marin-levanter 0.99__tar.gz

marin_levanter-0.99/PKG-INFO

@@ -0,0 +1,291 @@
+Metadata-Version: 2.3
+Name: marin-levanter
+Version: 0.99
+Summary: Scalable Training for Foundation Models with Named Tensors and JAX
+Author: David Hall, Jason Wang, Ahmed Ahmed, Ivan Zhou, Will Held, Virginia Adams
+Author-email: David Hall <dlwh@cs.stanford.edu>, Jason Wang <jsywang@cs.stanford.edu>, Ivan Zhou <ivanz@stanford.edu>
+Classifier: Programming Language :: Python :: 3
+Classifier: License :: OSI Approved :: Apache Software License
+Classifier: Operating System :: POSIX :: Linux
+Classifier: Operating System :: MacOS :: MacOS X
+Classifier: Development Status :: 4 - Beta
+Classifier: Intended Audience :: Science/Research
+Requires-Dist: marin-haliax==0.99
+Requires-Dist: equinox>=0.11.7,!=0.12.0
+Requires-Dist: jax>=0.9.2,<0.11
+Requires-Dist: marin-fray==0.99
+Requires-Dist: marin-rigging==0.99
+Requires-Dist: marin-zephyr==0.99
+Requires-Dist: einops
+Requires-Dist: jaxtyping>=0.2.34
+Requires-Dist: tokenizers>=0.15.2
+Requires-Dist: kitoken>=0.10.2
+Requires-Dist: transformers>=4.57.1,<5.0
+Requires-Dist: chex>=0.1.86
+Requires-Dist: optax>=0.1.9,<0.2.7
+Requires-Dist: wandb>0.24.0
+Requires-Dist: draccus>=0.11.5
+Requires-Dist: pyarrow>=11.0.0
+Requires-Dist: zstandard>=0.18.0
+Requires-Dist: datasets>=3.1.0,<5.0
+Requires-Dist: gcsfs>=2024.2,<2027
+Requires-Dist: braceexpand>=0.1.7
+Requires-Dist: chex>=0.1.86
+Requires-Dist: jmp>=0.0.3
+Requires-Dist: fsspec[http]>=2024.2,<2027
+Requires-Dist: tensorstore>=0.1.73,<0.1.82
+Requires-Dist: pytimeparse>=1.1.8
+Requires-Dist: humanfriendly==10.0
+Requires-Dist: safetensors[numpy]>=0.4.2,<0.7.0
+Requires-Dist: tblib>=1.7.0,<4.0.0
+Requires-Dist: dataclasses-json~=0.6.4
+Requires-Dist: pydantic<3
+Requires-Dist: filelock~=3.13
+Requires-Dist: async-lru~=2.0
+Requires-Dist: tqdm-loggable>=0.2
+Requires-Dist: deepdiff>=8.6.2
+Requires-Dist: lenses
+Requires-Dist: jinja2
+Requires-Dist: protobuf>=6
+Requires-Dist: immutabledict
+Requires-Dist: google-api-python-client>=2.175.0 ; extra == 'gcp'
+Requires-Dist: google-cloud-storage ; extra == 'gcp'
+Requires-Dist: google-cloud-storage-transfer ; extra == 'gcp'
+Requires-Dist: google-auth ; extra == 'gcp'
+Requires-Dist: jax[cuda13]==0.10.0 ; extra == 'gpu'
+Requires-Dist: nvidia-cublas>=13.2.0.9 ; sys_platform == 'linux' and extra == 'gpu'
+Requires-Dist: nvidia-nccl-cu13>=2.28.3 ; sys_platform == 'linux' and extra == 'gpu'
+Requires-Dist: tokamax ; extra == 'kernels'
+Requires-Dist: kitoken>=0.10.2 ; extra == 'kitoken'
+Requires-Dist: xprof ; extra == 'profiling'
+Requires-Dist: tensorboard>=2.16 ; extra == 'profiling'
+Requires-Dist: tensorboardx>=2.6 ; extra == 'profiling'
+Requires-Dist: fastapi>=0.100.0 ; extra == 'serve'
+Requires-Dist: uvicorn[standard]>=0.23.0 ; extra == 'serve'
+Requires-Dist: openai>=1.0.0 ; extra == 'serve'
+Requires-Dist: torch>=2.7.0 ; extra == 'torch-test'
+Requires-Dist: peft>=0.12.0 ; extra == 'torch-test'
+Requires-Dist: jax==0.9.2 ; extra == 'tpu'
+Requires-Dist: jaxlib==0.9.2 ; extra == 'tpu'
+Requires-Dist: libtpu==0.0.38 ; extra == 'tpu'
+Requires-Python: >=3.11
+Project-URL: Bug Tracker, https://github.com/stanford-crfm/levanter/issues
+Project-URL: Homepage, https://github.com/stanford-crfm/levanter
+Provides-Extra: gcp
+Provides-Extra: gpu
+Provides-Extra: kernels
+Provides-Extra: kitoken
+Provides-Extra: lm-eval
+Provides-Extra: profiling
+Provides-Extra: serve
+Provides-Extra: torch-test
+Provides-Extra: tpu
+Description-Content-Type: text/markdown
+
+# Levanter
+
+Levanter is developed and released from the
+[marin-community/marin](https://github.com/marin-community/marin) monorepo
+(`lib/levanter`) and published to PyPI as **`marin-levanter`**. This README
+documents the package as shipped from that monorepo.
