Mirrors
/
hf_text-generation-inference
mirror of https://github.com/huggingface/text-generation-inference.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

euff

This commit is contained in:
Mohit Sharma 2024-09-18 12:03:52 +00:00
parent 4ba9210f91 ce85efa968
commit e6d07a6d34
31 changed files with 4275 additions and 579 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
.github/workflows/build.yaml vendored
View File

@ -45,7 +45,7 @@ jobs:
export dockerfile="Dockerfile"
export label_extension=""
export docker_devices=""
export runs_on="aws-g6-12xlarge-plus-priv"
export runs_on="aws-g6-12xl-plus-priv-cache"
export platform=""
;;
rocm)

2
.github/workflows/nix_tests.yaml vendored
View File

@ -38,4 +38,4 @@ jobs:
env:
HF_TOKEN: ${{ secrets.HF_TOKEN }}
- name: Rust tests.
run: nix develop .#test --command cargo test
run: nix build .#checks.$(nix eval --impure --raw --expr 'builtins.currentSystem').rust -L

419
Cargo.lock generated
View File

File diff suppressed because it is too large Load Diff

2
Dockerfile
View File

@ -258,7 +258,7 @@ COPY server/Makefile server/Makefile
RUN cd server && \
make gen-server && \
pip install -r requirements_cuda.txt && \
pip install ".[bnb, accelerate, marlin, quantize, peft, outlines]" --no-cache-dir && \
pip install ".[bnb, accelerate, marlin, moe, quantize, peft, outlines]" --no-cache-dir && \
pip install nvidia-nccl-cu12==2.22.3
ENV LD_PRELOAD=/opt/conda/lib/python3.11/site-packages/nvidia/nccl/lib/libnccl.so.2

63
Dockerfile_intel
View File

@ -44,9 +44,35 @@ RUN cargo build --profile release-opt
# Text Generation Inference base image for Intel
FROM intel/intel-extension-for-pytorch:2.1.30-xpu AS xpu
FROM intel/intel-extension-for-pytorch:2.3.110-xpu AS xpu
USER root
ARG MAMBA_VERSION=23.1.0-1
ARG PYTHON_VERSION='3.11.10'
# Automatically set by buildx
ARG TARGETPLATFORM
ENV PATH /opt/conda/bin:$PATH
# TGI seem to require libssl.so.1.1 instead of libssl.so.3 so we can't use ubuntu 22.04. Ubuntu 20.04 has python==3.8, and TGI requires python>=3.9, hence the need for miniconda.
# Install mamba
# translating Docker's TARGETPLATFORM into mamba arches
RUN case ${TARGETPLATFORM} in \
"linux/arm64") MAMBA_ARCH=aarch64 ;; \
*) MAMBA_ARCH=x86_64 ;; \
esac && \
curl -fsSL -v -o ~/mambaforge.sh -O "https://github.com/conda-forge/miniforge/releases/download/${MAMBA_VERSION}/Mambaforge-${MAMBA_VERSION}-Linux-${MAMBA_ARCH}.sh"
RUN chmod +x ~/mambaforge.sh && \
bash ~/mambaforge.sh -b -p /opt/conda && \
rm ~/mambaforge.sh
RUN case ${TARGETPLATFORM} in \
"linux/arm64") exit 1 ;; \
*) /opt/conda/bin/conda update -y conda && \
/opt/conda/bin/conda install -y "python=${PYTHON_VERSION}" ;; \
esac && \
/opt/conda/bin/conda clean -ya
# libssl.so.1.1 is not installed on Ubuntu 22.04 by default, install it
RUN wget http://nz2.archive.ubuntu.com/ubuntu/pool/main/o/openssl/libssl1.1_1.1.1f-1ubuntu2_amd64.deb && \
dpkg -i ./libssl1.1_1.1.1f-1ubuntu2_amd64.deb
@ -56,7 +82,7 @@ RUN wget -qO - https://repositories.intel.com/gpu/intel-graphics.key | gpg --dea
RUN wget -O- https://apt.repos.intel.com/intel-gpg-keys/GPG-PUB-KEY-INTEL-SW-PRODUCTS.PUB \
| gpg --dearmor | tee /usr/share/keyrings/oneapi-archive-keyring.gpg > /dev/null && echo "deb [signed-by=/usr/share/keyrings/oneapi-archive-keyring.gpg] https://apt.repos.intel.com/oneapi all main" | tee /etc/apt/sources.list.d/oneAPI.list
RUN apt-get update && apt install -y intel-basekit xpu-smi cmake python3-dev ninja-build pciutils
RUN apt-get update && DEBIAN_FRONTEND=noninteractive apt install -y intel-basekit xpu-smi cmake ninja-build pciutils
# Text Generation Inference base env
ENV HF_HOME=/data \
@ -65,9 +91,7 @@ ENV HF_HOME=/data \
WORKDIR /usr/src
RUN wget https://intel-extension-for-pytorch.s3.amazonaws.com/ipex_dev/xpu/torch-2.1.0.post1%2Bcxx11.abi-cp310-cp310-linux_x86_64.whl && pip install torch-2.1.0.post1+cxx11.abi-cp310-cp310-linux_x86_64.whl
RUN pip install https://github.com/intel/intel-xpu-backend-for-triton/releases/download/v2.1.0/triton-2.1.0-cp310-cp310-manylinux_2_17_x86_64.manylinux2014_x86_64.whl
RUN git clone https://github.com/intel/intel-extension-for-pytorch && cd intel-extension-for-pytorch && git checkout -b distributed origin/dev/distributed
RUN pip install torch==2.3.1+cxx11.abi torchvision==0.18.1+cxx11.abi torchaudio==2.3.1+cxx11.abi intel-extension-for-pytorch==2.3.110+xpu oneccl_bind_pt==2.3.100+xpu --extra-index-url https://pytorch-extension.intel.com/release-whl/stable/xpu/us/ --no-cache-dir
# Install server
COPY proto proto
@ -82,14 +106,12 @@ ENV CCL_ROOT=/opt/intel/oneapi/ccl/latest
ENV I_MPI_ROOT=/opt/intel/oneapi/mpi/latest
ENV FI_PROVIDER_PATH=/opt/intel/oneapi/mpi/latest/opt/mpi/libfabric/lib/prov:/usr/lib/x86_64-linux-gnu/libfabric
ENV LIBRARY_PATH=/opt/intel/oneapi/mpi/latest/lib:/opt/intel/oneapi/ccl/latest/lib/:/opt/intel/oneapi/mkl/latest/lib/:/opt/intel/oneapi/compiler/latest/lib
ENV LD_LIBRARY_PATH=/opt/intel/oneapi/ccl/latest/lib/:/opt/intel/oneapi/mpi/latest/opt/mpi/libfabric/lib:/opt/intel/oneapi/mpi/latest/lib:/opt/intel/oneapi/mkl/latest/lib:/opt/intel/oneapi/compiler/latest/opt/compiler/lib:/opt/intel/oneapi/compiler/latest/lib:/opt/intel/oneapi/lib:/opt/intel/oneapi/lib/intel64:
ENV PATH=/opt/intel/oneapi/mpi/latest/opt/mpi/libfabric/bin:/opt/intel/oneapi/mpi/latest/bin:/opt/intel/oneapi/mpi/latest/opt/mpi/libfabric/bin:/opt/intel/oneapi/mkl/latest/bin/:/opt/intel/oneapi/compiler/latest/bin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin
ENV LD_LIBRARY_PATH=/opt/intel/oneapi/ccl/latest/lib/:/opt/intel/oneapi/mpi/latest/opt/mpi/libfabric/lib:/opt/intel/oneapi/mpi/latest/lib:/opt/intel/oneapi/mkl/latest/lib:/opt/intel/oneapi/compiler/latest/opt/compiler/lib:/opt/intel/oneapi/compiler/latest/lib:/opt/intel/oneapi/lib:/opt/intel/oneapi/lib/intel64:/opt/conda/lib
ENV PATH=/opt/conda/bin:/opt/intel/oneapi/mpi/latest/opt/mpi/libfabric/bin:/opt/intel/oneapi/mpi/latest/bin:/opt/intel/oneapi/mpi/latest/opt/mpi/libfabric/bin:/opt/intel/oneapi/mkl/latest/bin/:/opt/intel/oneapi/compiler/latest/bin:/usr/local/sbin:/usr/local/bin:/usr/sbin:/usr/bin:/sbin:/bin
ENV CCL_ZE_IPC_EXCHANGE=sockets
ENV CMAKE_PREFIX_PATH=/opt/intel/oneapi/mkl/latest/lib/cmake:/opt/intel/oneapi/compiler/latest
ENV CPATH=/opt/intel/oneapi/mpi/latest/include:/opt/intel/oneapi/ccl/latest/include:/opt/intel/oneapi/mkl/latest/include
RUN pip uninstall -y intel-extension-for-pytorch && cd intel-extension-for-pytorch && git submodule update --init --recursive && USE_AOT_DEVLIST='pvc' BUILD_SEPARATE_OPS=OFF BUILD_WITH_CPU=OFF USE_XETLA=ON python setup.py install && rm -rf /usr/src/intel-extension-for-pytorch
# Install benchmarker
COPY --from=builder /usr/src/target/release-opt/text-generation-benchmark /usr/local/bin/text-generation-benchmark
# Install router
@ -133,12 +155,19 @@ RUN chmod +x ~/mambaforge.sh && \
bash ~/mambaforge.sh -b -p /opt/conda && \
rm ~/mambaforge.sh
RUN case ${TARGETPLATFORM} in \
"linux/arm64") exit 1 ;; \
*) /opt/conda/bin/conda update -y conda && \
/opt/conda/bin/conda install -y "python=${PYTHON_VERSION}" ;; \
esac && \
/opt/conda/bin/conda clean -ya
RUN conda install -c conda-forge gperftools mkl
RUN pip install https://download.pytorch.org/whl/nightly/cpu/torch-2.4.0.dev20240612%2Bcpu-cp310-cp310-linux_x86_64.whl
RUN pip install https://download.pytorch.org/whl/nightly/cpu/torchvision-0.19.0.dev20240612%2Bcpu-cp310-cp310-linux_x86_64.whl
RUN pip install https://download.pytorch.org/whl/nightly/cpu/torchaudio-2.4.0.dev20240612%2Bcpu-cp310-cp310-linux_x86_64.whl
RUN pip install triton numa
RUN pip install https://download.pytorch.org/whl/nightly/cpu/torch-2.4.0.dev20240612%2Bcpu-cp311-cp311-linux_x86_64.whl
RUN pip install https://download.pytorch.org/whl/nightly/cpu/torchvision-0.19.0.dev20240612%2Bcpu-cp311-cp311-linux_x86_64.whl
RUN pip install https://download.pytorch.org/whl/nightly/cpu/torchaudio-2.4.0.dev20240612%2Bcpu-cp311-cp311-linux_x86_64.whl
RUN pip install triton py-libnuma
WORKDIR /usr/src
@ -151,10 +180,10 @@ RUN cd intel-extension-for-pytorch && git submodule sync && git submodule update
RUN cd torch-ccl && git submodule sync && git submodule update --init --recursive && pip install .
ENV LD_PRELOAD=/opt/conda/lib/libtcmalloc.so
ENV CCL_ROOT=/opt/conda/lib/python3.10/site-packages/oneccl_bindings_for_pytorch
ENV I_MPI_ROOT=/opt/conda/lib/python3.10/site-packages/oneccl_bindings_for_pytorch
ENV FI_PROVIDER_PATH=/opt/conda/lib/python3.10/site-packages/oneccl_bindings_for_pytorch/opt/mpi/libfabric/lib/prov:/usr/lib64/libfabric
ENV LD_LIBRARY_PATH=/opt/conda/lib/python3.10/site-packages/oneccl_bindings_for_pytorch/opt/mpi/libfabric/lib:/opt/conda/lib/python3.10/site-packages/oneccl_bindings_for_pytorch/lib
ENV CCL_ROOT=/opt/conda/lib/python3.11/site-packages/oneccl_bindings_for_pytorch
ENV I_MPI_ROOT=/opt/conda/lib/python3.11/site-packages/oneccl_bindings_for_pytorch
ENV FI_PROVIDER_PATH=/opt/conda/lib/python3.11/site-packages/oneccl_bindings_for_pytorch/opt/mpi/libfabric/lib/prov:/usr/lib64/libfabric
ENV LD_LIBRARY_PATH=/opt/conda/lib/python3.11/site-packages/oneccl_bindings_for_pytorch/opt/mpi/libfabric/lib:/opt/conda/lib/python3.11/site-packages/oneccl_bindings_for_pytorch/lib
ENV LD_LIBRARY_PATH="$LD_LIBRARY_PATH:/opt/conda/lib/"
# Install server

