model : 범용 DeepSeek Sparse Attention (DSA) 구현을 통한 DeepseekV32ForCausalLM 지원 ([#2

model : 범용 DeepSeek Sparse Attention (DSA) 구현을 통한 DeepseekV32ForCausalLM 지원 (#23346)

• llama : DeepSeek V3.2 모델 제품군 지원 (DSA lightning indexer 포함)

• convert : DeepseekV32ForCausalLM 아키텍처 처리

• ggml : f16 GGML_OP_FILL 지원

• memory : llama_kv_cache 생성자에서 hparams 인자 분리

• memory : llama_kv_cache_dsa 메모리 추가 (KV cache + lightning indexer cache)

• llama : LLM_ARCH_DEEPSEEK32 지원

• model : llama_model_deepseek32 구현

• model : DSA lightning indexer 구현에서 두 개의 스케일 연산을 하나로 병합

• chore : 사용되지 않는 코드 제거

• model : DeepSeek V3.2에서 NVFP4 지원

Co-authored-by: Sigbjørn Skjæret sigbjorn.skjaeret@scala.com

• memory : 리팩토링 TODO

Co-authored-by: ggerganov ggerganov@users.noreply.github.com

---

Co-authored-by: Stanisław Szymczyk sszymczy@gmail.com
Co-authored-by: Sigbjørn Skjæret sigbjorn.skjaeret@scala.com
Co-authored-by: ggerganov ggerganov@users.noreply.github.com

macOS/iOS:

• macOS Apple Silicon (arm64)
• macOS Apple Silicon (arm64, KleidiAI 활성화됨) 비활성화됨
• macOS Intel (x64)
• iOS XCFramework

Linux:

<strong>Linux:</strong>

• Ubuntu x64 (CPU)
• Ubuntu arm64 (CPU)
• Ubuntu s390x (CPU)
• Ubuntu x64 (Vulkan)
• Ubuntu arm64 (Vulkan)
• Ubuntu x64 (ROCm 7.2)
• Ubuntu x64 (OpenVINO)
• Ubuntu x64 (SYCL FP32) DISABLED

<strong>Android:</strong>

• Android arm64 (CPU)

<strong>Windows:</strong>

<strong>Windows:</strong>

• Windows x64 (CPU)
• Windows arm64 (CPU)
• Windows x64 (CUDA 12) - CUDA 12.4 DLLs
• Windows x64 (CUDA 13) - CUDA 13.3 DLLs
• Windows x64 (Vulkan)
• Windows x64 (SYCL) DISABLED
• Windows x64 (HIP)

openEuler:

• DISABLED
• openEuler x86 (310p)
• openEuler x86 (910b, ACL Graph)
• openEuler aarch64 (310p)
• openEuler aarch64 (910b, ACL Graph)

UI:

• UI