tarry.singh/Radix
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Remove example workspace and related documentation files.

Browse Source
This commit is contained in:
Foke Singh 2025-06-20 13:23:06 +00:00
parent e789e26008
commit a540564744
11 changed files with 155 additions and 292 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

30
docs/howtos/add_weights.md
View File

@ -5,7 +5,7 @@ Our [first model](../tutorials/write_first_model.md) did not need any weights fi
We just created weights and biases at runtime.
But real-world models typically need weights files, and maybe some other
supporting files.
supporting files.
We recommend, for easy deployments, you upload those files. In many instances,
you will use a site like [🤗 Hugging Face](https://huggingface.co).
@ -14,21 +14,14 @@ We also recommend to add a `weights.bzl` file to your project root directory, so
you don't "pollute" your build file with long URLs and SHAs:
```python
load("@bazel_tools//tools/build_defs/repo:http.bzl", "http_file")
load("@bazel_tools//tools/build_defs/repo:http.bzl", "http_archive")
def _weights_impl(mctx):
http_file(
name = "com_github_zml_cdn_mnist",
downloaded_file_path = "mnist.pt",
sha256 = "d8a25252e28915e147720c19223721f0f53e3317493727ca754a2dd672450ba9",
url = "https://github.com/ggerganov/ggml/raw/18703ad600cc68dbdb04d57434c876989a841d12/examples/mnist/models/mnist/mnist_model.state_dict",
)
http_file(
name = "com_github_zml_cdn_mnist_data",
downloaded_file_path = "mnist.ylc",
sha256 = "0fa7898d509279e482958e8ce81c8e77db3f2f8254e26661ceb7762c4d494ce7",
url = "https://github.com/ggerganov/ggml/raw/18703ad600cc68dbdb04d57434c876989a841d12/examples/mnist/models/mnist/t10k-images.idx3-ubyte",
http_archive(
name = "mnist",
sha256 = "075905e433ea0cce13c3fc08832448ab86225d089b5d412be67f59c29388fb19",
url = "https://mirror.zml.ai/data/mnist.tar.zst",
build_file_content = """exports_files(glob(["**"]), visibility = ["//visibility:public"])""",
)
return mctx.extension_metadata(
@ -54,12 +47,12 @@ the following way:
zig_cc_binary(
name = "mnist",
args = [
"$(location @com_github_zml_cdn_mnist//file)",
"$(location @com_github_zml_cdn_mnist_data//file)",
"$(location @mnist//:mnist.pt)",
"$(location @mnist//:t10k-images.idx3-ubyte)",
],
data = [
"@com_github_zml_cdn_mnist//file",
"@com_github_zml_cdn_mnist_data//file",
"@mnist//:mnist.pt",
"@mnist//:t10k-images.idx3-ubyte",
],
main = "mnist.zig",
deps = [
@ -74,4 +67,3 @@ See how:
- we use `data = [ ... ]` to reference the files in `weights.bzl`
- we use `args = [ ... ]` to pass the files as command-line arguments to the
MNIST executable at runtime, automatically.

8
docs/howtos/deploy_on_server.md
View File

@ -24,8 +24,7 @@ So, to run the Llama model from above **on your development machine**
housing an NVIDIA GPU, run the following:
```
cd examples
bazel run --config=release //llama --@zml//runtimes:cuda=true -- --hf-model-path=$HOME/Llama-3.2-1B-Instruct
bazel run --config=release //examples/llama --@zml//runtimes:cuda=true -- --hf-model-path=$HOME/Llama-3.2-1B-Instruct
```
@ -41,8 +40,7 @@ architectures:
As an example, here is how you build above Llama for CUDA on Linux X86_64:
```
cd examples
bazel build --config=release //llama \
bazel build --config=release //examples/llama \
--@zml//runtimes:cuda=true \
--@zml//runtimes:cpu=false \
--platforms=@zml//platforms:linux_amd64
@ -83,7 +81,6 @@ tar(
... and then build the TAR archive:
```
# cd examples
bazel build --config=release //mnist:archive \
--@zml//runtimes:cuda=true \
--@zml//runtimes:cpu=false \
@ -134,4 +131,3 @@ INFO: Running command line: bazel-bin/mnist/mnist ../_main~_repo_rules~com_githu
You see the command line right up there. On the server, you just need to replace
`../` with the 'runfiles' directory of your TAR.

