Views
No views yet
TQ3_4S is a 3.5-bit Walsh-Hadamard-transform weight format with four per-8 scales per 32-weight block.TQ3_4S GGUF quantization of Jackrong/Qwopus3.5-27B-v3, which is itself derived from the Qwen3.5-27B family.Jackrong/Qwopus3.5-27B-v3Qwen/Qwen3.5-27BQwopus3.5-27B-v3-f16.gguf1./build/bin/llama-quantize \
2 /path/to/Qwopus3.5-27B-v3-f16.gguf \
3 /path/to/Qwopus3.5-27B-v3-TQ3_4S.gguf \
4 TQ3_4S \
5 8wiki.test.raw, c=2048:Final PPL = 6.3433 +/- 0.03999Median chunk PPL = 6.1953llama.cpp-tq3 main:62eb27dceWrite ONLY the word ok.ok1./build/bin/llama-server \
2 -m /path/to/Qwopus3.5-27B-v3-TQ3_4S.gguf \
3 -a qwopus35-27b-v3-tq3_4s \
4 --host 127.0.0.1 --port 8080 \
5 -ngl 99 -c 4096 -np 1 \
6 -ctk q8_0 -ctv q8_0 -fa on \
7 --no-warmup --jinja \
8 --reasoning off --reasoning-budget 0 --reasoning-format deepseekTQ3_4S runtime is provided by:
turbo-tan/llama.cpp-tq3