Alex Mittell

amittell's activity

• All
• Models
• Datasets
• Spaces
• Buckets
• Papers
• Collections
• Community
• Posts
• Upvotes
• Likes
• Articles

New activity in RedHatAI/Qwen3.8-27B-speculator.dspark about 1 month ago

Prefix-cache hits on a hybrid Qwen3.8 target return garbage with DSpark in vLLM; how was the >262K MRCR eval served?

3

#2 opened about 1 month ago by

amittell

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

Serving notes: prefix cache on the hybrid attention (1584-token blocks, lazy materialization) + thinking field gotchas

👍 3

1

#169 opened 2 months ago by

fogf

New activity in unsloth/Qwen3.8-27B-NVFP4 about 1 month ago

Any chance of Unsloth's Qwen3.8 27B NVFP4 DSPARK/DFLASH2?

13

#20 opened 3 months ago by

MaMo7x

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

vLLM + hybrid Qwen3.8 target: prefix-cache hits return garbage (fixed by vLLM PR 55601), reuse lags a turn, and notes for >262K

❤️ 1

#11 opened about 1 month ago by

amittell

New activity in microsoft/Florence-2-large-ft 10 months ago

Fix KV cache compatibility with transformers 4.50+

#42 opened 10 months ago by

amittell