Hi, thanks for maintaining this great library — our codebase is built on TSlib and it was really helpful throughout the project.
I'd like to propose adding our model T1 for the imputation task.
T1 is a CNN-Transformer hybrid that uses Channel-Head Binding, where each CNN channel is directly paired with one attention head for selective cross-variable information transfer. It achieves
state-of-the-art on the standard imputation benchmarks (ETTh1/h2, ETTm1/m2, Weather, Electricity, Exchange, ILI).
Our code follows the TSlib structure, so integration should be straightforward. Happy to submit a PR if you're interested.
Thanks!
Hi, thanks for maintaining this great library — our codebase is built on TSlib and it was really helpful throughout the project.
I'd like to propose adding our model T1 for the imputation task.
T1 is a CNN-Transformer hybrid that uses Channel-Head Binding, where each CNN channel is directly paired with one attention head for selective cross-variable information transfer. It achieves
state-of-the-art on the standard imputation benchmarks (ETTh1/h2, ETTm1/m2, Weather, Electricity, Exchange, ILI).
Our code follows the TSlib structure, so integration should be straightforward. Happy to submit a PR if you're interested.
Thanks!