🤗 Serve any model with Inference Endpoints + Custom Handlers
•
2
Alvaro Bartolome
alvarobartt
AI & ML interests
☁️ cloud machine learning @huggingface and open source passionate
Recent Activity
New activity
1 day ago
hf-vision/object_detection_leaderboard:🚩 Leaderboard down!
updated
a collection
2 days ago
🤗 Custom Inference Endpoints Handlers
Reacted to
danielhanchen's
post
with 🔥
5 days ago
Vision finetuning is in 🦥Unsloth! You can now finetune Llama 3.2, Qwen2 VL, Pixtral and all Llava variants up to 2x faster and with up to 70% less VRAM usage! Colab to finetune Llama 3.2: https://colab.research.google.com/drive/1j0N4XTY1zXXy7mPAhOC1_gMYZ2F2EBlk?usp=sharing
Articles
Organizations
alvarobartt's activity
Reacted to
danielhanchen's
post with 🔥
5 days ago
Post
1198
Vision finetuning is in 🦥Unsloth! You can now finetune Llama 3.2, Qwen2 VL, Pixtral and all Llava variants up to 2x faster and with up to 70% less VRAM usage! Colab to finetune Llama 3.2: https://colab.research.google.com/drive/1j0N4XTY1zXXy7mPAhOC1_gMYZ2F2EBlk?usp=sharing
- 1 reply
Reacted to
clem's
post with 🔥
about 1 month ago
Post
4405
This is no Woodstock AI but will be fun nonetheless haha. I’ll be hosting a live workshop with team members next week about the Enterprise Hugging Face hub.
1,000 spots available first-come first serve with some surprises during the stream!
You can register and add to your calendar here: https://streamyard.com/watch/JS2jHsUP3NDM
1,000 spots available first-come first serve with some surprises during the stream!
You can register and add to your calendar here: https://streamyard.com/watch/JS2jHsUP3NDM
- 4 replies
posted
an
update
3 months ago
Post
2753
🤗 Serving Meta Llama 3.1 405B on Google Cloud is now possible via the Hugging Face Deep Learning Containers (DLCs) for Text Generation Inference (TGI)
In this post, we showcase how to deploy https://huggingface.co/meta-llama/Meta-Llama-3.1-405B-Instruct-FP8 on an A3 instance with 8 x H100 GPUs on Vertex AI
Thanks to the Hugging Face DLCs for TGI and Google Cloud Vertex AI, deploying a high-performance text generation container for serving Large Language Models (LLMs) has never been easier. And we’re not going to stop here – stay tuned as we enable more experiences to build AI with open models on Google Cloud!
Read the full post at https://huggingface.co/blog/llama31-on-vertex-ai
In this post, we showcase how to deploy https://huggingface.co/meta-llama/Meta-Llama-3.1-405B-Instruct-FP8 on an A3 instance with 8 x H100 GPUs on Vertex AI
Thanks to the Hugging Face DLCs for TGI and Google Cloud Vertex AI, deploying a high-performance text generation container for serving Large Language Models (LLMs) has never been easier. And we’re not going to stop here – stay tuned as we enable more experiences to build AI with open models on Google Cloud!
Read the full post at https://huggingface.co/blog/llama31-on-vertex-ai
Reacted to
merve's
post with 🔥
4 months ago
Post
4961
Idefics3-Llama is out! 💥💥
Model: HuggingFaceM4/Idefics3-8B-Llama3
Demo: HuggingFaceM4/idefics3
It's a multimodal model based on Llama 3.1 that accepts an arbitrary number of interleaved images with text with a huge context window (10k tokens!) ✨
Supported by Hugging Face transformers 🤗
Model: HuggingFaceM4/Idefics3-8B-Llama3
Demo: HuggingFaceM4/idefics3
It's a multimodal model based on Llama 3.1 that accepts an arbitrary number of interleaved images with text with a huge context window (10k tokens!) ✨
Supported by Hugging Face transformers 🤗
- 2 replies
Reacted to
dvilasuero's
post with 🤗🚀🔥
6 months ago
Post
7973
Today is a huge day in Argilla’s history. We couldn’t be more excited to share this with the community: we’re joining Hugging Face!
We’re embracing a larger mission, becoming part of a brilliant and kind team and a shared vision about the future of AI.
Over the past year, we’ve been collaborating with Hugging Face on countless projects: launching partner of Docker Spaces, empowering the community to clean Alpaca translations into Spanish and other languages, launching argilla/notus-7b-v1 building on Zephyr’s learnings, the Data is Better Together initiative with hundreds of community contributors, or releasing argilla/OpenHermesPreferences, one of the largest open preference tuning datasets
After more than 2,000 Slack messages and over 60 people collaborating for over a year, it already felt like we were part of the same team, pushing in the same direction. After a week of the smoothest transition you can imagine, we’re now the same team.
To those of you who’ve been following us, this won’t be a huge surprise, but it will be a big deal in the coming months. This acquisition means we’ll double down on empowering the community to build and collaborate on high quality datasets, we’ll bring full support for multimodal datasets, and we’ll be in a better place to collaborate with the Open Source AI community. For enterprises, this means that the Enterprise Hub will unlock highly requested features like single sign-on and integration with Inference Endpoints.
As a founder, I am proud of the Argilla team. We're now part of something bigger and a larger team but with the same values, culture, and goals. Grateful to have shared this journey with my beloved co-founders Paco and Amélie.
Finally, huge thanks to the Chief Llama Officer @osanseviero for sparking this and being such a great partner during the acquisition process.
Would love to answer any questions you have so feel free to add them below!
We’re embracing a larger mission, becoming part of a brilliant and kind team and a shared vision about the future of AI.
Over the past year, we’ve been collaborating with Hugging Face on countless projects: launching partner of Docker Spaces, empowering the community to clean Alpaca translations into Spanish and other languages, launching argilla/notus-7b-v1 building on Zephyr’s learnings, the Data is Better Together initiative with hundreds of community contributors, or releasing argilla/OpenHermesPreferences, one of the largest open preference tuning datasets
After more than 2,000 Slack messages and over 60 people collaborating for over a year, it already felt like we were part of the same team, pushing in the same direction. After a week of the smoothest transition you can imagine, we’re now the same team.
To those of you who’ve been following us, this won’t be a huge surprise, but it will be a big deal in the coming months. This acquisition means we’ll double down on empowering the community to build and collaborate on high quality datasets, we’ll bring full support for multimodal datasets, and we’ll be in a better place to collaborate with the Open Source AI community. For enterprises, this means that the Enterprise Hub will unlock highly requested features like single sign-on and integration with Inference Endpoints.
As a founder, I am proud of the Argilla team. We're now part of something bigger and a larger team but with the same values, culture, and goals. Grateful to have shared this journey with my beloved co-founders Paco and Amélie.
Finally, huge thanks to the Chief Llama Officer @osanseviero for sparking this and being such a great partner during the acquisition process.
Would love to answer any questions you have so feel free to add them below!
- 28 replies
Reacted to
tomaarsen's
post with 🔥❤️
6 months ago
Post
1942
‼️Sentence Transformers v3.0 is out! You can now train and finetune embedding models with multi-GPU training, bf16 support, loss logging, callbacks & much more. I also release 50+ datasets to train on.
1️⃣ Training Refactor
Embedding models can now be trained using an extensive trainer with a lot of powerful features:
- MultiGPU Training (Data Parallelism (DP) and Distributed Data Parallelism (DDP))
- bf16 training support; loss logging
- Evaluation datasets + evaluation loss
- Improved callback support + an excellent Weights & Biases integration
- Gradient checkpointing, gradient accumulation
- Improved model card generation
- Resuming from a training checkpoint without performance loss
- Hyperparameter Optimization
and much more!
Read my detailed blogpost to learn about the components that make up this new training approach: https://huggingface.co/blog/train-sentence-transformers
2️⃣ Similarity Score
Not sure how to compare embeddings? Don't worry, you can now use
3️⃣ Additional Kwargs
Sentence Transformers relies on various Transformers instances (AutoModel, AutoTokenizer, AutoConfig), but it was hard to provide valuable keyword arguments to these (like 'torch_dtype=torch.bfloat16' to load a model a lower precision for 2x inference speedup). This is now easy!