8
backends/v3/src/queue.rs
View File

@ -364,7 +364,7 @@ impl State {
// Add it back to the front
tracing::debug!("Over budget: not enough free blocks");
self.entries.push_front((id, entry));
break;
continue;
}
Some(block_allocation) => {
tracing::debug!("Allocation: {block_allocation:?}");
@ -436,6 +436,12 @@ impl State {
batch_entries.insert(id, entry);
}
// Empty batch
if batch_requests.is_empty() {
tracing::debug!("Filterered out all entries");
return None;
}
// Final batch size
let size = batch_requests.len() as u32;
next_batch_span.record("batch_size", size);

76
backends/v3/src/radix.rs
View File

@ -1,10 +1,22 @@
use crate::block_allocator::{Allocator, BlockAllocation};
use slotmap::{DefaultKey, SlotMap};
use std::hash::{Hash, Hasher};
use std::{
collections::{BTreeSet, HashMap},
sync::Arc,
};
fn hash(slice: &[u32]) -> u64 {
assert!(!slice.is_empty());
if slice.len() == 1 {
slice[0] as u64
} else {
let mut s = std::hash::DefaultHasher::new();
slice.hash(&mut s);
s.finish()
}
}
pub struct RadixAllocator {
allocation_id: u64,
@ -44,6 +56,10 @@ impl RadixAllocator {
// the free list if we cannot allocate enough blocks. This is only
// temporary, the trie needs to be able to report whether it can
// allocate the requested amount. Just not implemented yet.
tracing::debug!(
"Free blocks {} need {n_blocks_needed}",
self.free_blocks.len()
);
self.free_blocks.extend(
self.cache_blocks
.evict(n_blocks_needed - self.free_blocks.len()),
@ -94,6 +110,9 @@ impl Allocator for RadixAllocator {
match self.alloc_or_reclaim(suffix_blocks as usize) {
Some(suffix_blocks) => blocks.extend(suffix_blocks),
None => {
tracing::debug!("Cannot allocate {:?}", self.cache_blocks);
tracing::debug!("Found {prefix_len} prefix tokens need {suffix_blocks} suffix blocks for {tokens} tokens");
tracing::debug!("Block size {}", self.block_size);
self.cache_blocks
.decref(prefix_node)
.expect("Failed to decrement refcount");
@ -211,7 +230,6 @@ struct RadixAllocation {
pub enum TrieError {
InvalidNodeId,
RefCountUnderflow,
BlockTokenCountMismatch,
}
pub type NodeId = DefaultKey;
@ -268,16 +286,19 @@ impl RadixTrie {
fn find_(&mut self, mut node_id: NodeId, key: &[u32], blocks: &mut Vec<u32>) -> NodeId {
let node = &self.nodes[node_id];
if let Some(&child_id) = node.children.get(&key[0]) {
self.update_access_time(child_id);
let child = self.nodes.get(child_id).expect("Invalid child identifier");
let shared_prefix_len = shared_prefix(&child.key, key, self.block_size);
assert_eq!(shared_prefix_len % self.block_size, 0);
blocks.extend(&child.blocks[..shared_prefix_len / self.block_size]);
if key.len() >= self.block_size {
let node_key = hash(&key[..self.block_size]);
if let Some(&child_id) = node.children.get(&node_key) {
self.update_access_time(child_id);
let child = self.nodes.get(child_id).expect("Invalid child identifier");
let shared_prefix_len = shared_prefix(&child.key, key, self.block_size);
assert_eq!(shared_prefix_len % self.block_size, 0);
blocks.extend(&child.blocks[..shared_prefix_len / self.block_size]);
let key = &key[shared_prefix_len..];
if !key.is_empty() {
node_id = self.find_(child_id, key, blocks);
let key = &key[shared_prefix_len..];
if !key.is_empty() {
node_id = self.find_(child_id, key, blocks);
}
}
}
@ -344,9 +365,11 @@ impl RadixTrie {
// evict n_blocks and return `None` if we can't. We are now needlessly
// evicting prefixes from the cache in such a case.
let mut evicted = Vec::new();
tracing::debug!("Evicting in search of {n_blocks}");
while let Some((last_access, node_id)) = self.leaves.pop_first() {
let blocks_needed = n_blocks - evicted.len();
let blocks_needed = n_blocks.saturating_sub(evicted.len());
tracing::debug!("Evicting node {node_id:?} ");
let node = self.nodes.get(node_id).expect("Leave does not exist");
assert_eq!(
@ -368,8 +391,11 @@ impl RadixTrie {
// the required number of blocks and leave the remaining blocks
// untouched.
let node = self.nodes.get_mut(node_id).expect("Leave does not exist");
node.key.truncate(node.blocks.len() - blocks_needed);
evicted.extend(node.blocks.split_off(node.blocks.len() - blocks_needed));
let truncate_blocks = node.blocks.len() - blocks_needed;
let truncate_tokens = truncate_blocks * self.block_size;
node.key.truncate(truncate_tokens);
evicted.extend(node.blocks.split_off(truncate_blocks));
self.leaves.insert((last_access, node_id));
break;
}
@ -400,11 +426,10 @@ impl RadixTrie {
// the part of the prefix that is already in the trie to detect
// mismatches.
if tokens.len() != blocks.len() * self.block_size {
return Err(TrieError::BlockTokenCountMismatch);
}
assert_eq!(tokens.len(), blocks.len() * self.block_size);
if let Some(&child_id) = self.nodes[node_id].children.get(&tokens[0]) {
let node_key = hash(&tokens[..self.block_size]);
if let Some(&child_id) = self.nodes[node_id].children.get(&node_key) {
self.update_access_time(child_id);
let child = self
.nodes
@ -452,14 +477,15 @@ impl RadixTrie {
.get_mut(node_id)
.expect("Node to-be split does not exist");
let mut parent_key = node.key.split_off(prefix_len);
let mut parent_blocks = node.blocks.split_off(prefix_len);
let prefix_blocks = prefix_len / self.block_size;
let mut parent_blocks = node.blocks.split_off(prefix_blocks);
// Move first part of the prefix to the parent. We swap to avoid
// an allocation + copy for both splits of the key/blocks.
std::mem::swap(&mut node.key, &mut parent_key);
std::mem::swap(&mut node.blocks, &mut parent_blocks);
let node_key = node.key[0];
let node_key = hash(&node.key[..self.block_size]);
let grandparent_id = node.parent.expect("Node does not have a parent");
let parent_id = self.add_node(grandparent_id, parent_key, parent_blocks);
@ -484,7 +510,7 @@ impl RadixTrie {
) -> NodeId {
let key = key.into();
let blocks = blocks.into();
let first = key[0];
let first = hash(&key[..self.block_size]);
let child = TrieNode::new(key, blocks, self.time, Some(parent_id));
let child_id = self.nodes.insert(child);
@ -496,10 +522,10 @@ impl RadixTrie {
}
/// Add a node to the parent.
fn add_node_to_parent(&mut self, parent_id: NodeId, first: u32, child_id: NodeId) {
fn add_node_to_parent(&mut self, parent_id: NodeId, hash: u64, child_id: NodeId) {
// Unwrap here, passing in an unknown id is a programming error.
let parent = self.nodes.get_mut(parent_id).expect("Unknown parent node");
if parent.children.insert(first, child_id).is_none() {
if parent.children.insert(hash, child_id).is_none() {
// Only increase reference count if child does not replace another child.
self.incref(parent_id)
.expect("Failed to increase parent refcount");
@ -517,7 +543,9 @@ impl RadixTrie {
);
let parent_id = node.parent.expect("Attempted to remove root node");
let parent = self.nodes.get_mut(parent_id).expect("Unknown parent node");
parent.children.remove(&node.key[0]);
let node_key = hash(&node.key[..self.block_size]);
parent.children.remove(&node_key);
self.decref(parent_id)
.expect("Failed to decrease parent refcount");
node
@ -571,7 +599,7 @@ impl RadixTrie {
#[derive(Debug)]
struct TrieNode {
blocks: Vec<u32>,
children: HashMap<u32, NodeId>,
children: HashMap<u64, NodeId>,
key: Vec<u32>,
last_accessed: u64,
parent: Option<NodeId>,

