The GitHub Actions job "Teams" on tvm.git/main has succeeded. Run started by GitHub user tlopex (triggered by tlopex).
Head commit for run: fb6453a817264efd5b2e19c8b3a118e6b383725b / Soowon Jeong <[email protected]> [Relax][TFLite] Fix `MIRROR_PAD`/`ONE_HOT` converters and add tests for `PAD`, `PADV2`, `MIRROR_PAD`, `TOPK_V2`, `ONE_HOT` (#19373) Part of #18971 Two bugs in the TFLite Relax frontend converters are fixed, and unit tests are added for the **Padding / Sparse / Other** category operators claimed in that tracking issue. ## Bug fixes ### `convert_mirror_pad` Called `relax.op.nn.mirror_pad` which does not exist in the Relax op namespace. Replaced with `relax.op.nn.pad` using `pad_mode="reflect"` for REFLECT mode (the modes are semantically equivalent). SYMMETRIC mode raises `OpAttributeUnImplemented` as there is no direct Relax equivalent. ### `convert_one_hot` - `on_value` and `off_value` were passed as `Expr` (constant tensor nodes), but `relax.op.one_hot` requires `PrimValue` arguments. - An extra `dtype` positional argument was passed, which the function signature does not accept. Fixed by extracting the scalar from each tensor buffer and wrapping it in `relax.PrimValue` with the correct dtype via `tvm.tirx.FloatImm` / `tvm.tirx.IntImm`. ## Tests added Each test uses the `verify()` + `tf.Module` pattern and includes an explicit expected IRModule verified with `tvm.ir.assert_structural_equal`. | Operator | TFLite op | Notes | |----------|-----------|-------| | `test_pad` | `PAD` | constant zero padding | | `test_pad_v2` | `PADV2` | explicit `constant_values=5.0` | | `test_mirror_pad` | `MIRROR_PAD` | REFLECT mode | | `test_topk_v2` | `TOPK_V2` | returns top-k values | | `test_one_hot` | `ONE_HOT` | float32 on/off values, depth=4 | Report URL: https://github.com/apache/tvm/actions/runs/24207151255 With regards, GitHub Actions via GitBox --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
