crutcher / bimm / #182

Builds Branch Commit Type Ran Committer Via Coverage
#182 [no commit message] push
72.23
#181 main feat: add CLI options for DropBlock, drop path, and early stopping in resnet_tiny - Introduced new arguments (`drop_block_prob`, `drop_path_prob`, `patience`) in `resnet_tiny` example for better training configurability. - Integrated `MetricEarly... push crutcher
75.24
#180 main refactor: adjust ResNet config defaults and remove unused bottleneck flags - Increased `batch_size` default in `resnet_tiny` from 32 to 512 for better workload handling. - Updated `stochastic_depth_drop_path_rate` to 0.3 for improved training reg... push crutcher
75.24
#179 main refactor: remove hardcoded pretrained classes from ResNet example and improve weight loading - Eliminated `pretrained_classes` argument from `resnet_tiny` example for simplified configuration. - Updated ResNet I/O to dynamically infer output clas... push crutcher
75.22
#178 main feat: add model weight caching and support for loading pretrained ResNet weights - Introduced `cache` module with `fetch_model_weights` function for caching and downloading model weights. - Enabled loading pretrained weights in ResNet models with... push crutcher
75.29
#177 main refactor: relocate and deduplicate weight copying functions in ResNet IO - Moved `copy_downsample_weights` and `copy_conv_norm_weights` functions to consolidate their positions and improve modularity. - Resolved redundant implementations by ensur... push crutcher
75.77
#176 main feat: add PyTorch weights loader and expand ResNet IO support - Introduced `load_pytorch_weights` method for ResNet, enabling seamless integration with PyTorch weight files. - Added new module `resnet_io` with comprehensive stub implementations f... push crutcher
75.77
#175 main feat: add `InputStemConfig` with serialization support to ResNet - Introduced `InputStemConfig` enum for configurable ResNet input stem convolutions. - Added `Serialize` and `Deserialize` traits for seamless configuration handling. - Updated `Res... push crutcher
76.96
#174 main refactor: replace `basic_config` and refine ResNet initialization - Introduced `ResNetAbstractConfig` for high-level model configuration with flexible options. - Deprecated `new_basic_config` and replaced it with `to_structure` for better clarity... push crutcher
77.02
#173 main feat: enhance ResNet API and resnet_tiny example with DropBlock and stochastic depth support - Added new API methods to ResNet for setting DropBlock options and stochastic depth configurations (`with_standard_drop_block_prob`, `with_stochastic_de... push crutcher
77.09
#172 main refactor: update `resnet_tiny` defaults and improve ResNet API - Increased `batch_size` default from 32 to 512 for enhanced performance with larger workloads. - Adjusted `learning_rate` default from `1.0e-4` to `1.0e-3` for faster convergence. - ... push crutcher
78.37
#171 main Tune resnet example. push crutcher
78.02
#170 main refactor: replace `legacy_new` with `resnet18` and enhance ResNet config handling - Deprecated `legacy_new` in favor of the new `resnet18` method for ResNet initialization in `resnet_tiny`. - Introduced pre-defined block depths (e.g., `RESNET18_B... push crutcher
78.02
#169 main refactor: standardize `ConvNorm2d` usage across ResNet modules - Replaced `Conv2dNormBlock` with `ConvNorm2d` for consistency across `ResNet` layers, including `BasicBlock`, `BottleneckBlock`, and `ConvDownsample`. - Updated associated structs, c... push crutcher
78.04
#168 main refactor: update `resnet_tiny` defaults and image preprocessing settings - Adjusted `batch_size` to 32 for improved flexibility in smaller workloads. - Updated `learning_rate`, `lr_gamma`, and `artifact_dir` defaults for better alignment with use... push crutcher
78.15