4
benchmark/Cargo.toml
View File

@ -16,7 +16,7 @@ path = "src/main.rs"
[dependencies]
average = "0.14"
clap = { version = "4.4.5", features = ["derive", "env"] }
crossterm = "0.27"
crossterm = "0.28.1"
float-ord = "0.3.2"
serde = {version = "1.0.188", features = ["derive"]}
serde_json = "1.0"
@ -25,7 +25,7 @@ text-generation-client = { path = "../backends/client" }
thiserror = "1.0.48"
tokenizers = { workspace = true }
tokio = { version = "1.32.0", features = ["rt", "rt-multi-thread", "parking_lot", "signal", "sync", "macros"] }
tui = {package = "ratatui", version = "0.23", default-features = false, features = ["crossterm"]}
ratatui = { version = "0.28.1", default-features = false, features = ["crossterm"] }
tracing = "0.1.37"
tracing-subscriber = { version = "0.3.17", features = ["json", "env-filter"] }
hf-hub = { workspace = true }

21
benchmark/src/app.rs
View File

@ -1,16 +1,15 @@
/// Inspired by https://github.com/hatoo/oha/blob/bb989ea3cd77727e7743e7daa60a19894bb5e901/src/monitor.rs
use crate::generation::{Decode, Message, Prefill};
use crossterm::event::{KeyCode, KeyEvent, KeyModifiers};
use text_generation_client::ClientError;
use tokio::sync::mpsc;
use tui::backend::Backend;
use tui::layout::{Alignment, Constraint, Direction, Layout};
use tui::style::{Color, Modifier, Style};
use tui::text::{Line, Span};
use tui::widgets::{
use ratatui::layout::{Alignment, Constraint, Direction, Layout};
use ratatui::style::{Color, Modifier, Style};
use ratatui::text::{Line, Span};
use ratatui::widgets::{
Axis, BarChart, Block, Borders, Chart, Dataset, Gauge, GraphType, Paragraph, Tabs,
};
use tui::{symbols, Frame};
use ratatui::{symbols, Frame};
use text_generation_client::ClientError;
use tokio::sync::mpsc;
/// TUI powered App
pub(crate) struct App {
@ -153,7 +152,7 @@ impl App {
}
/// Render frame
pub fn render<B: Backend>(&mut self, f: &mut Frame<'_, B>) {
pub fn render(&mut self, f: &mut Frame) {
let batch_progress =
(self.completed_batch as f64 / self.data.batch_size.len() as f64).clamp(0.0, 1.0);
let run_progress =
@ -172,7 +171,7 @@ impl App {
]
.as_ref(),
)
.split(f.size());
.split(f.area());
// Top row horizontal layout
let top = Layout::default()
@ -239,7 +238,7 @@ impl App {
f.render_widget(helper, row5[0]);
// Batch tabs
let titles = self
let titles: Vec<Line> = self
.data
.batch_size
.iter()

4
benchmark/src/lib.rs
View File

@ -7,12 +7,12 @@ mod utils;
use crate::app::App;
use crate::event::Event;
use crossterm::ExecutableCommand;
use ratatui::backend::CrosstermBackend;
use ratatui::Terminal;
use std::io;
use text_generation_client::v3::{GrammarType, NextTokenChooserParameters, ShardedClient};
use tokenizers::Tokenizer;
use tokio::sync::{broadcast, mpsc};
use tui::backend::CrosstermBackend;
use tui::Terminal;
/// Run benchmarking app
#[allow(clippy::too_many_arguments)]

12
flake.lock
View File

@ -853,11 +853,11 @@
]
},
"locked": {
"lastModified": 1726021481,
"narHash": "sha256-4J4E+Fh+77XIYnq2RVtg+ENWXpu6t74P0jKN/f2RQmI=",
"lastModified": 1726280639,
"narHash": "sha256-YfLRPlFZWrT2oRLNAoqf7G3+NnUTDdlIJk6tmBU7kXM=",
"owner": "oxalica",
"repo": "rust-overlay",
"rev": "1c2c120246c51a644c20ba2a36a33d3bd4860d70",
"rev": "e9f8641c92f26fd1e076e705edb12147c384171d",
"type": "github"
},
"original": {
@ -978,11 +978,11 @@
"nixpkgs": "nixpkgs_6"
},
"locked": {
"lastModified": 1725950569,
"narHash": "sha256-nJHA1SvIQbXySpL2ueNbzQOhnkQASa5tOLz/kdW0PWA=",
"lastModified": 1726229792,
"narHash": "sha256-9xsLmjc9nr7a4PTddKv2DOi82ompTtJNyjO6R67y5tE=",
"owner": "danieldk",
"repo": "tgi-nix",
"rev": "d40f3c22e9bcc5e16c94d4605cf6a7d74dd07f46",
"rev": "1a902f4818e94c3f8d95f6000db17bc3fadd0ce7",
"type": "github"
},
"original": {

22
flake.nix
View File

@ -69,7 +69,29 @@
server = pkgs.python3.pkgs.callPackage ./nix/server.nix { inherit nix-filter; };
in
{
checks = {
rust = with pkgs; rustPlatform.buildRustPackage {
name = "rust-checks";
src = ./.;
cargoLock = {
lockFile = ./Cargo.lock;
};
buildInputs = [ openssl.dev ];
nativeBuildInputs = [ clippy pkg-config protobuf python3 rustfmt ];
buildPhase = ''
cargo check
'';
checkPhase = ''
cargo fmt -- --check
cargo test -j $NIX_BUILD_CORES
cargo clippy
'';
installPhase = "touch $out";
} ;
};
formatter = pkgs.nixfmt-rfc-style;
devShells = with pkgs; rec {
default = pure;

7
integration-tests/conftest.py
View File

@ -342,6 +342,7 @@ def launcher(event_loop):
max_total_tokens: Optional[int] = None,
lora_adapters: Optional[List[str]] = None,
cuda_graphs: Optional[List[int]] = None,
attention: Optional[str] = None,
):
port = random.randint(8000, 10_000)
master_port = random.randint(10_000, 20_000)
@ -401,6 +402,8 @@ def launcher(event_loop):
if not use_flash_attention:
env["USE_FLASH_ATTENTION"] = "false"
if attention is not None:
env["ATTENTION"] = attention
with tempfile.TemporaryFile("w+") as tmp:
# We'll output stdout/stderr to a temporary file. Using a pipe
@ -437,6 +440,7 @@ def launcher(event_loop):
max_total_tokens: Optional[int] = None,
lora_adapters: Optional[List[str]] = None,
cuda_graphs: Optional[List[int]] = None,
attention: Optional[str] = None,
):
port = random.randint(8000, 10_000)
@ -491,6 +495,8 @@ def launcher(event_loop):
}
if not use_flash_attention:
env["USE_FLASH_ATTENTION"] = "false"
if attention is not None:
env["ATTENTION"] = attention
if HF_TOKEN is not None:
env["HF_TOKEN"] = HF_TOKEN
@ -522,6 +528,7 @@ def launcher(event_loop):
devices=devices,
volumes=volumes,
ports={"80/tcp": port},
healthcheck={"timeout": int(10 * 1e9)},
shm_size="1G",
)

