Search results

Bug: qk_norm_q_scale / qk_norm_k_scale not applied in forward pass — attn_qk_norm_dim_scale=True silently has no effect

Status: Open.

#352 In lucidrains/x-transformers;

· lukewys opened on Apr 3, 2026

muP

Status: Open.

#349 In lucidrains/x-transformers;

· Aceticia opened on Feb 2, 2026

Polar pos emb doubles the output dimension

Status: Open.

#348 In lucidrains/x-transformers;

· inspirit opened on Dec 30, 2025

How to create something similar to t5/seq2seq?

Status: Open.

#342 In lucidrains/x-transformers;

· KlausikPL opened on Oct 12, 2025

Inference Time

Status: Open.

#338 In lucidrains/x-transformers;

· JLenzy opened on Jul 29, 2025

Is this project tensorrt compatible

Status: Open.

#337 In lucidrains/x-transformers;

· JohnHerry opened on Jul 17, 2025

Question: prepend_embeds & prepend_mask + KV caching

Status: Open.

#332 In lucidrains/x-transformers;

· pradeep-pyro opened on Jun 26, 2025

Request: Skewed relative position encoding (Music Transformer)

Status: Open.

#328 In lucidrains/x-transformers;

· JLenzy opened on May 29, 2025

Masked decoder use leads to excessive memory usage (mask=mask)

Status: Open.

#325 In lucidrains/x-transformers;

· andyschuster opened on May 21, 2025

Supporting BatchNorm

Status: Open.

#320 In lucidrains/x-transformers;

· egretholier opened on Apr 18, 2025

Adding additional embeddings after the Q and K are projected, but before the RoPE

Status: Open.

#319 In lucidrains/x-transformers;

· JavaStudentAlex opened on Apr 16, 2025

Multi GPU Support

Status: Open.

#312 In lucidrains/x-transformers;

· ghost opened on Mar 5, 2025