Hugging Face's logo

Code Llama

company
Activity Feed

AI & ML interests

None defined yet.

Recent Activity

lvwerraย  authored a paper about 2 months ago
SelfCodeAlign: Self-Alignment for Code Generation
lvwerraย  authored a paper 6 months ago
The FineWeb Datasets: Decanting the Web for the Finest Text Data at Scale
loubnabnlย  authored a paper 6 months ago
The FineWeb Datasets: Decanting the Web for the Finest Text Data at Scale
View all activity

codellama's activity

Narsilย 
posted an update 14 days ago
view post
Post
946
Performance leap: TGI v3 is out. Processes 3x more tokens, 13x faster than vLLM on long prompts. Zero config !



3x more tokens.

By reducing our memory footprint, weโ€™re able to ingest many more tokens and more dynamically than before. A single L4 (24GB) can handle 30k tokens on llama 3.1-8B, while vLLM gets barely 10k. A lot of work went into reducing the footprint of the runtime and its effect are best seen on smaller constrained environments.
13x faster

On long prompts (200k+ tokens) conversation replies take 27.5s in vLLM, while it takes only 2s in TGI. How so ? We keep the initial conversation around, so when a new reply comes in, we can answer almost instantly. The overhead of the lookup is ~5us. Thanks @Dani รซl de Kok for the beast data structure.
Zero config

Thatโ€™s it. Remove all the flags your are using and youโ€™re likely to get the best performance. By evaluating the hardware and model, TGI carefully selects automatic values to give best performance. In production, we donโ€™t have any flags anymore in our deployments. We kept all existing flags around, they may come in handy in niche scenarios.

Read more: https://huggingface.co/docs/text-generation-inference/conceptual/chunking
loubnabnlย 
posted an update about 1 month ago
view post
Post
1645
Making SmolLM2 reproducible: open-sourcing our training & evaluation toolkit ๐Ÿ› ๏ธ https://github.com/huggingface/smollm/

- Pre-training code with nanotron
- Evaluation suite with lighteval
- Synthetic data generation using distilabel (powers our new SFT dataset HuggingFaceTB/smoltalk)
- Post-training scripts with TRL & the alignment handbook
- On-device tools with llama.cpp for summarization, rewriting & agents

Apache 2.0 licensed. V2 pre-training data mix coming soon!

Which other tools should we add next?
andrewrreedย 
posted an update about 1 month ago
view post
Post
950
Trace LLM calls with Arize AI's Phoenix observability dashboards on Hugging Face Spaces! ๐Ÿš€

โœจ I just added a new recipe to the Open-Source AI Cookbook that shows you how to:
1๏ธโƒฃ Deploy Phoenix on HF Spaces with persistent storage in a few clicks
2๏ธโƒฃ Configure LLM tracing with the ๐—ฆ๐—ฒ๐—ฟ๐˜ƒ๐—ฒ๐—ฟ๐—น๐—ฒ๐˜€๐˜€ ๐—œ๐—ป๐—ณ๐—ฒ๐—ฟ๐—ฒ๐—ป๐—ฐ๐—ฒ ๐—”๐—ฃ๐—œ
3๏ธโƒฃ Observe multi-agent application runs with the CrewAI integration

๐—ข๐—ฏ๐˜€๐—ฒ๐—ฟ๐˜ƒ๐—ฎ๐—ฏ๐—ถ๐—น๐—ถ๐˜๐˜† ๐—ถ๐˜€ ๐—ฐ๐—ฟ๐˜‚๐—ฐ๐—ถ๐—ฎ๐—น for building robust LLM apps.

Phoenix makes it easy to visualize trace data, evaluate performance, and track down issues. Give it a try!

๐Ÿ”— Cookbook recipe: https://huggingface.co/learn/cookbook/en/phoenix_observability_on_hf_spaces
๐Ÿ”— Phoenix docs: https://docs.arize.com/phoenix
ArthurZย 
posted an update about 1 month ago
view post
Post
2678
Native tensor parallel has landed in transformers!!! https://github.com/huggingface/transformers/pull/34184 thanks a lot to the torch team for their support!

Contributions are welcome to support more models! ๐Ÿ”ฅ
lvwerraย 
authored a paper about 2 months ago
lvwerraย 
authored a paper 6 months ago
loubnabnlย 
authored a paper 6 months ago
loubnabnlย 
posted an update 7 months ago
view post
Post
5340
๐Ÿท FineWeb technical report is out and so is ๐Ÿ“š FineWeb-Edu, a 1.3 trillion tokens dataset that outperforms all other open web datasets, with remarkable improvements on educational benchmarksย such as MMLU, ARC, and OpenBookQA.

Technical report: HuggingFaceFW/blogpost-fineweb-v1
Dataset: HuggingFaceFW/fineweb-edu

We used Llama 3 generations to train an educational quality classifier, filtering the 15 trillion tokens of FineWeb to select only those with high educational value (an approach also used in Llama 3 and Phi-3 training datasets). We're releasing both FineWeb-Edu and the classifier, along with a larger, less heavily filtered version containing 5.4 trillion tokens.

