Jan Rosecky
honza-rosecky
AI & ML interests
None yet
Recent Activity
updated a dataset 8 days ago
thinletter/longctx-agent-bench published a dataset 8 days ago
thinletter/longctx-agent-bench updated a Space 9 days ago
thinletter/READMEOrganizations
Prompt-K/V variants added: four files calibrated with the query prompt's keys and values in full precision
#5 opened about 1 month ago
by 
honza-rosecky
First Czech vector-quantised client: 237 MiB, 96.8 % of fp32 nDCG@10 on WebFAQ-cs, runs in the browser
#1 opened about 1 month ago
by honza-rosecky
The runtime is now open source: quantiser, .vqw compiler and WebGPU runtime (Apache-2.0)
#4 opened about 1 month ago
by honza-rosecky
The query prompt's K/V in full precision: +0.01–0.04 nDCG@10 at 1.6–1.8 bits per weight for 2 MB of data (measured, not yet in the files)
2
#3 opened about 1 month ago
by honza-rosecky
Vector-quantised harrier-0.6b query clients at 1.8–2.1 bits per weight (105–127 MiB) — summary and links
#2 opened about 1 month ago
by honza-rosecky
harrier-0.6b query clients: 192–235 MiB, 94–99 % of fp32 on the unchanged index — summary and links
#1 opened about 1 month ago
by honza-rosecky
bge-m3 query-side clients at 3–4 bits, calibrated on Czech text — summary and links
#2 opened about 1 month ago
by honza-rosecky
Query-side clients for Qwen3-Embedding-0.6B: keep your index, shrink the query encoder — summary and links
#3 opened about 1 month ago
by honza-rosecky
Which objective should a post-training quantizer optimize? — summary and links
#2 opened about 1 month ago
by honza-rosecky
links: moved to thinletter
1
#1 opened about 1 month ago
by honza-rosecky
links: moved to thinletter
1
#1 opened about 1 month ago
by honza-rosecky
model card: note on the file badge
1
#1 opened about 1 month ago
by honza-rosecky
model card: links to the thinletter organisation (+ measured English Q5_K_M note)
1
#1 opened about 1 month ago
by honza-rosecky
Q5_K_M + q4_0 English imatrix (385 MiB) and updated model card
2
#2 opened about 1 month ago
by honza-rosecky
model card: links to the thinletter organisation (+ measured English Q5_K_M note)
1
#1 opened about 1 month ago
by honza-rosecky
Q5_K_M + q4_0 English imatrix (385 MiB) and updated model card
2
#2 opened about 1 month ago
by honza-rosecky
model card: note on the file badge
1
#1 opened about 1 month ago
by honza-rosecky
model card: links to the thinletter organisation (+ measured English Q5_K_M note)
1
#1 opened about 1 month ago
by honza-rosecky
model card: links to the thinletter organisation (+ measured English Q5_K_M note)
1
#1 opened about 1 month ago
by honza-rosecky
links: moved to thinletter
1
#1 opened about 1 month ago
by honza-rosecky