2550
integration-tests/models/__snapshots__/test_flash_llama_prefix_flashdecoding/test_flash_llama_flashdecoding.json Normal file
View File

File diff suppressed because it is too large Load Diff

114
integration-tests/models/__snapshots__/test_flash_mixtral/test_flash_mixtral.json Normal file
View File

@ -0,0 +1,114 @@
{
"details": {
"best_of_sequences": null,
"finish_reason": "length",
"generated_tokens": 10,
"prefill": [
{
"id": 1,
"logprob": null,
"text": "<s>"
},
{
"id": 1824,
"logprob": -6.1445312,
"text": "What"
},
{
"id": 349,
"logprob": -1.4648438,
"text": "is"
},
{
"id": 21135,
"logprob": -13.6875,
"text": "gradient"
},
{
"id": 24871,
"logprob": -1.6005859,
"text": "descent"
},
{
"id": 28804,
"logprob": -0.39526367,
"text": "?"
},
{
"id": 13,
"logprob": -0.640625,
"text": "\n"
},
{
"id": 13,
"logprob": -0.18774414,
"text": "\n"
}
],
"seed": null,
"tokens": [
{
"id": 20910,
"logprob": -0.96484375,
"special": false,
"text": "Grad"
},
{
"id": 722,
"logprob": -0.003168106,
"special": false,
"text": "ient"
},
{
"id": 24871,
"logprob": -0.16540527,
"special": false,
"text": " descent"
},
{
"id": 349,
"logprob": -0.08886719,
"special": false,
"text": " is"
},
{
"id": 396,
"logprob": -0.75878906,
"special": false,
"text": " an"
},
{
"id": 18586,
"logprob": -0.5703125,
"special": false,
"text": " optimization"
},
{
"id": 9464,
"logprob": -0.11242676,
"special": false,
"text": " algorithm"
},
{
"id": 1307,
"logprob": -0.7939453,
"special": false,
"text": " used"
},
{
"id": 298,
"logprob": -0.17102051,
"special": false,
"text": " to"
},
{
"id": 26518,
"logprob": -0.34326172,
"special": false,
"text": " minimize"
}
],
"top_tokens": null
},
"generated_text": "Gradient descent is an optimization algorithm used to minimize"
}

99
integration-tests/models/__snapshots__/test_flash_mixtral/test_flash_mixtral_all_params.json Normal file
View File

@ -0,0 +1,99 @@
{
"details": {
"best_of_sequences": null,
"finish_reason": "length",
"generated_tokens": 10,
"prefill": [
{
"id": 1,
"logprob": null,
"text": "<s>"
},
{
"id": 24871,
"logprob": -17.234375,
"text": "descent"
},
{
"id": 28804,
"logprob": -7.4375,
"text": "?"
},
{
"id": 13,
"logprob": -0.8046875,
"text": "\n"
},
{
"id": 13,
"logprob": -0.33032227,
"text": "\n"
}
],
"seed": 0,
"tokens": [
{
"id": 1313,
"logprob": -2.3613281,
"special": false,
"text": "It"
},
{
"id": 3969,
"logprob": -0.7285156,
"special": false,
"text": " seems"
},
{
"id": 298,
"logprob": -1.3466797,
"special": false,
"text": " to"
},
{
"id": 528,
"logprob": 0.0,
"special": false,
"text": " me"
},
{
"id": 28725,
"logprob": -1.6757812,
"special": false,
"text": ","
},
{
"id": 369,
"logprob": 0.0,
"special": false,
"text": " that"
},
{
"id": 513,
"logprob": -1.1269531,
"special": false,
"text": " if"
},
{
"id": 368,
"logprob": 0.0,
"special": false,
"text": " you"
},
{
"id": 28742,
"logprob": -2.4921875,
"special": false,
"text": "'"
},
{
"id": 267,
"logprob": 0.0,
"special": false,
"text": "re"
}
],
"top_tokens": null
},
"generated_text": "What is gradient descent?\n\nIt seems to me, that if you're"
}

458
integration-tests/models/__snapshots__/test_flash_mixtral/test_flash_mixtral_load.json Normal file
View File

@ -0,0 +1,458 @@
[
{
"details": {
"best_of_sequences": null,
"finish_reason": "length",
"generated_tokens": 10,
"prefill": [
{
"id": 1,
"logprob": null,
"text": "<s>"
},
{
"id": 1824,
"logprob": -6.1445312,
"text": "What"
},
{
"id": 349,
"logprob": -1.4648438,
"text": "is"
},
{
"id": 21135,
"logprob": -13.6875,
"text": "gradient"
},
{
"id": 24871,
"logprob": -1.6005859,
"text": "descent"
},
{
"id": 28804,
"logprob": -0.39526367,
"text": "?"
},
{
"id": 13,
"logprob": -0.640625,
"text": "\n"
},
{
"id": 13,
"logprob": -0.18774414,
"text": "\n"
}
],
"seed": null,
"tokens": [
{
"id": 20910,
"logprob": -0.96484375,
"special": false,
"text": "Grad"
},
{
"id": 722,
"logprob": -0.003168106,
"special": false,
"text": "ient"
},
{
"id": 24871,
"logprob": -0.16369629,
"special": false,
"text": " descent"
},
{
"id": 349,
"logprob": -0.0881958,
"special": false,
"text": " is"
},
{
"id": 396,
"logprob": -0.76708984,
"special": false,
"text": " an"
},
{
"id": 18586,
"logprob": -0.57373047,
"special": false,
"text": " optimization"
},
{
"id": 9464,
"logprob": -0.11291504,
"special": false,
"text": " algorithm"
},
{
"id": 1307,
"logprob": -0.79589844,
"special": false,
"text": " used"
},
{
"id": 298,
"logprob": -0.1694336,
"special": false,
"text": " to"
},
{
"id": 26518,
"logprob": -0.34350586,
"special": false,
"text": " minimize"
}
],
"top_tokens": null
},
"generated_text": "Gradient descent is an optimization algorithm used to minimize"
},
{
"details": {
"best_of_sequences": null,
"finish_reason": "length",
"generated_tokens": 10,
"prefill": [
{
"id": 1,
"logprob": null,
"text": "<s>"
},
{
"id": 1824,
"logprob": -6.1445312,
"text": "What"
},
{
"id": 349,
"logprob": -1.4677734,
"text": "is"
},
{
"id": 21135,
"logprob": -13.6875,
"text": "gradient"
},
{
"id": 24871,
"logprob": -1.6015625,
"text": "descent"
},
{
"id": 28804,
"logprob": -0.39453125,
"text": "?"
},
{
"id": 13,
"logprob": -0.6435547,
"text": "\n"
},
{
"id": 13,
"logprob": -0.18713379,
"text": "\n"
}
],
"seed": null,
"tokens": [
{
"id": 20910,
"logprob": -0.9628906,
"special": false,
"text": "Grad"
},
{
"id": 722,
"logprob": -0.0032176971,
"special": false,
"text": "ient"
},
{
"id": 24871,
"logprob": -0.16540527,
"special": false,
"text": " descent"
},
{
"id": 349,
"logprob": -0.08898926,
"special": false,
"text": " is"
},
{
"id": 396,
"logprob": -0.765625,
"special": false,
"text": " an"
},
{
"id": 18586,
"logprob": -0.5708008,
"special": false,
"text": " optimization"
},
{
"id": 9464,
"logprob": -0.11401367,
"special": false,
"text": " algorithm"
},
{
"id": 1307,
"logprob": -0.7963867,
"special": false,
"text": " used"
},
{
"id": 298,
"logprob": -0.17028809,
"special": false,
"text": " to"
},
{
"id": 26518,
"logprob": -0.34326172,
"special": false,
"text": " minimize"
}
],
"top_tokens": null
},
"generated_text": "Gradient descent is an optimization algorithm used to minimize"
},
{
"details": {
"best_of_sequences": null,
"finish_reason": "length",
"generated_tokens": 10,
"prefill": [
{
"id": 1,
"logprob": null,
"text": "<s>"
},
{
"id": 1824,
"logprob": -6.140625,
"text": "What"
},
{
"id": 349,
"logprob": -1.4658203,
"text": "is"
},
{
"id": 21135,
"logprob": -13.6796875,
"text": "gradient"
},
{
"id": 24871,
"logprob": -1.5898438,
"text": "descent"
},
{
"id": 28804,
"logprob": -0.3955078,
"text": "?"
},
{
"id": 13,
"logprob": -0.64501953,
"text": "\n"
},
{
"id": 13,
"logprob": -0.18493652,
"text": "\n"
}
],
"seed": null,
"tokens": [
{
"id": 20910,
"logprob": -0.9580078,
"special": false,
"text": "Grad"
},
{
"id": 722,
"logprob": -0.0032176971,
"special": false,
"text": "ient"
},
{
"id": 24871,
"logprob": -0.16552734,
"special": false,
"text": " descent"
},
{
"id": 349,
"logprob": -0.08874512,
"special": false,
"text": " is"
},
{
"id": 396,
"logprob": -0.75878906,
"special": false,
"text": " an"
},
{
"id": 18586,
"logprob": -0.5703125,
"special": false,
"text": " optimization"
},
{
"id": 9464,
"logprob": -0.11236572,
"special": false,
"text": " algorithm"
},
{
"id": 1307,
"logprob": -0.79541016,
"special": false,
"text": " used"
},
{
"id": 298,
"logprob": -0.17102051,
"special": false,
"text": " to"
},
{
"id": 26518,
"logprob": -0.34326172,
"special": false,
"text": " minimize"
}
],
"top_tokens": null
},
"generated_text": "Gradient descent is an optimization algorithm used to minimize"
},
{
"details": {
"best_of_sequences": null,
"finish_reason": "length",
"generated_tokens": 10,
"prefill": [
{
"id": 1,
"logprob": null,
"text": "<s>"
},
{
"id": 1824,
"logprob": -6.1328125,
"text": "What"
},
{
"id": 349,
"logprob": -1.4658203,
"text": "is"
},
{
"id": 21135,
"logprob": -13.6796875,
"text": "gradient"
},
{
"id": 24871,
"logprob": -1.5947266,
"text": "descent"
},
{
"id": 28804,
"logprob": -0.39648438,
"text": "?"
},
{
"id": 13,
"logprob": -0.6464844,
"text": "\n"
},
{
"id": 13,
"logprob": -0.18688965,
"text": "\n"
}
],
"seed": null,
"tokens": [
{
"id": 20910,
"logprob": -0.9609375,
"special": false,
"text": "Grad"
},
{
"id": 722,
"logprob": -0.003168106,
"special": false,
"text": "ient"
},
{
"id": 24871,
"logprob": -0.16601562,
"special": false,
"text": " descent"
},
{
"id": 349,
"logprob": -0.088134766,
"special": false,
"text": " is"
},
{
"id": 396,
"logprob": -0.7597656,
"special": false,
"text": " an"
},
{
"id": 18586,
"logprob": -0.5708008,
"special": false,
"text": " optimization"
},
{
"id": 9464,
"logprob": -0.11291504,
"special": false,
"text": " algorithm"
},
{
"id": 1307,
"logprob": -0.7944336,
"special": false,
"text": " used"
},
{
"id": 298,
"logprob": -0.17102051,
"special": false,
"text": " to"
},
{
"id": 26518,
"logprob": -0.34399414,
"special": false,
"text": " minimize"
}
],
"top_tokens": null
},
"generated_text": "Gradient descent is an optimization algorithm used to minimize"
}
]

