Qwen/Qwen2.5-Coder-7B-Instruct
Text Generation
•
Updated
•
144k
•
373
Noob
noobmldude
AI & ML interests
Explainable AI
Recent Activity
liked
a model
about 1 month ago
Qwen/Qwen2.5-Coder-7B-Instruct
Organizations
None yet
noobmldude's activity
liked
a
model
about 1 month ago
liked
a
Space
2 months ago
upvoted
a
collection
3 months ago
upvoted
an
article
3 months ago
Article
FineVideo: behind the scenes
•
27
reacted to
merve's
post with 👍
3 months ago
Post
2377
NVIDIA just dropped NVEagle 🦅
Super impressive vision language model that comes in 7B, 13B and 13B fine-tuned on chat 💬
Model repositories: merve/nveagle-66d0705108582d73bb235c26
Try it: NVEagle/Eagle-X5-13B-Chat 💬 (works very well! 🤯)
This model essentially explores having different experts (MoE) for image encoder part of vision language model.
How? 🧐
The authors concatenate the vision encoder output tokens together, and they apply "pre-alignment" essentially fine-tune experts with frozen text encoder.
Then they freeze both experts and the decoder and just train the projection layer, and finally, they unfreeze everything for supervised fine-tuning ✨
In the paper, they explore different fusion strategies and vision encoders, extending basic CLIP encoder, and figure out simply concatenating visual tokens works well.
Rest of the architecture is quite similar to LLaVA. (see below the architecture)
Super impressive vision language model that comes in 7B, 13B and 13B fine-tuned on chat 💬
Model repositories: merve/nveagle-66d0705108582d73bb235c26
Try it: NVEagle/Eagle-X5-13B-Chat 💬 (works very well! 🤯)
This model essentially explores having different experts (MoE) for image encoder part of vision language model.
How? 🧐
The authors concatenate the vision encoder output tokens together, and they apply "pre-alignment" essentially fine-tune experts with frozen text encoder.
Then they freeze both experts and the decoder and just train the projection layer, and finally, they unfreeze everything for supervised fine-tuning ✨
In the paper, they explore different fusion strategies and vision encoders, extending basic CLIP encoder, and figure out simply concatenating visual tokens works well.
Rest of the architecture is quite similar to LLaVA. (see below the architecture)
reacted to
singhsidhukuldeep's
post with 👍
3 months ago
Post
2550
Good folks at Meta has just unveiled Llama 3.2, pushing the boundaries of language models and computer vision.
Even more interesting is how they trained this cutting-edge model:
1️⃣ Architecture:
Llama 3.2 uses an optimized transformer architecture with auto-regressive capabilities. The largest models (11B and 90B) now support multimodal inputs, integrating both text and images.
2️⃣ Training Pipeline:
• Started with pretrained Llama 3.1 text models
• Added image adapters and encoders
• Pretrained on large-scale noisy (image, text) pair data
• Fine-tuned on high-quality in-domain and knowledge-enhanced (image, text) pairs
3️⃣ Vision Integration:
• Trained adapter weights to integrate a pre-trained image encoder
• Used cross-attention layers to feed image representations into the language model
• Preserved text-only capabilities by not updating language model parameters during adapter training
4️⃣ Post-Training Alignment:
• Multiple rounds of supervised fine-tuning (SFT)
• Rejection sampling (RS)
• Direct preference optimization (DPO)
• Synthetic data generation using Llama 3.1 for Q&A augmentation
• Reward model ranking for high-quality fine-tuning data
5️⃣ Lightweight Models:
• Used pruning and distillation techniques for 1B and 3B models
• Structured pruning from Llama 3.1 8B model
• Knowledge distillation using Llama 3.1 8B and 70B as teachers
6️⃣ Context Length:
All models support an impressive 128K token context length.
7️⃣ Safety Measures:
Incorporated safety mitigation data to balance helpfulness and safety.
The result? A suite of models ranging from edge-friendly 1B parameters to powerful 90B parameter versions, capable of sophisticated reasoning across text and images. Llama 3.2 is set to revolutionize AI applications from mobile devices to enterprise-scale solutions.
What are your thoughts on these advancements? How do you see Llama 3.2 impacting your industry? Let's discuss in the comments!
Even more interesting is how they trained this cutting-edge model:
1️⃣ Architecture:
Llama 3.2 uses an optimized transformer architecture with auto-regressive capabilities. The largest models (11B and 90B) now support multimodal inputs, integrating both text and images.
2️⃣ Training Pipeline:
• Started with pretrained Llama 3.1 text models
• Added image adapters and encoders
• Pretrained on large-scale noisy (image, text) pair data
• Fine-tuned on high-quality in-domain and knowledge-enhanced (image, text) pairs
3️⃣ Vision Integration:
• Trained adapter weights to integrate a pre-trained image encoder
• Used cross-attention layers to feed image representations into the language model
• Preserved text-only capabilities by not updating language model parameters during adapter training
4️⃣ Post-Training Alignment:
• Multiple rounds of supervised fine-tuning (SFT)
• Rejection sampling (RS)
• Direct preference optimization (DPO)
• Synthetic data generation using Llama 3.1 for Q&A augmentation
• Reward model ranking for high-quality fine-tuning data
5️⃣ Lightweight Models:
• Used pruning and distillation techniques for 1B and 3B models
• Structured pruning from Llama 3.1 8B model
• Knowledge distillation using Llama 3.1 8B and 70B as teachers
6️⃣ Context Length:
All models support an impressive 128K token context length.
