Skip to content

GitLab

"ppocr/vscode:/vscode.git/clone" did not exist on "400989320b2495db733d45ff2703387b4721de89"
Commit e75bc9be authored by chenzk's avatar chenzk
Browse files

v1.0

parents
Hide whitespace changes
Inline Side-by-side
Showing with 2089 additions and 0 deletions
.gitignore 0 → 100644
View file @ e75bc9be
# Byte-compiled / optimized / DLL files
__pycache__/
*.py[cod]
*$py.class
# C extensions
*.so
# Distribution / packaging
.Python
build/
develop-eggs/
dist/
downloads/
eggs/
.eggs/
lib/
lib64/
parts/
sdist/
var/
wheels/
share/python-wheels/
*.egg-info/
.installed.cfg
*.egg
MANIFEST
# PyInstaller
# Usually these files are written by a python script from a template
# before PyInstaller builds the exe, so as to inject date/other infos into it.
*.manifest
*.spec
# Installer logs
pip-log.txt
pip-delete-this-directory.txt
# Unit test / coverage reports
htmlcov/
.tox/
.nox/
.coverage
.coverage.*
.cache
nosetests.xml
coverage.xml
*.cover
*.py,cover
.hypothesis/
.pytest_cache/
cover/
# Translations
*.mo
*.pot
# Django stuff:
*.log
local_settings.py
db.sqlite3
db.sqlite3-journal
# Flask stuff:
instance/
.webassets-cache
# Scrapy stuff:
.scrapy
# Sphinx documentation
docs/_build/
# PyBuilder
.pybuilder/
target/
# Jupyter Notebook
.ipynb_checkpoints
# IPython
profile_default/
ipython_config.py
# pyenv
# For a library or package, you might want to ignore these files since the code is
# intended to run in multiple environments; otherwise, check them in:
# .python-version
# pipenv
# According to pypa/pipenv#598, it is recommended to include Pipfile.lock in version control.
# However, in case of collaboration, if having platform-specific dependencies or dependencies
# having no cross-platform support, pipenv may install dependencies that don't work, or not
# install all needed dependencies.
#Pipfile.lock
# poetry
# Similar to Pipfile.lock, it is generally recommended to include poetry.lock in version control.
# This is especially recommended for binary packages to ensure reproducibility, and is more
# commonly ignored for libraries.
# https://python-poetry.org/docs/basic-usage/#commit-your-poetrylock-file-to-version-control
#poetry.lock
# pdm
# Similar to Pipfile.lock, it is generally recommended to include pdm.lock in version control.
#pdm.lock
# pdm stores project-wide configurations in .pdm.toml, but it is recommended to not include it
# in version control.
# https://pdm.fming.dev/latest/usage/project/#working-with-version-control
.pdm.toml
.pdm-python
.pdm-build/
# PEP 582; used by e.g. github.com/David-OConnor/pyflow and github.com/pdm-project/pdm
__pypackages__/
# Celery stuff
celerybeat-schedule
celerybeat.pid
# SageMath parsed files
*.sage.py
# Environments
.env
.venv
env/
venv/
ENV/
env.bak/
venv.bak/
# Spyder project settings
.spyderproject
.spyproject
# Rope project settings
.ropeproject
# mkdocs documentation
/site
# mypy
.mypy_cache/
.dmypy.json
dmypy.json
# Pyre type checker
.pyre/
# pytype static type analyzer
.pytype/
# Cython debug symbols
cython_debug/
# PyCharm
# JetBrains specific template is maintained in a separate JetBrains.gitignore that can
# be found at https://github.com/github/gitignore/blob/main/Global/JetBrains.gitignore
# and can be added to the global gitignore or merged into this file. For a more nuclear
# option (not recommended) you can uncomment the following to ignore the entire idea folder.
#.idea/
# Mac
*.DS_Store
LICENSE 0 → 100644
View file @ e75bc9be
Apache License
Version 2.0, January 2004
http://www.apache.org/licenses/
TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
1. Definitions.
"License" shall mean the terms and conditions for use, reproduction,
and distribution as defined by Sections 1 through 9 of this document.
"Licensor" shall mean the copyright owner or entity authorized by
the copyright owner that is granting the License.
"Legal Entity" shall mean the union of the acting entity and all
other entities that control, are controlled by, or are under common
control with that entity. For the purposes of this definition,
"control" means (i) the power, direct or indirect, to cause the
direction or management of such entity, whether by contract or
otherwise, or (ii) ownership of fifty percent (50%) or more of the
outstanding shares, or (iii) beneficial ownership of such entity.
"You" (or "Your") shall mean an individual or Legal Entity
exercising permissions granted by this License.
"Source" form shall mean the preferred form for making modifications,
including but not limited to software source code, documentation
source, and configuration files.
"Object" form shall mean any form resulting from mechanical
transformation or translation of a Source form, including but
not limited to compiled object code, generated documentation,
and conversions to other media types.
"Work" shall mean the work of authorship, whether in Source or
Object form, made available under the License, as indicated by a
copyright notice that is included in or attached to the work
(an example is provided in the Appendix below).
"Derivative Works" shall mean any work, whether in Source or Object
form, that is based on (or derived from) the Work and for which the
editorial revisions, annotations, elaborations, or other modifications
represent, as a whole, an original work of authorship. For the purposes
of this License, Derivative Works shall not include works that remain
separable from, or merely link (or bind by name) to the interfaces of,
the Work and Derivative Works thereof.
"Contribution" shall mean any work of authorship, including
the original version of the Work and any modifications or additions
to that Work or Derivative Works thereof, that is intentionally
submitted to Licensor for inclusion in the Work by the copyright owner
or by an individual or Legal Entity authorized to submit on behalf of
the copyright owner. For the purposes of this definition, "submitted"
means any form of electronic, verbal, or written communication sent
to the Licensor or its representatives, including but not limited to
communication on electronic mailing lists, source code control systems,
and issue tracking systems that are managed by, or on behalf of, the
Licensor for the purpose of discussing and improving the Work, but
excluding communication that is conspicuously marked or otherwise
designated in writing by the copyright owner as "Not a Contribution."
"Contributor" shall mean Licensor and any individual or Legal Entity
on behalf of whom a Contribution has been received by Licensor and
subsequently incorporated within the Work.
2. Grant of Copyright License. Subject to the terms and conditions of
this License, each Contributor hereby grants to You a perpetual,
worldwide, non-exclusive, no-charge, royalty-free, irrevocable
copyright license to reproduce, prepare Derivative Works of,
publicly display, publicly perform, sublicense, and distribute the
Work and such Derivative Works in Source or Object form.
3. Grant of Patent License. Subject to the terms and conditions of
this License, each Contributor hereby grants to You a perpetual,
worldwide, non-exclusive, no-charge, royalty-free, irrevocable
(except as stated in this section) patent license to make, have made,
use, offer to sell, sell, import, and otherwise transfer the Work,
where such license applies only to those patent claims licensable
by such Contributor that are necessarily infringed by their
Contribution(s) alone or by combination of their Contribution(s)
with the Work to which such Contribution(s) was submitted. If You
institute patent litigation against any entity (including a
cross-claim or counterclaim in a lawsuit) alleging that the Work
or a Contribution incorporated within the Work constitutes direct
or contributory patent infringement, then any patent licenses
granted to You under this License for that Work shall terminate
as of the date such litigation is filed.
4. Redistribution. You may reproduce and distribute copies of the
Work or Derivative Works thereof in any medium, with or without
modifications, and in Source or Object form, provided that You
meet the following conditions:
(a) You must give any other recipients of the Work or
Derivative Works a copy of this License; and
(b) You must cause any modified files to carry prominent notices
stating that You changed the files; and
(c) You must retain, in the Source form of any Derivative Works
that You distribute, all copyright, patent, trademark, and
attribution notices from the Source form of the Work,
excluding those notices that do not pertain to any part of
the Derivative Works; and
(d) If the Work includes a "NOTICE" text file as part of its
distribution, then any Derivative Works that You distribute must
include a readable copy of the attribution notices contained
within such NOTICE file, excluding those notices that do not
pertain to any part of the Derivative Works, in at least one
of the following places: within a NOTICE text file distributed
as part of the Derivative Works; within the Source form or
documentation, if provided along with the Derivative Works; or,
within a display generated by the Derivative Works, if and
wherever such third-party notices normally appear. The contents
of the NOTICE file are for informational purposes only and
do not modify the License. You may add Your own attribution
notices within Derivative Works that You distribute, alongside
or as an addendum to the NOTICE text from the Work, provided
that such additional attribution notices cannot be construed
as modifying the License.
You may add Your own copyright statement to Your modifications and
may provide additional or different license terms and conditions
for use, reproduction, or distribution of Your modifications, or
for any such Derivative Works as a whole, provided Your use,
reproduction, and distribution of the Work otherwise complies with
the conditions stated in this License.
5. Submission of Contributions. Unless You explicitly state otherwise,
any Contribution intentionally submitted for inclusion in the Work
by You to the Licensor shall be under the terms and conditions of
this License, without any additional terms or conditions.
Notwithstanding the above, nothing herein shall supersede or modify
the terms of any separate license agreement you may have executed
with Licensor regarding such Contributions.
6. Trademarks. This License does not grant permission to use the trade
names, trademarks, service marks, or product names of the Licensor,
except as required for reasonable and customary use in describing the
origin of the Work and reproducing the content of the NOTICE file.
7. Disclaimer of Warranty. Unless required by applicable law or
agreed to in writing, Licensor provides the Work (and each
Contributor provides its Contributions) on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
implied, including, without limitation, any warranties or conditions
of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
PARTICULAR PURPOSE. You are solely responsible for determining the
appropriateness of using or redistributing the Work and assume any
risks associated with Your exercise of permissions under this License.
