All issues

Search results

DBRX: attentions_input is the pre-norm residual (input of norm_attn_norm), unlike other architectures

Status: Open.

#56 In ndif-team/nnterp;

· Butanium opened on Aug 28, 2026

Sublayer output multipliers (Granite residual_multiplier, Falcon-H1 attention_out_multiplier) make attentions_output / mlps_output differ from the residual contribution by a constant factor

Status: Open.

#54 In ndif-team/nnterp;

· Butanium opened on Aug 28, 2026

attentions_output / mlps_output are not the residual contribution on post-sublayer-norm ("sandwich") architectures (Gemma-2/3, GLM-4, OLMo-2)

Status: Open.

#53 In ndif-team/nnterp;

· Butanium opened on Aug 28, 2026

generate is now on nnsight too

Status: Open.

#46 In ndif-team/nnterp;

· Butanium opened on Apr 30, 2026

[dev] current vllm .layers_output is very likely wrong

Status: Open.

#42 In ndif-team/nnterp;

· Butanium opened on Apr 14, 2026

Gemma 3n not supported: AltUp 4D hidden states break nnterp assumptions

Status: Open.

#35 In ndif-team/nnterp;

· Butanium opened on Mar 25, 2026

Should we add support for hybrid mamba models?

Status: Open.

#18 In ndif-team/nnterp;

· Butanium opened on Oct 6, 2025

generate docs from notebooks

Status: Open.

#13 In ndif-team/nnterp;

· Butanium opened on Aug 25, 2025

add test for caching

Status: Open.

#12 In ndif-team/nnterp;

· Butanium opened on Aug 25, 2025

Add down_proj renaming

good first issueGood for newcomersGood for newcomers

Status: Open.

#11 In ndif-team/nnterp;

· Butanium opened on Aug 23, 2025

vLLM support

Status: Open.

#6 In ndif-team/nnterp;

· Butanium opened on Feb 15, 2025