229
integration-tests/models/test_flash_llama_prefix_flashdecoding.py Normal file
View File

File diff suppressed because one or more lines are too long

75
integration-tests/models/test_flash_mixtral.py Normal file
View File

@ -0,0 +1,75 @@
import pytest
@pytest.fixture(scope="module")
def flash_mixtral_handle(launcher):
with launcher("mistralai/Mixtral-8x7B-v0.1", num_shard=8) as handle:
yield handle
@pytest.fixture(scope="module")
async def flash_mixtral(flash_mixtral_handle):
await flash_mixtral_handle.health(300)
return flash_mixtral_handle.client
@pytest.mark.skip(reason="requires > 4 shards")
@pytest.mark.asyncio
async def test_flash_mixtral(flash_mixtral, response_snapshot):
response = await flash_mixtral.generate(
"What is gradient descent?\n\n", max_new_tokens=10, decoder_input_details=True
)
assert response.details.generated_tokens == 10
assert (
response.generated_text
== "Gradient descent is an optimization algorithm used to minimize"
)
assert response == response_snapshot
@pytest.mark.skip(reason="requires > 4 shards")
@pytest.mark.asyncio
async def test_flash_mixtral_all_params(flash_mixtral, response_snapshot):
response = await flash_mixtral.generate(
"What is gradient descent?\n\n",
max_new_tokens=10,
repetition_penalty=1.2,
return_full_text=True,
stop_sequences=["test"],
temperature=0.5,
top_p=0.9,
top_k=10,
truncate=5,
typical_p=0.9,
watermark=True,
decoder_input_details=True,
seed=0,
)
assert response.details.generated_tokens == 10
assert (
response.generated_text
== "What is gradient descent?\n\nIt seems to me, that if you're"
)
assert response == response_snapshot
@pytest.mark.skip(reason="requires > 4 shards")
@pytest.mark.asyncio
async def test_flash_mixtral_load(flash_mixtral, generate_load, response_snapshot):
responses = await generate_load(
flash_mixtral, "What is gradient descent?\n\n", max_new_tokens=10, n=4
)
assert len(responses) == 4
assert responses[0].details.generated_tokens == 10
assert (
responses[0].generated_text
== "Gradient descent is an optimization algorithm used to minimize"
)
assert all(
[r.generated_text == responses[0].generated_text for r in responses]
), f"{[r.generated_text for r in responses]}"
assert responses == response_snapshot

2
nix/server.nix
View File

@ -21,6 +21,7 @@
loguru,
mamba-ssm,
marlin-kernels,
moe-kernels,
opentelemetry-api,
opentelemetry-exporter-otlp,
opentelemetry-instrumentation-grpc,
@ -88,6 +89,7 @@ buildPythonPackage {
loguru
mamba-ssm
marlin-kernels
moe-kernels
opentelemetry-api
opentelemetry-exporter-otlp
opentelemetry-instrumentation-grpc

9
router/src/server.rs
View File

@ -2136,9 +2136,12 @@ async fn start(
.unwrap();
// .set_buckets_for_metric(skipped_matcher, &skipped_buckets)
// .unwrap();
let prom_handle = builder
.install_recorder()
.expect("failed to install metrics recorder");
// See: https://github.com/metrics-rs/metrics/issues/467#issuecomment-2022755151
let (recorder, _) = builder
.build()
.expect("failed to build prometheus recorder");
let prom_handle = recorder.handle();
metrics::set_global_recorder(recorder).expect("Failed to set global recorder");
// Metrics descriptions
metrics::describe_counter!("tgi_request_success", "Number of successful requests");

88
server/poetry.lock generated
View File

@ -1242,6 +1242,82 @@ files = [
{file = "mdurl-0.1.2.tar.gz", hash = "sha256:bb413d29f5eea38f31dd4754dd7377d4465116fb207585f97bf925588687c1ba"},
]
[[package]]
name = "moe-kernels"
version = "0.2.2"
description = "MoE kernels"
optional = true
python-versions = ">=3.7"
files = [
{file = "moe_kernels-0.2.2+cu123torch2.4-cp310-cp310-linux_x86_64.whl", hash = "sha256:d268d818932ddcbca9bc71021dc63b008aae832827a7c0484cf206bd59cfc9ab"},
]
[package.dependencies]
nvidia-ml-py = "*"
torch = "*"
triton = "*"
[package.source]
type = "url"
url = "https://github.com/danieldk/moe-kernels/releases/download/v0.2.2/moe_kernels-0.2.2+cu123torch2.4-cp310-cp310-linux_x86_64.whl"
[[package]]
name = "moe-kernels"
version = "0.2.2"
description = "MoE kernels"
optional = true
python-versions = ">=3.7"
files = [
{file = "moe_kernels-0.2.2+cu123torch2.4-cp311-cp311-linux_x86_64.whl", hash = "sha256:614bbc3f41b707b0c40372f0bb00e218ad0842d306f90bef28ce8e98e7fcb7cb"},
]
[package.dependencies]
nvidia-ml-py = "*"
torch = "*"
triton = "*"
[package.source]
type = "url"
url = "https://github.com/danieldk/moe-kernels/releases/download/v0.2.2/moe_kernels-0.2.2+cu123torch2.4-cp311-cp311-linux_x86_64.whl"
[[package]]
name = "moe-kernels"
version = "0.2.2"
description = "MoE kernels"
optional = true
python-versions = ">=3.7"
files = [
{file = "moe_kernels-0.2.2+cu123torch2.4-cp312-cp312-linux_x86_64.whl", hash = "sha256:c2f48ed541353be03157d4015270dff797f7b7b8a664babdcbdf7414867d5abd"},
]
[package.dependencies]
nvidia-ml-py = "*"
torch = "*"
triton = "*"
[package.source]
type = "url"
url = "https://github.com/danieldk/moe-kernels/releases/download/v0.2.2/moe_kernels-0.2.2+cu123torch2.4-cp312-cp312-linux_x86_64.whl"
[[package]]
name = "moe-kernels"
version = "0.2.2"
description = "MoE kernels"
optional = true
python-versions = ">=3.7"
files = [
{file = "moe_kernels-0.2.2+cu123torch2.4-cp39-cp39-linux_x86_64.whl", hash = "sha256:d5f0339b73426c422872f7ff060433df6cd8e881451baf85ee7454e0e905f9d8"},
]
[package.dependencies]
nvidia-ml-py = "*"
torch = "*"
triton = "*"
[package.source]
type = "url"
url = "https://github.com/danieldk/moe-kernels/releases/download/v0.2.2/moe_kernels-0.2.2+cu123torch2.4-cp39-cp39-linux_x86_64.whl"
[[package]]
name = "mpmath"
version = "1.3.0"
@ -1600,6 +1676,17 @@ files = [
[package.dependencies]
nvidia-nvjitlink-cu12 = "*"
[[package]]
name = "nvidia-ml-py"
version = "12.560.30"
description = "Python Bindings for the NVIDIA Management Library"
optional = true
python-versions = "*"
files = [
{file = "nvidia-ml-py-12.560.30.tar.gz", hash = "sha256:f0254dc7400647680a072ee02509bfd46102b60bdfeca321576d4d4817e7fe97"},
{file = "nvidia_ml_py-12.560.30-py3-none-any.whl", hash = "sha256:fea371c94d63e38a611c17bbb85fe400e9c8ddb9e8684a9cd0e47786a4bc3c73"},
]
[[package]]
name = "nvidia-nccl-cu12"
version = "2.20.5"
@ -3638,6 +3725,7 @@ test = ["big-O", "importlib-resources", "jaraco.functools", "jaraco.itertools",
accelerate = ["accelerate"]
bnb = ["bitsandbytes"]
marlin = ["marlin-kernels", "marlin-kernels", "marlin-kernels", "marlin-kernels"]
moe = ["moe-kernels", "moe-kernels", "moe-kernels", "moe-kernels"]
outlines = ["outlines"]
peft = ["peft"]
quantize = ["accelerate", "datasets", "texttable"]

