Resources

Fix assistant content order before tool calls

#55 opened 4 days ago by

youmuorz

Add IFBench evaluation result

#54 opened 8 days ago by

SaylorTwift

Drop the trailing space after the last system text part

#53 opened 26 days ago by

qgallouedec

Up to 4.23x faster inference with a DSpark draft model for gemma-4-E2B-it (167 --> 709 tokens/s; faster than MTP on 7/9 benchmarks)

#52 opened about 1 month ago by

rasyosef

Up to 3.74x faster inference with DFlash2 speculative decoding draft model for gemma-4-E2B-it (167 --> 626 tokens/s)

#50 opened about 2 months ago by

rasyosef

3.7x faster inference with a DSpark (speculative decoding) draft model for gemma-4-E2B-it

#49 opened about 2 months ago by

rasyosef

[Exploration] Applying Native Dual-Regime & KV-Cache Handoff to this model variant?

#48 opened about 2 months ago by

savhascelik

Weak at generating structured output.

#47 opened about 2 months ago by

Duonglv

Add ExtractBench evaluation results

#46 opened 3 months ago by

boyang-runllama

Internal model structure preview

#45 opened 3 months ago by

svetoviz

fix: chat_template, keep tool-call reasoning across later user turns

#44 opened 4 months ago by

dimondev

Raise per-image vision soft-token budget from 280 to 1120

#41 opened 4 months ago by

lucianommartins

Add IFStruct v1.0 evaluation result

#39 opened 4 months ago by

SaylorTwift

Add Vaani-Benchmark-V1.0 eval results

#38 opened 5 months ago by

PranavdbhatArtpark

Chat template may re-inject prior-turn reasoning during multi-turn tool use → repetition loops

#36 opened 5 months ago by

ManniX-ITA

fun config

#33 opened 6 months ago by

baseten-admin

Add MMMU-Pro evaluation result

#32 opened 6 months ago by

SaylorTwift

How to best prompt for transcription results when sending audio input?

#30 opened 6 months ago by

DiffusionFanatic1

GGUF + pure-C++ runtime in CrispASR — Gemma-4-E2B (ASR)

#29 opened 6 months ago by

cstr

[Feature request] Eliminate pre-attention RMSNorm in Gemma 4 via scale invariance + weight folding

👍 1

#27 opened 6 months ago by

graefics

fix(chat template): better SI and tool call handling

#25 opened 7 months ago by

dougreid

[Bug] Chat template redacting reasoning traces for non-tool related use case

#24 opened 7 months ago by

vibhas09

Why k/v proj exist in layers where "is_kv_shared_layer = True"?

#23 opened 7 months ago by

jiayuanm

Issue loading google/gemma-4-E2B-it with AutoProcessor (Unrecognized processing class)

#22 opened 7 months ago by

dipakmeher51

regarding of supported image token budgets

#20 opened 7 months ago by

J22

Update README.md

#18 opened 7 months ago by

Highgroundbkk

Litert version after fine-tuning

#17 opened 7 months ago by

Abhijitmore

Welcome G4! Any new T5-G4 coming?

#16 opened 7 months ago by

iSurpriseMe

Request

#14 opened 7 months ago by

mlboydaisuke

fix: function calling formatting in chat template

#12 opened 7 months ago by

RyanMullins

Allucination parsing JSON responses containing numerical data such as 1234.00

#10 opened 7 months ago by

jeisonvendetta

Benchmarking low resource languages

👍 1

#9 opened 7 months ago by

indiaiquest

Embed chat_template in tokenizer_config.json

#8 opened 7 months ago by

piero-atelico

Welcome G4! Any new T5-G4 coming?

#5 opened 7 months ago by

JohnGenetica

Add AIME 2026 evaluation result

#4 opened 7 months ago by

SaylorTwift

Upload 2 files

#2 opened 7 months ago by

merve