_main distributed_main(cfg.distributed_training.device_id, main, cfg, kwargs) File "/home/dev/OFA-main/fairseq/fairseq/distributed/utils.py", line 348, in distributed_main main(cfg, **kwargs) File "/home/dev/OFA-main/run_scripts/snli_ve/../../evaluate.py", line 168, in main if torch.any(scores) and isinstance(scores[0], tuple): TypeError: any(): argument 'input' (position 1) must be Tensor, not list

may i ask what u have changed? the run_script?
i only changed a few parameters in the evaluate_refcoco.sh,and it comes
"RuntimeError: Boolean value of Tensor with more than one value is ambiguous"

[JJJYmmm]

e

When I modify the model path and run the evaluate_snli_ve_base.sh or evaluate_snli_ve.sh to test the performance, I meet the following error:
Traceback (most recent call last): File "/home/dev/OFA-main/run_scripts/snli_ve/../../evaluate.py", line 203, in cli_main() File "/home/dev/OFA-main/run_scripts/snli_ve/../../evaluate.py", line 193, in cli_main distributed_utils.call_main( File "/home/dev/OFA-main/fairseq/fairseq/distributed/utils.py", line 374, in call_main distributed_main(cfg.distributed_training.device_id, main, cfg, kwargs) File "/home/dev/OFA-main/fairseq/fairseq/distributed/utils.py", line 348, in distributed_main main(cfg, **kwargs) File "/home/dev/OFA-main/run_scripts/snli_ve/../../evaluate.py", line 168, in main if torch.any(scores) and isinstance(scores[0], tuple): TypeError: any(): argument 'input' (position 1) must be Tensor, not list

may i ask what u have changed? the run_script? i only changed a few parameters in the evaluate_refcoco.sh,and it comes "RuntimeError: Boolean value of Tensor with more than one value is ambiguous"

#441