7
server/pyproject.toml
View File

@ -46,6 +46,12 @@ marlin-kernels = [
{ url = "https://github.com/danieldk/marlin-kernels/releases/download/v0.2.0/marlin_kernels-0.2.0+cu123torch2.4-cp311-cp311-linux_x86_64.whl", python = "~3.11", optional = true },
{ url = "https://github.com/danieldk/marlin-kernels/releases/download/v0.2.0/marlin_kernels-0.2.0+cu123torch2.4-cp312-cp312-linux_x86_64.whl", python = "~3.12", optional = true },
]
moe-kernels = [
{ url = "https://github.com/danieldk/moe-kernels/releases/download/v0.2.2/moe_kernels-0.2.2+cu123torch2.4-cp39-cp39-linux_x86_64.whl", python = "~3.9", optional = true },
{ url = "https://github.com/danieldk/moe-kernels/releases/download/v0.2.2/moe_kernels-0.2.2+cu123torch2.4-cp310-cp310-linux_x86_64.whl", python = "~3.10", optional = true },
{ url = "https://github.com/danieldk/moe-kernels/releases/download/v0.2.2/moe_kernels-0.2.2+cu123torch2.4-cp311-cp311-linux_x86_64.whl", python = "~3.11", optional = true },
{ url = "https://github.com/danieldk/moe-kernels/releases/download/v0.2.2/moe_kernels-0.2.2+cu123torch2.4-cp312-cp312-linux_x86_64.whl", python = "~3.12", optional = true },
]
rich = "^13.7.1"
[tool.poetry.extras]
@ -53,6 +59,7 @@ torch = ["torch"]
accelerate = ["accelerate"]
bnb = ["bitsandbytes"]
marlin = ["marlin-kernels"]
moe = ["moe-kernels"]
peft = ["peft"]
quantize = ["texttable", "datasets", "accelerate"]
outlines = ["outlines"]

6
server/text_generation_server/layers/attention/ipex.py
View File

@ -22,9 +22,9 @@ def attention(
# We do not need to check window_size_left (not supported) here, so it is already checked ahead of time at model load.
ipex.llm.functional.varlen_attention(
q,
key_cache,
value_cache,
q.contiguous() if q.device.type == "xpu" else q,
key_cache.contiguous() if key_cache.device.type == "xpu" else key_cache,
value_cache.contiguous() if value_cache.device.type == "xpu" else value_cache,
out,
seqlen.cu_seqlen_q,
seqlen.cu_seqlen_q,

76
server/text_generation_server/layers/moe/__init__.py Normal file
View File

@ -0,0 +1,76 @@
from typing import Optional
import torch
import torch.nn as nn
from text_generation_server.layers.fp8 import HybridFP8UnquantLoader
from text_generation_server.layers.moe.unquantized import UnquantizedSparseMoELayer
from text_generation_server.utils.weights import (
DefaultWeightsLoader,
UnquantizedWeight,
Weights,
)
class SparseMoELayer(nn.Module):
"""
Layer for MoE that uses fused kernels to only apply the active experts
for each token (rather than applying all experts and selecting the
outputs of active experts).
"""
def __init__(
self,
*,
n_expert_group: Optional[int],
n_experts: int,
prefix: str,
renormalize: bool,
topk: int,
topk_group: Optional[int],
weights: Weights,
gate_proj_name: str = "gate_proj",
up_proj_name: str = "up_proj",
down_proj_name: str = "down_proj",
):
super().__init__()
if (
isinstance(weights.loader, DefaultWeightsLoader)
and isinstance(weights.loader.weight_class, UnquantizedWeight)
) or isinstance(weights.loader, HybridFP8UnquantLoader):
cls = UnquantizedSparseMoELayer
# Once we wire up GPTQ-Marlin MoE:
# elif isinstance(weights.loader, GPTQMarlinWeightsLoader) and weights.loader.sym:
# cls = GPTQMarlinSparseMoELayer
else:
raise ValueError(
f"Unsupported weights loader: {weights.loader}, sparse MoE is only supported for unquantized and GPTQ weights"
)
self.moe = cls(
n_expert_group=n_expert_group,
n_experts=n_experts,
prefix=prefix,
renormalize=renormalize,
topk=topk,
topk_group=topk_group,
weights=weights,
gate_proj_name=gate_proj_name,
up_proj_name=up_proj_name,
down_proj_name=down_proj_name,
)
def forward(self, x: torch.Tensor, *, gating_output: torch.Tensor) -> torch.Tensor:
return self.moe(x, gating_output=gating_output)
@staticmethod
def is_supported(weights: Weights) -> bool:
return (
(
isinstance(weights.loader, DefaultWeightsLoader)
and isinstance(weights.loader.weight_class, UnquantizedWeight)
)
or isinstance(weights.loader, HybridFP8UnquantLoader)
# Once we wire up GPTQ-Marlin MoE:
# or isinstance(weights.loader, GPTQMarlinWeightsLoader)
)

125
server/text_generation_server/layers/moe/unquantized.py Normal file
View File

@ -0,0 +1,125 @@
from typing import Optional
import torch
import torch.nn as nn
from text_generation_server.utils.import_utils import SYSTEM
from text_generation_server.utils.weights import UnquantizedWeight, Weights
if SYSTEM != "ipex":
from moe_kernels.fused_moe import fused_moe
class UnquantizedSparseMoELayer(nn.Module):
def __init__(
self,
*,
n_expert_group: Optional[int],
n_experts: int,
prefix: str,
renormalize: bool,
topk: int,
topk_group: Optional[int],
weights: Weights,
gate_proj_name: str = "gate_proj",
up_proj_name: str = "up_proj",
down_proj_name: str = "down_proj",
):
super().__init__()
assert (n_expert_group is None) == (
topk_group is None
), "n_expert_group and topk_group must both be None or have some value"
self.n_expert_group = n_expert_group
self.topk = topk
self.topk_group = topk_group
self.renormalize = renormalize
self.gate_up_proj = _load_expert_multi_weights_col(
prefix=prefix,
n_experts=n_experts,
gate_proj_name=gate_proj_name,
up_proj_name=up_proj_name,
weights=weights,
)
self.down_proj = _load_expert_weights_row(
prefix=prefix,
n_experts=n_experts,
name=down_proj_name,
weights=weights,
)
def forward(self, x: torch.Tensor, *, gating_output: torch.Tensor) -> torch.Tensor:
return fused_moe(
x,
w1=self.gate_up_proj,
w2=self.down_proj,
gating_output=gating_output,
topk=self.topk,
renormalize=self.renormalize,
inplace=True,
use_grouped_topk=self.n_expert_group is not None,
num_expert_group=self.n_expert_group,
topk_group=self.topk_group,
)
def _load_expert_multi_weights_col(
*,
prefix: str,
n_experts: int,
gate_proj_name: str,
up_proj_name: str,
weights: Weights,
) -> torch.Tensor:
all_weight = None
for i in range(n_experts):
weight = weights.get_multi_weights_col(
[f"{prefix}.{i}.{gate_proj_name}", f"{prefix}.{i}.{up_proj_name}"], 0
)
assert isinstance(weight, UnquantizedWeight)
if all_weight is None:
all_weight = torch.empty(
(n_experts,) + weight.weight.shape,
dtype=weight.weight.dtype,
device=weight.weight.device,
)
all_weight[i] = weight.weight
assert all_weight is not None
return all_weight
def _load_expert_weights_row(
*,
prefix: str,
n_experts: int,
name: str,
weights: Weights,
) -> torch.Tensor:
all_weight = None
for i in range(n_experts):
weight = weights.get_weights_row(
f"{prefix}.{i}.{name}",
)
assert isinstance(weight, UnquantizedWeight)
if all_weight is None:
all_weight = torch.empty(
(n_experts,) + weight.weight.shape,
dtype=weight.weight.dtype,
device=weight.weight.device,
)
all_weight[i] = weight.weight
assert all_weight is not None
return all_weight

