Skip to content

Releases: mosaicml/llm-foundry

v0.10.0

02 Jul 13:31
@b-chu b-chu
v0.10.0
742f340
This commit was signed with the committer’s verified signature.
dakinggg Daniel King
SSH Key Fingerprint: Jb5bzWxVcrTuIyarRLYeq58aneGbft0BV3CZ8qHca/U Learn about vigilant mode.
Compare
Choose a tag to compare
v0.10.0

🚀 LLM Foundry v0.10.0

New Features

Registry for ICL datasets (#1252)

ICL datasets have now been added as a registry.

Curriculum Learning Callback (#1256)

You can now switch dataloaders while training which enables curriculum learning.

train_loader:
  <dataloader parameters>
callback:
  curriculum_learning:
  - duration: <number>tok
    train_loader:  # matches top level train_loader
      <dataloader parameters>
  - duration: <number>tok
    train_loader:
      <dataloader parameters>
  - duration: <number>tok
    train_loader:
      <dataloader parameters>

[Experimental] Interweave Attention Layers (#1299)

You can now override default block configs for certain layers, allowing for different sliding window sizes, reusing the previous layer's kv cache, etc.

model:
    ...
    (usual model configs)
    ...
    block_overrides:
        order:
        - name: default
        - order:
          - name: sliding_window_layer
          - name: sliding_window_layer_reuse
          - name: sliding_window_layer
          - repeat: 2
            name: sliding_window_layer_reuse
          - name: reuse_kv_layer
          repeat: 2
        overrides:
            sliding_window_layer:
                attn_config:
                    sliding_window_size: 1024
            sliding_window_layer_reuse:
                attn_config:
                    sliding_window_size: 1024
                    reuse_kv_layer_idx: -1 # Relative index of the layer whose kv cache to reuse
            reuse_kv_layer:
                attn_config:
                    reuse_kv_layer_idx: -6 # Relative index of the layer whose kv cache to reuse

Bug fixes

What's Changed

New Contributors

Full Changelog: v0.9.1...v0.10.0

Contributors

sjawhar, j316chuck, and 11 other contributors
Assets 2
Loading

v0.9.1

24 Jun 23:00
@dakinggg dakinggg
v0.9.1
1402b7e
Compare
Choose a tag to compare
v0.9.1

🚀 LLM Foundry v0.9.1

This is a minor patch release to bump the minimum version of mlflow to make sure to buffer writes (mosaicml/composer#3401)

Whats changed

Full Changelog: v0.9.0...v0.9.1

Contributors

JackZ-db
Assets 2
Loading

v0.9.0

08 Jun 04:58
@KuuCi KuuCi
v0.9.0
e2d68f6
Compare
Choose a tag to compare
v0.9.0

🚀 LLM Foundry v0.9.0

New Features

More Token Encoding Types (#1254)

We've expanded the different ways to encode token IDs by allowing uint32 and uint16 formats, which saves significant space for datasets with smaller vocab sizes. We also extended ndarray type support for MDS dataset columns to the generic text dataset and updated conversion scripts accordingly.

Enforced Stricter Configs (#1254, #1225, #1202)

We've implemented stricter enforcement on our Train and Eval configs to further protect users from attempting to train with invalid configs. In conjunction with numerous other PRs, we have stronger error handling to help users use LLM Foundry smoothly.

Previously, this was allowed:

parameters:
   train_dataloader:
      ...
      seed: ${global_seed}
      random_other_key_that's_not_in_the_dataloader_constructor # this is not allowed
   ...
   global_seed: 17 # this is also not allowed

But we've added a variables section. Please do this instead:

parameters:
  variables:
    global_seed: 42
  ...
  train_dataloader:
    seed: ${variables.global_seed}

Chunked text to mds conversion (#1240)

We've updated our text to mds to convertion script to convert files to MDS in chunks. This protects us from loading entire large files at once (potentially causing OOMs), and drastically speeds up converting long sequences.

Breaking Changes and Deprecations

What's Changed

Full Changelog: v0.8.0...v0.9.0

Contributors

j316chuck, irenedea, and 10 other contributors
Assets 2
Loading

v0.8.0

08 May 01:36
@milocress milocress
v0.8.0
57cdd2a
Compare
Choose a tag to compare
v0.8.0

🚀 LLM Foundry v0.8.0

New Features

Megablocks support (#1102)

Support for training optimized MoE models at large scale.

