DATA

1 Followers · 0 Following

• Joined on 2026-02-13

Repositories

1

Projects Packages Public Activity Starred Repositories

DATA pushed to main at DATA/llama_qwen3.6_A3B 2026-07-25 16:04:46 +02:00

3ff4a5646f Move runner port 18008 -> 18010

DATA pushed to main at DATA/llama_qwen3.6_A3B 2026-07-25 15:32:53 +02:00

e2157e9226 Ship reasoning-budget 8192 as the compromise

DATA pushed to main at DATA/llama_qwen3.6_A3B 2026-07-25 15:22:50 +02:00

9daabf3223 Set reasoning-budget 16384; measurement shows it buys nothing here

DATA pushed to main at DATA/llama_qwen3.6_A3B 2026-07-25 15:03:51 +02:00

09230bc499 Measure --reasoning-budget: a real cap, unlike --predict

DATA pushed to main at DATA/llama_qwen3.6_A3B 2026-07-25 14:53:04 +02:00

e41dc2ba6f Raise --predict default to 32768 and document that it is not a cap

DATA pushed to main at DATA/llama_qwen3.6_A3B 2026-07-25 14:25:46 +02:00

e9a091c35f Benchmark A3B against dense Qwen3.6-27B on the same box

DATA pushed to main at DATA/llama_qwen3.6_A3B 2026-07-25 13:54:09 +02:00

f628eedbaa Measure speculative decoding: all ngram modes lose to plain decoding

DATA pushed to main at DATA/llama_qwen3.6_A3B 2026-07-25 13:49:35 +02:00

2c9cd04e96 llama.cpp CUDA runner for Qwen3.6-35B-A3B at 128k context

DATA created branch main in DATA/llama_qwen3.6_A3B 2026-07-25 13:49:35 +02:00

DATA created repository DATA/llama_qwen3.6_A3B 2026-07-25 13:35:37 +02:00