+
+<a href="https://levanter.readthedocs.io/en/latest/?badge=latest">
+    <img alt="Documentation Status" src="https://readthedocs.org/projects/levanter/badge/?version=latest">
+</a>
+<a href="https://pypi.org/project/marin-levanter/">
+    <img alt="PyPI" src="https://img.shields.io/pypi/v/marin-levanter?color=blue" />
+</a>
+
+
+<!--levanter-intro-start-->
+> *You could not prevent a thunderstorm, but you could use the electricity; you could not direct the wind, but you could trim your sail so as to propel your vessel as you pleased, no matter which way the wind blew.* <br/>
+> — Cora L. V. Hatch
+
+Levanter is a framework for training large language models (LLMs) and other foundation models that strives for legibility, scalability, and reproducibility:
+
+1. **Legible**: Levanter uses our named tensor library [Haliax](https://github.com/stanford-crfm/haliax) to write easy-to-follow, composable deep learning code, while still being high performance.
+2. **Scalable**: Levanter scales to large models, and to be able to train on a variety of hardware, including GPUs and TPUs.
+3. **Reproducible**: Levanter is bitwise deterministic, meaning that the same configuration will always produce the same results, even in the face of preemption and resumption.
+
+We built Levanter with [JAX](https://github.com/jax-ml/jax), [Equinox](https://github.com/patrick-kidger/equinox), and [Haliax](https://github.com/stanford-crfm/haliax).
+
+## Documentation
+
+Levanter's documentation is available at [levanter.readthedocs.io](https://levanter.readthedocs.io/en/latest/).
+Haliax's documentation is available at [haliax.readthedocs.io](https://haliax.readthedocs.io/en/latest/).
+
+## Features
+
+* **Distributed Training**: We support distributed training on TPUs and GPUs, including FSDP and tensor parallelism.
+* **Compatibility**: Levanter supports importing and exporting models to/from the Hugging Face ecosystem, including tokenizers, datasets, and models via [SafeTensors](https://github.com/huggingface/safetensors).
+* **Performance**: Levanter's performance rivals commercially-backed frameworks like MosaicML's Composer or Google's MaxText.
+* **Resilience**: Levanter supports fast, distributed checkpointing and fast resume from checkpoints with no data seek, making Levanter robust to preemption and hardware failure.
+* **Cached On-Demand Data Preprocessing**: We preprocess corpora online, but we cache the results of preprocessing so
+that resumes are much faster and so that subsequent runs are even faster. As soon as the first part of the cache is complete, Levanter will start training.
+* **Logging**: Levanter logs a rich and detailed set of metrics covering loss and performance. Levanter also supports a few different logging backends, including [WandB](https://wandb.ai/site) and [TensorBoard](https://www.tensorflow.org/tensorboard). (Adding a new logging backend is easy!) Levanter even exposes the ability
+to log inside of JAX `jit`-ted functions.
+* **Reproducibility**: On TPU, Levanter is bitwise deterministic, meaning that the same configuration will always produce the same results, even in the face of preemption and resumption.
+* **Distributed Checkpointing**: Distributed checkpointing is supported via Google's [TensorStore](https://google.github.io/tensorstore/) library. Training can even be resumed on a different number of hosts, though this breaks reproducibility for now.
+* **Optimization**: We support [Optax](https://github.com/deepmind/optax) for optimization with AdamW, as well as newer optimizers like Muon, SOAP, and more.
