👋 Open to Work

Kaloyan Nikolov

sleepyeldrazi

25 1 6

https://git.kokoham.com/sleepy

AI & ML interests

ML engineer working on LLM inference and training. Finalizing my M.Sc. Computer Science @ RWTH Aachen. Currently focusing on DGX Spark models/frameworks. Contact: konik98@gmail.com

Recent Activity

new activity 1 day ago

ISTA-DASLab/Qwen3.8-Flash-Next-GSQ-RCO-GGUF:Are the ngrams quanted to low quants as well or does it have q8 ngrams?

new activity 1 day ago

ISTA-DASLab/Qwen3.8-Flash-Next-GSQ-RCO-GGUF:Great quantization method, but model is too wild to use anywhere

new activity 4 days ago

sleepyeldrazi/Qwen3.8-Flash-Next-GSQ-RCO-Q8_0-PLE-GGUF:Does this increase the quality too or just changes the data format?

View all activity

Organizations

None yet

New activity in ISTA-DASLab/Qwen3.8-Flash-Next-GSQ-RCO-GGUF 1 day ago

Are the ngrams quanted to low quants as well or does it have q8 ngrams?

👍 4

#34 opened 3 days ago by

clover-supply

Great quantization method, but model is too wild to use anywhere

#37 opened 1 day ago by

krustik

New activity in sleepyeldrazi/Qwen3.8-Flash-Next-GSQ-RCO-Q8_0-PLE-GGUF 4 days ago

Does this increase the quality too or just changes the data format?

#1 opened 4 days ago by

GanjaSmoker

New activity in ISTA-DASLab/Qwen3.8-Flash-Next-GSQ-RCO-GGUF 4 days ago

higher precision PLE weights?

👍 3

#28 opened 6 days ago by

aba-aba

New activity in ISTA-DASLab/Qwen3.8-Flash-Next-GSQ-RCO-GGUF 13 days ago

Quick Comparison: ISTA-DASLab GSQ-RCO-IQ3_XXS vs unsloth UD-IQ4_XS on 4070Ti Super 16G + 64G RAM

#18 opened 15 days ago by

BipedalBit

New activity in sleepyeldrazi/Ornith-1.5-REAP-20B-A3B-GGUF 28 days ago

List of pruned experts

#1 opened about 1 month ago by

kdrapel

New activity in unsloth/Qwen3.8-Flash-Next-GGUF about 1 month ago

Early report running it on a DGX Spark, 4 concurrent @ 256k bf16 with SSD offload

👍 7

#31 opened about 1 month ago by

sleepyeldrazi

New activity in quimmedes/Qwen3.8-Flash-Next-MTP-GGUF about 1 month ago

BF16 Missing currently

#1 opened about 1 month ago by

sleepyeldrazi

New activity in unsloth/Qwen3.8-Flash-Next-GGUF about 1 month ago

NVME vs Optane?

#29 opened about 1 month ago by

AndyFaceHugger

Is it possible to offload n-gram to an NVMe SSD?

👀🚀 31

#11 opened about 1 month ago by

lingyezhixing

New activity in z-lab/Qwen3.8-27B-DFlash2 about 1 month ago

Yet to see any gains from DFlash

👍 11

#3 opened about 1 month ago by

Dnonmi

New activity in incoai/Qwen3.8-27B-DFlash2 about 1 month ago

FP8

#2 opened about 2 months ago by

vcerny

New activity in poolside/Laguna-S-2.1 about 1 month ago

BenchMAXXXED

➕ 2

#28 opened 2 months ago by

TawusGames

New activity in Qwen/Qwen3.8-27B about 2 months ago

Exploring ThinkingCap fine-tunes for Qwen 3.8: Is it still necessary with native `reasoning_effort`?

👍 4

#106 opened about 2 months ago by

icognito

How can I run this well on an RTX 5070 ti?

#36 opened about 2 months ago by

theonerm2

New activity in Qwen/Qwen3.8-2.4T-A95B about 2 months ago

Text-only? Plans for multimodal release?

👀😔 5

#7 opened about 2 months ago by

sleepyeldrazi

New activity in poolside/Laguna-S-2.1-NVFP4 2 months ago

Reproducible Laguna S 2.1 NVFP4 runaway reasoning: output budget scales 2K → 16K with 12/12 length exhaustion and no final answer

👀👍 8

#16 opened 2 months ago by

darkmatter2222

Still looping after update - details inside

👀 2

#10 opened 2 months ago by

cbert33

New activity in poolside/Laguna-S-2.1 2 months ago

Model is decent, its faster than Qwen and pretty close on real world tool calls

❤️👀 3

#22 opened 2 months ago by

Jakaryun

New activity in prism-ml/Ternary-Bonsai-27B-gguf 2 months ago

MOE or changing to A3B

#43 opened 2 months ago by

Tom-Neverwinter

Kaloyan Nikolov

AI & ML interests

Recent Activity

Organizations

sleepyeldrazi's activity

Are the ngrams quanted to low quants as well or does it have q8 ngrams?

Great quantization method, but model is too wild to use anywhere

Does this increase the quality too or just changes the data format?

higher precision PLE weights?

Quick Comparison: ISTA-DASLab GSQ-RCO-IQ3_XXS vs unsloth UD-IQ4_XS on 4070Ti Super 16G + 64G RAM

List of pruned experts

Early report running it on a DGX Spark, 4 concurrent @ 256k bf16 with SSD offload

BF16 Missing currently

NVME vs Optane?

Is it possible to offload n-gram to an NVMe SSD?

Yet to see any gains from DFlash

FP8

BenchMAXXXED

Exploring ThinkingCap fine-tunes for Qwen 3.8: Is it still necessary with native `reasoning_effort`?

How can I run this well on an RTX 5070 ti?

Text-only? Plans for multimodal release?

Reproducible Laguna S 2.1 NVFP4 runaway reasoning: output budget scales 2K → 16K with 12/12 length exhaustion and no final answer

Still looping after update - details inside

Model is decent, its faster than Qwen and pretty close on real world tool calls

MOE or changing to A3B