Check out the megablocks documentation for more information on building state of the art MoE models.

Expanded Registries (#1080, #1093, #1094, #1095, #1096, #1165)

We've expanded support for registries to include, dataloaders, FFN layers, attention layers, norms, and parameter initialization functions.

Check out the README for detailed instructions and code examples!

Support for ShareGPT chat format (#1098)

We now support the ShareGPT format for finetuning.

Breaking Changes and Deprecations

We have updated the minimum supported PyTorch version to torch 2.3 (#1152).

In Context Learning Code Evaluation (#1181)

We've removed the code_evaluation task from the allowed in context learning task types, and we've deleted the InContextLearningCodeEvaluationDataset and InContextLearningCodeEvalAccuracy classes.

Question-Answering

We've removed the question_answering task type. Please use the generation_task_with_answers task instead.

What's Changed

New Contributors

Read more

Contributors

lupesko, vchiley, and 22 other contributors
Assets 2
Loading

v0.7.0

27 Mar 05:12
@irenedea irenedea
v0.7.0
f044d6c
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Learn about vigilant mode.
Compare
Choose a tag to compare
v0.7.0

🚀 LLM Foundry v0.7.0

LLM Foundry is an efficient codebase for training, evaluating, and deploying Large Language Models (LLMs) and serves as the foundation for the MPT model series.

In addition to the usual bug fixes and performance improvements, we've made foundry more customizable and extensible!

New Features

Registerable Components (#975, #1043, #1052, #1057)

We've made key components of LLM Foundry registrable, such as models, loggers, and callbacks. You can use the registry to easily customize and extend your training workflows.

This means that you can register new options for these components, and then use them in your yaml config.

Check out the README for detailed instructions and code examples!

Breaking Changes and Deprecations

Deprecated Feature Removals (#1063)

We've removed support for deprecated features: triton attention, Prefix LMs, Llama attention patch, z-loss, and text denoising. These features were little used, and we removed them to focus on the core features that are heavily used.

If you were using these features please let us know how you were using them in a GitHub issue. We're happy to add things back that are in heavy usage.

What's Changed

New Contributors

Full Changelog: v0.6.0...v0.7.0

Contributors

dennyglee, sashaDoubov, and 6 other contributors
Assets 2
Loading

v0.6.0

12 Mar 20:22
@dakinggg dakinggg
v0.6.0
08359b5
Compare
Choose a tag to compare
v0.6.0

🚀 LLM Foundry v0.6.0

LLM Foundry is an efficient codebase for training, evaluating, and deploying Large Language Models (LLMs) and serves as the foundation for the MPT model series.

In addition to the usual bug fixes and performance improvements, we've added lots of new features!

New Features

Configurable loss for chat-formatted data (#985)

For chat-formatted data, you can now specify which tokens should be loss-generating in a configurable way.

This can be specified in the train_loader.dataset section of your yaml as follows:

...
train_loader:
  dataset:
    ...
    target_prompts: <FILL IN>
    target_reseponses: <FILL IN>

See the docstring for a description of the options.

Olmo support (#1016)

We've added support for the OLMo model from AI2.

To use OLMo, there are a few configuration parameters you need to set. First of all, you will need to install LLM Foundry with the extra package for OLMo (pip install .[gpu,olmo]).

Then you will need to adjust the tokenizer section of your config as follows:

tokenizer:
  name: allenai/OLMo-7B
  kwargs:
    revision: main
    model_max_length: 2048
    model_input_names:
    - input_ids
    - attention_mask
    trust_remote_code: true

Token accuracy (#983)

We've added a new, on-by-default metric to compute token accuracy in addition to cross entropy and perplexity.

Configurable activation checkpointing (#951)

More configurable activation checkpointing for MPT allows finer granular control over memory usage when training MPT. See the docstring for more details.

Finetuning with multiple streams, and pretokenized data (#933, #945, #946)

We've brought the finetuning dataloader up to speed with the pretraining dataloader to support mixing multiple streams, and pretokenizing finetuning data. See the yaml for a full example.

Eval Gauntlet v0.3 (#824)

We've release v0.3 of our Evaluation gauntlet. See the README for a full description.

