Skip to content

feat(onnx): add onnx runtime for better CPU perf #328

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 3 commits into from
Jul 5, 2024
Merged

feat(onnx): add onnx runtime for better CPU perf #328

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
91f8abf
feat(onnx): add onnx runtime for better CPU perf
OlivierDehaene Jul 5, 2024
aaecf3b
add support for sentence transformers onnx exports
OlivierDehaene Jul 5, 2024
fa30a2f
fix integration test model id for ort
OlivierDehaene Jul 5, 2024
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
83 changes: 83 additions & 0 deletions Cargo.lock
View file
Open in desktop

Some generated files are not rendered by default. Learn more about how customized files appear on GitHub.

1 change: 1 addition & 0 deletions Cargo.toml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2,6 +2,7 @@
members = [
"backends",
"backends/candle",
"backends/ort",
"backends/core",
"backends/python",
"backends/grpc-client",
Expand Down
38 changes: 4 additions & 34 deletions Dockerfile
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -28,22 +28,9 @@ ARG ACTIONS_CACHE_URL
ARG ACTIONS_RUNTIME_TOKEN
ARG SCCACHE_GHA_ENABLED

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 && DEBIAN_FRONTEND=noninteractive apt-get install -y --no-install-recommends \
intel-oneapi-mkl-devel=2024.0.0-49656 \
build-essential \
&& rm -rf /var/lib/apt/lists/*

RUN echo "int mkl_serv_intel_cpu_true() {return 1;}" > fakeintel.c && \
gcc -shared -fPIC -o libfakeintel.so fakeintel.c

COPY --from=planner /usr/src/recipe.json recipe.json

RUN cargo chef cook --release --features candle --features mkl-dynamic --no-default-features --recipe-path recipe.json && sccache -s
RUN cargo chef cook --release --features ort --no-default-features --recipe-path recipe.json && sccache -s

COPY backends backends
COPY core core
Expand All @@ -53,7 +40,7 @@ COPY Cargo.lock ./

FROM builder as http-builder

RUN cargo build --release --bin text-embeddings-router -F candle -F mkl-dynamic -F http --no-default-features && sccache -s
RUN cargo build --release --bin text-embeddings-router -F ort -F http --no-default-features && sccache -s

FROM builder as grpc-builder

Expand All @@ -65,35 +52,18 @@ RUN PROTOC_ZIP=protoc-21.12-linux-x86_64.zip && \

COPY proto proto

RUN cargo build --release --bin text-embeddings-router -F grpc -F candle -F mkl-dynamic --no-default-features && sccache -s
RUN cargo build --release --bin text-embeddings-router -F grpc -F ort --no-default-features && sccache -s

FROM debian:bookworm-slim as base

ENV HUGGINGFACE_HUB_CACHE=/data \
PORT=80 \
MKL_ENABLE_INSTRUCTIONS=AVX512_E4 \
RAYON_NUM_THREADS=8 \
LD_PRELOAD=/usr/local/libfakeintel.so \
LD_LIBRARY_PATH=/usr/local/lib
PORT=80

RUN apt-get update && DEBIAN_FRONTEND=noninteractive apt-get install -y --no-install-recommends \
libomp-dev \
ca-certificates \
libssl-dev \
curl \
&& rm -rf /var/lib/apt/lists/*

# Copy a lot of the Intel shared objects because of the mkl_serv_intel_cpu_true patch...
COPY --from=builder /opt/intel/oneapi/mkl/latest/lib/intel64/libmkl_intel_lp64.so.2 /usr/local/lib/libmkl_intel_lp64.so.2
COPY --from=builder /opt/intel/oneapi/mkl/latest/lib/intel64/libmkl_intel_thread.so.2 /usr/local/lib/libmkl_intel_thread.so.2
COPY --from=builder /opt/intel/oneapi/mkl/latest/lib/intel64/libmkl_core.so.2 /usr/local/lib/libmkl_core.so.2
COPY --from=builder /opt/intel/oneapi/mkl/latest/lib/intel64/libmkl_vml_def.so.2 /usr/local/lib/libmkl_vml_def.so.2
COPY --from=builder /opt/intel/oneapi/mkl/latest/lib/intel64/libmkl_def.so.2 /usr/local/lib/libmkl_def.so.2
COPY --from=builder /opt/intel/oneapi/mkl/latest/lib/intel64/libmkl_vml_avx2.so.2 /usr/local/lib/libmkl_vml_avx2.so.2
COPY --from=builder /opt/intel/oneapi/mkl/latest/lib/intel64/libmkl_vml_avx512.so.2 /usr/local/lib/libmkl_vml_avx512.so.2
COPY --from=builder /opt/intel/oneapi/mkl/latest/lib/intel64/libmkl_avx2.so.2 /usr/local/lib/libmkl_avx2.so.2
COPY --from=builder /opt/intel/oneapi/mkl/latest/lib/intel64/libmkl_avx512.so.2 /usr/local/lib/libmkl_avx512.so.2
COPY --from=builder /usr/src/libfakeintel.so /usr/local/libfakeintel.so

FROM base as grpc

Expand Down
92 changes: 0 additions & 92 deletions Dockerfile-arm64
View file
Open in desktop

This file was deleted.

