Harry Chen

Harry22615356's activity

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

New activity in Baekpica/Qwen3.8-Flash-Next-Uncensored-Mixed-Quant-SSD-PLE-GGUF 3 days ago

Premature generation stop due to <|im_end|> token during model reasoning/explanation with ds4-dfm-rs

👀 1

8

#1 opened 11 days ago by

Harry22615356

New activity in Baekpica/Qwen3.8-Flash-Next-Mixed-Quant-SSD-PLE-GGUF about 1 month ago

Feedback: Vision capabilities failing – Silent drops (hallucination) via LibreChat and "Corrupt JPEG" decode errors via curl

6

#6 opened about 1 month ago by

Harry22615356

Feedback: Disk-KV is amazing, but hitting token-mismatch during ChatML reconstruction (LibreChat v0.8.7)

👍 1

2

#3 opened about 1 month ago by

Harry22615356

Incredible SSD-PLE efficiency on DGX Spark (GB10), but facing severe Prefill bottlenecks (100 t/s) & KV Cache serialization errors for Qwen4Exp

👍 1

1

#1 opened about 1 month ago by

Harry22615356