ferrum-engine 0.8.3

Model orchestration engine for Ferrum LLM inference
Documentation
# THIS FILE IS AUTOMATICALLY GENERATED BY CARGO
#
# When uploading crates to the registry Cargo will automatically
# "normalize" Cargo.toml files for maximal compatibility
# with all versions of Cargo and also rewrite `path` dependencies
# to registry (e.g., crates.io) dependencies.
#
# If you are reading this file be aware that the original Cargo.toml
# will likely look very different (and much more reasonable).
# See Cargo.toml.orig for the original contents.

[package]
edition = "2021"
name = "ferrum-engine"
version = "0.8.3"
authors = ["Ferrum Team"]
build = false
autolib = false
autobins = false
autoexamples = false
autotests = false
autobenches = false
description = "Model orchestration engine for Ferrum LLM inference"
readme = "README.md"
license = "MIT"

[features]
accelerate = [
    "candle-core/accelerate",
    "candle-nn/accelerate",
    "ferrum-models/accelerate",
]
candle-cuda-compat = [
    "cuda",
    "candle-core/cuda",
    "candle-nn/cuda",
    "ferrum-models/candle-cuda-compat",
    "ferrum-kernels/candle-cuda-compat",
]
cuda = [
    "ferrum-models/cuda",
    "ferrum-kernels/cuda",
    "ferrum-quantization/cuda",
]
default = []
marlin = ["cuda"]
metal = [
    "candle-core/metal",
    "dep:candle-metal-kernels",
    "dep:metal",
    "dep:mpsgraph",
    "dep:objc2",
    "dep:objc2-foundation",
    "dep:bytemuck",
    "dep:ndarray",
    "ferrum-models/metal",
    "ferrum-kernels/metal",
]
tensor-parallel = ["cuda"]
triton-kernels = [
    "cuda",
    "ferrum-kernels/triton-kernels",
]

[lib]
name = "ferrum_engine"
path = "src/lib.rs"

[[test]]
name = "backend_safety_test"
path = "tests/backend_safety_test.rs"

[[test]]
name = "chunked_prefill_test"
path = "tests/chunked_prefill_test.rs"

[[test]]
name = "concurrency_stress_test"
path = "tests/concurrency_stress_test.rs"

[[test]]
name = "continuous_batch_test"
path = "tests/continuous_batch_test.rs"

[[test]]
name = "correctness_test"
path = "tests/correctness_test.rs"

[[test]]
name = "latency_profile_test"
path = "tests/latency_profile_test.rs"

[[test]]
name = "paged_attention_test"
path = "tests/paged_attention_test.rs"

[[test]]
name = "prefix_cache_test"
path = "tests/prefix_cache_test.rs"

[[test]]
name = "qwen_concurrency"
path = "tests/qwen_concurrency.rs"

[[test]]
name = "regex_guided_test"
path = "tests/regex_guided_test.rs"

[[test]]
name = "spec_decode_test"
path = "tests/spec_decode_test.rs"

[[test]]
name = "tiny_stack"
path = "tests/tiny_stack.rs"

[[bench]]
name = "engine_bench"
path = "benches/engine_bench.rs"
harness = false

[dependencies.anyhow]
version = "1.0"

[dependencies.async-trait]
version = "0.1"

[dependencies.bytesize]
version = "1.3"

[dependencies.candle-core]
version = "0.9.2"

[dependencies.candle-flash-attn]
version = "0.9.2"
optional = true

[dependencies.candle-metal-kernels]
version = "0.9.2"
optional = true

[dependencies.candle-nn]
version = "0.9.2"

[dependencies.candle-transformers]
version = "0.9.2"

[dependencies.chrono]
version = "0.4"
features = ["serde"]

[dependencies.crossbeam-channel]
version = "0.5"

[dependencies.crossbeam-utils]
version = "0.8"

[dependencies.ferrum-bench-core]
version = "0.8.3"

[dependencies.ferrum-interfaces]
version = "0.8.3"

[dependencies.ferrum-kernels]
version = "0.8.3"

[dependencies.ferrum-kv]
version = "0.8.3"

[dependencies.ferrum-models]
version = "0.8.3"

[dependencies.ferrum-quantization]
version = "0.8.3"

[dependencies.ferrum-sampler]
version = "0.8.3"

[dependencies.ferrum-scheduler]
version = "0.8.3"

[dependencies.ferrum-tokenizer]
version = "0.8.3"

[dependencies.ferrum-types]
version = "0.8.3"

[dependencies.futures]
version = "0.3"

[dependencies.half]
version = "2.5.0"
features = [
    "num-traits",
    "use-intrinsics",
    "rand_distr",
]

[dependencies.hf-hub]
version = "0.4.1"

[dependencies.metrics]
version = "0.22"

[dependencies.ndarray]
version = "0.16.1"
optional = true

[dependencies.once_cell]
version = "1.19"

[dependencies.parking_lot]
version = "0.12"

[dependencies.pin-project]
version = "1.1"

[dependencies.rand]
version = "0.9.0"

[dependencies.rand_distr]
version = "0.4.3"

[dependencies.reqwest]
version = "0.12"
features = [
    "rustls-tls",
    "socks",
    "stream",
    "json",
]
default-features = false

[dependencies.safetensors]
version = "0.4"

[dependencies.serde]
version = "1.0"
features = ["derive"]

[dependencies.serde_json]
version = "1.0"

[dependencies.sha2]
version = "0.10"

[dependencies.smallvec]
version = "1.11"

[dependencies.thiserror]
version = "1.0"

[dependencies.tokenizers]
version = "0.21"
features = ["onig"]

[dependencies.tokio]
version = "1.35"
features = [
    "full",
    "rt-multi-thread",
    "sync",
    "time",
]

[dependencies.tokio-stream]
version = "0.1"

[dependencies.tracing]
version = "0.1"

[dependencies.uuid]
version = "1.6"
features = [
    "v4",
    "serde",
]

[dev-dependencies.criterion]
version = "0.5"

[dev-dependencies.ferrum-models]
version = "0.8.3"
features = ["test-support"]

[dev-dependencies.ferrum-testkit]
version = "0.8.3"

[dev-dependencies.futures]
version = "0.3"

[dev-dependencies.tokio-test]
version = "0.4"

[target.'cfg(target_os = "ios")'.dependencies.bytemuck]
version = "1.23.0"
optional = true

[target.'cfg(target_os = "ios")'.dependencies.metal]
version = "0.27.0"
optional = true

[target.'cfg(target_os = "ios")'.dependencies.mpsgraph]
version = "0.2.0"
optional = true

[target.'cfg(target_os = "ios")'.dependencies.objc2]
version = "0.6.1"
optional = true

[target.'cfg(target_os = "ios")'.dependencies.objc2-foundation]
version = "0.3.1"
optional = true

[target.'cfg(target_os = "macos")'.dependencies.bytemuck]
version = "1.23.0"
optional = true

[target.'cfg(target_os = "macos")'.dependencies.metal]
version = "0.27.0"
optional = true

[target.'cfg(target_os = "macos")'.dependencies.mpsgraph]
version = "0.2.0"
optional = true

[target.'cfg(target_os = "macos")'.dependencies.objc2]
version = "0.6.1"
optional = true

[target.'cfg(target_os = "macos")'.dependencies.objc2-foundation]
version = "0.3.1"
optional = true