Breaking changes and deprecations

Flash attention v1 removal (#1023)

Support for flash attention v1 has now been removed.

Extra BOS token removed (#1003)

When tokenizing prompt/response and chat data, for some tokenizers, we were mistakenly adding an extra BOS token between the prompt and the response. This has now been removed.

Deprecation of triton flash attention, prefixLM, and text denoising (#1007)

We've deprecated use of the triton version of flash attention, prefixLM, and text denoising, as these features were not heavily used or actively maintained.

What's Changed

New Contributors

Full Changelog: v0.5.0...v0.6.0

Contributors

bigning, alextrott16, and 13 other contributors
Assets 2
Loading

v0.5.0

08 Feb 00:01
@irenedea irenedea
v0.5.0
a667ebf
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Learn about vigilant mode.
Compare
Choose a tag to compare
v0.5.0

🚀 LLM Foundry v0.5.0

LLM Foundry is an efficient codebase for training, evaluating, and deploying Large Language Models (LLMs) and serves as the foundation for the MPT model series.

In addition to the usual bug fixes and performance improvements, we've added lots of new features!

New Features

LoRA Support (with FSDP!) (#886)

LLM Foundry now supports LoRA via an integration with the PEFT library. Within LLM Foundry, run train.py, adding peft_config arguments to the model section of the config .yaml, like so:

model:
  ...
  peft_config:
      r: 16
      peft_type: LORA
      task_type: CAUSAL_LM
      lora_alpha: 32
      lora_dropout: 0.05
      target_modules:
      - q_proj
      - k_proj

Read more about it in the tutorial.

ALiBi for Flash Attention (#820)

We've added support for using ALiBi with Flash Attention (v2.4.2 or higher).

model:
     ...
     attn_config:
         attn_impl: flash
         alibi: True

Chat Data for Finetuning (#884)

We now support finetuning on chat data, with automatic formatting applied using Hugging Face tokenizer chat templates.

Each sample requires a single key "messages" that maps to an array of message objects. Each message object in the array represents a single message in the conversation and must contain the following keys:

  • role : A string indicating the author of the message. Possible values are "system" ,"user" , and "assistant" .
  • content : A string containing the text of the message.

We require that there must be at least one message with the role "assistant", and the last message in the "messages" array must have the role "assistant" .

Here's an example .jsonl with chat data:


{ "messages": [ { "role": "user", "content": "Hi, MPT!" }, { "role": "assistant", "content": "Hi, user!" } ]}
{ "messages": [ 
  { "role": "system": "A conversation between a user and a helpful and honest assistant"}
  { "role": "user", "content": "Hi, MPT!" }, 
  { "role": "assistant", "content": "Hi, user!" },
  { "role": "user", "content": "Is multi-turn chat supported?"},
  { "role": "assistant", "content": "Yes, we can chat for as long as my context length allows." }
]}
...

Safe Load for HuggingFace Datasets (#798)

We now provide a safe_load option when loading HuggingFace datasets for finetuning.

This restricts loaded files to .jsonl, .csv, or .parquet extensions to prevent arbitrary code execution.

To use, set safe_load to true in your dataset configuration:

  train_loader:
    name: finetuning
    dataset:
      safe_load: true
      ...

New PyTorch, Composer, Streaming, and Transformers versions

As always, we've updated to new versions of the core dependencies of LLM Foundry, bringing better performance, new features, and support for new models (mixtral in particular).

Deprecations

Support for Flash Attention v1 (#921)

Will be removed in v0.6.0.

Breaking Changes

Removed support for PyTorch versions before 2.1 (#787)

We no longer support PyTorch versions before 2.1.

Removed Deprecated Features (#948)

We've removed features that have been deprecated for at least one release.

What's Changed

Read more

Contributors

Skylion007, sashaDoubov, and 23 other contributors
Assets 2
Loading

v0.4.0

22 Nov 03:45
@dakinggg dakinggg
v0.4.0
4638092
Compare
Choose a tag to compare
v0.4.0

🚀 LLM Foundry v0.4.0

LLM Foundry is an efficient codebase for training, evaluating, and deploying Large Language Models (LLMs) and serves as the foundation for the MPT-7B and MPT-30B models.

In addition to the usual bug fixes and performance improvements, we've added lots of new features!

New Features

Automatic sequence packing (#683)

You can now specify packing_ratio: auto under your finetuning dataset, to automatically profile and select a good packing ratio to efficiently pack your sequences together on the fly during finetuning. This can dramatically reduce the amount of compute wasted on padding tokens.

Flash Attention 2 (#651, #666, #672)

We now support using Flash Attention 2 both in MPT and in any model that supports Flash Attention 2 via the Transformers library. See the training instructions to learn how to use the different versions of Flash Attention.

New PyTorch, Composer, Streaming, and Transformers versions (#648, #672, #736)

As always, we've updated to new versions of the core dependencies of LLM Foundry, bringing better performance, new features, and support for new models (codellama and mistral in particular).

Easy Databricks model deployment (#618)

We've made it much easier to go from a training run to a served model using Databricks model serving. To make use of this feature, you need to specify both an MLFlowLogger and a HuggingFaceCheckpointer for your run.

The MLFlowLogger should have a Unity Catalog model registry prefix in the form of catalog.schema. This specifies where to register your models to. For example,

loggers:
    mlflow:
        experiment_name: /Users/first.last@email.com/my_experiment_name
        tracking_uri: databricks
        model_registry_prefix: catalog.schema
        model_registry_uri: databricks-uc

The HuggingFaceCheckpointer should specify the name you want to register the model under. For example,

callbacks:
    hf_checkpointer:
        save_interval: 1ep # Save Hugging Face formatted checkpoints each epoch
        save_folder: s3://bucket/path/to/my/checkpoints
        mlflow_registered_model_name: my_model_name # Final model will be registered to catalog.schema.my_model_name

MPT model configurations

We've added a few new options when training with the MPT architecture in LLM Foundry.

  • Rotary embeddings (#675)
  • (Un)Tied word embeddings (#728)
  • Fine grained activation checkpointing (#720)

Evaluation Improvements

We've released v0.1 of our Eval Gauntlet (#674, #748)! This adds many new benchmarks, chain-of-thought, and a new safety category. Check out the README for full details!

In addition, we've made a few improvements to our evaluation options, with more to come!

  • Allow specifying multiple evaluation datasets to compute cross entropy and perplexity on during training (#603)
  • Easier versions of the HumanEval dataset, which can be useful for comparing smaller models (#645)
  • More options for averaging the results of the Eval Gauntlet (#640)

New pretraining benchmarks (#543)

Added H100 profiling results to our benchmarking table.

Quality of life improvements

  • Improved Generate callback with more logging options. Use the Generate callback to log generations from your model over the course of training. (#631)
  • Count number of tokens during training excluding padding tokens. Previously this count included padding tokens. (#676)
  • Use the PyTorch profiler to profile your training runs. (#678)
  • A convenience script for using the much faster Hugging Face snapshot_download to download models from the Hugging Face Hub. (#708)
  • New AWS specific Docker images with LLM Foundry dependencies pre-installed. (#731)

Experimental features

Inverse square root learning rate scheduler (#657)

We've added experimental support for the inverse square root learning rate scheduler.

Breaking changes

Updated Streaming defaults (#723)

We've upgraded to the latest Streaming version, including vastly improved default settings for partitioning and shuffling. This means that if you were using the defaults, you will get different results after upgrading. The new defaults should be more performant for the large majority of use cases. See the Streaming release notes for more details.

Removed support for PrefixLM for Bloom and OPT models (#704)

We occasionally remove unused experimental parts of the code base to focus on new features and better support for existing features, and we've removed support for PrefixLM applied to Bloom and OPT models in this release.

What's Changed

Read more

Contributors

mcarbin, dblalock, and 22 other contributors
Assets 2
Loading

v0.3.0

27 Sep 22:06
@dakinggg dakinggg
v0.3.0
54cbb8b
Compare
Choose a tag to compare
v0.3.0

🚀 LLM Foundry v0.3.0

LLM Foundry is an efficient codebase for training, evaluating, and deploying Large Language Models (LLMs) and serves as the foundation for the MPT model series. This release includes lots of bug fixes, stability improvements, and improved error messages, in addition to all the new features listed below!