4
docs/howtos/dockerize_models.md
View File

@ -133,8 +133,7 @@ platform_transition_filegroup(
And that's almost it! You can already build the image:
```
# cd examples
bazel build --config=release //simple_layer:image
bazel build --config=release //examples/simple_layer:image
INFO: Analyzed target //simple_layer:image (1 packages loaded, 8 targets configured).
INFO: Found 1 target...
@ -375,4 +374,3 @@ MNIST model, including weights and dataset, to the docker registry:
```
bazel run //mnist:push --@zml//runtimes:cuda=true -- --repository index.docker.io/my_org/zml_mnist
```

4
docs/howtos/huggingface_access_token.md
View File

@ -30,8 +30,6 @@ Now you're ready to download a gated model like `Meta-Llama-3.2-1b`!
```
# requires token in $HOME/.cache/huggingface/token, as created by the
# `huggingface-cli login` command, or the `HUGGINGFACE_TOKEN` environment variable.
cd examples
bazel run @zml//tools:hf -- download meta-llama/Llama-3.2-1B-Instruct --local-dir $HOME/Llama-3.2-1B-Instruct --exclude='*.pth'
bazel run --config=release //llama -- --hf-model-path=$HOME/Llama-3.2-1B-Instruct --prompt="What is the capital of France?"
bazel run --config=release //examples/llama -- --hf-model-path=$HOME/Llama-3.2-1B-Instruct --prompt="What is the capital of France?"
```

13
docs/tutorials/getting_started.md
View File

@ -49,7 +49,6 @@ compile it, and classify a randomly picked example from the test dataset.
On the command line:
```
cd examples
bazel run --config=release //mnist
```
@ -74,9 +73,8 @@ Once you've been granted access, you're ready to download a gated model like
```
# requires token in $HOME/.cache/huggingface/token, as created by the
# `huggingface-cli login` command, or the `HUGGINGFACE_TOKEN` environment variable.
cd examples
bazel run @zml//tools:hf -- download meta-llama/Llama-3.1-8B-Instruct --local-dir $HOME/Llama-3.1-8B-Instruct --exclude='*.pth'
bazel run --config=release //llama -- --hf-model-path=$HOME/Llama-3.1-8B-Instruct
bazel run --config=release //examples/llama -- --hf-model-path=$HOME/Llama-3.1-8B-Instruct
bazel run --config=release //llama -- --hf-model-path=$HOME/Llama-3.1-8B-Instruct --prompt="What is the capital of France?"
```
@ -88,9 +86,8 @@ Like the 8B model above, this model also requires approval. See
[here](https://huggingface.co/meta-llama/Llama-3.2-1B-Instruct) for access requirements.
```
cd examples
bazel run @zml//tools:hf -- download meta-llama/Llama-3.2-1B-Instruct --local-dir $HOME/Llama-3.2-1B-Instruct --exclude='*.pth'
bazel run --config=release //llama -- --hf-model-path=$HOME/Llama-3.2-1B-Instruct
bazel run --config=release //examples/llama -- --hf-model-path=$HOME/Llama-3.2-1B-Instruct
bazel run --config=release //llama -- --hf-model-path=$HOME/Llama-3.2-1B-Instruct --prompt="What is the capital of France?"
```
@ -121,9 +118,8 @@ So, to run the OpenLLama model from above on your host sporting an NVIDIA GPU,
run the following:
```
cd examples
bazel run --config=release //llama:Llama-3.2-1B-Instruct \
--@zml//runtimes:cuda=true \
bazel run --config=release //examples/llama:Llama-3.2-1B-Instruct \
--@zml//runtimes:cuda=true \
-- --prompt="What is the capital of France?"
