Jun 6

1. 18:09

   fix: use eager attention for MiniCPM4.1-8B to prevent SDPA crash, fix Nemotron Parse processor format

   Kunalv
   32bc132
2. 17:55

   fix: switch Core Router to AutoModelForCausalLM for generate support, fix Nemotron Parse inference with AutoProcessor

   Kunalv
   bb9bc8a
3. 17:50

   fix: block flash_attn at runtime with sys.modules=None to prevent FA2 on HF ZeroGPU where flash_attn is pre-installed

   Kunalv
   9fc5c52
4. 17:44

   fix: resolve infinite recursion in flash_attn patch by capturing original get_imports before patching

   Kunalv
   70f54cc
5. 17:40

   fix: replace flash_attn mock with get_imports patch to properly disable FA2 for MiniCPM4.1-8B and Nemotron Parse

   Kunalv
   d006ae4
6. 17:19

   fix: correct vision model dtype, Nemotron Parse model ID, and patch missing is_torch_fx_available

   Kunalv
   4e3a057
7. 16:43

   feat: add Parallel Vision Engine — MiniCPM-V and Nemotron Parse fire async with confidence-score arbiter

   Kunalv
   96f4123