You can find more details about the dataset and the experiments we ran in the FineWeb technical report, It's a 45-minute read but it contains all the secret sauce for building high quality web datasets.

Enjoy!
lvwerraย 
authored a paper 7 months ago
loubnabnlย 
authored a paper 7 months ago
Narsilย 
posted an update 7 months ago
andrewrreedย 
posted an update 8 months ago
view post
Post
2546
๐Ÿ”ฌ Open LLM Progress Tracker ๐Ÿ”ฌ

Inspired by the awesome work from @mlabonne , I created a Space to monitor the narrowing gap between open and proprietary LLMs as scored by the LMSYS Chatbot Arena ELO ratings ๐Ÿค—

The goal is to have a continuously updated place to easily visualize these rapidly evolving industry trends ๐Ÿš€

๐Ÿ”— Open LLM Progress Tracker: andrewrreed/closed-vs-open-arena-elo
๐Ÿ”— Source of Inspiration: https://www.linkedin.com/posts/maxime-labonne_arena-elo-graph-updated-with-new-models-activity-7187062633735368705-u2jB/
  • 2 replies
ยท
Narsilย 
posted an update 8 months ago
pcuenqย 
posted an update 8 months ago
view post
Post
4659
OpenELM in Core ML

Apple recently released a set of efficient LLMs in sizes varying between 270M and 3B parameters. Their quality, according to benchmarks, is similar to OLMo models of comparable size, but they required half the pre-training tokens because they use layer-wise scaling, where the number of attention heads increases in deeper layers.

I converted these models to Core ML, for use on Apple Silicon, using this script: https://gist.github.com/pcuenca/23cd08443460bc90854e2a6f0f575084. The converted models were uploaded to this community in the Hub for anyone that wants to integrate inside their apps: corenet-community/openelm-core-ml-6630c6b19268a5d878cfd194

The conversion was done with the following parameters:
- Precision: float32.
- Sequence length: fixed to 128.

With swift-transformers (https://github.com/huggingface/swift-transformers), I'm getting about 56 tok/s with the 270M on my M1 Max, and 6.5 with the largest 3B model. These speeds could be improved by converting to float16. However, there's some precision loss somewhere and generation doesn't work in float16 mode yet. I'm looking into this and will keep you posted! Or take a look at this issue if you'd like to help: https://github.com/huggingface/swift-transformers/issues/95

I'm also looking at optimizing inference using an experimental kv cache in swift-transformers. It's a bit tricky because the layers have varying number of attention heads, but I'm curious to see how much this feature can accelerate performance in this model family :)

Regarding the instruct fine-tuned models, I don't know the chat template that was used. The models use the Llama 2 tokenizer, but the Llama 2 chat template, or the default Alignment Handbook one that was used to train, are not recognized. Any ideas on this welcome!
ยท
andrewrreedย 
posted an update 8 months ago
view post
Post
2318
IMO, the "grounded generation" feature from Cohere's CommandR+ has flown under the radar...

For RAG use cases, responses directly include inline citations, making source attribution an inherent part of generation rather than an afterthought ๐Ÿ˜Ž

Who's working on an open dataset with this for the HF community to fine-tune with??

๐Ÿ”—CommandR+ Docs: https://docs.cohere.com/docs/retrieval-augmented-generation-rag

๐Ÿ”—Model on the ๐Ÿค— Hub: CohereForAI/c4ai-command-r-plus
  • 1 reply
ยท
ArthurZย 
authored a paper 9 months ago
philschmidย 
posted an update 9 months ago
view post
Post
6900
New state-of-the-art open LLM! ๐Ÿš€ Databricks just released DBRX, a 132B MoE trained on 12T tokens. Claiming to surpass OpenAI GPT-3.5 and is competitive with Google Gemini 1.0 Pro. ๐Ÿคฏ

TL;DR
๐Ÿงฎ 132B MoE with 16 experts with 4 active in generation
๐ŸชŸ 32 000 context window
๐Ÿ“ˆ Outperforms open LLMs on common benchmarks, including MMLU
๐Ÿš€ Up to 2x faster inference than Llama 2 70B
๐Ÿ’ป Trained on 12T tokens
๐Ÿ”ก Uses the GPT-4 tokenizer
๐Ÿ“œ Custom License, commercially useable

Collection: databricks/dbrx-6601c0852a0cdd3c59f71962
Demo: databricks/dbrx-instruct

Kudos to the Team at Databricks and MosaicML for this strong release in the open community! ๐Ÿค—
ยท
loubnabnlย 
posted an update 9 months ago
view post
Post
6401
We've just published a detailed blog post on the creation of Cosmopedia dataset. We hope this will provide insights about generating synthetic data at scale for pre-training.
https://huggingface.co/blog/cosmopedia

Here are some key takeaways:
๐ŸŽฏ Prompt curation is crucial: we want to cover many topics with few duplicates.
๐Ÿ“š You can leverage various resources for diversity: using different seed data, generation formats, and target audiences.
โš™๏ธ The importance of a good technical stack: for scalable generations with tools like llm-swarm and fast model training and evaluation.

Have a good read!
  • 1 reply
ยท
ArthurZย 
posted an update 10 months ago
Narsilย 
authored a paper 10 months ago