Skip to content

sometimes CI stuck because test_recognize_digits run forever #8168

@helinwang

Description

@helinwang

The loop termination condition (below some loss threshold) could never reach because the model reached a bad state with NaN loss:

[22:37:57] : [Step 1/1] 72/72 Test #351: test_recognize_digits ...................***Exception: Other2588.00 sec [22:37:57] : [Step 1/1] PassID 0, BatchID 0010, Test Loss 1.7, Acc 0.41 [22:37:57] : [Step 1/1] PassID 0, BatchID 0020, Test Loss 1.1, Acc 0.71 [22:37:57] : [Step 1/1] PassID 0, BatchID 0030, Test Loss 0.68, Acc 0.8 [22:37:57] : [Step 1/1] PassID 0, BatchID 0040, Test Loss 0.51, Acc 0.84 [22:37:57] : [Step 1/1] infer results: [[0.04630879 0.00167811 0.12129738 0.34336442 0.14744693 0.03341632 [22:37:57] : [Step 1/1] 0.01416888 0.00522202 0.27912304 0.00797407]] [22:37:57] : [Step 1/1] .PassID 0, BatchID 0010, Test Loss 1.4, Acc 0.59 [22:37:57] : [Step 1/1] PassID 0, BatchID 0020, Test Loss 0.86, Acc 0.72 [22:37:57] : [Step 1/1] PassID 0, BatchID 0030, Test Loss 0.59, Acc 0.83 [22:37:57] : [Step 1/1] .EEPassID 0, BatchID 0010, Test Loss 1.3, Acc 0.62 [22:37:57] : [Step 1/1] PassID 0, BatchID 0020, Test Loss 0.8, Acc 0.75 [22:37:57] : [Step 1/1] PassID 0, BatchID 0030, Test Loss 0.65, Acc 0.8 [22:37:57] : [Step 1/1] PassID 0, BatchID 0040, Test Loss 0.54, Acc 0.84 [22:37:57] : [Step 1/1] infer results: [[0.21086314 0.12968424 0.05096722 0.27543962 0.03747584 0.00835655 [22:37:57] : [Step 1/1] 0.09035194 0.00980698 0.14980537 0.03724904]] [22:37:57] : [Step 1/1] .PassID 0, BatchID 0010, Test Loss 1.2, Acc 0.65 [22:37:57] : [Step 1/1] PassID 0, BatchID 0020, Test Loss 0.77, Acc 0.78 [22:37:57] : [Step 1/1] PassID 0, BatchID 0030, Test Loss 0.57, Acc 0.85 [22:37:57] : [Step 1/1] .PassID 0, BatchID 0010, Test Loss 1.3, Acc 0.65 [22:37:57] : [Step 1/1] PassID 0, BatchID 0020, Test Loss 0.79, Acc 0.76 [22:37:57] : [Step 1/1] PassID 0, BatchID 0030, Test Loss 0.63, Acc 0.82 [22:37:57] : [Step 1/1] PassID 0, BatchID 0040, Test Loss 0.52, Acc 0.85 [22:37:57] : [Step 1/1] PassID 0, BatchID 0050, Test Loss 0.51, Acc 0.84 [22:37:57] : [Step 1/1] .PassID 0, BatchID 0010, Test Loss nan, Acc 0.028 [22:37:57] : [Step 1/1] PassID 0, BatchID 0020, Test Loss nan, Acc 0.028 [22:37:57] : [Step 1/1] PassID 0, BatchID 0030, Test Loss nan, Acc 0.028 [22:37:57] : [Step 1/1] PassID 0, BatchID 0040, Test Loss nan, Acc 0.028 [22:37:57] : [Step 1/1] PassID 0, BatchID 0050, Test Loss nan, Acc 0.028 [22:37:57] : [Step 1/1] PassID 0, BatchID 0060, Test Loss nan, Acc 0.028 [22:37:57] : [Step 1/1] PassID 0, BatchID 0070, Test Loss nan, Acc 0.028 [22:37:57] : [Step 1/1] PassID 0, BatchID 0080, Test Loss nan, Acc 0.028 

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions