Skip to content

Commit

Permalink
Update README.md (#112)
Browse files Browse the repository at this point in the history 
Update readme

Update README.md (#113)

update README.md

Update README.md (#114)

Update README.md (#115)

Update Readme.md

Update README.md (#116)

Update README.md

Update README.md (#118)

Update README.md

Update README.md (#121)

Update REAME based on #107

Update README.md (#123)

Add information about params-path to README, update spelling of torchat
  • Loading branch information
@mikekgfb @malfet
mikekgfb authored and malfet committed Jul 17, 2024
1 parent a0bb484 commit 4b32284
Showing 1 changed file with 125 additions and 21 deletions.
146 changes: 125 additions & 21 deletions README.md
View file
Original file line number Diff line number Diff line change
Expand Up @@ -4,10 +4,10 @@
items that are not factual. If you find an item that is incorrect, please tag as an issue, so we can triage and determine whether to fix,
or drop from our initial release.*

# TorchAt *NORTHSTAR*
# torchat *NORTHSTAR*
A repo for building and using llama on servers, desktops and mobile.

The TorchAt repo enables model inference of llama models (and other LLMs) on servers, desktop and mobile devices.
The torchat repo enables model inference of llama models (and other LLMs) on servers, desktop and mobile devices.
For a list of devices, see below, under *SUPPORTED SYSTEMS*.

A goal of this repo, and the design of the PT2 components was to offer seamless integration and consistent workflows.
Expand All @@ -29,12 +29,12 @@ Featuring:
and backend-specific mobile runtimes ("delegates", such as CoreML and Hexagon).

The model definition (and much more!) is adopted from gpt-fast, so we support the same models. As new models are supported by gpt-fast,
bringing them into TorchAt should be straight forward. In addition, we invite community contributions
bringing them into torchat should be straight forward. In addition, we invite community contributions

# Getting started

Follow the `gpt-fast` [installation instructions](https://github.com/pytorch-labs/gpt-fast?tab=readme-ov-file#installation).
Because TorchAt was designed to showcase the latest and greatest PyTorch 2 features for Llama (and related llama-style) models, many of the features used in TorchAt are hot off the press. [Download PyTorch nightly](https://pytorch.org/get-started/locally/) with the latest steaming hot PyTorch 2 features.
Because torchat was designed to showcase the latest and greatest PyTorch 2 features for Llama (and related llama-style) models, many of the features used in torchat are hot off the press. [Download PyTorch nightly](https://pytorch.org/get-started/locally/) with the latest steaming hot PyTorch 2 features.


Install sentencepiece and huggingface_hub
Expand Down Expand Up @@ -67,6 +67,10 @@ export MODEL_DOWNLOAD=meta-llama/Llama-2-7b-chat-hf
While we strive to support a broad range of models, we can't test all models. Consequently, we classify supported models as tested ✅,
work in progress 🚧 and not tested. We invite community contributions of both new models, as well as test reports.

Some common models are recognized by torchat based on their filename (`Transformer.from_name()`). For models not recognized based
on the filename, you can construct a model by initializing the `ModelArgs` dataclass that controls model construction from a parameter json
specified using the `params-path ${PARAMS_PATH}` containing the appropriate model parameters.

| Model | tested | eager | torch.compile | AOT Inductor | ET Runtime | Fits on Mobile |
|-----|--------|-------|-----|-----|-----|-----|
tinyllamas/stories15M | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ |
Expand All @@ -81,6 +85,7 @@ codellama/CodeLlama-34b-Python-hf | -| ✅ | ✅ | ✅ | ✅ | ❌ |
mistralai/Mistral-7B-v0.1 | 🚧 | ✅ | ✅ | ✅ | ✅ | ❹ |
mistralai/Mistral-7B-Instruct-v0.1 | - | ✅ | ✅ | ✅ | ✅ | ❹ |
mistralai/Mistral-7B-Instruct-v0.2 | - | ✅ | ✅ | ✅ | ✅ | ❹ |
Llama3 | 🚧 | ✅ | ✅ | ✅ | ✅ | ❹ |

*Key:* ✅ works correctly; 🚧 work in progress; ❌ not supported; ❹ requires 4bit groupwise quantization; 📵 not on mobile phone (may fit some high-end devices such as tablets);

Expand All @@ -89,10 +94,10 @@ mistralai/Mistral-7B-Instruct-v0.2 | - | ✅ | ✅ | ✅ | ✅ | ❹ |
### More downloading


First cd into TorchAt. We first create a directory for stories15M and download the model and tokenizers.
First cd into torchat. We first create a directory for stories15M and download the model and tokenizers.
We show how to download @Andrej Karpathy's stories15M tiny llama-style model that were used in llama2.c. Advantageously,
stories15M is both a great example and quick to download and run across a range of platforms, ideal for introductions like this
README and for [testing](https://github.com/pytorch-labs/TorchAt/blob/main/.github/workflows). We will be using it throughout
README and for [testing](https://github.com/pytorch-labs/torchat/blob/main/.github/workflows). We will be using it throughout
this introduction as our running example.

```
Expand Down Expand Up @@ -122,11 +127,11 @@ We use several variables in this example, which may be set as a preparatory step
name of the directory holding the files for the corresponding model. You *must* follow this convention to
ensure correct operation.

* `MODEL_OUT` is the location where we store model and tokenizer information for a particular model. We recommend `checkpoints/${MODEL_NAME}`
* `MODEL_DIR` is the location where we store model and tokenizer information for a particular model. We recommend `checkpoints/${MODEL_NAME}`
or any other directory you already use to store model information.

* `MODEL_PATH` describes the location of the model. Throughput the description
herein, we will assume that MODEL_PATH starts with a subdirectory of the TorchAt repo
herein, we will assume that MODEL_PATH starts with a subdirectory of the torchat repo
named checkpoints, and that it will contain the actual model. In this case, the MODEL_PATH will thus
be of the form ${MODEL_OUT}/model.{pt,pth}. (Both the extensions `pt` and `pth`
are used to describe checkpoints. In addition, model may be replaced with the name of the model.)
Expand All @@ -143,7 +148,7 @@ You can set these variables as follows for the exemplary model15M model from And
MODEL_NAME=stories15M
MODEL_DIR=checkpoints/${MODEL_NAME}
MODEL_PATH=${MODEL_OUT}/stories15M.pt
MODEL_OUT=~/TorchAt-exports
MODEL_OUT=~/torchat-exports
```

When we export models with AOT Inductor for servers and desktops, and Executorch for mobile and edge devices,
Expand Down Expand Up @@ -179,13 +184,20 @@ environment:
./run ${MODEL_OUT}/model.{so,pte} -z ${MODEL_OUT}/tokenizer.bin
```

### llama3 tokenizer

Add option to load tiktoken
```
--tiktoken
```

# Generate Text

## Eager Execution

Model definition in model.py, generation code in generate.py. The
model checkpoint may have extensions `pth` (checkpoint and model definition) or `pt` (model checkpoint).
At present, we always use the TorchAt model for export and import the checkpoint into this model definition
At present, we always use the torchat model for export and import the checkpoint into this model definition
because we have tested that model with the export descriptions described herein.

```
Expand Down Expand Up @@ -223,7 +235,7 @@ quantization to achieve this, as described below.

We export the model with the export.py script. Running this script requires you first install executorch with pybindings, see [here](#setting-up-executorch-and-runner-et).
At present, when exporting a model, the export command always uses the
xnnpack delegate to export. (Future versions of TorchAt will support additional
xnnpack delegate to export. (Future versions of torchat will support additional
delegates such as Vulkan, CoreML, MPS, HTP in addition to Xnnpack as they are released for Executorch.)


Expand All @@ -250,8 +262,32 @@ device supported by Executorch, most models need to be compressed to
fit in the target device's memory. We use quantization to achieve this.


# llama3 support

How to obtain snapshot (to be filled in when published by Meta, we use internal snapshot]

enable llama3 tokenizer with option `--tiktoken` (see also discussion under tokenizer)

Enable all export options for llama3 as described below

Identify and enable a runner/run.cpp with a binary tiktoken optimizer. (May already be available in OSS)
we cannot presently run runner/run.cpp with llama3, until we have a C/C++ tokenizer im[plementation
(initial tiktoken is python)

# Optimizing your model for server, desktop and mobile devices

## Model precision (dtype precision setting)_

You can generate models (for both export and generate, with eager, torch.compile, AOTI, ET, for all backends - mobile at present will primarily support fp32, with all options)
specify the precision of the model with
```
python generate.py --dtype [bf16 | fp16 | fp32] ...
python export.py --dtype [bf16 | fp16 | fp32] ...
```

Unlike gpt-fast which uses bfloat16 as default, Torch@ uses float32 as the default. As a consequence you will have to set to `--dtype bf16` or `--dtype fp16` on server / desktop for best performance.