8. Limitation of Liability. In no event and under no legal theory,
whether in tort (including negligence), contract, or otherwise,
unless required by applicable law (such as deliberate and grossly
negligent acts) or agreed to in writing, shall any Contributor be
liable to You for damages, including any direct, indirect, special,
incidental, or consequential damages of any character arising as a
result of this License or out of the use or inability to use the
Work (including but not limited to damages for loss of goodwill,
work stoppage, computer failure or malfunction, or any and all
other commercial damages or losses), even if such Contributor
has been advised of the possibility of such damages.
9. Accepting Warranty or Additional Liability. While redistributing
the Work or Derivative Works thereof, You may choose to offer,
and charge a fee for, acceptance of support, warranty, indemnity,
or other liability obligations and/or rights consistent with this
License. However, in accepting such obligations, You may act only
on Your own behalf and on Your sole responsibility, not on behalf
of any other Contributor, and only if You agree to indemnify,
defend, and hold each Contributor harmless for any liability
incurred by, or claims asserted against, such Contributor by reason
of your accepting any such warranty or additional liability.
END OF TERMS AND CONDITIONS
APPENDIX: How to apply the Apache License to your work.
To apply the Apache License to your work, attach the following
boilerplate notice, with the fields enclosed by brackets "[]"
replaced with your own identifying information. (Don't include
the brackets!) The text should be enclosed in the appropriate
comment syntax for the file format. We also recommend that a
file or class name and description of purpose be included on the
same "printed page" as the copyright notice for easier
identification within third-party archives.
Copyright [yyyy] [name of copyright owner]
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
http://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.
README.md 0 → 100644
View file @ e75bc9be
# SmolLM2
端侧小模型新星SmolLM2 1.7B击败Qwen 2.5 1.5B和Llama 3.2 1B，可以在不到2GB的VRAM上运行。
## 论文
`未发表`
## 模型结构
SmolLM2对Llama结构进行了瘦身，认为深-窄结构具有更好的梯度流和参数效率，仍为Decoder-only结构。
<div align=center>
<img src="./doc/llama3.png"/>
</div>
## 算法原理
Llama将输入embedding（将语句根据词汇量和词的位置、属性转换成数字化矩阵）后放入attention+ffn等提取特征，最后利用Softmax将解码器最后一层产生的未经归一化的分数向量（logits）转换为概率分布，其中每个元素表示生成对应词汇的概率，这使得模型可以生成一个分布，并从中选择最可能的词作为预测结果，然后一个字一个预测出来就是咱们看到的对话生成效果。
<div align=center>
<img src="./doc/algorithm.png"/>
</div>
## 环境配置
```
mv smollm_pytorch smollm # 去框架名后缀
```
### Docker（方法一）
```
docker pull image.sourcefind.cn:5000/dcu/admin/base/pytorch:2.3.0-ubuntu22.04-dtk24.04.2-py3.10
# <your IMAGE ID>为以上拉取的docker的镜像ID替换，本镜像为：83714c19d308
docker run -it --shm-size=64G -v $PWD/smollm:/home/smollm -v /opt/hyhal:/opt/hyhal:ro --privileged=true --device=/dev/kfd --device=/dev/dri/ --group-add video --name smollm2 <your IMAGE ID> bash
```
### Dockerfile（方法二）
```
cd cd /home/smollm/docker
docker build --no-cache -t smollm2:latest .
docker run --shm-size=64G --name smollm2 -v /opt/hyhal:/opt/hyhal:ro --privileged=true --device=/dev/kfd --device=/dev/dri/ --group-add video -v $PWD/../../smollm:/home/smollm -it smollm2 bash
# 若遇到Dockerfile启动的方式安装环境需要长时间等待，可注释掉里面的pip安装，启动容器后再安装python库：pip install -r requirements.txt。
cd /home/smollm
pip install -r finetuning/requirements.txt
```
### Anaconda（方法三）
1、关于本项目DCU显卡所需的特殊深度学习库可从光合开发者社区下载安装：
- https://developer.hpccube.com/tool/
```
DTK驱动:dtk24.04.2
python:python3.10
torch:2.3.0
torchvision:0.18.1
torchaudio:2.1.2
triton:2.1.0
flash-attn:2.0.4
deepspeed:0.14.2
apex:1.3.0
xformers:0.0.25
```
`Tips：以上dtk驱动、python、torch等DCU相关工具版本需要严格一一对应。`
2、其它非特殊库参照requirements.txt安装
```
cd /home/smollm
pip install -r finetuning/requirements.txt
```
## 数据集
[bigcode/the-stack-smol](http://113.200.138.88:18080/aimodels/bigcode/the-stack-smol.git)
本项目已提供迷你数据集[`the-stack-smol/data/python`](./finetuning/bigcode/the-stack-smol/data/python/data.json)，可直接试用。
数据完整目录结构如下：
```
/home/smollm/finetuning/bigcode/the-stack-smol
├── README.md
├── data_generation.py
└── data
├── assembly
├── python
...
└── visual-basic
```
更多资料可参考源项目的[`README_origin`](./README_origin.md)
## 训练
```
关闭 wandb
wandb disabled
wandb offline
cd /home/smollm
sh train.sh
```
更多资料可参考源项目的[`README`](./finetuning/README.md)
## 推理
```
cd /home/smollm
# 方法一
python infer.py
# 方法二
trl chat --model_name_or_path HuggingFaceTB/SmolLM2-1.7B-Instruct --device cpu
```
更多资料可参考源项目的[`README`](./finetuning/README.md)
## result
推理方法一效果示例：
`输入: `
```
user:"Write a 100-word article on 'Benefits of Open-Source in AI research"
```
`输出:`
```
<|im_start|>system
You are a helpful AI assistant named SmolLM, trained by Hugging Face<|im_end|>
<|im_start|>user
Write a 100-word article on 'Benefits of Open-Source in AI research<|im_end|>
<|im_start|>assistant
Open-source in AI research offers numerous benefits. Firstly, it fosters collaboration and community engagement, accelerating the development of AI technologies. Secondly, it promotes transparency and accountability, as open-source projects are subject to peer review and scrutiny
```
### 精度
DCU与GPU精度一致，推理框架：pytorch。
## 应用场景
### 算法类别
`对话问答`
### 热点应用行业
`制造,广媒,金融,能源,医疗,家居,教育`
## 预训练权重
预训练权重快速下载中心：[SCNet AIModels](http://113.200.138.88:18080/aimodels) ，项目中的预训练权重可从快速下载通道下载：[SmolLM2-1.7B-Instruct](http://113.200.138.88:18080/aimodels/huggingfacetb/SmolLM2-1.7B-Instruct.git) 。
Hugging Face下载地址为：[HuggingFaceTB/SmolLM2-1.7B-Instruct](https://huggingface.co/HuggingFaceTB/SmolLM2-1.7B-Instruct)
## 源码仓库及问题反馈
- http://developer.sourcefind.cn/codes/modelzoo/smollm_pytorch.git
## 参考资料
- https://github.com/huggingface/smollm.git
README_origin.md 0 → 100644
View file @ e75bc9be
# SmolLM2
SmolLM2 is a family of compact language models available in three size: 135M, 360M, and 1.7B parameters. They are capable of solving a wide range of tasks while being lightweight enough to run on-device.
You can find our most capable model **🤏 SmolLM2-1.7B-Instruct** [here](https://huggingface.co/HuggingFaceTB/SmolLM2-1.7B-Instruct).
**New: Introducing [SmolTalk](https://huggingface.co/datasets/HuggingFaceTB/smoltalk), the SFT dataset of SmolLM2 🚀**
<img src="https://cdn-uploads.huggingface.co/production/uploads/61c141342aac764ce1654e43/y45hIMNREW7w_XpHYB_0q.png" alt="Evaluation Results" width="600">
## Table of Contents
1. [Usage](#usage)
- [Transformers](#transformers)
- [Chat in TRL](#chat-in-trl)
- [Local inference](#local-inference)
2. [Smol-tools](#smol-tools)
3. [Pre-training](#pre-training)
4. [Fine-tuning](#fine-tuning)
5. [Evaluation](#evaluation)
6. [Synthetic data pipelines](#synthetic-data-pipelines)
## Usage
Our most powerful model is `SmolLM2-1.7B-Instruct`, which you can use as an assistant with `transformers`, `trl`, or using quantized versions with tools like `llama.cpp`, `MLX`, and `transformers.js`. For lighter applications, you can also use the smaller models `SmolLM2-360M` and`SmolLM2-135M`, which are suitable for on-device usage and can be integrated similarly.
All available in this [collection](https://huggingface.co/collections/HuggingFaceTB/smollm2-6723884218bcda64b34d7db9).
### Transformers
```bash
pip install transformers
```
```python
from transformers import AutoModelForCausalLM, AutoTokenizer
checkpoint = "HuggingFaceTB/SmolLM2-1.7B-Instruct"
device = "cuda" # for GPU usage or "cpu" for CPU usage
tokenizer = AutoTokenizer.from_pretrained(checkpoint)
# for multiple GPUs install accelerate and do `model = AutoModelForCausalLM.from_pretrained(checkpoint, device_map="auto")`
model = AutoModelForCausalLM.from_pretrained(checkpoint).to(device)
messages = [{"role": "user", "content": "Write a 100-word article on 'Benefits of Open-Source in AI research"}]
input_text=tokenizer.apply_chat_template(messages, tokenize=False)
inputs = tokenizer.encode(input_text, return_tensors="pt").to(device)
outputs = model.generate(inputs, max_new_tokens=50, temperature=0.2, top_p=0.9, do_sample=True)
print(tokenizer.decode(outputs[0]))
```
### Chat in TRL
You can also use the TRL CLI to chat with the model from the terminal:
```bash
pip install trl
trl chat --model_name_or_path HuggingFaceTB/SmolLM2-1.7B-Instruct --device cpu
```
You can find more details on how to leverage the model for use cases such as text summarization, text rewriting and function calling in the model card: https://huggingface.co/HuggingFaceTB/SmolLM2-1.7B-Instruct
### Local inference
You can use the models locally with frameworks like `llama.cpp`, `MLX`, `MLC` and `transformers.js`. You can find the instructions to run SmolLM2 with these frameworks at [local-inference](local_inference/README.md).