8 changes: 4 additions & 4 deletions README.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -72,9 +72,9 @@ Below are some examples of the currently supported models:
| MTEB Rank | Model Size | Model Type | Model ID |
|-----------|---------------------|-------------|--------------------------------------------------------------------------------------------------|
| 1 | 7B (Very Expensive) | Mistral | [Salesforce/SFR-Embedding-2_R](https://hf.co/Salesforce/SFR-Embedding-2_R) |
| 2 | 7B (Very Expensive) | Qwen2 | [Alibaba-NLP/gte-Qwen2-7B-instruct](https://hf.co/Alibaba-NLP/gte-Qwen2-7B-instruct) |
| 9 | 1.5B (Expensive) | Qwen2 | [Alibaba-NLP/gte-Qwen2-1.5B-instruct](https://hf.co/Alibaba-NLP/gte-Qwen2-1.5B-instruct) |
| 15 | 0.4B | Alibaba GTE | [Alibaba-NLP/gte-large-en-v1.5](Alibaba-NLP/gte-large-en-v1.5) |
| 2 | 7B (Very Expensive) | Qwen2 | [Alibaba-NLP/gte-Qwen2-7B-instruct](https://hf.co/Alibaba-NLP/gte-Qwen2-7B-instruct) |
| 9 | 1.5B (Expensive) | Qwen2 | [Alibaba-NLP/gte-Qwen2-1.5B-instruct](https://hf.co/Alibaba-NLP/gte-Qwen2-1.5B-instruct) |
| 15 | 0.4B | Alibaba GTE | [Alibaba-NLP/gte-large-en-v1.5](https://hf.co/Alibaba-NLP/gte-large-en-v1.5) |
| 20 | 0.3B | Bert | [WhereIsAI/UAE-Large-V1](https://hf.co/WhereIsAI/UAE-Large-V1) |
| 24 | 0.5B | XLM-RoBERTa | [intfloat/multilingual-e5-large-instruct](https://hf.co/intfloat/multilingual-e5-large-instruct) |
| N/A | 0.1B | NomicBert | [nomic-ai/nomic-embed-text-v1](https://hf.co/nomic-ai/nomic-embed-text-v1) |
Expand Down Expand Up @@ -568,7 +568,7 @@ supported via Docker. As such inference will be CPU bound and most likely pretty
M1/M2 ARM CPU.

```
docker build . -f Dockerfile-arm64 --platform=linux/arm64
docker build . -f Dockerfile --platform=linux/arm64
```

## Examples
Expand Down
4 changes: 4 additions & 0 deletions backends/Cargo.toml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -7,15 +7,19 @@ homepage.workspace = true

[dependencies]
clap = { workspace = true, optional = true }
hf-hub = { workspace = true }
serde_json = { workspace = true }
text-embeddings-backend-core = { path = "core" }
text-embeddings-backend-python = { path = "python", optional = true }
text-embeddings-backend-candle = { path = "candle", optional = true }
text-embeddings-backend-ort = { path = "ort", optional = true }
tokio = { workspace = true }
tracing = { workspace = true }

[features]
clap = ["dep:clap", "text-embeddings-backend-core/clap"]
python = ["dep:text-embeddings-backend-python"]
ort = ["dep:text-embeddings-backend-ort"]
candle = ["dep:text-embeddings-backend-candle"]
cuda = ["text-embeddings-backend-candle?/cuda"]
metal = ["text-embeddings-backend-candle?/metal"]
Expand Down
17 changes: 17 additions & 0 deletions backends/ort/Cargo.toml
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,17 @@
[package]
name = "text-embeddings-backend-ort"
version.workspace = true
edition.workspace = true
authors.workspace = true
homepage.workspace = true

[dependencies]
anyhow = { workspace = true }
nohash-hasher = { workspace = true }
ndarray = "0.15.6"
ort = { version = "2.0.0-rc.2", default-features = false, features = ["download-binaries", "half", "onednn", "ndarray"] }
text-embeddings-backend-core = { path = "../core" }
tracing = { workspace = true }
thiserror = { workspace = true }
serde = { workspace = true }
serde_json = { workspace = true }
Loading
Loading