mirror of
https://github.com/hanzoai/ml.git
synced 2026-11-19 23:50:51 +00:00
CUDA kernels index with signed 32-bit integers, so a tensor of more than i32::MAX elements miscomputes with no error. Allocation, host copies, zeros, rand, broadcast_as, reshape and to_device onto CUDA now return CudaError::TooManyElements instead. Shape products use checked multiplication so a wrapping shape cannot pass as empty. Claude-Session: https://claude.ai/code/session_015XYKLix9PSyn6Jw2HWKprk
47 lines
1.5 KiB
TOML
47 lines
1.5 KiB
TOML
[package]
|
|
name = "hanzo-transformers"
|
|
version = "0.11.20"
|
|
edition = "2021"
|
|
description = "Multi-backend tensor & ML framework for Rust — part of the Hanzo ML stack."
|
|
repository = "https://github.com/hanzoai/ml"
|
|
keywords = ["blas", "tensor", "machine-learning"]
|
|
categories = ["science"]
|
|
license = "MIT OR Apache-2.0"
|
|
readme = "README.md"
|
|
|
|
[dependencies]
|
|
accelerate-src = { version = "0.3.2", optional = true }
|
|
byteorder = "1.4.3"
|
|
hanzo-ml = { path = "../hanzo-ml", package = "hanzo-ml", version = "0.11.115" }
|
|
hanzo-nn = { path = "../hanzo-nn", version = "0.11.50" }
|
|
fancy-regex = "0.19.0"
|
|
intel-mkl-src = { version = "0.8.1", features = ["mkl-static-lp64-iomp"], optional = true }
|
|
num-traits = "0.2.15"
|
|
rand = "0.9.0"
|
|
rayon = "1.7.0"
|
|
serde = { version = "1.0.171", features = ["derive"] }
|
|
serde_json = "1.0.99"
|
|
serde_plain = "1.0.2"
|
|
tracing = "0.1.37"
|
|
|
|
[features]
|
|
default = []
|
|
accelerate = ["dep:accelerate-src", "hanzo-ml/accelerate", "hanzo-nn/accelerate"]
|
|
cuda = ["hanzo-ml/cuda", "hanzo-nn/cuda"]
|
|
cudnn = ["hanzo-ml/cudnn", "hanzo-nn/cudnn"]
|
|
flash-attn = ["cuda"]
|
|
mkl = ["dep:intel-mkl-src", "hanzo-ml/mkl", "hanzo-nn/mkl"]
|
|
metal = ["hanzo-ml/metal", "hanzo-nn/metal"]
|
|
rocm = ["hanzo-ml/rocm", "hanzo-nn/rocm"]
|
|
vulkan = ["hanzo-ml/vulkan", "hanzo-nn/vulkan"]
|
|
metal-debug-labels = ["metal", "hanzo-ml/metal-debug-labels", "hanzo-nn/metal-debug-labels"]
|
|
|
|
[lints.clippy]
|
|
explicit_counter_loop = "allow"
|
|
unnecessary_unwrap = "allow"
|
|
manual_checked_ops = "allow"
|
|
needless_borrow = "allow"
|
|
large_const_arrays = "allow"
|
|
needless_range_loop = "allow"
|
|
too_many_arguments = "allow"
|