279
server/text_generation_server/models/custom_modeling/flash_deepseek_v2_modeling.py
View File

@ -13,9 +13,10 @@
# See the License for the specific language governing permissions and
# limitations under the License.
from typing import Any, Dict, List, Optional, Tuple
from typing import List, Optional, Tuple
from text_generation_server.models.globals import PAGED_KV
from moe_kernels.fused_moe import grouped_topk
import torch
import torch.distributed
from text_generation_server.layers import (
@ -33,6 +34,7 @@ from text_generation_server.layers.attention import (
Seqlen,
)
from text_generation_server.layers.layernorm import FastRMSNorm
from text_generation_server.layers.moe import SparseMoELayer
from text_generation_server.layers.rotary import PositionRotaryEmbedding, get_mscale
from text_generation_server.utils.import_utils import SYSTEM
from text_generation_server.utils.weights import Weights
@ -153,44 +155,6 @@ class DeepseekV2Config(PretrainedConfig):
)
def _load_experts(config, prefix: str, mat: str, weights: Weights):
if config.quantize is not None:
raise NotImplementedError(
"Deepseek V2 does not support weight quantization yet."
)
assert mat in ["gate_proj", "up_proj", "down_proj"]
world_size = weights.process_group.size()
rank = weights.process_group.rank()
assert (
config.moe_intermediate_size % world_size == 0
), f"The chosen size {config.moe_intermediate_size} is not compatible with sharding on {world_size} shards"
block_size = config.moe_intermediate_size // world_size
start = rank * block_size
stop = (rank + 1) * block_size
tensor = torch.empty(
(config.n_routed_experts * block_size, config.hidden_size),
dtype=weights.dtype,
device=weights.device,
)
for i in range(config.n_routed_experts):
slice_ = weights._get_slice(f"{prefix}.{i}.{mat}.weight")
if mat == "down_proj":
expert_slice = slice_[:, start:stop].t().contiguous()
else:
expert_slice = slice_[start:stop]
tensor[i * block_size : (i + 1) * block_size] = expert_slice.to(
dtype=weights.dtype
).to(device=weights.device)
return tensor
class DeepseekV2Attention(torch.nn.Module):
def __init__(
self,
@ -454,33 +418,21 @@ class BlockSparseMoE(nn.Module):
self.moe_intermediate_size = (
config.moe_intermediate_size // weights.process_group.size()
)
self.n_routed_experts = config.n_routed_experts
self.n_expert_group = config.n_group
self.topk_group = config.topk_group
self.top_k = config.num_experts_per_tok
self.norm_topk_prob = config.norm_topk_prob
self.routed_scaling_factor = config.routed_scaling_factor
gate_proj = _load_experts(
config, f"{prefix}.experts", "gate_proj", weights
).view(self.n_routed_experts, self.moe_intermediate_size, self.hidden_dim)
up_proj = _load_experts(config, f"{prefix}.experts", "up_proj", weights).view(
self.n_routed_experts, self.moe_intermediate_size, self.hidden_dim
)
self.gate_up_proj = torch.cat([gate_proj, up_proj], dim=1)
self.down_proj = (
_load_experts(config, f"{prefix}.experts", "down_proj", weights)
.view(self.n_routed_experts, self.moe_intermediate_size, self.hidden_dim)
.transpose(1, 2)
.contiguous()
)
# Gating
self.gate = FastLinear.load(config, f"{prefix}.gate", weights, bias=False)
self.moe_layer = SparseMoELayer(
prefix=f"{prefix}.experts",
n_experts=config.n_routed_experts,
n_expert_group=config.n_group,
renormalize=config.norm_topk_prob,
topk=config.num_experts_per_tok,
topk_group=config.topk_group,
weights=weights,
)
if config.n_shared_experts is not None:
self.shared_experts = DeepseekV2MLP(
prefix=f"{prefix}.shared_experts",
@ -501,25 +453,8 @@ class BlockSparseMoE(nn.Module):
shared_output = None
router_logits = self.gate(x)
topk_weights, topk_ids = grouped_topk(
x,
router_logits,
self.top_k,
renormalize=self.norm_topk_prob,
num_expert_group=self.n_expert_group,
topk_group=self.topk_group,
)
out = (
fused_experts(
x,
self.gate_up_proj,
self.down_proj,
topk_weights,
topk_ids,
inplace=True,
)
* self.routed_scaling_factor
)
out = self.moe_layer(x, gating_output=router_logits)
if shared_output is not None:
out = out + shared_output
@ -637,7 +572,9 @@ class DeepseekV2Layer(nn.Module):
and layer_id >= config.first_k_dense_replace
and layer_id % config.moe_layer_freq == 0
):
moe_cls = BlockSparseMoE if config.quantize is None else DenseMoE
moe_cls = (
BlockSparseMoE if SparseMoELayer.is_supported(weights) else DenseMoE
)
self.mlp = moe_cls(f"{prefix}.mlp", config, weights)
else:
self.mlp = DeepseekV2MLP(
@ -801,183 +738,3 @@ class FlashDeepseekV2ForCausalLM(torch.nn.Module):
hidden_states = hidden_states[lm_head_indices]
logits, speculative_logits = self.lm_head(hidden_states)
return logits, speculative_logits
# Functions below are from vLLM:
#
# https://github.com/vllm-project/vllm/blob/f7160d946a0a07703e72d81ba9ecf3913f192605/vllm/model_executor/layers/fused_moe/fused_moe.py#L397
#
# Remove after we have synced our version with upstream.
def grouped_topk(
hidden_states: torch.Tensor,
gating_output: torch.Tensor,
topk: int,
renormalize: bool,
num_expert_group: int = 0,
topk_group: int = 0,
) -> Tuple[torch.Tensor, torch.Tensor]:
scores = torch.softmax(gating_output, dim=-1)
num_token = scores.shape[0]
group_scores = (
scores.view(num_token, num_expert_group, -1).max(dim=-1).values
) # [n, n_group]
group_idx = torch.topk(group_scores, k=topk_group, dim=-1, sorted=False)[
1
] # [n, top_k_group]
group_mask = torch.zeros_like(group_scores) # [n, n_group]
group_mask.scatter_(1, group_idx, 1) # [n, n_group]
score_mask = (
group_mask.unsqueeze(-1)
.expand(num_token, num_expert_group, scores.shape[-1] // num_expert_group)
.reshape(num_token, -1)
) # [n, e]
tmp_scores = scores.masked_fill(~score_mask.bool(), 0.0) # [n, e]
topk_weights, topk_ids = torch.topk(tmp_scores, k=topk, dim=-1, sorted=False)
if renormalize:
topk_weights = topk_weights / topk_weights.sum(dim=-1, keepdim=True)
return topk_weights, topk_ids
def get_default_config(
M: int,
E: int,
N: int,
K: int,
topk: int,
dtype: Optional[str],
) -> Dict[str, int]:
config = {
"BLOCK_SIZE_M": 64,
"BLOCK_SIZE_N": 64,
"BLOCK_SIZE_K": 32,
"GROUP_SIZE_M": 8,
}
if M <= E:
config = {
"BLOCK_SIZE_M": 16,
"BLOCK_SIZE_N": 32,
"BLOCK_SIZE_K": 64,
"GROUP_SIZE_M": 1,
}
return config
def fused_experts(
hidden_states: torch.Tensor,
w1: torch.Tensor,
w2: torch.Tensor,
topk_weights: torch.Tensor,
topk_ids: torch.Tensor,
inplace: bool = False,
override_config: Optional[Dict[str, Any]] = None,
use_fp8: bool = False,
w1_scale: Optional[torch.Tensor] = None,
w2_scale: Optional[torch.Tensor] = None,
a1_scale: Optional[torch.Tensor] = None,
a2_scale: Optional[torch.Tensor] = None,
):
# Check constraints.
assert hidden_states.shape[1] == w1.shape[2], "Hidden size mismatch"
assert topk_weights.shape == topk_ids.shape, "topk shape mismatch"
assert hidden_states.is_contiguous(), "Hidden_states must be contiguous"
assert w1.is_contiguous(), "Expert weights1 must be contiguous"
assert w2.is_contiguous(), "Expert weights2 must be contiguous"
assert hidden_states.dtype in [torch.float32, torch.float16, torch.bfloat16]
import triton.language as tl
from vllm import _custom_ops as ops
from vllm.model_executor.layers.fused_moe.fused_moe import (
get_moe_configs,
invoke_fused_moe_kernel,
moe_align_block_size,
)
M, _ = hidden_states.shape
E, N, _ = w1.shape
if override_config:
config = override_config
else:
# First try to load optimal config from the file
configs = get_moe_configs(E, w2.shape[2], "float8" if use_fp8 else None)
if configs:
# If an optimal configuration map has been found, look up the
# optimal config
config = configs[min(configs.keys(), key=lambda x: abs(x - M))]
else:
# Else use the default config
config = get_default_config(
M, E, N, w1.shape[2], topk_ids.shape[1], "float8" if use_fp8 else None
)
intermediate_cache1 = torch.empty(
(M, topk_ids.shape[1], N),
device=hidden_states.device,
dtype=hidden_states.dtype,
)
intermediate_cache2 = torch.empty(
(M * topk_ids.shape[1], N // 2),
device=hidden_states.device,
dtype=hidden_states.dtype,
)
intermediate_cache3 = torch.empty(
(M, topk_ids.shape[1], w2.shape[1]),
device=hidden_states.device,
dtype=hidden_states.dtype,
)
sorted_token_ids, expert_ids, num_tokens_post_padded = moe_align_block_size(
topk_ids, config["BLOCK_SIZE_M"], E
)
compute_type = tl.bfloat16 if hidden_states.dtype == torch.bfloat16 else tl.float16
invoke_fused_moe_kernel(
hidden_states,
w1,
intermediate_cache1,
a1_scale,
w1_scale,
topk_weights,
topk_ids,
sorted_token_ids,
expert_ids,
num_tokens_post_padded,
False,
topk_ids.shape[1],
config,
compute_type=compute_type,
use_fp8=use_fp8,
)
ops.silu_and_mul(intermediate_cache2, intermediate_cache1.view(-1, N))
invoke_fused_moe_kernel(
intermediate_cache2,
w2,
intermediate_cache3,
a2_scale,
w2_scale,
topk_weights,
topk_ids,
sorted_token_ids,
expert_ids,
num_tokens_post_padded,
True,
1,
config,
compute_type=compute_type,
use_fp8=use_fp8,
)
if inplace:
return torch.sum(
intermediate_cache3.view(*intermediate_cache3.shape),
dim=1,
out=hidden_states,
)
return torch.sum(intermediate_cache3.view(*intermediate_cache3.shape), dim=1)