Features

Llama-2 (#485, #520, #533)

Adds support for training Llama-2 models with optimized flash attention. To enable flash attention, set the attention_patch_type in your yaml like so:

model:
    ...
    attention_patch_type: triton
    ...

See the example yaml for a full example of how to finetune Llama-2 on the MosaicML platform.

8-bit Lion (#514)

We have implemented an 8-bit version of the Lion optimizer. This reduces the memory needed per parameter from 12 bits to 9 bits. To switch from Lion to 8-bit Lion, simply change the optimizer name from decoupled_lionw to decoupled_lionw_8b!

Checkpoint conversion (#526, #519, #594)

We've greatly improved our utilities for checkpoint conversion, including generalizing the Composer to Hugging Face conversion script to support all causal LMs, adding a callback to perform the conversion to Hugging Face format during the training job, and support for Faster Transformer conversion from a Composer MPT checkpoint.

To enable the new callback, add the hf_checkpointer callback to your yaml like so:

callbacks:
    ...
    hf_checkpointer:
        # Save a Hugging Face formatted checkpoint at the end of each epoch
        save_interval: 1ep
        # The Hugging Face formatted checkpoints will be saved inside a subfolder called huggingface, 
        # so this folder will likely be the same as your overall save_folder
        save_folder: ./{run_name}/checkpoints 
        # Set the precision you want the checkpoint saved in
        precision: bfloat16

Code evaluation (#587)

We have added support for running HumanEval (code evaluation) using LLM Foundry! See the evaluation readme for a more detailed description and the tasks yaml for an ICL yaml that can be used to run the HumanEval evaluation task.

Transformer Engine support (#432)

Adds support for using NVIDIA's Transformer Enginer to enable FP8 training. To enable, set fc_type='te' and/or ffn_config['ffn_type']='te_ln_mlp' and precision='amp_fp8'.

MLFlow (#475)

Adds support for using MLFlow as an experiment tracker. To enable, simply add mlflow to the loggers section of your yaml. See the Composer docs for more configuration options for MLFlow. Stay tuned for automatic model logging to MLFlow for easy deployment.

Updated streaming version/defaults (#503, #573, #580, #602)

Updates to the latest release of MosaicML Streaming and sets better defaults for improved shuffling quality and training throughput. Check out the Streaming release notes for the full details of all the new options!

Grouped Query Attention (#492)

Implements Grouped Query Attention, which can strike a good balance between the quality of Multi Head Attention and the speed of Multi Query Attention. To enable, set attn_config['attn_type']='grouped_query_attention' and attn_config['kv_n_heads'] to the desired number of kv heads.

MPT quality of life improvements (#559, #599)

Thanks to @tdoublep and @lorabit110 for making MPT a bit easier to use with other parts of the NLP ecosystem!

Eval gauntlet during training, inference API eval wrapper (#501, #494)

Improvements to our evaluation setup, including the ability to run the eval gauntlet during training, and a wrapper to allow using inference APIs with our eval gauntlet. The ICL tasks and gauntlet can be specified as shown [here](https://github.com/mosaicml/llm-foundry/blob/fd36398dad5ac9fde085af679514189ce9439be4/scripts/eval/yamls/hf_eval.yaml#L46-L47.

tiktoken support (#610)

We have enabled training with tiktoken tokenizers with a thin wrapper around the tiktoken library for compatibility with all the tooling built around Hugging Face tokenizers. You can enable this with a simple change to the tokenizer section of your yaml:

tokenizer:
    name: tiktoken
    kwargs:
        model_name: gpt-4

LoRA eval (#515)

Allows the use of our evaluation script with a model trained using LoRA. Coming soon, full support for LoRA with FSDP! See this yaml for an example of evaluating a model trained using LoRA. Stay tuned for full LoRA support with FSDP!

Finetuning API

Lastly, we are building a finetuning API on top of LLM Foundry, Composer, and Streaming. Please reach out if you might be interested in using this API as a customer!