+* **Flexible**: Levanter supports tuning data mixtures without having to retokenize or shuffle data.
+
+<!--levanter-intro-end-->
+
+Levanter was created by [Stanford's Center for Research on Foundation Models (CRFM)](https://crfm.stanford.edu/)'s research engineering team.
+You can also find us in the #levanter channel on the unofficial [Jax LLM Discord](https://discord.gg/CKazXcbbBm)
+
+## Getting Started
+
+Here is a small set of examples to get you started. For more information about the various configuration options,
+please see the [Getting Started](./docs/Getting-Started-Training.md) guide or the [In-Depth Configuration Guide](docs/reference/Configuration.md).
+You can also use `--help` or poke around other configs to see all the options available to you.
+
+
+### Installing Levanter
+
+<!--levanter-installation-start-->
+
+After [installing JAX](https://github.com/google/jax/blob/main/README.md#installation) with the appropriate configuration
+for your platform, install Levanter from PyPI:
+
+```bash
+pip install marin-levanter
+wandb login  # optional, we use wandb for logging
+```
+
+For development, clone the marin monorepo and use `uv sync` to install Levanter
+alongside its sibling packages (Haliax, Iris, etc.) in editable form:
+
+```bash
+git clone https://github.com/marin-community/marin.git
+cd marin
+uv sync
+```
+
+<!--levanter-installation-end-->
+
+Please refer to the [Installation Guide](docs/Installation.md) for more information on how to install Levanter.
+
+If you're using a TPU, more complete documentation for setting that up is available [here](docs/Getting-Started-TPU-VM.md). GPU support is still in-progress; documentation is available [here](docs/Getting-Started-GPU.md).
+
+<!--levanter-user-guide-start-->
+
+### Training a GPT2-nano
+
+As a kind of hello world, here's how you can train a GPT-2 "nano"-sized model on a small dataset.
+
+```bash
+python -m levanter.main.train_lm --config_path config/gpt2_nano.yaml
+
+# alternatively, if you didn't use -e and are in a different directory
+python -m levanter.main.train_lm --config_path gpt2_nano
+```
+
+This will train a GPT2-nano model on the [WikiText-103](https://huggingface.co/datasets/Salesforce/wikitext) dataset.
+
+### Training a Llama-small on your own data
+
+You can also change the dataset by changing the `dataset` field in the config file.
+If your dataset is a [Hugging Face dataset](https://huggingface.co/docs/datasets/loading_datasets.html), you can use the `data.id` field to specify it:
+
+```bash
+python -m levanter.main.train_lm --config_path config/llama_small_fast.yaml --data.id openwebtext
+
+# optionally, you may specify a tokenizer and/or a cache directory, which may be local or on gcs
+python -m levanter.main.train_lm --config_path config/llama_small_fast.yaml --data.id openwebtext --data.tokenizer "NousResearch/Llama-2-7b-hf" --data.cache_dir "gs://path/to/cache/dir"
+```
+
+If instead your data is a list of URLs, you can use the `data.train_urls` and `data.validation_urls` fields to specify them.
+Data URLS can be local files, gcs files, or http(s) URLs, or anything that [fsspec](https://filesystem-spec.readthedocs.io/en/latest/) supports.
+Levanter (really, fsspec) will automatically uncompress `.gz` and `.zstd` files, and probably other formats too.
+
+```bash
+python -m levanter.main.train_lm --config_path config/llama_small_fast.yaml --data.train_urls ["https://path/to/train/data_*.jsonl.gz"] --data.validation_urls ["https://path/to/val/data_*.jsonl.gz"]
+```
+
+### Customizing a Config File
+
+You can modify the config file to change the model, the dataset, the training parameters, and more. Here's
+the `llama_small_fast.yaml` file:
+
+```yaml
+data:
+  train_urls:
+      - "gs://pubmed-mosaic/openwebtext-sharded/openwebtext_train.{1..128}-of-128.jsonl.gz"
+  validation_urls:
+      - "gs://pubmed-mosaic/openwebtext-sharded/openwebtext_val.{1..8}-of-8.jsonl.gz"
+  cache_dir: "gs://pubmed-mosaic/tokenized/openwebtext/"
+model:
+  type: llama
+  hidden_dim: 768
+  intermediate_dim: 2048
+  num_heads: 12
+  num_kv_heads: 12
+  num_layers: 12
+  seq_len: 1024
+  gradient_checkpointing: true
+trainer:
+  tracker:
+    type: wandb
+    project: "levanter"
+    tags: [ "openwebtext", "llama" ]
+
+  mp: p=f32,c=bfloat16
+  mesh:
+    axes: {data: -1, replica: 1, model: 1}   # inherited defaults; override if you need TP
+  per_device_parallelism: 4
+
+  train_batch_size: 512
+optimizer:
+  learning_rate: 6E-4
+  weight_decay: 0.1
+  min_lr_ratio: 0.1
+```
+
+### Other Architectures
+
+Currently, we support the following architectures:
+
+* GPT-2
+* [LLama](https://ai.meta.com/llama/), including Llama 1, 2 and 3
+* [Gemma](https://ai.google.dev/gemma), including Gemma 1, 2 and Gemma 3.
