mirror of
https://github.com/hanzoai/minutes.git
synced 2026-11-02 14:16:28 +00:00
Mohammed Safvan
2177dbfbad
Refs #462 Addresses L-1, L-9 Replace the deprecated token_to_bytes usage with token_to_piece_bytes while preserving special-token decoding behavior and retrying when the initial token buffer is too small. Pin llama-cpp-2 to the locked 0.1.146 version so the helper manifest matches the resolved dependency used by the workspace.
23 lines
432 B
TOML
23 lines
432 B
TOML
[package]
|
|
name = "llama-helper"
|
|
version = "0.1.0"
|
|
edition = "2021"
|
|
|
|
[dependencies]
|
|
anyhow = "1.0"
|
|
serde = { version = "1.0", features = ["derive"] }
|
|
serde_json = "1.0"
|
|
llama-cpp-2 = "=0.1.146"
|
|
encoding_rs = "0.8"
|
|
|
|
[features]
|
|
default = []
|
|
metal = ["llama-cpp-2/metal"]
|
|
cuda = ["llama-cpp-2/cuda"]
|
|
vulkan = ["llama-cpp-2/vulkan"]
|
|
|
|
[profile.release]
|
|
codegen-units = 1
|
|
lto = true
|
|
opt-level = "s" # Optimize for size for faster loading
|