```
@ -140,4 +136,3 @@ You might also want to check out the
[documentation](../README.md), start
[writing your first model](../tutorials/write_first_model.md), or read about more
high-level [ZML concepts](../learn/concepts.md).

3
docs/tutorials/write_first_model.md
View File

@ -381,8 +381,7 @@ With everything in place now, running the model is easy:
```
# run release (--config=release)
cd examples
bazel run --config=release //simple_layer
bazel run --config=release //examples/simple_layer
# compile and run debug version
bazel run //simple_layer

14
examples/benchmark/BUILD.bazel
View File

@ -1,11 +1,17 @@
load("@zml//bazel:zig.bzl", "zig_cc_binary")
load("@bazel_skylib//rules:build_test.bzl", "build_test")
load("//bazel:zig.bzl", "zig_cc_binary")
zig_cc_binary(
name = "benchmark",
main = "main.zig",
deps = [
"@zml//async",
"@zml//stdx",
"@zml//zml",
"//async",
"//stdx",
"//zml",
],
)
build_test(
name = "test",
targets = [":benchmark"],
)

146
examples/llama/BUILD.bazel
View File

@ -1,11 +1,8 @@
load("@aspect_bazel_lib//lib:expand_template.bzl", "expand_template")
load("@aspect_bazel_lib//lib:tar.bzl", "mtree_spec", "tar")
load("@aspect_bazel_lib//lib:transitions.bzl", "platform_transition_filegroup")
load("@bazel_skylib//rules:native_binary.bzl", "native_test")
load("@bazel_skylib//rules:write_file.bzl", "write_file")
load("@rules_cc//cc:cc_binary.bzl", "cc_binary")
load("@bazel_skylib//rules:build_test.bzl", "build_test")
load("@rules_oci//oci:defs.bzl", "oci_image", "oci_load", "oci_push")
load("@zml//bazel:zig.bzl", "zig_cc_binary")
load("//bazel:zig.bzl", "zig_cc_binary")
zig_cc_binary(
name = "llama",
@ -14,65 +11,58 @@ zig_cc_binary(
],
main = "main.zig",
deps = [
"//async",
"//stdx",
"//zml",
"@com_github_hejsil_clap//:clap",
"@zml//async",
"@zml//stdx",
"@zml//zml",
],
)
zig_cc_binary(
name = "test-implementation",
srcs = ["llama.zig"],
args = [
"--weights=$(location @Meta-Llama-3.1-8B-Instruct//:model.safetensors.index.json)",
"--config=$(location @Meta-Llama-3.1-8B-Instruct//:config.json)",
],
data = [
"@Meta-Llama-3.1-8B-Instruct//:config.json",
"@Meta-Llama-3.1-8B-Instruct//:model.safetensors.index.json",
],
main = "test.zig",
tags = [
"manual",
],
deps = [
"@zml//async",
"@zml//stdx",
"@zml//zml",
],
)