+* [Qwen2](https://huggingface.co/Qwen/Qwen2.5-7B)
+* [Qwen3](https://huggingface.co/Qwen/Qwen3-8B)
+* [Mistral](https://huggingface.co/mistralai/Mistral-7B-Instruct-v0.3)
+* [Mixtral](https://huggingface.co/mistralai/Mixtral-8x7B-Instruct-v0.1)
+* [Olmo2](https://huggingface.co/allenai/Olmo-2-1124-7B)
+
+We plan to add more in the future.
+
+For speech, we currently only support [Whisper](https://huggingface.co/openai/whisper-large-v3).
+
+#### Continued Pretraining with Llama
+
+Here's an example of how to continue pretraining a Llama 1 or Llama 2 model on the OpenWebText dataset:
+
+```bash
+python -m levanter.main.train_lm --config_path config/llama2_7b_continued.yaml
+```
+
+
+## Distributed and Cloud Training
+
+### Training on a TPU Cloud VM
+
+Please see the [TPU Getting Started](docs/Getting-Started-TPU-VM.md) guide for more information on how to set up a TPU Cloud VM and run Levanter there.
+
+### Training with CUDA
+
+Please see the [CUDA Getting Started](docs/Getting-Started-GPU.md) guide for more information on how to set up a CUDA environment and run Levanter there.
+
+<!--levanter-user-guide-end-->
+
+## Contributing
+
+We welcome contributions! Please see [CONTRIBUTING.md](CONTRIBUTING.md) for
+more information. Issues and pull requests are tracked at
+[marin-community/marin](https://github.com/marin-community/marin/issues).
+
+## License
+
+Levanter is licensed under the Apache License, Version 2.0. See [LICENSE](LICENSE) for the full license text.

marin_levanter-0.99/README.md

@@ -0,0 +1,207 @@
+# Levanter
+
+Levanter is developed and released from the
+[marin-community/marin](https://github.com/marin-community/marin) monorepo
+(`lib/levanter`) and published to PyPI as **`marin-levanter`**. This README
+documents the package as shipped from that monorepo.
+
+<a href="https://levanter.readthedocs.io/en/latest/?badge=latest">
+    <img alt="Documentation Status" src="https://readthedocs.org/projects/levanter/badge/?version=latest">
+</a>
+<a href="https://pypi.org/project/marin-levanter/">
+    <img alt="PyPI" src="https://img.shields.io/pypi/v/marin-levanter?color=blue" />
+</a>
+
+
+<!--levanter-intro-start-->
+> *You could not prevent a thunderstorm, but you could use the electricity; you could not direct the wind, but you could trim your sail so as to propel your vessel as you pleased, no matter which way the wind blew.* <br/>
+> — Cora L. V. Hatch
+
+Levanter is a framework for training large language models (LLMs) and other foundation models that strives for legibility, scalability, and reproducibility:
+
+1. **Legible**: Levanter uses our named tensor library [Haliax](https://github.com/stanford-crfm/haliax) to write easy-to-follow, composable deep learning code, while still being high performance.
+2. **Scalable**: Levanter scales to large models, and to be able to train on a variety of hardware, including GPUs and TPUs.
+3. **Reproducible**: Levanter is bitwise deterministic, meaning that the same configuration will always produce the same results, even in the face of preemption and resumption.
+
+We built Levanter with [JAX](https://github.com/jax-ml/jax), [Equinox](https://github.com/patrick-kidger/equinox), and [Haliax](https://github.com/stanford-crfm/haliax).
+
+## Documentation
+
+Levanter's documentation is available at [levanter.readthedocs.io](https://levanter.readthedocs.io/en/latest/).