4️⃣ Hyperparameter Optimization
Sentence Transformers now ships with HPO, allowing you to effectively choose your hyperparameters for your data and task.
5️⃣ Dataset Release
To help you out with finetuning models, I've released 50+ ready-to-go datasets that can be used with training or finetuning embedding models: sentence-transformers/embedding-model-datasets-6644d7a3673a511914aa7552
Full release notes: https://github.com/UKPLab/sentence-transformers/releases/tag/v3.0.0
1️⃣ Training Refactor
Embedding models can now be trained using an extensive trainer with a lot of powerful features:
- MultiGPU Training (Data Parallelism (DP) and Distributed Data Parallelism (DDP))
- bf16 training support; loss logging
- Evaluation datasets + evaluation loss
- Improved callback support + an excellent Weights & Biases integration
- Gradient checkpointing, gradient accumulation
- Improved model card generation
- Resuming from a training checkpoint without performance loss
- Hyperparameter Optimization
and much more!
Read my detailed blogpost to learn about the components that make up this new training approach: https://huggingface.co/blog/train-sentence-transformers
2️⃣ Similarity Score
Not sure how to compare embeddings? Don't worry, you can now use
model.similarity(embeddings1, embeddings2) and you'll get your similarity scores immediately. Model authors can specify their desired similarity score, so you don't have to worry about it anymore!3️⃣ Additional Kwargs
Sentence Transformers relies on various Transformers instances (AutoModel, AutoTokenizer, AutoConfig), but it was hard to provide valuable keyword arguments to these (like 'torch_dtype=torch.bfloat16' to load a model a lower precision for 2x inference speedup). This is now easy!
4️⃣ Hyperparameter Optimization
Sentence Transformers now ships with HPO, allowing you to effectively choose your hyperparameters for your data and task.
5️⃣ Dataset Release
To help you out with finetuning models, I've released 50+ ready-to-go datasets that can be used with training or finetuning embedding models: sentence-transformers/embedding-model-datasets-6644d7a3673a511914aa7552
Full release notes: https://github.com/UKPLab/sentence-transformers/releases/tag/v3.0.0
Reacted to
merve's
post with 🚀
6 months ago
Post
1756
New open Vision Language Model by
@Google
: PaliGemma 💙🤍
📝 Comes in 3B, pretrained, mix and fine-tuned models in 224, 448 and 896 resolution
🧩 Combination of Gemma 2B LLM and SigLIP image encoder
🤗 Supported in transformers
PaliGemma can do..
🧩 Image segmentation and detection! 🤯
📑 Detailed document understanding and reasoning
🙋 Visual question answering, captioning and any other VLM task!
Read our blog 🔖 hf.co/blog/paligemma
Try the demo 🪀 hf.co/spaces/google/paligemma
Check out the Spaces and the models all in the collection 📚 google/paligemma-release-6643a9ffbf57de2ae0448dda
Collection of fine-tuned PaliGemma models google/paligemma-ft-models-6643b03efb769dad650d2dda
📝 Comes in 3B, pretrained, mix and fine-tuned models in 224, 448 and 896 resolution
🧩 Combination of Gemma 2B LLM and SigLIP image encoder
🤗 Supported in transformers
PaliGemma can do..
🧩 Image segmentation and detection! 🤯
📑 Detailed document understanding and reasoning
🙋 Visual question answering, captioning and any other VLM task!
Read our blog 🔖 hf.co/blog/paligemma
Try the demo 🪀 hf.co/spaces/google/paligemma
Check out the Spaces and the models all in the collection 📚 google/paligemma-release-6643a9ffbf57de2ae0448dda
Collection of fine-tuned PaliGemma models google/paligemma-ft-models-6643b03efb769dad650d2dda
- 13 replies
posted
an
update
7 months ago
Post
3001
🔥 Prometheus 2 was recently released by Kaist AI as an alternative and closely mirroring both human and GPT-4 evaluation, and surpassing the former Prometheus!
prometheus-eval/prometheus-7b-v2.0
prometheus-eval/prometheus-8x7b-v2.0
🌬️Fine-tuned on top of mistralai/Mistral-7B-Instruct-v0.2 and mistralai/Mixtral-8x7B-Instruct-v0.1
🗂️The datasets used for fine-tuning have been publicly released i.e. prometheus-eval/Feedback-Collection and prometheus-eval/Preference-Collection
🤝🏻Unified LM evaluator for absolute (a single prompt-completion pair) and relative (two completions for a given prompt) due to model merging
❌No longer needs a mandatory reference / golden answer, but can still be provided optionally
🔝Surpasses the former version of Prometheus, and has a high correlation with human, GPT-4, and Claude 3 Opus scores when evaluating LMs
📝Apache 2.0 license
Long-story short, an amazing job from Kaist AI bridging the gap with LLM evaluators other than proprietary and bigger models!
This week at Argilla, we decided to add a new task to use Prometheus 2 as an LLM evaluator using
😱 Using
Find the generated dataset and the code at distilabel-internal-testing/instruction-dataset-prometheus
prometheus-eval/prometheus-7b-v2.0
prometheus-eval/prometheus-8x7b-v2.0
🌬️Fine-tuned on top of mistralai/Mistral-7B-Instruct-v0.2 and mistralai/Mixtral-8x7B-Instruct-v0.1
🗂️The datasets used for fine-tuning have been publicly released i.e. prometheus-eval/Feedback-Collection and prometheus-eval/Preference-Collection
🤝🏻Unified LM evaluator for absolute (a single prompt-completion pair) and relative (two completions for a given prompt) due to model merging
❌No longer needs a mandatory reference / golden answer, but can still be provided optionally
🔝Surpasses the former version of Prometheus, and has a high correlation with human, GPT-4, and Claude 3 Opus scores when evaluating LMs
📝Apache 2.0 license
Long-story short, an amazing job from Kaist AI bridging the gap with LLM evaluators other than proprietary and bigger models!
This week at Argilla, we decided to add a new task to use Prometheus 2 as an LLM evaluator using
distilabel, so we implemented PrometheusEval.😱 Using
PrometheusEval running their 7B variant with vLLM in a single L40 on top of
HuggingFaceH4/instruction-dataset, we got the 327 existing prompt-completion pairs evaluated and pushed to the Hub in less than 2 minutes!Find the generated dataset and the code at distilabel-internal-testing/instruction-dataset-prometheus
- 1 reply
Reacted to
davanstrien's
post with 🤗
7 months ago
Post
1670
As part of the Data is Better Together MPEP project, we are now at the point where some translation efforts have successfully translated 500 highly ranked prompts into a new target language (amazing work from
@Rijgersberg
et al!)
Our next step is to use these translated prompts to evaluate the performance of LLMs for non English languages.
Does LLM, as a judge, work outside of English?
Ideally, it would be compelling to leverage LLMs to judge models for non-English since this significantly lowers the barrier to evaluating models (although it doesn't remove this barrier altogether).
What we want to know is:
- does auto/LLM eval work in general for a particular language
- which model(s) works best as a judge
- do LLMs' judgments of non-English models match human preferences?
We're starting to think about how to approach this. If you have any ideas of possible approaches feel free to comment or join the discussion here: https://github.com/huggingface/data-is-better-together/issues/61
Other ideas...
Could an approach like Replacing Judges with Juries: Evaluating LLM Generations with a Panel of Diverse Models (2404.18796) with the SOA models for a particular language work? i.e., choose 4 of the best open LLMs for Arabic and use those at the pool of raters rather than relying on one powerful judge LLM?
Our next step is to use these translated prompts to evaluate the performance of LLMs for non English languages.
Does LLM, as a judge, work outside of English?
Ideally, it would be compelling to leverage LLMs to judge models for non-English since this significantly lowers the barrier to evaluating models (although it doesn't remove this barrier altogether).
What we want to know is:
- does auto/LLM eval work in general for a particular language
- which model(s) works best as a judge
- do LLMs' judgments of non-English models match human preferences?
We're starting to think about how to approach this. If you have any ideas of possible approaches feel free to comment or join the discussion here: https://github.com/huggingface/data-is-better-together/issues/61
Other ideas...
