nisten
/

qwenv2-7b-inst-imatrix-gguf

Model card Files Files and versions Community

qwenv2-7b-inst-imatrix-gguf

1 contributor

History: 19 commits

nisten's picture

best speed/perplexity quant for mobile devices with 8bit acceleration

d2b704a verified about 1 month ago

.gitattributes

3.07 kB

best speed/perplexity quant for mobile devices with 8bit acceleration about 1 month ago
8bitimatrix.dat

4.54 MB
LFS

calculated imatrix in 8bit, was jsut as good as f16 imatrix about 1 month ago
README.md

1.55 kB

Update README.md about 1 month ago
qwen7bv2inst_Iq4xs_output6k.gguf

4.22 GB
LFS

Standard IQ4XS quantizing down from full bf16 ( not from f16) about 1 month ago
qwen7bv2inst_iq4xs_embedding8_outputq8.gguf
4.64 GB
LFS

great quant if your chip has 8bit acceleration, slightly better than 4k embedding about 1 month ago
qwen7bv2inst_iq4xs_output8bit.gguf
4.35 GB
LFS

best speed/perplexity quant for mobile devices with 8bit acceleration about 1 month ago
qwen7bv2inst_q4km_embeddingf16_outputf16.gguf
6.11 GB
LFS

Good speed reference quant for older CPUs, however not much improvement from f16 embedding about 1 month ago
qwen7bv2inst_q4km_output8bit.gguf
4.82 GB
LFS

very good quant for speed/perplexity, embedding is at q4k about 1 month ago
qwen7bv2instruct_bf16.gguf
15.2 GB
LFS

Rename qwen7bf16.gguf to qwen7bv2instruct_bf16.gguf about 1 month ago
qwen7bv2instruct_q5km.gguf
5.58 GB
LFS

standard q5km conversions with 8bit output for reference. about 1 month ago
qwen7bv2instruct_q8.gguf
8.1 GB
LFS

Best q8 conversion down from bf16 with slightly better perplexity than f16 based quants about 1 month ago