## Smol-tools
A collection of lightweight AI-powered tools built with LLaMA.cpp and small language models. These tools are designed to run locally on your machine without requiring expensive GPU resources.
Further instructions on how to use the tools can be found in the [smol-tools README](smol_tools/README.md).
## Pre-training
You can find scripts for launching pre-training with [nanotron](https://github.com/huggingface/nanotron/) under [pre-training](pre-training/README.md), we share the exact configs for training SmolLM1 and will upload SmolLM2's configs soon.
## Fine-tuning
You can find an example script to finetune SmolLM2 using `TRL` and `PEFT` in the `finetuning` folder. We also link to our post-training scripts for SmolLM2 using the alignment handbook.
## Evaluation
![image/png](https://cdn-uploads.huggingface.co/production/uploads/61c141342aac764ce1654e43/T-cHJVA7FBaI0cgDApzEj.png)
You can find more detailed evaluation of each model size in the model cards in this [collection](https://huggingface.co/collections/HuggingFaceTB/smollm2-6723884218bcda64b34d7db9).
We use [lighteval](https://github.com/huggingface/lighteval) for all our evaluations, for more details refer to the [evaluation README](evaluation/README.md).
## Synthetic data pipelines
We released [SmolTalk](https://huggingface.co/datasets/HuggingFaceTB/smoltalk) the SFT dataset used for building SmolLM2 instruct models. It was created with [distilabel](https://github.com/argilla-io/distilabel) and you can check and execute the synthetic data pipelines in [distilabel_pipelines README](distilabel_pipelines/README.md)
<div align="center">
<img src="https://cdn-uploads.huggingface.co/production/uploads/61c141342aac764ce1654e43/JLTEbnsBQ_qY032mxFzgC.png" width="800"/>
<p><em>Comparison of models finetuned on SmolTalk and Orca AgentInstruct 1M. For more details, refer to the <a href="https://huggingface.co/datasets/HuggingFaceTB/smoltalk" target="_blank">dataset card</a>.</em></p>
</div>
distilabel_pipelines/README.md 0 → 100644
View file @ e75bc9be
# distilabel pipelines
We used [`distilabel`](https://github.com/argilla-io/distilabel) to create several pipelines for generating instruction-following and multi-turn datasets for the post-training of SmolLM2.
> [!NOTE]
> This section is still in WIP. We will upload the rest of the pipelines soon. Thanks for your patience!
distilabel_pipelines/magpie_ultra_v1/README.md 0 → 100644
View file @ e75bc9be
# MagPie Ultra v1.0
This [`distilabel`](https://github.com/argilla-io/distilabel) was used to generate the [magpie-ultra-v1.0](https://huggingface.co/datasets/argilla/magpie-ultra-v1.0) dataset. The dataset follows the [MagPie](https://magpie-align.github.io) pipeline recipe to generate a multi-turn conversation dataset using [meta-llama/Llama-3.1-405B-Instruct-FP8](https://huggingface.co/meta-llama/Llama-3.1-405B-Instruct-FP8).
## Setup
You will need to install `distilabel` with a few extra dependencies to be able to execute the pipeline:
```bash
pip install distilabel[ray,vllm,sentence-transformers,faiss-cpu,hf-transformers]
```
\ No newline at end of file
distilabel_pipelines/magpie_ultra_v1/pipeline.py 0 → 100644
View file @ e75bc9be
import faiss
import json
from typing import Union, Dict, Any, Literal, List, TYPE_CHECKING
from distilabel.llms import vLLM
from distilabel.pipeline import Pipeline
from distilabel.steps.tasks import MagpieGenerator, Task, ChatGeneration
from distilabel.steps import (
step,
StepInput,
EmbeddingGeneration,
FaissNearestNeighbour,
RewardModelScore,
CombineOutputs,
)
from distilabel.embeddings import SentenceTransformerEmbeddings
if TYPE_CHECKING:
from distilabel.steps.tasks.typing import ChatType
from distilabel.steps.typing import StepOutput
INFORMATION_SEEKING_PROMPT = (
"You are an AI assistant designed to provide accurate and concise information on a wide"
" range of topics."
" The user will engage in a multi-round conversation with you, asking initial questions and following up with additional related questions."
" Your purpose is to assist users in finding specific facts,"
" explanations, or details about various subjects. Provide clear, factual responses and,"
" when appropriate, offer additional context or related information that might be useful"
" to the user."
"\n\nUser inputs will typically be direct questions seeking factual information, explanations"
" of concepts, or details about specific topics. Users may ask about historical events,"
" scientific phenomena, current affairs, or any subject requiring factual knowledge."
"\n\nImportant: Be concise in your responses. Do not use bold text, enumerations, or lists of"
" steps unless specifically requested by the user. Avoid verbosity and focus on providing"
" clear, direct answers in a flowing, narrative format."
)
REASONING_PROMPT = (
"You are an AI assistant specialized in logical thinking and problem-solving."
" The user will engage in a multi-round conversation with you, asking initial questions and following up with additional related questions."
" Your purpose is to help users work through complex ideas, analyze situations, and draw"
" conclusions based on given information. Approach each query with structured thinking,"
" break down problems into manageable parts, and guide users through the reasoning"
" process in a clear, narrative format."
"\n\nUser inputs will often present complex scenarios, logical puzzles, or arguments that"
" require analysis. Users may ask for help in identifying logical fallacies, solving"
" riddles, or evaluating the pros and cons of different situations. Inputs may be"
" lengthy and require careful consideration of multiple factors."
"\n\nImportant: Provide concise, clear reasoning. Avoid unnecessary formatting like bold"
" text, enumerations, or lists of steps unless specifically requested by the user. Focus on delivering"
" structured, efficient explanations in a flowing, narrative format without excessive elaboration."
)
PLANNING_PROMPT = (
"You are an AI assistant focused on helping users create effective plans and strategies."
" The user will engage in a multi-round conversation with you, asking initial questions and following up with additional related questions."
" Your purpose is to assist in organizing thoughts, setting goals, and developing"
" actionable approaches for various projects or activities. Offer structured ideas,"
" consider potential challenges, and provide tips for efficient execution of plans."
"\n\nUser inputs will typically describe a goal or project that requires planning. This could"
" range from personal activities like planning a trip, to professional tasks like"
" launching a new product. Users may provide some initial ideas or constraints and will"
" expect guidance on creating a structured, actionable plan."
"\n\nImportant: Present plans concisely and clearly in a narrative format. Use formatting like bold text or"
" enumerations only when specifically requested by the user. Avoid verbose explanations and"
" focus on delivering actionable, efficient plans in a flowing, paragraph-based structure."
)
EDITING_PROMPT = (
"You are an AI assistant specialized in editing and improving written content."
" The user will engage in a multi-round conversation with you, asking initial questions and following up with additional related questions."
" Your purpose is to help users refine their writing by offering suggestions for grammar,"
" style, clarity, and overall structure. Provide constructive feedback, explain your"
" edits, and offer alternative phrasings when appropriate."
"\n\nUser inputs will usually consist of written text that needs improvement. This could be"
" anything from a single sentence to a full essay or article. Users may ask for general"
" editing, specific focus on grammar or style, or help in making their writing more"
" concise or impactful."
"\n\nImportant: Offer edits and suggestions concisely in a narrative format. Use formatting like bold text or"
" enumerations only when specifically requested by the user. Focus on providing clear, efficient"
" feedback without unnecessary elaboration or step-by-step breakdowns unless asked."
)
CODING_DEBUGGING_PROMPT = (
"You are an AI assistant designed to help with programming tasks. "
" The user will engage in a multi-round conversation with you, asking initial questions and following up with additional related questions."
"Your purpose is to"
" assist users in writing, reviewing, and debugging code across various programming"
" languages. Provide clear explanations, offer best practices, and help troubleshoot"
" issues. When appropriate, suggest optimizations or alternative approaches to coding"
" problems."
"\n\nUser inputs will typically involve code snippets, error messages, or descriptions of"
" programming challenges. Users may ask for help in debugging specific issues, optimizing"
" code performance, or understanding certain programming concepts. Inputs may span"
" various programming languages and complexity levels."
"\n\nImportant: Provide coding assistance concisely. Use formatting like bold text or"
" enumerations only when specifically requested by the user or necessary for code structure. Focus on clear,"
" efficient explanations and solutions without verbose commentary or step-by-step breakdowns unless asked."
)
MATH_SYSTEM_PROMPT = (
"You are an AI assistant specializing in mathematics, capable of addressing questions "
"across a wide spectrum of mathematical disciplines. "
" The user will engage in a multi-round conversation with you, asking initial questions and following up with additional related questions."
" Your expertise spans from foundational "
"concepts to advanced topics, including but not limited to:"
"\n\n- Arithmetic and Number Theory"
"\n- Algebra (Linear, Abstract, Commutative)"
"\n- Geometry (Euclidean, Non-Euclidean, Algebraic)"
"\n- Calculus and Analysis (Real, Complex, Functional)"
"\n- Topology and Differential Geometry"
"\n- Probability and Statistics"
"\n- Discrete Mathematics and Combinatorics"
"\n- Numerical Analysis and Computational Mathematics"
"\n- Mathematical Logic and Set Theory"
"\n- Applied Mathematics (including Physics and Engineering applications)"
"\n\nWhen formulating problems or questions, strive for elegance and clarity. Prefer "
"problems that showcase the beauty and interconnectedness of mathematics. Avoid overly "
"contrived scenarios or those leading to unwieldy calculations or solutions."
"\n\nIn your responses:"
"\n- Provide clear, concise explanations of concepts and problem-solving strategies in a narrative format."
"\n- Use a flowing, paragraph-based approach for solutions, emphasizing logical progression and key insights."
"\n- Highlight connections between different areas of mathematics when relevant."
"\n- Use mathematical notation judiciously, ensuring it enhances rather than obscures understanding."
"\n- When possible, discuss multiple approaches or interpretations of a problem within the narrative."
"\n- For abstract or theoretical questions, balance rigor with intuitive explanations."
"\n\nImportant: Provide mathematical explanations concisely. Avoid using formatting like bold "
"text, enumerations, or step-by-step breakdowns unless specifically requested by the user or absolutely essential for mathematical notation. "
"Focus on clear, efficient problem-solving without unnecessary elaboration or formatting."
"\n\nYour goal is to not just solve problems, but to cultivate a deeper appreciation "
"for the elegance and power of mathematical thinking, while maintaining a clean and "
"uncluttered presentation style."
)
ROLE_PLAYING_PROMPT = (
"You are an AI assistant capable of engaging in various role-playing scenarios."
" The user will engage in a multi-round conversation with you, asking initial questions and following up with additional related questions."
" Your purpose is to adopt different personas or characters as requested by the user. Maintain"
" consistency with the chosen role, respond in character, and help create immersive and"
" interactive experiences for the user."
"\n\nUser inputs will typically begin with a request to assume a specific role or character."
" Following this, users will engage in dialogue or present scenarios consistent with the"
" chosen role-play setting. Inputs may vary widely depending on the nature of the"
" role-playing scenario."
"\n\nImportant: Engage in role-play concisely and effectively. Use formatting like bold text"
" or enumerations only when specifically requested by the user or when it significantly enhances the role-play experience. Focus on immersive,"
" character-appropriate responses without unnecessary verbosity or structured breakdowns."
)
DATA_ANALYSIS_PROMPT = (
"You are an AI assistant specialized in data analysis and interpretation. "
" The user will engage in a multi-round conversation with you, asking initial questions and following up with additional related questions."
" Your purpose is"
" to help users understand and derive insights from data sets, statistics, and analytical"
" tasks. Offer clear explanations of data trends, assist with statistical calculations,"
" and provide guidance on data visualization and interpretation techniques."
"\n\nUser inputs will often involve questions about data interpretation, statistical analysis,"
" or data visualization. Users may present datasets, ask for help in understanding"
" statistical concepts, or seek guidance on how to best analyze or present their data."
" Inputs may range from simple data queries to complex analytical challenges."
"\n\nImportant: Provide data analysis and insights concisely in a narrative format. Use formatting like bold text"
" or enumerations only when specifically requested by the user or necessary for data presentation. Focus on clear,"
" efficient explanations of data trends and analytical techniques without excessive detail or step-by-step breakdowns unless asked."
)
CREATIVE_WRITING_PROMPT = (
"You are an AI assistant designed to support creative writing endeavors. "
" The user will engage in a multi-round conversation with you, asking initial questions and following up with additional related questions."
" Your purpose is"
" to help users craft engaging stories, poems, and other creative texts. Offer"
" suggestions for plot development, character creation, dialogue writing, and other"
" aspects of creative composition. Provide constructive feedback and inspire creativity."
"\n\nUser inputs will typically seek assistance with various aspects of creative writing."
" This may include requests for story ideas, character development tips, help with"
" dialogue or descriptive passages, or feedback on written pieces. Users may provide"
" partial works or ideas and ask for help in expanding or improving them."
"\n\nImportant: Offer creative writing assistance concisely in a flowing, narrative format. Use formatting like bold text"
" or enumerations only when specifically requested by the user or when it significantly enhances the creative process. Focus on providing clear,"
" inspiring suggestions without unnecessary elaboration or structured breakdowns."
)
ADVICE_SEEKING_PROMPT = (
"You are an AI assistant focused on providing thoughtful advice and guidance."
" The user will engage in a multi-round conversation with you, asking initial questions and following up with additional related questions."
" Your purpose is to help users navigate various personal or professional issues by offering"
" balanced perspectives, considering potential outcomes, and suggesting practical"
" solutions. Encourage users to think critically about their situations while providing"
" supportive and constructive advice."
"\n\nUser inputs will generally describe personal or professional situations where advice is"
" needed. These could range from career decisions and interpersonal relationships to"
" personal development challenges. Users may provide context about their situation and"
" ask for guidance or potential solutions."
"\n\nImportant: Provide advice concisely and effectively in a narrative format. Use formatting like bold text or"
" enumerations only when specifically requested by the user. Focus on offering clear,"
" practical guidance without excessive elaboration or step-by-step breakdowns unless asked."
)
BRAINSTORMING_PROMPT = (
"You are an AI assistant specialized in generating ideas and facilitating creative"
" thinking."
" The user will engage in a multi-round conversation with you, asking initial questions and following up with additional related questions."
" Your purpose is to help users explore possibilities, think outside the box,"
" and develop innovative concepts. Encourage free-flowing thoughts, offer diverse"
" perspectives, and help users build upon and refine their ideas."
"\n\nUser inputs will typically present a problem or area where creative ideas are needed."
" This could be for business innovations, artistic projects, problem-solving, or any"
" situation requiring novel thinking. Users may provide some initial thoughts or"
" constraints and expect a range of creative suggestions or conceptual explorations."
"\n\nImportant: Generate and present ideas concisely in a flowing, narrative format. Use formatting like bold text or"
" enumerations only when specifically requested by the user. Focus on providing"
" clear, innovative concepts without unnecessary verbosity or structured breakdowns unless asked."
)
CATEGORIES_SYSTEM_PROMPTS = {
"information-seeking": (INFORMATION_SEEKING_PROMPT, 0.05),
"reasoning": (REASONING_PROMPT, 0.125),
"planning": (PLANNING_PROMPT, 0.05),
"editing": (EDITING_PROMPT, 0.10),
"coding": (CODING_DEBUGGING_PROMPT, 0.125),
"math": (MATH_SYSTEM_PROMPT, 0.125),
"role-playing": (ROLE_PLAYING_PROMPT, 0.10),
"data-analysis": (DATA_ANALYSIS_PROMPT, 0.125),
"creative-writing": (CREATIVE_WRITING_PROMPT, 0.10),
"advice-seeking": (ADVICE_SEEKING_PROMPT, 0.05),
"brainstorming": (BRAINSTORMING_PROMPT, 0.05),
}
INPUT_DIFFICULTY_RATING_TEMPLATE = """
# Instruction
You first need to identify the given user intent and then label the difficulty level of the user query based on the content of the user query.
## User Query
```
{input}
```
## Output Format
Given the user query, in your output, you first need to identify the user intent and the knowledge needed to solve the task in the user query.
Then, rate the difficulty level of the user query as `very easy`, `easy`, `medium`, `hard`, or `very hard`.
Now, please output the user intent and difficulty level below in a json format by filling in the placeholders in []:
```
{{
"intent": "The user wants to [....]",
"knowledge": "To solve this problem, the models need to know [....]",
"difficulty": "[very easy/easy/medium/hard/very hard]"
}}
```
""".lstrip()
OUTPUT_DIFFICULTY_JSON_SCHEMA = {
"type": "object",
"properties": {
"intent": {"type": "string"},
"knowledge": {"type": "string"},
"difficulty": {
"type": "string",
"enum": ["very easy", "easy", "medium", "hard", "very hard"],
},
},
"required": ["intent", "knowledge", "difficulty"],
}
INPUT_QUALITY_RATING_TEMPLATE = """
# Instruction
You need to rate the quality of the user query based on its clarity, specificity, and coherence.
The rating scale is as follows:
- very poor: The query is unclear, vague, or incoherent. It lacks essential information and context.
- poor: The query is somewhat unclear or lacks important details. It requires significant clarification.
- average: The query is moderately clear and specific. It may require some additional information for a complete understanding.
- good: The query is clear, specific, and mostly well-formed. It provides sufficient context for understanding the user's intent.
- excellent: The query is very clear, specific, and well-articulated. It contains all the necessary information and context for providing a comprehensive response.
## User Query
```
{input}
```
## Output Format
Given the user query, you first need to give an assesement, highlighting the strengths and/or weaknesses of the user query.
Then, you need to output a rating from very poor to excellent by filling in the placeholders in [...]:
```
{{
"explanation": "[...]",
"quality": "[very poor/poor/average/good/excellent]"
}}
```
""".lstrip()
OUTPUT_QUALITY_JSON_SCHEMA = {
"type": "object",
"properties": {
"explanation": {"type": "string"},
"quality": {
"type": "string",
"enum": ["very poor", "poor", "average", "good", "excellent"],
},
},
"required": ["explanation", "quality"],
}
INPUT_CLASSIFICATION_TEMPLATE = """
# Instruction
Please label the task tags for the user query.
## User Query
```
{input}
```
## Tagging the user input
Please label the task tags for the user query. You will need to analyze the user query and select the most relevant task tag from the list below.
all_task_tags = [
"Information seeking", # Users ask for specific information or facts about various topics.
"Reasoning", # Queries require logical thinking, problem-solving, or processing of complex ideas.
"Planning", # Users need assistance in creating plans or strategies for activities and projects.
"Editing", # Involves editing, rephrasing, proofreading, or other tasks related to the composition of general written content.
"Coding & Debugging", # Users seek help with writing, reviewing, or fixing code in programming.
"Math", # Queries related to mathematical concepts, problems, and calculations.
"Role playing", # Users engage in scenarios requiring ChatGPT to adopt a character or persona.
"Data analysis", # Requests involve interpreting data, statistics, or performing analytical tasks.