+Haliax's documentation is available at [haliax.readthedocs.io](https://haliax.readthedocs.io/en/latest/).
+
+## Features
+
+* **Distributed Training**: We support distributed training on TPUs and GPUs, including FSDP and tensor parallelism.
+* **Compatibility**: Levanter supports importing and exporting models to/from the Hugging Face ecosystem, including tokenizers, datasets, and models via [SafeTensors](https://github.com/huggingface/safetensors).
+* **Performance**: Levanter's performance rivals commercially-backed frameworks like MosaicML's Composer or Google's MaxText.
+* **Resilience**: Levanter supports fast, distributed checkpointing and fast resume from checkpoints with no data seek, making Levanter robust to preemption and hardware failure.
+* **Cached On-Demand Data Preprocessing**: We preprocess corpora online, but we cache the results of preprocessing so
+that resumes are much faster and so that subsequent runs are even faster. As soon as the first part of the cache is complete, Levanter will start training.
+* **Logging**: Levanter logs a rich and detailed set of metrics covering loss and performance. Levanter also supports a few different logging backends, including [WandB](https://wandb.ai/site) and [TensorBoard](https://www.tensorflow.org/tensorboard). (Adding a new logging backend is easy!) Levanter even exposes the ability
+to log inside of JAX `jit`-ted functions.
+* **Reproducibility**: On TPU, Levanter is bitwise deterministic, meaning that the same configuration will always produce the same results, even in the face of preemption and resumption.
+* **Distributed Checkpointing**: Distributed checkpointing is supported via Google's [TensorStore](https://google.github.io/tensorstore/) library. Training can even be resumed on a different number of hosts, though this breaks reproducibility for now.
+* **Optimization**: We support [Optax](https://github.com/deepmind/optax) for optimization with AdamW, as well as newer optimizers like Muon, SOAP, and more.
+* **Flexible**: Levanter supports tuning data mixtures without having to retokenize or shuffle data.
+
+<!--levanter-intro-end-->
+
+Levanter was created by [Stanford's Center for Research on Foundation Models (CRFM)](https://crfm.stanford.edu/)'s research engineering team.
+You can also find us in the #levanter channel on the unofficial [Jax LLM Discord](https://discord.gg/CKazXcbbBm)
+
+## Getting Started
+
+Here is a small set of examples to get you started. For more information about the various configuration options,
+please see the [Getting Started](./docs/Getting-Started-Training.md) guide or the [In-Depth Configuration Guide](docs/reference/Configuration.md).
+You can also use `--help` or poke around other configs to see all the options available to you.
+
+
+### Installing Levanter
+
+<!--levanter-installation-start-->
+
+After [installing JAX](https://github.com/google/jax/blob/main/README.md#installation) with the appropriate configuration
+for your platform, install Levanter from PyPI:
+
+```bash
+pip install marin-levanter
+wandb login  # optional, we use wandb for logging
+```
+
+For development, clone the marin monorepo and use `uv sync` to install Levanter
+alongside its sibling packages (Haliax, Iris, etc.) in editable form:
+
+```bash
+git clone https://github.com/marin-community/marin.git
+cd marin
+uv sync
+```
+
+<!--levanter-installation-end-->
+
+Please refer to the [Installation Guide](docs/Installation.md) for more information on how to install Levanter.
+
+If you're using a TPU, more complete documentation for setting that up is available [here](docs/Getting-Started-TPU-VM.md). GPU support is still in-progress; documentation is available [here](docs/Getting-Started-GPU.md).
+
+<!--levanter-user-guide-start-->
+
+### Training a GPT2-nano
+
+As a kind of hello world, here's how you can train a GPT-2 "nano"-sized model on a small dataset.
+
+```bash
+python -m levanter.main.train_lm --config_path config/gpt2_nano.yaml
+
+# alternatively, if you didn't use -e and are in a different directory
+python -m levanter.main.train_lm --config_path gpt2_nano
+```
+
+This will train a GPT2-nano model on the [WikiText-103](https://huggingface.co/datasets/Salesforce/wikitext) dataset.
+
+### Training a Llama-small on your own data
+
+You can also change the dataset by changing the `dataset` field in the config file.