# TODO(steeve): Re-enable when rules_zig linkmode is fixed.
#
# zig_cc_binary(
# name = "test-implementation",
# srcs = ["llama.zig"],
# args = [
# "--weights=$(location @Meta-Llama-3.1-8B-Instruct//:model.safetensors.index.json)",
# "--config=$(location @Meta-Llama-3.1-8B-Instruct//:config.json)",
# ],
# data = [
# "@Meta-Llama-3.1-8B-Instruct//:config.json",
# "@Meta-Llama-3.1-8B-Instruct//:model.safetensors.index.json",
# ],
# main = "test.zig",
# deps = [
# "//async",
# "//stdx",
# "//zml",
# ],
# )
native_test(
name = "test_tokenizer",
src = "@zml//zml/tokenizer:main",
# Note: all Llama-3.x tokenizers are the same,
# but using the 3.2-1B version because downloading the tokenizer triggers downloading the model.
args = [
"--tokenizer=$(location @Meta-Llama-3.2-1B-Instruct//:tokenizer.json)",
"""--prompt='Examples of titles:
📉 Stock Market Trends
🍪 Perfect Chocolate Chip Recipe
Evolution of Music Streaming
Remote Work Productivity Tips
Artificial Intelligence in Healthcare
🎮 Video Game Development Insights
'""",
# this correspond to encoding with HF tokenizers, with bos=False
"--expected=41481,315,15671,512,9468,241,231,12937,8152,50730,198,9468,235,103,24118,39520,32013,26371,198,35212,3294,315,10948,45910,198,25732,5664,5761,1968,26788,198,9470,16895,22107,304,39435,198,9468,236,106,8519,4140,11050,73137,198",
],
data = ["@Meta-Llama-3.2-1B-Instruct//:tokenizer.json"],
tags = [
"manual",
],
)
# native_test(
# name = "test_tokenizer",
# src = "//zml/tokenizer:main",
# # Note: all Llama-3.x tokenizers are the same,
# # but using the 3.2-1B version because downloading the tokenizer triggers downloading the model.
# args = [
# "--tokenizer=$(location @Meta-Llama-3.2-1B-Instruct//:tokenizer.json)",
# """--prompt='Examples of titles:
# 📉 Stock Market Trends
# 🍪 Perfect Chocolate Chip Recipe
# Evolution of Music Streaming
# Remote Work Productivity Tips
# Artificial Intelligence in Healthcare
# 🎮 Video Game Development Insights
# '""",
# # this correspond to encoding with HF tokenizers, with bos=False
# "--expected=41481,315,15671,512,9468,241,231,12937,8152,50730,198,9468,235,103,24118,39520,32013,26371,198,35212,3294,315,10948,45910,198,25732,5664,5761,1968,26788,198,9470,16895,22107,304,39435,198,9468,236,106,8519,4140,11050,73137,198",
# ],
# data = ["@Meta-Llama-3.2-1B-Instruct//:tokenizer.json"],
# )
mtree_spec(
name = "mtree",
srcs = [":llama"],
tags = [
"manual",
],
)
tar(
@ -80,54 +70,48 @@ tar(
srcs = [":llama"],
args = [
"--options",
"zstd:compression-level=9",
",".join([
"zstd:compression-level=9",
"zstd:threads=16",
]),
],
compress = "zstd",
mtree = ":mtree",
tags = [
"manual",
],
)
oci_image(
name = "image_",
base = "@distroless_cc_debian12_debug",
entrypoint = ["./{}/llama".format(package_name())],
tags = [
"manual",
],
tars = [
"@zml//runtimes:layers",
":archive",
base = "@distroless_cc_debian12",
entrypoint = ["/{}/llama".format(package_name())],
target_compatible_with = [
"@platforms//os:linux",
],
tars = [":archive"],
)
platform_transition_filegroup(
name = "image",
srcs = [":image_"],
tags = [
"manual",
],
target_platform = "@zml//platforms:linux_amd64",
tags = ["manual"],
target_platform = "//platforms:linux_amd64",
)
oci_load(
name = "load",
image = ":image",
repo_tags = [
"distroless/llama:latest",
],
tags = [
"manual",
],
repo_tags = ["zmlai/llama:latest"],