Could an approach like Replacing Judges with Juries: Evaluating LLM Generations with a Panel of Diverse Models (2404.18796) with the SOA models for a particular language work? i.e., choose 4 of the best open LLMs for Arabic and use those at the pool of raters rather than relying on one powerful judge LLM?
Reacted to
davanstrien's
post with 🤗
7 months ago
Post
1535
Can you create domain-specific synthetic datasets in under 20 minutes?
@burtenshaw recently launched the Domain Specific Dataset Project as part of Data is Better Together. As part of this, Ben created a Space that you can use to define some key perspectives and concepts from a domain. This seed dataset can then be used to generate a synthetic dataset for a particular domain.
In less than 30 minutes this afternoon, I created a domain-specific dataset focused on data-centric machine learning using these tools: davanstrien/data-centric-ml-sft.
You can create your own domain specific datasets using this approach. Find the steps to follow here: https://github.com/huggingface/data-is-better-together/blob/main/domain-specific-datasets/README.md
@burtenshaw recently launched the Domain Specific Dataset Project as part of Data is Better Together. As part of this, Ben created a Space that you can use to define some key perspectives and concepts from a domain. This seed dataset can then be used to generate a synthetic dataset for a particular domain.
In less than 30 minutes this afternoon, I created a domain-specific dataset focused on data-centric machine learning using these tools: davanstrien/data-centric-ml-sft.
You can create your own domain specific datasets using this approach. Find the steps to follow here: https://github.com/huggingface/data-is-better-together/blob/main/domain-specific-datasets/README.md
-
- 1 reply
Reacted to
tomaarsen's
post with 🔥
7 months ago
Post
3161
🚀 Sentence Transformers v2.7.0 is out! Featuring a new loss function, easier Matryoshka model inference & evaluation, CrossEncoder improvements & Intel Gaudi2 Accelerator support. Details:
1️⃣ A new loss function: CachedGISTEmbedLoss
This loss function is a combination of CachedMultipleNegativesRankingLoss and the GISTEmbedLoss, both of which are already excellent. The caching mechanism allows for much higher batch sizes with constant memory usage, which boosts training performance. The GIST part introduces a guide model to guide the in-batch negative sample selection. This prevents false negatives, resulting in a stronger training signal.
2️⃣ Automatic Matryoshka model truncation
Matryoshka models produce embeddings that are still useful after truncation. However, this truncation always had to be done manually, until now! We've added a
3️⃣ Additionally, you can now specify
4️⃣ CrossEncoder improvements
The CrossEncoder now supports 'push_to_hub' to upload trained reranker models to Hugging Face. Additionally, CrossEncoders now support
5️⃣ Inference on Intel Gaudi2
If you have an Intel Gaudi2 Accelerator, Sentence Transformers now uses it automatically for even faster inference. No changes are necessary to your code, the device is automatically detected!
Check out the release notes for all of the details: https://github.com/UKPLab/sentence-transformers/releases/tag/v2.7.0
I'm very excited for the upcoming releases: I'm making great progress with a notable v3 refactor that should heavily improve the training process for embedding models!
1️⃣ A new loss function: CachedGISTEmbedLoss
This loss function is a combination of CachedMultipleNegativesRankingLoss and the GISTEmbedLoss, both of which are already excellent. The caching mechanism allows for much higher batch sizes with constant memory usage, which boosts training performance. The GIST part introduces a guide model to guide the in-batch negative sample selection. This prevents false negatives, resulting in a stronger training signal.
2️⃣ Automatic Matryoshka model truncation
Matryoshka models produce embeddings that are still useful after truncation. However, this truncation always had to be done manually, until now! We've added a
truncate_dim option to the Sentence Transformer constructor. This also allows truncation when using HuggingFaceEmbeddings from LlamaIndex or LangChain.3️⃣ Additionally, you can now specify
truncate_dim in evaluators to get the performance after truncation. (Hint: it's surprisingly good, even for models not trained with MatryoshkaLoss, and it can speed up e.g. clustering, retrieval, etc.)4️⃣ CrossEncoder improvements
The CrossEncoder now supports 'push_to_hub' to upload trained reranker models to Hugging Face. Additionally, CrossEncoders now support
trust_remote_code to load models with custom modelling code.5️⃣ Inference on Intel Gaudi2
If you have an Intel Gaudi2 Accelerator, Sentence Transformers now uses it automatically for even faster inference. No changes are necessary to your code, the device is automatically detected!