What's Changed

Read more

Contributors

dblalock, sashaDoubov, and 23 other contributors
Assets 2
Loading

v0.2.0

04 Jul 05:36
@vchiley vchiley
v0.2.0
d0efe55
This commit was created on GitHub.com and signed with GitHub’s verified signature. The key has expired.
GPG key ID: 4AEE18F83AFDEB23
Expired
Learn about vigilant mode.
Compare
Choose a tag to compare
v0.2.0

🚀 LLM Foundry v0.2.0

LLM Foundry is an efficient codebase for training, evaluating, and deploying Large Language Models (LLM). LLM Foundry serves as the efficient training codebase for the MPT-7B and MPT-30B models. Our emphasis is on efficiency, scalability, and ease-of-use, to enable fast iteration and prototyping.

We are excited to share the release of v0.2.0, packed with support for new hardware, features, and tutorials.

📖 Tutorials

We have released new tutorial content and helper scripts for dataset preparation, pre-training, fine-tuning, and inference!

To start off, a basic walkthrough and answers to FAQs can be found in our Basic Tutorial.

Next, detailed guides for different workflows are linked below:

Training

  1. Part 1: LLM Pretraining
    1. Installation
    2. Dataset Preparation
    3. How to start single and multi-node pretraining
  2. Part 2: LLM Finetuning
    1. Using a dataset on the HuggingFace Hub
    2. Using a local dataset
    3. Using a StreamingDataset (MDS) formatted dataset locally or in an object store

In addition, for a more advanced and self-contained example of finetuning the MPT-7B model, see Finetune Example.

Inference

The inference tutorials cover several new features we've added that improve integration with HuggingFace and FasterTransformer libraries:

Major Features

LLM Foundry now uses Composer v0.15.0 and Streaming v0.5.1 as minimum requirements. For more details, see their release notes for Composer and Streaming for all the improvements.

⚠️ The new Streaming release includes a few API changes, see the Streaming v0.5 release notes for more details. Our API have also been changed to reflect these API modifications.

  1. 🆕 Torch 2.0 support

    LLM Foundry is now Torch 2.0 compatible!

    Note: we have not tested torch.compile, but do not expect significant performance improvements.

  2. ⚡ H100 Support

    We now support NVIDIA H100 systems! See our blog post on Benchmarking LLMs on H100 GPUs for initial performance and convergence details.

    To run LLM Foundry with NVIDIA H100 systems, be sure to use a docker images that has CUDA 11.8+ and PyTorch 2.0+ versions.

    For example, mosaicml/pytorch:2.0.1_cu118-python3.10-ubuntu20.04 from our dockerhub has been tested with NVIDIA H100 systems.

    No code changes should be required.

  3. 📈 AMD MI250 GPU Support

    With the release of PyTorch 2.0 and ROCm 5.4+, excited to share that LLM training now works out of the box on AMD Datacenter GPUs! Read our blog post on Training LLMs with AMD MI250 GPUs for more details.

    Running with our stack was straightforward: use the ROCm 5.4 docker image rocm/dev-ubuntu-20.04:5.4.3-complete; and install PyTorch for ROCm 5.4 and install Flash Attention.

    Modify your configuration settings:

    • attn_impl=flash instead of the default triton
      • Note: ALiBi is currently not supported with attn_impl=flash.
    • loss_fn=torch_crossentropy instead of the default fused_crossentropy.

  4. 🚧 LoRA finetuning (Preview)

    We have included a preview release of Low Rank Adaptation (LoRA) support for memory-efficient fine-tuning of LLMs (Shen et al, 2021).

    To use LoRA, follow the instructions found here.

    Note: This is a preview feature, please let us know any feedback! The API and support is subject to change.

  5. 🔎 Evaluation Refactor (#308)

    Our evaluation suite has been significantly refactored into our Model Gauntlet approach. This includes a number of breaking API changes to support multiple models:

    • Instead of model, use the models keyword and provide a list of models.
    • tokenizer is now model-specific.

    For example, to run the gauntlet of various eval tasks with mosaicml/mpt-7b:

    cd llm-foundry/scripts
composer eval/eval.py eval/yamls/hf_eval.yaml
    model_name_or_path=mosaicml/mpt-7b

    This release also makes evaluation deterministic even on different number of GPUs.

    For more details on all these changes, see #308

  6. ⏱️ Benchmarking Inference

    To better support the deployment of LLMs, we have included inference benchmarking suite and results across different hardware and other LLM models.

PR List

Read more

Contributors

patrickhwood, dwyatte, and 29 other contributors
Assets 2
Loading