Files
zeekay 925e62aae4 release(v1.7.43): Vulkan DSL MoE + demand-sized KV cache; hanzo-ml 0.11.55
- Demand-sized unified-memory KV cache: size to max_seq_len*max_batch_size
  (context*concurrency) bounded by the post-model/OS-reserve ceiling, instead
  of grabbing all free RAM. Fixes the ROCm/APU thrash where an ~88 GB grab on a
  128 GB unified box never returned; 30B-A3B now 88 GB -> 24 GB, elastic with
  concurrency. (hanzo-engine/src/paged_attention/mod.rs)
- Vulkan paged-attention wired (is_vulkan branch, opt-in via --paged-attn on):
  runs and eliminates the naive-Sdpa layout-copy churn, but the v1 scalar kernel
  regresses decode on this APU, so default OFF until v2 + MoE-glue fusion land.
  i64/i32 paged-attn index metadata now binds directly (backend coerces to u32).
- hanzo-quant: Vulkan/Wgpu storage arms in the distributed pipeline.
- Drop the temporary [patch.crates-io] hanzo-ml redirect; pin hanzo-ml 0.11.55.
- Manifest hygiene: version on path deps (enso, ffi, server-core, zen3-serving).
2026-07-14 17:26:58 -07:00

23 lines
821 B
TOML

[package]
name = "hanzo-engine-ffi"
version = "0.1.0"
edition = "2021"
publish = false
description = "C ABI over the Hanzo native inference engine, for cgo/FFI callers (Go EVM precompiles)."
[lib]
name = "hanzo_engine_ffi"
# cdylib for cgo dynamic linking; staticlib for fully-static embedding.
crate-type = ["cdylib", "staticlib"]
[dependencies]
hanzo-engine = { version = "1.7.41", path = "../hanzo-engine" }
hanzo-server-core = { version = "1.7.41", path = "../hanzo-server-core", default-features = false }
tokio = { workspace = true }
[features]
# Forward GPU/CPU backends to the engine. The Go bench builds with `accelerate`.
accelerate = ["hanzo-engine/accelerate", "hanzo-server-core/accelerate"]
metal = ["hanzo-engine/metal", "hanzo-server-core/metal"]
cuda = ["hanzo-engine/cuda", "hanzo-server-core/cuda"]