## Making your models fit and execute fast!

Next, we'll show you how to optimize your model for mobile execution
Expand All @@ -260,7 +296,7 @@ AOTI). The basic model build for mobile surfaces two issues: Models
quickly run out of memory and execution can be slow. In this section,
we show you how to fit your models in the limited memory of a mobile
device, and optimize execution speed -- both using quantization. This
is the `TorchAt` repo after all!
is the `torchat` repo after all!

For high-performance devices such as GPUs, quantization provides a way
to reduce the memory bandwidth required to and take advantage of the
Expand All @@ -274,6 +310,9 @@ We can specify quantization parameters with the --quantize option. The
quantize option takes a JSON/dictionary with quantizers and
quantization options.

generate and export (for both ET and AOTI) can both accept quantization options. We only show a subset of the combinations
to avoid combinatorial explosion.

#### Embedding quantization (8 bit integer, channelwise & groupwise)

*Channelwise quantization*:
Expand Down Expand Up @@ -390,27 +429,58 @@ not been optimized for CUDA and CPU targets where the best
performnance requires a group-wise quantized mixed dtype linear
operator.

#### 4-bit integer quantization (int4)
To compress your model even more, 4-bit integer quantization may be used. To achieve good accuracy, we recommend the use
of groupwise quantization where (small to mid-sized) groups of int4 weights share a scale.
```
python export.py --checkpoint-path ${MODEL_PATH} -d fp32 --quant "{'linear:int4': {'group_size' : 32} }" [ --output-pte-path ${MODEL_OUT}/${MODEL_NAME}_int4-gw32.pte | --output-dso-path ${MODEL_OUT}/${MODEL_NAME}_int4-gw32.dso]
```

Now you can run your model with the same command as before:
```
python generate.py [ --pte-path ${MODEL_OUT}/${MODEL_NAME}_int4-gw32.pte | --dso-path ${MODEL_OUT}/${MODEL_NAME}_int4-gw32.dso] --prompt "Hello my name is"
```

#### 4-bit integer quantization (8da4w)
To compress your model even more, 4-bit integer quantization may be used. To achieve good accuracy, we recommend the use
of groupwise quantization where (small to mid-sized) groups of int4 weights share a scale. We also quantize activations to 8-bit, giving
this scheme its name (8da4w = 8b dynamically quantized activations with 4b weights), and boost performance.
```
python export.py --checkpoint-path ${MODEL_PATH} -d fp32 --quant "{'linear:8da4w': {'group_size' : 7} }" --output-pte-path ${MODEL_OUT}/${MODEL_NAME}_8da4w.pte
python export.py --checkpoint-path ${MODEL_PATH} -d fp32 --quant "{'linear:8da4w': {'group_size' : 7} }" [ --output-pte-path ${MODEL_OUT}/${MODEL_NAME}_8da4w.pte | ...dso... ]
```

Now you can run your model with the same command as before:
```
python generate.py --pte-path ${MODEL_OUT}/${MODEL_NAME}_8da4w.pte --prompt "Hello my name is"
python generate.py [ --pte-path ${MODEL_OUT}/${MODEL_NAME}_8da4w.pte | ...dso...] --prompt "Hello my name is"
```

#### Quantization with GPTQ (gptq)

```
python export.py --checkpoint-path ${MODEL_PATH} -d fp32 --quant "{'linear:gptq': {'group_size' : 32} }" [ --output-pte-path ${MODEL_OUT}/${MODEL_NAME}_gptq.pte | ...dso... ] # may require additional options, check with AO team
```

#### Quantization with GPTQ (8da4w-gptq)
TBD.
Now you can run your model with the same command as before:
```
python generate.py [ --pte-path ${MODEL_OUT}/${MODEL_NAME}_gptq.pte | ...dso...] --prompt "Hello my name is"
```

#### Adding additional quantization schemes
#### Adding additional quantization schemes (hqq)
We invite contributors to submit established quantization schemes, with accuracy and performance results demonstrating soundness.


# Loading GGUF models

GGUF is a nascent industry standard format and will will read fp32, fp16 and some quantized formats (q4_0 and whatever is necessary to read llama2_78_q4_0.gguf)

```
--load_gguf <gguf_filename> # all other options as described elsewhere, works for generate and export, for all backends, but cannot be used with --quantize
```

```
--dequantize_gguf <gguf_filename # all other options as described elsewhere, works for generate and export, for all backends, and be used with --quantize
```

# Standalone Execution

In addition to running the exported and compiled models for server, desktop/laptop and mobile/edge devices by loading them in a PyTorch environment under the Python interpreter,
Expand Down Expand Up @@ -468,10 +538,13 @@ To run your pte model, use the following command (assuming you already generated

### Android

Check out the [tutorial on how to build an Android app running your PyTorch models with Executorch](https://pytorch.org/executorch/main/llm/llama-demo-android.html), and give your TorchAt models a spin.
Check out the [tutorial on how to build an Android app running your PyTorch models with Executorch](https://pytorch.org/executorch/main/llm/llama-demo-android.html), and give your torchat models a spin.