+If your dataset is a [Hugging Face dataset](https://huggingface.co/docs/datasets/loading_datasets.html), you can use the `data.id` field to specify it:
+
+```bash
+python -m levanter.main.train_lm --config_path config/llama_small_fast.yaml --data.id openwebtext
+
+# optionally, you may specify a tokenizer and/or a cache directory, which may be local or on gcs
+python -m levanter.main.train_lm --config_path config/llama_small_fast.yaml --data.id openwebtext --data.tokenizer "NousResearch/Llama-2-7b-hf" --data.cache_dir "gs://path/to/cache/dir"
+```
+
+If instead your data is a list of URLs, you can use the `data.train_urls` and `data.validation_urls` fields to specify them.
+Data URLS can be local files, gcs files, or http(s) URLs, or anything that [fsspec](https://filesystem-spec.readthedocs.io/en/latest/) supports.
+Levanter (really, fsspec) will automatically uncompress `.gz` and `.zstd` files, and probably other formats too.
+
+```bash
+python -m levanter.main.train_lm --config_path config/llama_small_fast.yaml --data.train_urls ["https://path/to/train/data_*.jsonl.gz"] --data.validation_urls ["https://path/to/val/data_*.jsonl.gz"]
+```
+
+### Customizing a Config File
+
+You can modify the config file to change the model, the dataset, the training parameters, and more. Here's
+the `llama_small_fast.yaml` file:
+
+```yaml
+data:
+  train_urls:
+      - "gs://pubmed-mosaic/openwebtext-sharded/openwebtext_train.{1..128}-of-128.jsonl.gz"
+  validation_urls:
+      - "gs://pubmed-mosaic/openwebtext-sharded/openwebtext_val.{1..8}-of-8.jsonl.gz"
+  cache_dir: "gs://pubmed-mosaic/tokenized/openwebtext/"
+model:
+  type: llama
+  hidden_dim: 768
+  intermediate_dim: 2048
+  num_heads: 12
+  num_kv_heads: 12
+  num_layers: 12
+  seq_len: 1024
+  gradient_checkpointing: true
+trainer:
+  tracker:
+    type: wandb
+    project: "levanter"
+    tags: [ "openwebtext", "llama" ]
+
+  mp: p=f32,c=bfloat16
+  mesh:
+    axes: {data: -1, replica: 1, model: 1}   # inherited defaults; override if you need TP
+  per_device_parallelism: 4
+
+  train_batch_size: 512
+optimizer:
+  learning_rate: 6E-4
+  weight_decay: 0.1
+  min_lr_ratio: 0.1
+```
+
+### Other Architectures
+
+Currently, we support the following architectures:
+
+* GPT-2
+* [LLama](https://ai.meta.com/llama/), including Llama 1, 2 and 3
+* [Gemma](https://ai.google.dev/gemma), including Gemma 1, 2 and Gemma 3.
+* [Qwen2](https://huggingface.co/Qwen/Qwen2.5-7B)
+* [Qwen3](https://huggingface.co/Qwen/Qwen3-8B)
+* [Mistral](https://huggingface.co/mistralai/Mistral-7B-Instruct-v0.3)
+* [Mixtral](https://huggingface.co/mistralai/Mixtral-8x7B-Instruct-v0.1)
+* [Olmo2](https://huggingface.co/allenai/Olmo-2-1124-7B)
+
+We plan to add more in the future.
+
+For speech, we currently only support [Whisper](https://huggingface.co/openai/whisper-large-v3).
+
+#### Continued Pretraining with Llama
+
+Here's an example of how to continue pretraining a Llama 1 or Llama 2 model on the OpenWebText dataset:
+
+```bash
+python -m levanter.main.train_lm --config_path config/llama2_7b_continued.yaml
+```
+
+
+## Distributed and Cloud Training
+
+### Training on a TPU Cloud VM
+
+Please see the [TPU Getting Started](docs/Getting-Started-TPU-VM.md) guide for more information on how to set up a TPU Cloud VM and run Levanter there.
+
+### Training with CUDA
+
+Please see the [CUDA Getting Started](docs/Getting-Started-GPU.md) guide for more information on how to set up a CUDA environment and run Levanter there.
+
+<!--levanter-user-guide-end-->
+
+## Contributing
+
+We welcome contributions! Please see [CONTRIBUTING.md](CONTRIBUTING.md) for
+more information. Issues and pull requests are tracked at
+[marin-community/marin](https://github.com/marin-community/marin/issues).
+
+## License
+
+Levanter is licensed under the Apache License, Version 2.0. See [LICENSE](LICENSE) for the full license text.