"Creative writing", # Users seek assistance with crafting stories, poems, or other creative texts.
"Advice seeking", # Users ask for recommendations or guidance on various personal or professional issues.
"Brainstorming", # Involves generating ideas, creative thinking, or exploring possibilities.
"Others" # Any queries that do not fit into the above categories or are of a miscellaneous nature.
]
## Output Format:
Note that you can only select a single primary tag. Other applicable tags can be added to the list of other tags.
Now, please output your tags below in a json format by filling in the placeholders in <...>:
```
{{
"primary_tag": "<primary tag>",
"other_tags": ["<tag 1>", "<tag 2>", ... ]
}}
```
"""
OUTPUT_CLASSIFICATION_JSON_SCHEMA = {
"type": "object",
"properties": {
"primary_tag": {
"type": "string",
"enum": [
"Information seeking",
"Reasoning",
"Planning",
"Editing",
"Coding & Debugging",
"Math",
"Role playing",
"Data analysis",
"Creative writing",
"Advice seeking",
"Brainstorming",
"Others",
],
},
"other_tags": {
"type": "array",
"items": {
"type": "string",
"enum": [
"Information seeking",
"Reasoning",
"Planning",
"Editing",
"Coding & Debugging",
"Math",
"Role playing",
"Data analysis",
"Creative writing",
"Advice seeking",
"Brainstorming",
"Others",
],
},
},
},
"required": ["primary_tag", "other_tags"],
}
@step(inputs=["conversation"], outputs=["instruction"])
def GetInstruction(inputs: StepInput) -> "StepOutput":
for input in inputs:
input["instruction"] = input["conversation"][0]["content"]
yield inputs
class AssignTags(Task):
mission: Literal["difficulty", "quality", "classification"]
@property
def inputs(self) -> List[str]:
return ["instruction"]
def format_input(self, input: Dict[str, Any]) -> "ChatType":
instruction = input["instruction"]
if self.mission == "difficulty":
input_message = INPUT_DIFFICULTY_RATING_TEMPLATE.format(input=instruction)
elif self.mission == "quality":
input_message = INPUT_QUALITY_RATING_TEMPLATE.format(input=instruction)
else:
input_message = INPUT_CLASSIFICATION_TEMPLATE.format(input=instruction)
return [{"role": "user", "content": input_message}]
@property
def outputs(self) -> List[str]:
if self.mission == "difficulty":
return ["intent", "knowledge", "difficulty", "model_name"]
if self.mission == "quality":
return ["explanation", "quality", "model_name"]
return ["primary_tag", "other_tags", "model_name"]
def _impute_output(self) -> Dict[str, None]:
if self.mission == "difficulty":
return {"intent": None, "knowledge": None, "difficulty": None}
if self.mission == "quality":
return {"explanation": None, "quality": None}
return {"primary_tag": None, "other_tags": None}
def format_output(
self,
output: Union[str, None],
input: Union[Dict[str, Any], None] = None,
) -> Dict[str, Any]:
if output is None:
return self._impute_output()
return json.loads(output)
# https://github.com/magpie-align/magpie/blob/b08a80193c92ea7ec329dd9c23d6c23450c283b5/exp/gen_ins.py#L134
def de_md_logits_processor_for_llama3_1(token_ids, logits):
# Only process the initial logits
if len(token_ids) == 0:
logits[2] = -9999.999 # "#": 2,
logits[567] = -9999.999 # "##": 567,
logits[14711] = -9999.999 # "###": 14711,
logits[827] = -9999.999 # "####": 827,
logits[334] = -9999.999 # "**": 334
logits[3146] = -9999.999 # " **": 3146
logits[96618] = -9999.99 # "**:": 96618
return logits
with Pipeline(name="magpie-ultra-v1.0") as pipeline:
generate_instructions = MagpieGenerator(
llm=vLLM(
model="meta-llama/Meta-Llama-3.1-405B-Instruct-FP8",
tokenizer="meta-llama/Meta-Llama-3.1-405B-Instruct-FP8",
magpie_pre_query_template="llama3",
extra_kwargs={
"tensor_parallel_size": 8,
"max_model_len": 8192,
"enable_prefix_caching": True,
},
generation_kwargs={
"temperature": 0.8,
"top_p": 1.0,
"max_new_tokens": 1024,
"stop": [
"<|eot_id|>",
"<|end_of_text|>",
"<|start_header_id|>",
"<|end_header_id|>",
],
"stop_token_ids": [
128009,
128001,
128006,
128007,
],
"logits_processors": [de_md_logits_processor_for_llama3_1],
},
),
system_prompt=CATEGORIES_SYSTEM_PROMPTS,
batch_size=250,
n_turns=3,
)
get_instruction = GetInstruction(input_batch_size=5000)
assign_difficulty = AssignTags(
mission="difficulty",
llm=vLLM(
model="meta-llama/Meta-Llama-3.1-8B-Instruct",
extra_kwargs={
"tensor_parallel_size": 1,
},
structured_output={
"format": "json",
"schema": OUTPUT_DIFFICULTY_JSON_SCHEMA,
},
),
output_mappings={"model_name": "model_name_difficulty"},
input_batch_size=1000,
)
assign_quality = AssignTags(
mission="quality",
llm=vLLM(
model="meta-llama/Meta-Llama-3.1-8B-Instruct",
extra_kwargs={
"tensor_parallel_size": 1,
},
structured_output={
"format": "json",
"schema": OUTPUT_QUALITY_JSON_SCHEMA,
},
),
output_mappings={"model_name": "model_name_quality"},
input_batch_size=1000,
)
assign_classification = AssignTags(
mission="classification",
llm=vLLM(
model="meta-llama/Meta-Llama-3.1-8B-Instruct",
extra_kwargs={
"tensor_parallel_size": 1,
},
structured_output={
"format": "json",
"schema": OUTPUT_CLASSIFICATION_JSON_SCHEMA,
},
),
output_mappings={"model_name": "model_name_classification"},
input_batch_size=1000,
)
embeddings = EmbeddingGeneration(
embeddings=SentenceTransformerEmbeddings(
model="Alibaba-NLP/gte-large-en-v1.5",
device="cuda",
trust_remote_code=True,
),
input_mappings={"text": "instruction"},
output_mappings={"model_name": "model_name_embeddings"},
input_batch_size=50,
)
reward_model_score = RewardModelScore(
model="RLHFlow/ArmoRM-Llama3-8B-v0.1",
device_map="auto",
trust_remote_code=True,
input_batch_size=20,
)
combine_outputs = CombineOutputs()
guard = ChatGeneration(
llm=vLLM(
model="meta-llama/Llama-Guard-3-8B",
extra_kwargs={
"tensor_parallel_size": 1,
},
structured_output={
"format": "regex",
"schema": r"\n\n(?:safe|unsafe\n(?:S(?:[1-9]|1[0-4])))",
},
),
input_mappings={"messages": "conversation"},
output_mappings={"generation": "guard", "model_name": "model_name_guard"},
input_batch_size=1000,
)
nearest_neighbours = FaissNearestNeighbour(
metric_type=faiss.METRIC_INNER_PRODUCT, k=5
)
(
generate_instructions
>> get_instruction
>> [
assign_difficulty,
assign_quality,
assign_classification,
embeddings,
reward_model_score,
guard,
]
>> combine_outputs
>> nearest_neighbours
)
if __name__ == "__main__":
distiset = pipeline.run(
parameters={
generate_instructions.name: {"num_rows": 1000000, "resources": {"gpus": 8}},
assign_difficulty.name: {
"llm": {
"generation_kwargs": {"max_new_tokens": 512, "temperature": 0.0}
},
"resources": {"gpus": 1},
},
assign_quality.name: {
"llm": {
"generation_kwargs": {"max_new_tokens": 512, "temperature": 0.0}
},
"resources": {"gpus": 1},
},
assign_classification.name: {
"llm": {
"generation_kwargs": {"max_new_tokens": 512, "temperature": 0.0}
},
"resources": {"gpus": 1},
},
embeddings.name: {
"resources": {"gpus": 1},
},
reward_model_score.name: {"resources": {"gpus": 1, "replicas": 3}},
guard.name: {
"llm": {
"generation_kwargs": {"max_new_tokens": 128, "temperature": 0.0}
},
"resources": {"gpus": 1},
},
},
)
distiset.push_to_hub("argilla/magpie-ultra-v1.0")
docker/Dockerfile 0 → 100644
View file @ e75bc9be
FROM image.sourcefind.cn:5000/dcu/admin/base/pytorch:2.3.0-ubuntu22.04-dtk24.04.2-py3.10
ENV DEBIAN_FRONTEND=noninteractive
# RUN yum update && yum install -y git cmake wget build-essential
# RUN source /opt/dtk-24.04.2/env.sh
# # 安装pip相关依赖
COPY requirements.txt requirements.txt
RUN pip3 install -r requirements.txt -i http://mirrors.aliyun.com/pypi/simple/ --trusted-host mirrors.aliyun.com
docker/requirements.txt 0 → 100644
View file @ e75bc9be
transformers
trl
peft
accelerate
datasets
scipy
wandb # wandb offline & wandb disabled
bitsandbytes
\ No newline at end of file
docker_start.sh 0 → 100644
View file @ e75bc9be
docker run -it --shm-size=64G -v $PWD/smollm:/home/smollm -v /public/DL_DATA/AI:/home/AI -v /opt/hyhal:/opt/hyhal:ro --privileged=true --device=/dev/kfd --device=/dev/dri/ --group-add video --name smollm 83714c19d308 bash
# python -m torch.utils.collect_env
evaluation/README.md 0 → 100644
View file @ e75bc9be
# SmolLM evaluation scripts
We're using the [LightEval](https://github.com/huggingface/lighteval/) library to benchmark our models.
Check out the [quick tour](https://github.com/huggingface/lighteval/wiki/Quicktour) to configure it to your own hardware and tasks.
