Open Issues Need Help
View All on GitHubMachine learning metrics for distributed, scalable PyTorch applications.
Machine learning metrics for distributed, scalable PyTorch applications.
Machine learning metrics for distributed, scalable PyTorch applications.
Machine learning metrics for distributed, scalable PyTorch applications.
Machine learning metrics for distributed, scalable PyTorch applications.
Machine learning metrics for distributed, scalable PyTorch applications.
Machine learning metrics for distributed, scalable PyTorch applications.
Machine learning metrics for distributed, scalable PyTorch applications.
Machine learning metrics for distributed, scalable PyTorch applications.
Machine learning metrics for distributed, scalable PyTorch applications.
Machine learning metrics for distributed, scalable PyTorch applications.
Machine learning metrics for distributed, scalable PyTorch applications.
Machine learning metrics for distributed, scalable PyTorch applications.
Machine learning metrics for distributed, scalable PyTorch applications.
Machine learning metrics for distributed, scalable PyTorch applications.
Machine learning metrics for distributed, scalable PyTorch applications.
Machine learning metrics for distributed, scalable PyTorch applications.
Machine learning metrics for distributed, scalable PyTorch applications.
Machine learning metrics for distributed, scalable PyTorch applications.
Machine learning metrics for distributed, scalable PyTorch applications.
Machine learning metrics for distributed, scalable PyTorch applications.
Machine learning metrics for distributed, scalable PyTorch applications.
Machine learning metrics for distributed, scalable PyTorch applications.
Machine learning metrics for distributed, scalable PyTorch applications.
Machine learning metrics for distributed, scalable PyTorch applications.
Machine learning metrics for distributed, scalable PyTorch applications.
Machine learning metrics for distributed, scalable PyTorch applications.
Machine learning metrics for distributed, scalable PyTorch applications.
Machine learning metrics for distributed, scalable PyTorch applications.
Machine learning metrics for distributed, scalable PyTorch applications.
Machine learning metrics for distributed, scalable PyTorch applications.
Machine learning metrics for distributed, scalable PyTorch applications.
Machine learning metrics for distributed, scalable PyTorch applications.
Machine learning metrics for distributed, scalable PyTorch applications.
Machine learning metrics for distributed, scalable PyTorch applications.
Machine learning metrics for distributed, scalable PyTorch applications.
Machine learning metrics for distributed, scalable PyTorch applications.
AI Summary: The `MetricCollection._compute_groups_create_state_ref` method is incorrectly linking cached computations (`_computed` attribute) between metrics within the same compute group. This is problematic because, while metrics in a compute group share update state, they are expected to have distinct `compute` methods and thus should cache their own unique computed values, not share them. The current implementation's linking of `_computed` attributes is identified as the source of this bug.
Machine learning metrics for distributed, scalable PyTorch applications.
Machine learning metrics for distributed, scalable PyTorch applications.
Machine learning metrics for distributed, scalable PyTorch applications.
Machine learning metrics for distributed, scalable PyTorch applications.
Machine learning metrics for distributed, scalable PyTorch applications.
Machine learning metrics for distributed, scalable PyTorch applications.
AI Summary: Debug and fix a shape inconsistency in the `_pearson_corrcoef_update` function of the TorchMetrics library that causes errors during distributed training when the number of observations is 1 and the number of outputs is greater than 1. The fix should ensure correct shape handling of `max_abs_dev_x` to prevent issues with metric aggregation across multiple GPUs.
Machine learning metrics for distributed, scalable PyTorch applications.
AI Summary: The task is to fix a bug in the `StructuralSimilarityIndexMeasure` and `MultiScaleStructuralSimilarityIndexMeasure` classes within the TorchMetrics library. The bug causes an error when using distributed training with `reduction='none'` due to incorrect distributed reduction (`dist_reduce_fx='cat'`) of the `similarity` state. The solution involves changing the `dist_reduce_fx` to `None` for the `similarity` state when `reduction` is `None`.
Machine learning metrics for distributed, scalable PyTorch applications.
AI Summary: The task is to debug a bug in the `PearsonCorrCoef` metric within the TorchMetrics library. The metric returns `NaN` when input values have small ranges, which is inconsistent with the results from SciPy's `pearsonr`. The solution involves investigating the numerical stability of the TorchMetrics implementation and potentially improving it to handle small-range inputs more robustly.
Machine learning metrics for distributed, scalable PyTorch applications.
AI Summary: The task is to modify the `PSNR` function in the TorchMetrics library to require a `data_range` argument instead of estimating it from the input data. This involves removing the data-dependent estimation logic and potentially adding default `data_range` values based on data type (e.g., assuming a range of 0-1 for floats and 0-255 for uint8). The goal is to improve the accuracy and consistency of the PSNR calculation, preventing misleading results when the `data_range` is not explicitly specified.
Machine learning metrics for distributed, scalable PyTorch applications.