Check out the release notes for all of the details: https://github.com/UKPLab/sentence-transformers/releases/tag/v2.7.0
I'm very excited for the upcoming releases: I'm making great progress with a notable v3 refactor that should heavily improve the training process for embedding models!
- 2 replies
Reacted to
sayakpaul's
post with 🤗🔥
7 months ago
Post
2890
We're introducing experimental support for
If you have multiple GPUs you want to use to distribute the pipeline models, you can do so. Additionally, this becomes more useful when you have multiple low-VRAM GPUs.
Documentation:
https://huggingface.co/docs/diffusers/main/en/training/distributed_inference#device-placement
🚨 Currently, only "balanced" device mapping strategy is supported.
device_map in Diffusers 🤗If you have multiple GPUs you want to use to distribute the pipeline models, you can do so. Additionally, this becomes more useful when you have multiple low-VRAM GPUs.
Documentation:
https://huggingface.co/docs/diffusers/main/en/training/distributed_inference#device-placement
🚨 Currently, only "balanced" device mapping strategy is supported.
posted
an
update
7 months ago
Post
2760
🦫 We have just released
argilla/Capybara-Preferences in collaboration with Kaist AI (
@JW17
,
@nlee-208
) and Hugging Face (
@lewtun
)
A new synthetic preference dataset built using
The current dataset combines the already generated alternative completions from argilla/distilabel-capybara-dpo-7k-binarized, while also adding the remaining ones using the same approach!
Here are some key features on how we built it:
- 🧹 Duplicate removal, keeping the conversation besides the last assistant response, and some slight pre-processing
- 🤖 Generation of alternative completions for the existing conversations (last turn only) with: mlabonne/NeuralBeagle14-7B, argilla/notus-7b-v1, and teknium/OpenHermes-2.5-Mistral-7B
- 👨🏻🏫 Running UltraFeedback via GPT-4 to generate the critique i.e. ratings and rationales, for the last assistant responses
- 🎉 Finally, we selected the chosen and rejected responses based on their UltraFeedback score, and applied some slight post-processing!
Sounds simple right? Start building your own synthetic datasets with https://github.com/argilla-io/distilabel already!
A new synthetic preference dataset built using
distilabel on top of the awesome
LDJnr/Capybara from
@LDJnr
The current dataset combines the already generated alternative completions from argilla/distilabel-capybara-dpo-7k-binarized, while also adding the remaining ones using the same approach!
Here are some key features on how we built it:
- 🧹 Duplicate removal, keeping the conversation besides the last assistant response, and some slight pre-processing
- 🤖 Generation of alternative completions for the existing conversations (last turn only) with: mlabonne/NeuralBeagle14-7B, argilla/notus-7b-v1, and teknium/OpenHermes-2.5-Mistral-7B
- 👨🏻🏫 Running UltraFeedback via GPT-4 to generate the critique i.e. ratings and rationales, for the last assistant responses
- 🎉 Finally, we selected the chosen and rejected responses based on their UltraFeedback score, and applied some slight post-processing!
Sounds simple right? Start building your own synthetic datasets with https://github.com/argilla-io/distilabel already!
Reacted to
davanstrien's
post with 🚀
8 months ago
Post
2032
Could more DPO-style preference data be crucial for enhancing open LLMs across different languages?
Leveraging a 7k preference dataset Argilla ( @alvarobartt ), Hugging Face ( @lewtun ) and Kaist AI ( @JW17 & @nlee-208 )
utilized Kaist AI's recently introduced ORPO technique ORPO: Monolithic Preference Optimization without Reference Model (2403.07691) with the latest MistralAI MOE model to create a very high-performing open LLM: HuggingFaceH4/zephyr-orpo-141b-A35b-v0.1
Since ORPO doesn't require a separate SFT stage, all that is needed is a strong base model + high-quality DPO-style datasets.