## Setup
Use conda/venv with `python>=3.10`.
Adjust the pytorch installation according to your environment:
```bash
pip install torch==2.4.0 --index-url https://download.pytorch.org/whl/cu121
```
For reproducibility, we recommend fixed versions of the libraries:
```bash
pip install -r requirements.txt
```
## Running the evaluations
### SmolLM2 base models
```bash
lighteval accelerate \
--model_args "pretrained=HuggingFaceTB/SmolLM2-1.7B,revision=main,dtype=bfloat16,vllm,gpu_memory_utilisation=0.8,max_model_length=2048" \
--custom_tasks "tasks.py" --tasks "smollm2_base.txt" --output_dir "./evals" --save_details
```
### SmolLM2 instruction-tuned models
(note the `--use_chat_template` flag)
```bash
lighteval accelerate \
--model_args "pretrained=HuggingFaceTB/SmolLM2-1.7B-Instruct,revision=main,dtype=bfloat16,vllm,gpu_memory_utilisation=0.8,max_model_length=2048" \
--custom_tasks "tasks.py" --tasks "smollm2_instruct.txt" --use_chat_template --output_dir "./evals" --save_details
```
### MATH and other extra tasks
```bash
lighteval accelerate \
--model_args "pretrained=HuggingFaceTB/SmolLM2-1.7B-Instruct,revision=main,dtype=bfloat16,vllm,gpu_memory_utilisation=0.8,max_model_length=4096" \
--custom_tasks "tasks.py" --tasks "custom|math|4|1" --use_chat_template --output_dir "./evals" --save_details
```
evaluation/math_utils.py 0 → 100644
View file @ e75bc9be
import re
from typing import Optional
def math_normalizer(text: str) -> str:
def last_boxed_only_string(string: str) -> Optional[str]:
idx = string.rfind("\\boxed")
if "\\boxed " in string:
return "\\boxed " + string.split("\\boxed ")[-1].split("$")[0]
if idx < 0:
idx = string.rfind("\\fbox")
if idx < 0:
return None
i = idx
right_brace_idx = None
num_left_braces_open = 0
while i < len(string):
if string[i] == "{":
num_left_braces_open += 1
if string[i] == "}":
num_left_braces_open -= 1
if num_left_braces_open == 0:
right_brace_idx = i
break
i += 1
if right_brace_idx is None:
retval = None
else:
retval = string[idx: right_brace_idx + 1]
return retval
def remove_boxed(s: str) -> str:
if "\\boxed " in s:
left = "\\boxed "
#assert s[: len(left)] == left
return s[len(left):]
left = "\\boxed{"
#assert s[: len(left)] == left
#assert s[-1] == "}"
return s[len(left): -1]
SUBSTITUTIONS = [
("an ", ""),
("a ", ""),
(".$", "$"),
("\\$", ""),
(r"\ ", ""),
(" ", ""),
("mbox", "text"),
(",\\text{and}", ","),
("\\text{and}", ","),
("\\text{m}", "\\text{}"),
]
REMOVED_EXPRESSIONS = [
"square",
"ways",
"integers",
"dollars",
"mph",
"inches",
"ft",
"hours",
"km",
"units",
"\\ldots",
"sue",
"points",
"feet",
"minutes",
"digits",
"cents",
"degrees",
"cm",
"gm",
"pounds",
"meters",
"meals",
"edges",
"students",
"childrentickets",
"multiples",
"\\text{s}",
"\\text{.}",
"\\text{\ns}",
"\\text{}^2",
"\\text{}^3",
"\\text{\n}",
"\\text{}",
r"\mathrm{th}",
r"^\circ",
r"^{\circ}",
r"\;",
r",\!",
"{,}",
'"',
"\\dots",
]
def normalize_final_answer(final_answer: str) -> str:
"""
Normalize a final answer to a quantitative reasoning question.
Copied character for character from appendix D of Lewkowycz et al. (2022)
"""
final_answer = final_answer.split("=")[-1]
for before, after in SUBSTITUTIONS:
final_answer = final_answer.replace(before, after)
for expr in REMOVED_EXPRESSIONS:
final_answer = final_answer.replace(expr, "")
# Extract answer that is in LaTeX math, is bold,
# is surrounded by a box, etc.
final_answer = re.sub(r"(.*?)(\$)(.*?)(\$)(.*)", "$\\3$", final_answer)
final_answer = re.sub(r"(\\text\{)(.*?)(\})", "\\2", final_answer)
final_answer = re.sub(r"(\\textbf\{)(.*?)(\})", "\\2", final_answer)
final_answer = re.sub(r"(\\overline\{)(.*?)(\})", "\\2", final_answer)
final_answer = re.sub(r"(\\boxed\{)(.*)(\})", "\\2", final_answer)
# Normalize shorthand TeX:
# \fracab -> \frac{a}{b}
# \frac{abc}{bef} -> \frac{abc}{bef}
# \fracabc -> \frac{a}{b}c
# \sqrta -> \sqrt{a}
# \sqrtab -> sqrt{a}b
final_answer = re.sub(r"(frac)([^{])(.)", "frac{\\2}{\\3}", final_answer)
final_answer = re.sub(r"(sqrt)([^{])", "sqrt{\\2}", final_answer)
final_answer = final_answer.replace("$", "")
# Normalize 100,000 -> 100000
if final_answer.replace(",", "").isdigit():
final_answer = final_answer.replace(",", "")
return final_answer
if text is None:
return ""
boxed_string = last_boxed_only_string(text)
if boxed_string is None:
return ""
return normalize_final_answer(remove_boxed(boxed_string))
\ No newline at end of file
evaluation/requirements.txt 0 → 100644
View file @ e75bc9be
lighteval[accelerate,extended_tasks,vllm] @ git+https://github.com/huggingface/lighteval@ea46419a93fb390e8f694f7c6c64c1e684487c9d
fsspec>=2024.3.0
\ No newline at end of file
evaluation/smollm2_base.txt 0 → 100644
View file @ e75bc9be
custom|hellaswag|0|1
custom|arc|0|1
custom|piqa|0|1
custom|mmlu_pro|0|1
custom|commonsense_qa|0|1
custom|trivia_qa|0|1
custom|winogrande|0|1
custom|openbook_qa|0|1
custom|gsm8k|5|1
\ No newline at end of file
evaluation/smollm2_instruct.txt 0 → 100644
View file @ e75bc9be
extended|ifeval|0|0
custom|hellaswag|0|1
custom|arc|0|1
custom|piqa|0|1
custom|mmlu_pro|0|1
custom|bbh|3|1
custom|gsm8k|5|1
\ No newline at end of file
evaluation/tasks.py 0 → 100644
View file @ e75bc9be
import re
import numpy as np
from lighteval.tasks.lighteval_task import LightevalTaskConfig
from lighteval.tasks.requests import Doc
from lighteval.metrics.metrics import Metrics, SampleLevelMetric, MetricCategory, MetricUseCase, ExactMatches
import lighteval.tasks.default_prompts as prompt
from .math_utils import math_normalizer
def prompt_hellaswag(line, task_name: str = None):
def preprocess(text):
"""Comes from AiHarness"""