tags = ["manual"],
)
oci_push(
name = "push",
image = ":image",
remote_tags = ["latest"],
repository = "index.docker.io/steeve/llama",
tags = [
"manual",
],
repository = "index.docker.io/zmlai/llama",
tags = ["manual"],
)
build_test(
name = "test",
targets = [":llama"],
)

106
examples/mnist/BUILD.bazel
View File

@ -1,108 +1,24 @@
load("@aspect_bazel_lib//lib:expand_template.bzl", "expand_template")
load("@aspect_bazel_lib//lib:tar.bzl", "mtree_spec", "tar")
load("@aspect_bazel_lib//lib:transitions.bzl", "platform_transition_filegroup")
load("@rules_oci//oci:defs.bzl", "oci_image", "oci_load", "oci_push")
load("@zml//bazel:zig.bzl", "zig_cc_binary")
load("@bazel_skylib//rules:build_test.bzl", "build_test")
load("//bazel:zig.bzl", "zig_cc_binary")
# Executable
zig_cc_binary(
name = "mnist",
args = [
"$(location @com_github_ggerganov_ggml_mnist//file)",
"$(location @com_github_ggerganov_ggml_mnist_data//file)",
"$(location @mnist//:mnist.pt)",
"$(location @mnist//:t10k-images.idx3-ubyte)",
],
data = [
"@com_github_ggerganov_ggml_mnist//file",
"@com_github_ggerganov_ggml_mnist_data//file",
"@mnist//:mnist.pt",
"@mnist//:t10k-images.idx3-ubyte",
],
main = "mnist.zig",
deps = [
"@zml//async",
"@zml//zml",
"//async",
"//zml",
],
)
mtree_spec(
name = "mtree",
srcs = [":mnist"],
)
tar(
name = "archive",
srcs = [":mnist"],
args = [
"--options",
"zstd:compression-level=9",
],
compress = "zstd",
mtree = ":mtree",
)
expand_template(
name = "entrypoint",
data = [
":mnist",
"@com_github_ggerganov_ggml_mnist//file",
"@com_github_ggerganov_ggml_mnist_data//file",
],
substitutions = {
":model": "$(rlocationpath @com_github_ggerganov_ggml_mnist//file)",
":data": "$(rlocationpath @com_github_ggerganov_ggml_mnist_data//file)",
},
template = [
"./{}/mnist".format(package_name()),
"./{}/mnist.runfiles/:model".format(package_name()),
"./{}/mnist.runfiles/:data".format(package_name()),
],
)
oci_image(
name = "image_",
base = "@distroless_cc_debian12",
entrypoint = ":entrypoint",
target_compatible_with = [
"@platforms//os:linux",
],
tars = [":archive"],
)
platform_transition_filegroup(
name = "image",
srcs = [":image_"],
target_compatible_with = [
"@platforms//os:linux",
],
target_platform = "@zml//platforms:linux_amd64",
)
oci_load(
name = "load",
image = ":image",
repo_tags = [
"distroless/mnist:latest",
],
target_compatible_with = [
"@platforms//os:linux",
],
)
oci_push(
name = "push",
image = ":image",
remote_tags = ["latest"],
repository = "index.docker.io/steeve/mnist",
target_compatible_with = [
"@platforms//os:linux",
],
)
oci_load(
name = "debug_image",
image = ":image",
repo_tags = [
"distroless/mnist:latest",
],
target_compatible_with = [
"@platforms//os:linux",
],
build_test(
name = "test",
targets = [":mnist"],
)

84
examples/modernbert/BUILD.bazel
View File

@ -1,70 +1,46 @@
load("@rules_cc//cc:cc_binary.bzl", "cc_binary")
load("@zml//bazel:zig.bzl", "zig_cc_binary")
load("@bazel_skylib//rules:build_test.bzl", "build_test")
load("//bazel:zig.bzl", "zig_cc_binary")
zig_cc_binary(
name = "modernbert",
srcs = ["modernbert.zig"],
main = "main.zig",
deps = [
"@com_github_hejsil_clap//:clap",
"@zml//async",
"@zml//stdx",
"@zml//zml",
],
)
cc_binary(
name = "ModernBERT-base",
args = [
"--model=$(location @ModernBERT-base//:model.safetensors)",