Currently, there is a significant lack of non-English DPO datasets. Filling this gap could significantly improve open LLMs in various languages.
You can get an overview of the current state of DPO datasets across different languages here: https://huggingface.co/spaces/DIBT/preference_data_by_language
Leveraging a 7k preference dataset Argilla ( @alvarobartt ), Hugging Face ( @lewtun ) and Kaist AI ( @JW17 & @nlee-208 )
utilized Kaist AI's recently introduced ORPO technique ORPO: Monolithic Preference Optimization without Reference Model (2403.07691) with the latest MistralAI MOE model to create a very high-performing open LLM: HuggingFaceH4/zephyr-orpo-141b-A35b-v0.1
Since ORPO doesn't require a separate SFT stage, all that is needed is a strong base model + high-quality DPO-style datasets.
Currently, there is a significant lack of non-English DPO datasets. Filling this gap could significantly improve open LLMs in various languages.
You can get an overview of the current state of DPO datasets across different languages here: https://huggingface.co/spaces/DIBT/preference_data_by_language
Reacted to
lewtun's
post with 🤗❤️
8 months ago
Post
4790
Introducing Zephyr 141B-A35B 🪁:
HuggingFaceH4/zephyr-orpo-141b-A35b-v0.1
Yesterday, Mistral released their latest base model (via magnet link of course 😅) and the community quickly converted it to transformers format and pushed it to the Hub: mistral-community/Mixtral-8x22B-v0.1
Early evals of this model looked extremely strong, so we teamed up with Argilla and KAIST AI to cook up a Zephyr recipe with a few new alignment techniques that came out recently:
🧑🍳 Align the base model with Odds Ratio Preference Optimisation (ORPO). This novel algorithm developed by @JW17 and @nlee-208 and @j6mes and does not require an SFT step to achieve high performance and is thus much more computationally efficient than methods like DPO and PPO.
🦫 Use a brand new dataset of 7k high-quality, multi-turn preferences that has been developed by our friends at Argilla. To create this dataset, they took the excellent Capybara SFT dataset from @LDJnr LDJnr/Capybara and converted it into a preference dataset by augmenting the final turn with responses from new LLMs that were then ranked by GPT-4.
What we find especially neat about this approach is that training on 7k samples only takes ~1.3h on 4 H100 nodes, yet produces a model that is very strong on chat benchmarks like IFEval and BBH.
Kudos to @alvarobartt @JW17 and @nlee-208 for this very nice and fast-paced collab!
For more details on the paper and dataset, checkout our collection: HuggingFaceH4/zephyr-orpo-6617eba2c5c0e2cc3c151524
HuggingFaceH4/zephyr-orpo-141b-A35b-v0.1
Yesterday, Mistral released their latest base model (via magnet link of course 😅) and the community quickly converted it to transformers format and pushed it to the Hub: mistral-community/Mixtral-8x22B-v0.1
Early evals of this model looked extremely strong, so we teamed up with Argilla and KAIST AI to cook up a Zephyr recipe with a few new alignment techniques that came out recently:
🧑🍳 Align the base model with Odds Ratio Preference Optimisation (ORPO). This novel algorithm developed by @JW17 and @nlee-208 and @j6mes and does not require an SFT step to achieve high performance and is thus much more computationally efficient than methods like DPO and PPO.
🦫 Use a brand new dataset of 7k high-quality, multi-turn preferences that has been developed by our friends at Argilla. To create this dataset, they took the excellent Capybara SFT dataset from @LDJnr LDJnr/Capybara and converted it into a preference dataset by augmenting the final turn with responses from new LLMs that were then ranked by GPT-4.
What we find especially neat about this approach is that training on 7k samples only takes ~1.3h on 4 H100 nodes, yet produces a model that is very strong on chat benchmarks like IFEval and BBH.
Kudos to @alvarobartt @JW17 and @nlee-208 for this very nice and fast-paced collab!
For more details on the paper and dataset, checkout our collection: HuggingFaceH4/zephyr-orpo-6617eba2c5c0e2cc3c151524