7️⃣ Safety Measures:
Incorporated safety mitigation data to balance helpfulness and safety.
The result? A suite of models ranging from edge-friendly 1B parameters to powerful 90B parameter versions, capable of sophisticated reasoning across text and images. Llama 3.2 is set to revolutionize AI applications from mobile devices to enterprise-scale solutions.
What are your thoughts on these advancements? How do you see Llama 3.2 impacting your industry? Let's discuss in the comments!
liked
a
model
3 months ago
upvoted
a
collection
4 months ago
upvoted
an
article
4 months ago
Article
XetHub is joining Hugging Face!
•
81
upvoted
an
article
5 months ago
Article
🪆 Introduction to Matryoshka Embedding Models
•
62
liked
a
Space
5 months ago
upvoted
a
paper
6 months ago
upvoted
an
article
6 months ago
Article
BigCodeBench: Benchmarking Large Language Models on Solving Practical and Challenging Programming Tasks
•
43
liked
a
model
8 months ago
reacted to
merve's
post with ❤️
8 months ago
Post
3842
just landed at Hugging Face Hub: community-led computer vision course 📖🤍
learn from fundamentals to details of the bleeding edge vision transformers!
learn from fundamentals to details of the bleeding edge vision transformers!
- 1 reply
upvoted
an
article
8 months ago
Article
How to Finetune phi-3 on MacBook Pro
By
•
•
64
liked
a
model
8 months ago
reacted to
fdaudens's
post with 🔥
8 months ago
Post
2411
5 interesting news stories today:
An AI startup made a hyperrealistic deepfake of me that’s so good it’s scary
- "'I think we might just have to say goodbye to finding out about the truth in a quick way,” says Sandra Wachter, a professor at the Oxford Internet Institute"
- "Synthesia uses both large language models and diffusion models to do this. Sees itself as a platform for businesses. Its bet is this: As people spend more time watching videos on YouTube and TikTok, there will be more demand for video content."
- "Synthesia’s policy is to not create avatars of people without their explicit consent. But it hasn’t been immune from abuse."
https://www.technologyreview.com/2024/04/25/1091772/new-generative-ai-avatar-deepfake-synthesia/
WIRED found thousands of ads running on Meta's social platforms promoting sexually explicit "Al girlfriend" apps.
- "Some human sex workers say the platform unfairly polices their own posts more harshly."
- "Many of the virtual women seen in ads reviewed by WIRED are lifelike—if somewhat uncanny—young, and stereotypically pornographic."
https://www.wired.com/story/ads-for-explicit-ai-girlfriends-swarming-facebook-and-instagram/
Wall Street’s Patience for a Costly A.I. Arms Race Is Waning
- "A sell-off in Meta’s stock after the company disclosed huge investments in the technology may be a sign of investor fears about tech giants’ spending."
- "The company plans to spend $35 billion to $40 billion this year — much of that on the technology."
https://www.nytimes.com/2024/04/25/business/dealbook/meta-artificial-intelligence-spending.html
Saudia Arabia Spends Big to Become an A.I. Superpower
https://www.nytimes.com/2024/04/25/technology/to-the-future-saudi-arabia-spends-big-to-become-an-ai-superpower.html
UK competition watchdog steps up scrutiny of big tech’s role in AI startups
https://www.theguardian.com/technology/2024/apr/24/uk-competition-watchdog-steps-up-scrutiny-of-big-techs-role-in-ai-startups-cma-microsoft-amazon
An AI startup made a hyperrealistic deepfake of me that’s so good it’s scary
- "'I think we might just have to say goodbye to finding out about the truth in a quick way,” says Sandra Wachter, a professor at the Oxford Internet Institute"
- "Synthesia uses both large language models and diffusion models to do this. Sees itself as a platform for businesses. Its bet is this: As people spend more time watching videos on YouTube and TikTok, there will be more demand for video content."
- "Synthesia’s policy is to not create avatars of people without their explicit consent. But it hasn’t been immune from abuse."
https://www.technologyreview.com/2024/04/25/1091772/new-generative-ai-avatar-deepfake-synthesia/
WIRED found thousands of ads running on Meta's social platforms promoting sexually explicit "Al girlfriend" apps.
- "Some human sex workers say the platform unfairly polices their own posts more harshly."
- "Many of the virtual women seen in ads reviewed by WIRED are lifelike—if somewhat uncanny—young, and stereotypically pornographic."
https://www.wired.com/story/ads-for-explicit-ai-girlfriends-swarming-facebook-and-instagram/
Wall Street’s Patience for a Costly A.I. Arms Race Is Waning
- "A sell-off in Meta’s stock after the company disclosed huge investments in the technology may be a sign of investor fears about tech giants’ spending."
- "The company plans to spend $35 billion to $40 billion this year — much of that on the technology."
https://www.nytimes.com/2024/04/25/business/dealbook/meta-artificial-intelligence-spending.html
Saudia Arabia Spends Big to Become an A.I. Superpower
https://www.nytimes.com/2024/04/25/technology/to-the-future-saudi-arabia-spends-big-to-become-an-ai-superpower.html
UK competition watchdog steps up scrutiny of big tech’s role in AI startups
https://www.theguardian.com/technology/2024/apr/24/uk-competition-watchdog-steps-up-scrutiny-of-big-techs-role-in-ai-startups-cma-microsoft-amazon
liked
a
model
8 months ago
updated
a
collection
10 months ago