![Screenshot](https://pytorch.org/executorch/main/_static/img/android_llama_app.png "Android app running Llama model")

Detailed step by step in conjunction with ET Android build, to run on simulator for Android. `scripts/android_example.sh` for running a model on an Android simulator (on Mac)


### iOS

Open the iOS Llama Xcode project at https://github.com/pytorch/executorch/tree/main/examples/demo-apps/apple_ios/LLaMA/LLaMA.xcodeproj in Xcode and click Run.
Expand All @@ -482,6 +555,9 @@ Once you can run the app on you device,
2 - copy the model and tokenizer.bin to the iOS Llama app
3 - select the tokenizer and model with the `(...)` control (bottom left of screen, to the left of the text entrybox)


Detailed step by step in conjunction with ET iOS build, to run on simulator for iOS.

# Supported Systems

PyTorch and the mobile Executorch backend support a broad range of devices for running PyTorch with Python (using either eager or eager + `torch.compile`) or using a Python-free environment with AOT Inductor, as well as runtimes for executing exported models.
Expand Down Expand Up @@ -527,6 +603,25 @@ PyTorch and the mobile Executorch backend support a broad range of devices for r
| Raspberry Pi 4/5 | Android | ? | ? | ? | ? |
| ARM 32b (up to v7) | any | | ? | ? | ? | ? |

## Runtime performance with Llama3, in tokens per second (4b quantization)

| Hardware | OS | eager | eager + compile | AOT compile | ET Runtime |
|-----|------|-----|-----|-----|-----|
| x86 | Linux | ? | ? | ? | ? |
| x86 | macOS | ? | ? | ? | ? |
| aarch64 | Linux | ? | ? | ? | ? |
| aarch64 | macOS | ? | ? | ? | ? |
| AMD GPU | Linux | ? | ? | ? | ? |
| Nvidia GPU | Linux | ? | ? | ? | ? |
| MPS | macOS | ? | ? | ? | ? |
| MPS | iOS | ? | ? | ? | ? |
| aarch64 | Android | ? | ? | ? | ? |
| Mobile GPU (Vulkan) | Android | ? | ? | ? | ? |
| CoreML | iOS | | ? | ? | ? | ? |
| Hexagon DSP | Android | | ? | ? | ? | ? |
| Raspberry Pi 4/5 | Raspbian | ? | ? | ? | ? |
| Raspberry Pi 4/5 | Android | ? | ? | ? | ? |
| ARM 32b (up to v7) | any | | ? | ? | ? | ? |

## Installation Instructions

Expand All @@ -544,23 +639,23 @@ Alternatively, you can also find libraries here: https://mac.r-project.org/openm
macOS running on x86 is reaching end-of-life. To use PyTorch on x86 running macOS, you can download prebuilt binaries up to PyTorch 2.2. You can download recent PyTorch releases and
install them from source.

### iOS CoreML and MPS
### iOS CoreML, Vulkan, MPS

List dependencies for these backends

### Setting up ExecuTorch and runner-et
Set up ExecuTorch by following the instructions [here](https://pytorch.org/executorch/stable/getting-started-setup.html#setting-up-executorch).
For convenience, we provide a script that does this for you.

From the TorchAt root directory, run the following
From the torchat root directory, run the following
```
export LLAMA_FAST_ROOT=${PWD}
./scripts/install_et.sh
```

This will create a build directory, git clone ExecuTorch to ./build/src, applies some patches to the ExecuTorch source code, install the ExecuTorch python libraries with pip, and install the required ExecuTorch C++ libraries to ./build/install. This will take a while to complete.

After ExecuTorch is installed, you can build runner-et from the TorchAt root directory with the following
After ExecuTorch is installed, you can build runner-et from the torchat root directory with the following

```
export LLAMA_FAST_ROOT=${PWD}
Expand All @@ -570,6 +665,15 @@ cmake --build ./build/cmake-out

The built executable is located at ./build/cmake-out/runner-et.

### Tiktoken instructions & instructions for running llama3 without a python environment

for mobile and runner, if we can get a C/C++ tokenizer


### Raspberry Pi 5 instructions

Expanded version of digant's note.

# Acknowledgements

A big thank you to
Expand Down

0 comments on commit 4b32284

Please sign in to comment.