57
server/text_generation_server/models/custom_modeling/flash_mixtral_modeling.py
View File

@ -24,10 +24,7 @@ import torch.distributed
from torch import nn
from text_generation_server.utils.import_utils import SYSTEM
if SYSTEM != "ipex":
from vllm.model_executor.layers.fused_moe import fused_moe
from transformers.activations import ACT2FN
from transformers.configuration_utils import PretrainedConfig
from typing import Optional, List, Tuple
@ -46,6 +43,7 @@ from text_generation_server.layers import (
SpeculativeHead,
get_linear,
)
from text_generation_server.layers.moe import SparseMoELayer
from text_generation_server.layers.layernorm import (
FastRMSNorm,
)
@ -320,40 +318,21 @@ def round_up(x: torch.Tensor, value: int):
class BlockSparseMoE(nn.Module):
def __init__(self, prefix, config: MixtralConfig, weights):
super().__init__()
self.hidden_dim = config.hidden_size
self.ffn_dim = config.intermediate_size // weights.process_group.size()
self.num_experts = config.num_local_experts
self.top_k = config.num_experts_per_tok
act = config.hidden_act
if "gelu" in act:
self.act = lambda x: torch.nn.functional.gelu(
x,
approximate=(
"tanh" if act in ["gelu_fast", "gelu_pytorch_tanh"] else "none"
),
)
elif "silu" in act:
self.act = torch.nn.functional.silu
else:
self.act = ACT2FN[act]
# gating
self.gate = FastLinear.load(config, f"{prefix}.gate", weights, bias=False)
# merged expert weights, all of size (n_experts * ffn_dim, hidden_dim)
w1 = _load_experts(config, f"{prefix}.experts", "w1", weights).view(
self.num_experts, self.ffn_dim, self.hidden_dim
)
w3 = _load_experts(config, f"{prefix}.experts", "w3", weights).view(
self.num_experts, self.ffn_dim, self.hidden_dim
)
self.w13 = torch.cat([w1, w3], dim=1)
self.w2 = (
_load_experts(config, f"{prefix}.experts", "w2", weights)
.view(self.num_experts, self.ffn_dim, self.hidden_dim)
.transpose(1, 2)
.contiguous()
self.moe = SparseMoELayer(
n_expert_group=None,
n_experts=config.num_local_experts,
prefix=f"{prefix}.experts",
renormalize=True,
topk=config.num_experts_per_tok,
topk_group=None,
weights=weights,
gate_proj_name="w1",
up_proj_name="w3",
down_proj_name="w2",
)
self.process_group = weights.process_group
@ -361,15 +340,7 @@ class BlockSparseMoE(nn.Module):
def forward(self, x: torch.Tensor) -> torch.Tensor:
# router_logits: (num_tokens, n_experts)
router_logits = self.gate(x)
out = fused_moe(
x,
self.w13,
self.w2,
router_logits,
self.top_k,
renormalize=True,
inplace=True,
)
out = self.moe(x, gating_output=router_logits)
# Reduce sum
if self.process_group.size() > 1:
@ -476,7 +447,7 @@ class MixtralLayer(nn.Module):
prefix=f"{prefix}.self_attn", config=config, weights=weights
)
moe_cls = BlockSparseMoE if config.quantize is None else DenseMoE
moe_cls = BlockSparseMoE if SparseMoELayer.is_supported(weights) else DenseMoE
self.moe = moe_cls(f"{prefix}.block_sparse_moe", config, weights)
self.input_layernorm = FastRMSNorm.load(

22
server/text_generation_server/models/flash_causal_lm.py
View File

@ -82,7 +82,7 @@ def init_cpu_threads_env(rank_id: int, world_size: int):
import numa
import psutil
nodes = numa.get_max_node() + 1
nodes = numa.info.get_max_node() + 1
rank_per_node = math.ceil(world_size / nodes)
num_cpus_per_nodes = int(psutil.cpu_count(logical=False) / nodes)
node_id = int(rank_id / rank_per_node)
@ -91,18 +91,22 @@ def init_cpu_threads_env(rank_id: int, world_size: int):
num_cpus_per_rank = max(int(num_cpus_per_nodes / rank_per_node), 1)
else:
num_cpus_per_rank = int(os.getenv("OMP_NUM_THREADS"))
if len(numa.get_membind()) == nodes:
numa.set_membind([node_id])
if len(numa.memory.get_membind_nodes()) == nodes:
numa.memory.set_membind_nodes((node_id))
torch.set_num_threads(num_cpus_per_rank)
if len(numa.get_affinity(0)) == psutil.cpu_count(logical=True):
if len(numa.schedule.get_affinitive_cpus(0)) == psutil.cpu_count(logical=True):
cpu_start = num_cpus_per_rank * rank_offset_per_node
numa.set_affinity(
numa.schedule.run_on_cpus(
0,
list(numa.node_to_cpus(node_id))[
cpu_start : cpu_start + num_cpus_per_rank
],
*(
numa.info.node_to_cpus(node_id)[
cpu_start : cpu_start + num_cpus_per_rank
]
),
)
logger.info(f"affinity={numa.get_affinity(0)}, membind = {numa.get_membind()}")
logger.info(
f"affinity={numa.schedule.get_affinitive_cpus(0)}, membind = {numa.memory.get_membind_nodes()}"
)
@dataclass

11
server/text_generation_server/utils/adapter.py
View File

@ -77,12 +77,12 @@ def load_and_merge_adapters(
) -> Tuple["ModuleMap", "AdapterConfig", Set[str], PreTrainedTokenizer]:
if len(adapter_parameters.adapter_info) == 1:
adapter_info = next(iter(adapter_parameters.adapter_info))
adapter = next(iter(adapter_parameters.adapter_info))
return load_module_map(
model_id,
adapter_info.revision,
adapter_info.id,
adapter_info.path,
adapter.revision,
adapter.id,
adapter.path,
weight_names,
trust_remote_code,
)
@ -90,7 +90,6 @@ def load_and_merge_adapters(
adapter_params = AdapterParametersContainer(adapter_parameters, adapter_index)
return _load_and_merge(
model_id,
adapter_params.revision,
adapter_params,
weight_names,
trust_remote_code,
@ -109,7 +108,6 @@ class AdapterParametersContainer:
@lru_cache(maxsize=32)
def _load_and_merge(
model_id: str,
revision: str,
adapter_params: AdapterParametersContainer,
weight_names: Tuple[str],
trust_remote_code: bool = False,
@ -126,6 +124,7 @@ def _load_and_merge(
module_map, adapter_config, adapter_weight_names, adapter_tokenizer = (
load_module_map(
model_id,
adapter.revision,
adapter.id,
adapter.path,
weight_names,

5
server/text_generation_server/utils/weights.py
View File

@ -120,7 +120,6 @@ class DefaultWeightsLoader(WeightsLoader):
prefix: str,
block_sizes: Union[int, List[int]],
):
return self.weight_class(
weights.get_packed_sharded(
f"{prefix}.weight", dim=0, block_sizes=block_sizes
@ -405,6 +404,10 @@ class Weights:
finally:
self.weights_loader = old_loader
@property
def loader(self):
return self.weights_loader
def _blocks_to_block_sizes(total_size: int, blocks: Union[int, List[int]]) -> List[int]:
"""