vllm
clip-retrieval
vllm | clip-retrieval | |
---|---|---|
31 | 11 | |
19,344 | 2,152 | |
12.6% | - | |
9.9 | 7.7 | |
1 day ago | 29 days ago | |
Python | Jupyter Notebook | |
Apache License 2.0 | MIT License |
Stars - the number of stars that a project has on GitHub. Growth - month over month growth in stars.
Activity is a relative number indicating how actively a project is being developed. Recent commits have higher weight than older ones.
For example, an activity of 9.0 indicates that a project is amongst the top 10% of the most actively developed projects that we are tracking.
vllm
-
AI leaderboards are no longer useful. It's time to switch to Pareto curves
I guess the root cause of my claim is that OpenAI won't tell us whether or not GPT-3.5 is an MoE model, and I assumed it wasn't. Since GPT-3.5 is clearly nondeterministic at temp=0, I believed the nondeterminism was due to FPU stuff, and this effect was amplified with GPT-4's MoE. But if GPT-3.5 is also MoE then that's just wrong.
What makes this especially tricky is that small models are truly 100% deterministic at temp=0 because the relative likelihoods are too coarse for FPU issues to be a factor. I had thought 3.5 was big enough that some of its token probabilities were too fine-grained for the FPU. But that's probably wrong.
On the other hand, it's not just GPT, there are currently floating-point difficulties in vllm which significantly affect the determinism of any model run on it: https://github.com/vllm-project/vllm/issues/966 Note that a suggested fix is upcasting to float32. So it's possible that GPT-3.5 is using an especially low-precision float and introducing nondeterminism by saving money on compute costs.
Sadly I do not have the money[1] to actually run a test to falsify any of this. It seems like this would be a good little research project.
[1] Or the time, or the motivation :) But this stuff is expensive.
-
Mistral AI Launches New 8x22B Moe Model
The easiest is to use vllm (https://github.com/vllm-project/vllm) to run it on a Couple of A100's, and you can benchmark this using this library (https://github.com/EleutherAI/lm-evaluation-harness)
- FLaNK AI for 11 March 2024
-
Show HN: We got fine-tuning Mistral-7B to not suck
Great question! scheduling workloads onto GPUs in a way where VRAM is being utilised efficiently was quite the challenge.
What we found was the IO latency for loading model weights into VRAM will kill responsiveness if you don't "re-use" sessions (i.e. where the model weights remain loaded and you run multiple inference sessions over the same loaded weights).
Obviously projects like https://github.com/vllm-project/vllm exist but we needed to build out a scheduler that can run a fleet of GPUs for a matrix of text/image vs inference/finetune sessions.
disclaimer: I work on Helix
-
Mistral CEO confirms 'leak' of new open source AI model nearing GPT4 performance
FYI, vLLM also just added experiment multi-lora support: https://github.com/vllm-project/vllm/releases/tag/v0.3.0
Also check out the new prefix caching, I see huge potential for batch processing purposes there!
- VLLM Sacrifices Accuracy for Speed
- Easy, fast, and cheap LLM serving for everyone
- vllm
- Mixtral Expert Parallelism
- Mixtral 8x7B Support
clip-retrieval
- FLaNK AI for 11 March 2024
-
[D] data for handwriting recognition
The tool clip-retreival lets you filter those 400 million images to whatever subsets you're interested in --- for example, 10,000 images of (mostly) handwriting.
- Stable Attribution
-
Same.energy: Image Search by Similarity
Hehe, well you know, PR welcome, the front end is 500 lines https://github.com/rom1504/clip-retrieval/blob/main/front/sr...
Other people have done a few alternate front ends already
This one is meant to be functional, but could sure be made prettier
-
Is there a way to use clip or blip to search a massive collection of images for specific things within the picture?
This might work: https://github.com/rom1504/clip-retrieval .
-
Ai art
HaveIBeenTrained uses clip retrieval to search the Laion-5B and Laion-400M image datasets. These are currently the largest public text-to-image datsets, and they are used to train models like Stable Diffusion, Imagen, among many others.
- Image Similarity Score using transfer learning
-
Exploring 12M of the 2.3B Images Used to Train Stable Diffusion
Done https://github.com/rom1504/clip-retrieval/commit/53e3383f58b...
Using clip for searching is better than direct text indexing for a variety of reasons but here for example because it matches better what stable diffusion sees
-
Semantic and Similarity Image Search Engine
Based on OpenAI's CLIP and the clip-retrieval library (https://github.com/rom1504/clip-retrieval), I've built an end-to-end demo for a semantic and similarity image search engine. It's incredibly powerful for finding similar images amongst large image datasets, or just submitting text/natural language queries and finding the most relevant images in your dataset. Really useful tool for introspection into large datasets before annotation or ML work begins. This could potentially be used to filter or downsize your datasets by several orders of magnitude and make annotation and ML work easier and less costly.
Checkout the demo here:
http://ec2-52-39-251-116.us-west-2.compute.amazonaws.com/
And you can checkout our website or email me for updates and email list, etc.:
https://machineperception.co
-
What every software engineer should know about search
Assuming you have an NVIDIA GPU, you can build a semantic search engine by indexing CLIP embeds (image or text).
https://github.com/rom1504/clip-retrieval
What are some alternatives?
TensorRT - NVIDIA® TensorRT™ is an SDK for high-performance deep learning inference on NVIDIA GPUs. This repository contains the open source components of TensorRT.
Typesense - Open Source alternative to Algolia + Pinecone and an Easier-to-Use alternative to ElasticSearch ⚡ 🔍 ✨ Fast, typo tolerant, in-memory fuzzy Search Engine for building delightful search experiences
CTranslate2 - Fast inference engine for Transformer models
MoTIS - [NAACL 2022]Mobile Text-to-Image search powered by multimodal semantic representation models(e.g., OpenAI's CLIP)
lmdeploy - LMDeploy is a toolkit for compressing, deploying, and serving LLMs.
laion-aesthetic-datasette - Use Datasette to explore LAION improved_aesthetics_6plus training data used by Stable DIffusion
Llama-2-Onnx
open_clip - An open source implementation of CLIP.
tritony - Tiny configuration for Triton Inference Server
clip-italian - CLIP (Contrastive Language–Image Pre-training) for Italian
faster-whisper - Faster Whisper transcription with CTranslate2
Queryable - Run OpenAI's CLIP model on iOS to search photos.