"--tokenizer=$(location @ModernBERT-base//:tokenizer)",
"--tokenizer=$(location @ModernBERT-base//:tokenizer.json)",
"--num-attention-heads=12",
"--tie-word-embeddings=true",
],
data = [
"@ModernBERT-base//:model.safetensors",
"@ModernBERT-base//:tokenizer",
],
tags = [
"manual",
],
deps = [":modernbert_lib"],
)
cc_binary(
name = "ModernBERT-large",
args = [
"--model=$(location @ModernBERT-large//:model.safetensors)",
"--tokenizer=$(location @ModernBERT-large//:tokenizer)",
"--num-attention-heads=16",
"--tie-word-embeddings=true",
],
data = [
"@ModernBERT-large//:model.safetensors",
"@ModernBERT-large//:tokenizer",
],
tags = [
"manual",
],
deps = [":modernbert_lib"],
)
zig_cc_binary(
name = "test-implementation",
srcs = ["modernbert.zig"],
args = [
"--model=$(location @ModernBERT-base//:model.safetensors)",
],
data = [
"@ModernBERT-base//:model.safetensors",
],
main = "test.zig",
tags = [
"manual",
"@ModernBERT-base//:tokenizer.json",
],
main = "main.zig",
deps = [
"//async",
"//stdx",
"//zml",
"@com_github_hejsil_clap//:clap",
"@zml//async",
"@zml//zml",
],
)
# zig_cc_binary(
# name = "test-implementation",
# srcs = ["modernbert.zig"],
# args = [
# "--model=$(location @ModernBERT-base//:model.safetensors)",
# ],
# data = [
# "@ModernBERT-base//:model.safetensors",
# ],
# main = "test.zig",
# deps = [
# "//async",
# "//zml",
# "@com_github_hejsil_clap//:clap",
# ],
# )
build_test(
name = "test",
targets = [":modernbert"],
)

35
examples/simple_layer/BUILD.bazel
View File

@ -1,14 +1,15 @@
load("@aspect_bazel_lib//lib:tar.bzl", "mtree_spec", "tar")
load("@aspect_bazel_lib//lib:transitions.bzl", "platform_transition_filegroup")
load("@bazel_skylib//rules:build_test.bzl", "build_test")
load("@rules_oci//oci:defs.bzl", "oci_image", "oci_load", "oci_push")
load("@zml//bazel:zig.bzl", "zig_cc_binary")
load("//bazel:zig.bzl", "zig_cc_binary")
zig_cc_binary(
name = "simple_layer",
main = "main.zig",
deps = [
"@zml//async",
"@zml//zml",
"//async",
"//zml",
],
)
@ -24,7 +25,10 @@ tar(
srcs = [":simple_layer"],
args = [
"--options",
"zstd:compression-level=9",
",".join([
"zstd:compression-level=9",
"zstd:threads=16",
]),
],
compress = "zstd",
mtree = ":mtree",
@ -34,7 +38,7 @@ tar(
oci_image(
name = "image_",
base = "@distroless_cc_debian12",
entrypoint = ["./{}/simple_layer".format(package_name())],
entrypoint = ["/{}/simple_layer".format(package_name())],
target_compatible_with = [
"@platforms//os:linux",
],
@ -45,10 +49,8 @@ oci_image(
platform_transition_filegroup(
name = "image",
srcs = [":image_"],
target_compatible_with = [
"@platforms//os:linux",
],
target_platform = "@zml//platforms:linux_amd64",
tags = ["manual"],
target_platform = "//platforms:linux_amd64",
)
# Load will immediatly load the image (eg: docker load)
@ -58,9 +60,7 @@ oci_load(
repo_tags = [
"distroless/simple_layer:latest",
],
target_compatible_with = [
"@platforms//os:linux",
],
tags = ["manual"],
)
# Bazel target for pushing the Linux image to the docker registry
@ -69,8 +69,11 @@ oci_push(
image = ":image",
remote_tags = ["latest"],
# override with -- --repository foo.bar/org/image
repository = "index.docker.io/renerocksai/simple_layer",
target_compatible_with = [
"@platforms//os:linux",
],
repository = "index.docker.io/zmlai/simple_layer",
tags = ["manual"],
)
build_test(
name = "test",
targets = [":simple_layer"],
)