moonshotai/Kimi-K2-Instruct

Update tokenization_kimi.py

#56 opened 27 days ago by

lfu

Quality compare in IQ4_NL (582Gb RAM) with Q5_K_XLARGE (735Gb RAM) on $150 ancient Xeon PC from 2014

1

#55 opened about 2 months ago by

krustik

Question Regarding Muon optimizer over AdamW

#54 opened about 2 months ago by

jahidhasan

Please, someone distill this model!

#53 opened about 2 months ago by

treehugg3

Any plan to support MTP?

2

#52 opened about 2 months ago by

1000Xia

Update chat_template.jinja

1

#51 opened about 2 months ago by

AdvancedMage

Kimi-K2 Open-Source Model Tool Call Output Format Anomaly: Non-Standard tool_call_id Triggers Parsing Failures Compared with Official Mode

9

#48 opened about 2 months ago by

liopen

Please considering training a smaller dense variant in the same style.

👍 4

#46 opened about 2 months ago by

drmcbride

Function Call Format Error in Multi-turn Dialogues

3

#45 opened about 2 months ago by

chaos778

Chinese AI company true open source champion!

🔥 3

#43 opened 2 months ago by

OrlandoHugBot

Can Kimi-k2 tokenizer please be added to transformers.js ??

#42 opened 2 months ago by

lebronjames6michealjordan23

Function calling always outputs the <|tool_calls_section_begin|> tags.

4

#41 opened 2 months ago by

CaryH

Issue with Kimi K2 Model Support

2

#40 opened 2 months ago by

jerin-scalers-ai

Model seems to reason even though it's an instruct model?

#37 opened 2 months ago by

steve2972

Amazing quality in Q4 on 2014 ANCIENT Xeon CPU with just shy 582Gb RAM

🚀 3

2

#34 opened 2 months ago by

krustik

Really appreciate the work you put into this.🤍

🤗 🔥 8

#32 opened 2 months ago by

deep-div

What actually is the EOS token for this model?

4

#31 opened 2 months ago by

jukofyork

web_search built-in tool via API

1

#29 opened 2 months ago by

PriNova

K2 reasoning

1

#27 opened 2 months ago by

ccocks-deca

vllm nightly build + H200 only achieve Avg generation throughput: 7.2 tokens/

3

#25 opened 2 months ago by

doramonk

Good work - Great potential and current results

👍 6

1

#24 opened 3 months ago by

app-31

Model halucinate

1

#23 opened 3 months ago by

hamidtech

Why not use PreTrainedTokenizerFast

1

#22 opened 3 months ago by

zymu

Bug in multi run function-call

5

#21 opened 3 months ago by

judycc

I may not know the maybe obvious answer to this but I'm curious.

1

#20 opened 3 months ago by

drmcbride

Kimi K2 Great model, but performance vs. resource tradeoff?

2

#19 opened 3 months ago by

Hussain2050

How about hyper-fragmented sparse MoE ?

🔥 1

1

#18 opened 3 months ago by

0x76F5ee

Add notebook examples for structured outputs and function calling

2

#17 opened 3 months ago by

burtenshaw

Question

2

#16 opened 3 months ago by

PrincelyEndeavor

Adjust number of reserved tokens to match the model

#15 opened 3 months ago by

dzhulgakov

Run 1T-param on A100/H100(80G)x8 using FP4

🚀 🔥 5

7

#9 opened 3 months ago by

ghostplant

Any plan to release a Vision enabled version with the same or near the same base and instruct model?

❤️ 🤗 2

8

#7 opened 3 months ago by

drmcbride

Synthetic data generation

🔥 1

4

#6 opened 3 months ago by

ahycl

Multimodality

3

#5 opened 3 months ago by

Dampfinchen

Thorough Testing Video of Kimi K2 - Step by Step

❤️ 15

1

#4 opened 3 months ago by

fahdmirzac

Can you provide Machine Specs

11

#2 opened 3 months ago by

kingabzpro