# text = text.strip()
# NOTE: Brackets are artifacts of the WikiHow dataset portion of HellaSwag.
text = text.replace(" [title]", ". ")
text = re.sub("\\[.*?\\]", "", text)
text = text.replace(" ", " ")
return text
ctx = f"{line['ctx_a']} {line['ctx_b'].capitalize()} "
return Doc(
task_name=task_name,
query=preprocess(line["activity_label"] + ": " + ctx),
choices=[" " + preprocess(ending) for ending in line["endings"]],
gold_index=int(line["label"]) if line["label"] != "" else -1, # -1 for test
)
def prompt_commonsense_qa(line, task_name: str = None):
return Doc(
task_name=task_name,
query=line["question"],
choices=[f" {c}" for c in line["choices"]["text"]],
gold_index=line["choices"]["label"].index(line["answerKey"].strip()),
instruction="",
)
def mmlu_pro_mc_prompt(line, task_name: str = None):
options = line["options"]
letters = [chr(ord("A") + i) for i in range(len(options))]
topic = line["category"].replace('_', ' ')
query = f"The following are multiple choice questions (with answers) about {topic}.\n\n"
query += line["question"] + "\n"
query += "".join([f"{letter}. {choice}\n" for letter, choice in zip(letters, options)])
query += "Answer:"
return Doc(
task_name=task_name,
query=query,
choices=letters,
gold_index=line["answer_index"],
instruction=f"The following are multiple choice questions (with answers) about {topic}.\n\n",
)
def bbh_prompt(line, task_name: str = None):
return Doc(
task_name=task_name,
query="Question: " + line["input"] + "\nAnswer: ",
choices=[line["target"]],
gold_index=0,
)
def prompt_math(line, task_name: str = None):
return Doc(
task_name=task_name,
query=f"Problem:\n{line['problem']}\n\nSolution:\n",
gold_index=0,
choices=[f"{line['solution']}\n"],
)
TASKS_TABLE = [
LightevalTaskConfig(
name="arc:easy",
prompt_function=prompt.arc,
suite=["custom"],
hf_repo="ai2_arc",
hf_revision="210d026faf9955653af8916fad021475a3f00453",
hf_subset=f"ARC-Easy",
evaluation_splits=["test"],
metric=[Metrics.loglikelihood_acc_norm_nospace],
),
LightevalTaskConfig(
name="arc:challenge",
prompt_function=prompt.arc,
suite=["custom"],
hf_repo="ai2_arc",
hf_revision="210d026faf9955653af8916fad021475a3f00453",
hf_subset=f"ARC-Challenge",
evaluation_splits=["test"],
metric=[Metrics.loglikelihood_acc_norm_nospace],
),
LightevalTaskConfig(
name="openbook_qa",
prompt_function=prompt.openbookqa,
suite=["custom"],
hf_repo="allenai/openbookqa",
hf_subset="main",
hf_revision="388097ea7776314e93a529163e0fea805b8a6454",
metric=[Metrics.loglikelihood_acc_norm_nospace],
),
LightevalTaskConfig(
name="hellaswag",
prompt_function=prompt_hellaswag,
suite=["custom"],
hf_repo="Rowan/hellaswag",
hf_subset="default",
hf_revision="6002345709e0801764318f06bf06ce1e7d1a1fe3",
trust_dataset=True,
metric=[Metrics.loglikelihood_acc_norm_nospace],
),
LightevalTaskConfig(
name="commonsense_qa",
prompt_function=prompt_commonsense_qa,
suite=["custom"],
hf_repo="tau/commonsense_qa",
hf_subset="default",
hf_revision="94630fe30dad47192a8546eb75f094926d47e155",
metric=[Metrics.loglikelihood_acc_norm_nospace],
),
LightevalTaskConfig(
name="winogrande",
prompt_function=prompt.winogrande,
suite=["custom"],
hf_repo="allenai/winogrande",
hf_subset="winogrande_xl",
hf_revision="85ac5b5a3b7a930e22d590176e39460400d19e41",
trust_dataset=True,
metric=[Metrics.loglikelihood_acc_norm_nospace],
),
LightevalTaskConfig(
name="piqa",
prompt_function=prompt.piqa_harness,
suite=["custom"],
hf_repo="ybisk/piqa",
hf_subset="plain_text",
hf_revision="2e8ac2dffd59bac8c3c6714948f4c551a0848bb0",
trust_dataset=True,
metric=[Metrics.loglikelihood_acc_norm_nospace],
),
LightevalTaskConfig(
name="trivia_qa",
prompt_function=prompt.triviaqa,
suite=["custom"],
hf_repo="mandarjoshi/trivia_qa",
hf_subset="rc.nocontext",
hf_revision="0f7faf33a3908546c6fd5b73a660e0f8ff173c2f",
hf_avail_splits=["train", "validation"],
evaluation_splits=["validation"],
metric=[Metrics.quasi_exact_match_triviaqa],
generation_size=20,
trust_dataset=True,
stop_sequence=["Question:", "Question"],
few_shots_select="random_sampling_from_train",
),
LightevalTaskConfig(
name="mmlu_pro",
prompt_function=mmlu_pro_mc_prompt,
suite=["custom"],
hf_repo="TIGER-Lab/MMLU-Pro",
hf_subset="default",
hf_revision="3373e0b32277875b8db2aa555a333b78a08477ea",
metric=[Metrics.loglikelihood_acc_norm_nospace],
evaluation_splits=["test"],
few_shots_split="validation",
),
LightevalTaskConfig(
name="gsm8k",
prompt_function=prompt.gsm8k,
suite=["custom"],
hf_repo="openai/gsm8k",
hf_subset="main",
hf_revision="e53f048856ff4f594e959d75785d2c2d37b678ee",
hf_avail_splits=["train", "test"],
evaluation_splits=["test"],
metric=[Metrics.quasi_exact_match_gsm8k],
generation_size=256,
stop_sequence=["Question:", "Question"],
few_shots_select="random_sampling_from_train",
)
]
BBH_TASKS = [
LightevalTaskConfig(
name=f"bbh:{subset}",
prompt_function=bbh_prompt,
suite=["custom"],
hf_repo="lighteval/big_bench_hard",
hf_subset=subset,
hf_revision="80610173426f05e6f1448f047e2db4840a7dd899",
metric=[Metrics.exact_match],
hf_avail_splits=["train"],
# this is the only split available, obviously not used in training
evaluation_splits=["train"],
few_shots_split="train",
trust_dataset=True,
stop_sequence=["Question:", "Question"],
)
for subset in [
"boolean_expressions",
"causal_judgement",
"date_understanding",
"disambiguation_qa",
"dyck_languages",
"formal_fallacies",
"geometric_shapes",
"hyperbaton",
"logical_deduction_five_objects",
"logical_deduction_seven_objects",
"logical_deduction_three_objects",
"movie_recommendation",
"multistep_arithmetic_two",
"navigate",
"object_counting",
"penguins_in_a_table",
"reasoning_about_colored_objects",
"ruin_names",
"salient_translation_error_detection",
"snarks",
"sports_understanding",
"temporal_sequences",
"tracking_shuffled_objects_five_objects",
"tracking_shuffled_objects_seven_objects",
"tracking_shuffled_objects_three_objects",
"web_of_lies",
"word_sorting",
]
]
TASKS_TABLE.extend(BBH_TASKS)
quasi_exact_match_math = SampleLevelMetric(
metric_name="qem",
sample_level_fn=ExactMatches(
strip_strings=True, normalize_pred=math_normalizer, normalize_gold=math_normalizer
).compute,
category=MetricCategory.GENERATIVE,
use_case=MetricUseCase.MATH,
corpus_level_fn=np.mean,
higher_is_better=True,
)
MATH_TASKS = [
LightevalTaskConfig(
name=f"math:{subset}",
prompt_function=prompt_math,
suite=["custom"],
hf_repo="HuggingFaceTB/MATH",
hf_subset=subset,
hf_revision="140a673f1f7182daf7923fdc7108e8cdbf97df46",
hf_avail_splits=["train", "test"],
evaluation_splits=["test"],
metric=[quasi_exact_match_math],
generation_size=1024,
stop_sequence=["Problem:"],
few_shots_split="fewshot",
few_shots_select="sequential",
trust_dataset=True,
)
for subset in [
"algebra",
"counting_and_probability",
"geometry",
"intermediate_algebra",
"number_theory",
"prealgebra",
"precalculus",
]
]
TASKS_TABLE.extend(MATH_TASKS)
if __name__ == "__main__":
print(t.name for t in TASKS_TABLE)
print(len(TASKS_TABLE))
\ No newline at end of file
finetuning/Dockerfile 0 → 100644
View file @ e75bc9be
# base image: CUDA 12.1
FROM nvidia/cuda:12.1.1-cudnn8-runtime-ubuntu22.04
WORKDIR /app
# install necessary packages
RUN apt-get update && apt-get install -y \
git \
wget \
curl \
ca-certificates \
libglib2.0-0 \
libsm6 \
libxrender1 \
libxext6 \
libssl-dev \
libffi-dev \
python3 \
python3-pip \
&& rm -rf /var/lib/apt/lists/*
# set python3 as default python
RUN update-alternatives --install /usr/bin/python python /usr/bin/python3 1
RUN pip install --upgrade pip setuptools
RUN pip install torch torchvision torchaudio --extra-index-url https://download.pytorch.org/whl/nightly/cu121
COPY requirements.txt .
RUN pip install -r requirements.txt
COPY . .
ENV PYTHONUNBUFFERED=1
CMD ["bash"]
\ No newline at end of file
finetuning/HuggingFaceTB/SmolLM2-1.7B-Instruct/README.md 0 → 100644
View file @ e75bc9be
---
library_name: transformers
license: apache-2.0
language:
- en
---
# SmolLM2
![image/png](https://cdn-uploads.huggingface.co/production/uploads/61c141342aac764ce1654e43/y45hIMNREW7w_XpHYB_0q.png)
## Table of Contents
1. [Model Summary](#model-summary)
2. [Evaluation](#evaluation)
3. [Examples](#examples)
4. [Limitations](#limitations)
5. [Training](#training)
6. [License](#license)
7. [Citation](#citation)
## Model Summary
SmolLM2 is a family of compact language models available in three size: 135M, 360M, and 1.7B parameters. They are capable of solving a wide range of tasks while being lightweight enough to run on-device.
The 1.7B variant demonstrates significant advances over its predecessor SmolLM1-1.7B, particularly in instruction following, knowledge, reasoning, and mathematics. It was trained on 11 trillion tokens using a diverse dataset combination: FineWeb-Edu, DCLM, The Stack, along with new mathematics and coding datasets that we curated and will release soon. We developed the instruct version through supervised fine-tuning (SFT) using a combination of public datasets and our own curated datasets. We then applied Direct Preference Optimization (DPO) using [UltraFeedback](https://huggingface.co/datasets/HuggingFaceH4/ultrafeedback_binarized).
The instruct model additionally supports tasks such as text rewriting, summarization and function calling thanks to datasets developed by [Argilla](https://huggingface.co/argilla) such as [Synth-APIGen-v0.1](https://huggingface.co/datasets/argilla/Synth-APIGen-v0.1).
### How to use
### Transformers
```bash
pip install transformers
```
```python
from transformers import AutoModelForCausalLM, AutoTokenizer
checkpoint = "HuggingFaceTB/SmolLM2-1.7B-Instruct"
device = "cuda" # for GPU usage or "cpu" for CPU usage
tokenizer = AutoTokenizer.from_pretrained(checkpoint)
# for multiple GPUs install accelerate and do `model = AutoModelForCausalLM.from_pretrained(checkpoint, device_map="auto")`
model = AutoModelForCausalLM.from_pretrained(checkpoint).to(device)
messages = [{"role": "user", "content": "What is the capital of France."}]
input_text=tokenizer.apply_chat_template(messages, tokenize=False)
inputs = tokenizer.encode(input_text, return_tensors="pt").to(device)
outputs = model.generate(inputs, max_new_tokens=50, temperature=0.2, top_p=0.9, do_sample=True)
print(tokenizer.decode(outputs[0]))
```
### Chat in TRL
You can also use the TRL CLI to chat with the model from the terminal:
```bash
pip install trl
trl chat --model_name_or_path HuggingFaceTB/SmolLM2-1.7B-Instruct --device cpu
```
## Evaluation
In this section, we report the evaluation results of SmolLM2. All evaluations are zero-shot unless stated otherwise, and we use [lighteval](https://github.com/huggingface/lighteval) to run them.
