Repository navigation

refactor(rai): run Qwen3, Qwen3.5 and Gemma 4 as corelib's own models - #775

Closed

chizamd wants to merge 10 commits into

chizamd:feature/rai-model-api

Closed

refactor(rai): run Qwen3, Qwen3.5 and Gemma 4 as corelib's own models#775
chizamd wants to merge 10 commits into
ROCm:mainfrom
chizamd:feature/rai-model-api

Commits on Oct 11, 2026
- feat(rai): move to ryzenai-corelib 0.26, add Qwen3.6-35B-A3B, run Gemma 4 at 16k
  
  chizamd
  and
  cursoragent
  committed
  abcfa3b
- feat(rai): add one engine for every model corelib runs whole
  chizamd
  committed
  f4c49ab
- refactor(rai): run Qwen3.6 through the shared corelib model backend
  chizamd
  committed
  b09b432
- refactor(rai): run Qwen3 on corelib's qwen3 model
  
  chizamd
  and
  cursoragent
  committed
  5e873ad
- perf(rai): leave a seekable model's KV cache alone on clear_context
  
  chizamd
  and
  cursoragent
  committed
  bda670d
- refactor(rai): run Qwen3.5 on corelib's qwen35 model
  
  chizamd
  and
  cursoragent
  committed
  cce6233
- refactor(rai): run Gemma 4 on corelib's gemma4 model
  
  chizamd
  and
  cursoragent
  committed
  1a8190a
- refactor(rai): drop the GGUF reader and weight cache no engine uses
  
  chizamd
  and
  cursoragent
  committed
  9bcc027
- refactor(rai): bind only the corelib entry points FastFlowLM calls
  
  chizamd
  and
  cursoragent
  committed
  866aa2b
- refactor(rai): tidy after the move to corelib's model API
  
  chizamd
  and
  cursoragent
  committed
  1a75461