## Base Pre-Trained Model
| Metric | SmolLM2-1.7B | Llama-1B | Qwen2.5-1.5B | SmolLM1-1.7B |
|------------------|--------------|-------------|---------------|--------------|
| HellaSwag | **68.7** | 61.2 | 66.4 | 62.9 |
| ARC (Average) | **60.5** | 49.2 | 58.5 | 59.9 |
| PIQA | **77.6** | 74.8 | 76.1 | 76.0 |
| MMLU-Pro (MCF) | **19.4** | 11.7 | 13.7 | 10.8 |
| CommonsenseQA | **43.6** | 41.2 | 34.1 | 38.0 |
| TriviaQA | **36.7** | 28.1 | 20.9 | 22.5 |
| Winogrande | **59.4** | 57.8 | 59.3 | 54.7 |
| OpenBookQA | 42.2 | 38.4 | 40.0 | **42.4** |
| GSM8K (5-shot) | 31.0 | 7.2 | **61.3** | 5.5 |
## Instruction Model
| Metric | SmolLM2-1.7B-Instruct | Llama-1B-Instruct | Qwen2.5-1.5B-Instruct | SmolLM1-1.7B-Instruct |
|:-----------------------------|:---------------------:|:-----------------:|:----------------------:|:----------------------:|
| IFEval (Average prompt/inst) | **56.7** | 53.5 | 47.4 | 23.1 |
| MT-Bench | 6.13 | 5.48 | **6.52** | 4.33 |
| OpenRewrite-Eval (micro_avg RougeL) | 44.9 | 39.2 | **46.9** | NaN |
| HellaSwag | **66.1** | 56.1 | 60.9 | 55.5 |
| ARC (Average) | **51.7** | 41.6 | 46.2 | 43.7 |
| PIQA | **74.4** | 72.3 | 73.2 | 71.6 |
| MMLU-Pro (MCF) | 19.3 | 12.7 | **24.2** | 11.7 |
| BBH (3-shot) | 32.2 | 27.6 | **35.3** | 25.7 |
| GSM8K (5-shot) | **48.2** | 26.8 | 42.8 | 4.62 |
## Examples
Below are some system and instruct prompts that work well for special tasks
### Text rewriting
```python
system_prompt_rewrite = "You are an AI writing assistant. Your task is to rewrite the user's email to make it more professional and approachable while maintaining its main points and key message. Do not return any text other than the rewritten message."
user_prompt_rewrite = "Rewrite the message below to make it more friendly and approachable while maintaining its main points and key message. Do not add any new information or return any text other than the rewritten message\nThe message:"
messages = [{"role": "system", "content": system_prompt_rewrite}, {"role": "user", "content":f"{user_prompt_rewrite} The CI is failing after your last commit!"}]
input_text=tokenizer.apply_chat_template(messages, tokenize=False)
inputs = tokenizer.encode(input_text, return_tensors="pt").to(device)
outputs = model.generate(inputs, max_new_tokens=50, temperature=0.2, top_p=0.9, do_sample=True)
print(tokenizer.decode(outputs[0]))
```
```
Hey there! I noticed that the CI isn't passing after your latest commit. Could you take a look and let me know what's going on? Thanks so much for your help!
```
### Summarization
```python
system_prompt_summarize = "Provide a concise, objective summary of the input text in up to three sentences, focusing on key actions and intentions without using second or third person pronouns."
messages = [{"role": "system", "content": system_prompt_rewrite}, {"role": "user", "content": INSERT_LONG_EMAIL]
input_text=tokenizer.apply_chat_template(messages, tokenize=False)
inputs = tokenizer.encode(input_text, return_tensors="pt").to(device)
outputs = model.generate(inputs, max_new_tokens=50, temperature=0.2, top_p=0.9, do_sample=True)
print(tokenizer.decode(outputs[0]))
```
### Function calling
SmolLM2-1.7B-Instruct can handle function calling, it scores 27% on the [BFCL Leaderboard](https://gorilla.cs.berkeley.edu/blogs/8_berkeley_function_calling_leaderboard.html). Here's how you can leverage it:
```python
import json
import re
from typing import Optional
from jinja2 import Template
import torch
from transformers import AutoModelForCausalLM, AutoTokenizer
from transformers.utils import get_json_schema
system_prompt = Template("""You are an expert in composing functions. You are given a question and a set of possible functions.
Based on the question, you will need to make one or more function/tool calls to achieve the purpose.
If none of the functions can be used, point it out and refuse to answer.
If the given question lacks the parameters required by the function, also point it out.
You have access to the following tools:
<tools>{{ tools }}</tools>
The output MUST strictly adhere to the following format, and NO other text MUST be included.
The example format is as follows. Please make sure the parameter type is correct. If no function call is needed, please make the tool calls an empty list '[]'.
<tool_call>[
{"name": "func_name1", "arguments": {"argument1": "value1", "argument2": "value2"}},
... (more tool calls as required)
]</tool_call>""")
def prepare_messages(
query: str,
tools: Optional[dict[str, any]] = None,
history: Optional[list[dict[str, str]]] = None
) -> list[dict[str, str]]:
"""Prepare the system and user messages for the given query and tools.
Args:
query: The query to be answered.
tools: The tools available to the user. Defaults to None, in which case if a
list without content will be passed to the model.
history: Exchange of messages, including the system_prompt from
the first query. Defaults to None, the first message in a conversation.
"""
if tools is None:
tools = []
if history:
messages = history.copy()
messages.append({"role": "user", "content": query})
else:
messages = [
{"role": "system", "content": system_prompt.render(tools=json.dumps(tools))},
{"role": "user", "content": query}
]
return messages
def parse_response(text: str) -> str | dict[str, any]:
"""Parses a response from the model, returning either the
parsed list with the tool calls parsed, or the
model thought or response if couldn't generate one.
Args:
text: Response from the model.
"""
pattern = r"<tool_call>(.*?)</tool_call>"
matches = re.findall(pattern, text, re.DOTALL)
if matches:
return json.loads(matches[0])
return text
model_name_smollm = "HuggingFaceTB/SmolLM2-1.7B-Instruct"
model = AutoModelForCausalLM.from_pretrained(model_name_smollm, device_map="auto", torch_dtype="auto", trust_remote_code=True)
tokenizer = AutoTokenizer.from_pretrained(model_name_smollm)
from datetime import datetime
import random
def get_current_time() -> str:
"""Returns the current time in 24-hour format.
Returns:
str: Current time in HH:MM:SS format.
"""
return datetime.now().strftime("%H:%M:%S")
def get_random_number_between(min: int, max: int) -> int:
"""
Gets a random number between min and max.
Args:
min: The minimum number.
max: The maximum number.
Returns:
A random number between min and max.
"""
return random.randint(min, max)
tools = [get_json_schema(get_random_number_between), get_json_schema(get_current_time)]
toolbox = {"get_random_number_between": get_random_number_between, "get_current_time": get_current_time}
query = "Give me a number between 1 and 300"
messages = prepare_messages(query, tools=tools)
inputs = tokenizer.apply_chat_template(messages, add_generation_prompt=True, return_tensors="pt").to(model.device)
outputs = model.generate(inputs, max_new_tokens=512, do_sample=False, num_return_sequences=1, eos_token_id=tokenizer.eos_token_id)
result = tokenizer.decode(outputs[0][len(inputs[0]):], skip_special_tokens=True)
tool_calls = parse_response(result)
# [{'name': 'get_random_number_between', 'arguments': {'min': 1, 'max': 300}}
# Get tool responses
tool_responses = [toolbox.get(tc["name"])(*tc["arguments"].values()) for tc in tool_calls]
# [63]
# For the second turn, rebuild the history of messages:
history = messages.copy()
# Add the "parsed response"
history.append({"role": "assistant", "content": result})
query = "Can you give me the hour?"
history.append({"role": "user", "content": query})
inputs = tokenizer.apply_chat_template(history, add_generation_prompt=True, return_tensors="pt").to(model.device)
outputs = model.generate(inputs, max_new_tokens=512, do_sample=False, num_return_sequences=1, eos_token_id=tokenizer.eos_token_id)
result = tokenizer.decode(outputs[0][len(inputs[0]):], skip_special_tokens=True)
tool_calls = parse_response(result)
tool_responses = [toolbox.get(tc["name"])(*tc["arguments"].values()) for tc in tool_calls]
# ['07:57:25']
```
More details such as parallel function calls and tools not available can be found [here](https://huggingface.co/HuggingFaceTB/SmolLM2-1.7B-Instruct/blob/main/instructions_function_calling.md)
## Limitations
SmolLM2 models primarily understand and generate content in English. They can produce text on a variety of topics, but the generated content may not always be factually accurate, logically consistent, or free from biases present in the training data. These models should be used as assistive tools rather than definitive sources of information. Users should always verify important information and critically evaluate any generated content.
## Training
### Model
- **Architecture:** Transformer decoder
- **Pretraining tokens:** 11T
- **Precision:** bfloat16
### Hardware
- **GPUs:** 256 H100
### Software
- **Training Framework:** [nanotron](https://github.com/huggingface/nanotron/tree/main)
- **Alignement Handbook** [alignement-handbook](https://github.com/huggingface/alignment-handbook/)
## License
[Apache 2.0](https://www.apache.org/licenses/LICENSE-2.0)
## Citation
```bash
@misc{allal2024SmolLM2,
title={SmolLM2 - with great data, comes great performance},
author={Loubna Ben Allal and Anton Lozhkov and Elie Bakouch and Gabriel Martín Blázquez and Lewis Tunstall and Agustín Piqueres and Andres Marafioti and Cyril Zakka and Leandro von Werra and Thomas Wolf},
year={2024},
}
```
\ No newline at end of file
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment