TheDrunkenSnail/Dirty-Muse-Writer-v01-Uncensored-Erotica-NSFW-Q6_K-GGUF
Updated
•
30
Danny
TheDrunkenSnail
AI & ML interests
None yet
Recent Activity
updated
a model
1 day ago
TheDrunkenSnail/Dirty-Muse-Writer-v01-Uncensored-Erotica-NSFW-Q6_K-GGUF
published
a model
1 day ago
TheDrunkenSnail/Dirty-Muse-Writer-v01-Uncensored-Erotica-NSFW-Q6_K-GGUF
updated
a model
1 day ago
TheDrunkenSnail/Dirty-Muse-Writer-v01-Uncensored-Erotica-NSFW-Q4_K_M-GGUF
Organizations
TheDrunkenSnail's activity
updated
a
model
1 day ago
published
a
model
1 day ago
updated
a
model
1 day ago
published
a
model
1 day ago
reacted to
eaddario's
post with 👍
15 days ago
Post
2727
Squeezing out tensor bits?
I have been tinkering with quantization and pruning to reduce model sizes. So far, I've had modest success in producing, on average, 8% smaller versions with negligible loss of quality, and I think further reductions in the 10-15% range are realistic, but I've come across a behaviour I wasn't expecting!
Part of the process I'm following consists of quantizing the embedding and output layers aggressively. Since the embedding layer is more about lookup than complex computation, the vectors representing the relative distances between embeddings are usually preserved well enough making this layer fairly robust to quantization. So far, so good.
The output layer, on the other hand, maps the final hidden state to the vocabulary logits and therefore, small changes in these logits could lead to a different probability distribution over the vocabulary, resulting in incorrect word predictions, or so I thought.
Surprisingly, I'm finding that even at Q2_K the loss of overall capability is minimal. Was this to be expected? or am I missing something?
I have published a version with all the test results if you want to give it a try: eaddario/DeepSeek-R1-Distill-Qwen-7B-GGUF
I'll upload other models as time allows.
Any ideas / clarifications / suggestions are very much welcomed!
I have been tinkering with quantization and pruning to reduce model sizes. So far, I've had modest success in producing, on average, 8% smaller versions with negligible loss of quality, and I think further reductions in the 10-15% range are realistic, but I've come across a behaviour I wasn't expecting!
Part of the process I'm following consists of quantizing the embedding and output layers aggressively. Since the embedding layer is more about lookup than complex computation, the vectors representing the relative distances between embeddings are usually preserved well enough making this layer fairly robust to quantization. So far, so good.
The output layer, on the other hand, maps the final hidden state to the vocabulary logits and therefore, small changes in these logits could lead to a different probability distribution over the vocabulary, resulting in incorrect word predictions, or so I thought.
Surprisingly, I'm finding that even at Q2_K the loss of overall capability is minimal. Was this to be expected? or am I missing something?
I have published a version with all the test results if you want to give it a try: eaddario/DeepSeek-R1-Distill-Qwen-7B-GGUF
I'll upload other models as time allows.
Any ideas / clarifications / suggestions are very much welcomed!
- 4 replies
liked
a
model
20 days ago
reacted to
kadirnar's
post with 👀
about 1 month ago
Post
3925
Researchers developed Sonic AI enabling precise facial animation from speech cues 🎧 Decouples head/expression control via audio tone analysis + time-aware fusion for natural long-form synthesis
- 1 reply
liked
a
model
about 1 month ago
reacted to
mkurman's
post with 👍
about 1 month ago
Post
2068
Blurred-Thoughts Supervised Fine-Tuning (BT-SFT) 🤖
Can we teach a model to think completely on its own without reinforcement learning? Actually, yes.
We can do straightforward supervised fine-tuning using a relatively simple trick: blurring a part of CoT thoughts. But why is this effective?
We observed that various models differ in their thinking processes, and fine-tuning one model on another model’s thoughts (CoT) can sometimes be inefficient—often resulting in the model simply memorizing reasoning rather than learning how to actually think.
I discovered that this process can still be efficient if we clearly indicate when the model should start and stop thinking and uncover only a part of CoT and the expected answer, blurring the other part of CoT. This approach allows the model to learn only a portion of the thought process while still arriving at an expected answer.
To demonstrate this, you can watch my experimental BT-SFT on meditsolutions/Llama-3.2-SUN-2.5B-chat model, which was fine-tuned on 151 million tokens from the Magpie-Align/Magpie-Reasoning-V2-250K-CoT-Deepseek-R1-Llama-70B dataset.
Enjoy! 🚀
PS. If you were curious enough to read this, leave me a comment. It's always nice to chat with open-minded and intelligent ppl.
Can we teach a model to think completely on its own without reinforcement learning? Actually, yes.
We can do straightforward supervised fine-tuning using a relatively simple trick: blurring a part of CoT thoughts. But why is this effective?
We observed that various models differ in their thinking processes, and fine-tuning one model on another model’s thoughts (CoT) can sometimes be inefficient—often resulting in the model simply memorizing reasoning rather than learning how to actually think.
I discovered that this process can still be efficient if we clearly indicate when the model should start and stop thinking and uncover only a part of CoT and the expected answer, blurring the other part of CoT. This approach allows the model to learn only a portion of the thought process while still arriving at an expected answer.
To demonstrate this, you can watch my experimental BT-SFT on meditsolutions/Llama-3.2-SUN-2.5B-chat model, which was fine-tuned on 151 million tokens from the Magpie-Align/Magpie-Reasoning-V2-250K-CoT-Deepseek-R1-Llama-70B dataset.
Enjoy! 🚀
PS. If you were curious enough to read this, leave me a comment. It's always nice to chat with open-minded and intelligent ppl.
- 3 replies
reacted to
Kseniase's
post with 🚀
about 2 months ago
Post
3106
7 Open-source Methods to Improve Video Generation and Understanding
AI community is making great strides toward achieving the full potential of multimodality in video generation and understanding. Last week studies showed that working with videos is now one of the main focuses for improving AI models. Another highlight of the week is that open source, once again, proves its value. For those who were impressed by DeepSeek-R1, we’re with you!
Today, we’re combining these two key focuses and bringing you a list of open-source methods for better video generation and understanding:
1. VideoLLaMA 3 model: Excels in various video and image tasks thanks to vision-centric training approach. VideoLLaMA 3: Frontier Multimodal Foundation Models for Image and Video Understanding (2501.13106)
2. FILMAGENT framework assigns roles to multiple AI agents, like a director, screenwriter, actor, and cinematographer, to automate the filmmaking process in 3D virtual environments. FilmAgent: A Multi-Agent Framework for End-to-End Film Automation in Virtual 3D Spaces (2501.12909)
3. Improving Video Generation with Human Feedback (2501.13918) proposes a new VideoReward Model and approach that uses human feedback to refine video generation models.
4. DiffuEraser video inpainting model, based on stable diffusion, is designed to fill in missing areas with detailed, realistic content and to ensure consistent structures across frames. DiffuEraser: A Diffusion Model for Video Inpainting (2501.10018)
5. MAGI is a hybrid video gen model that combines masked and casual modeling. Its key innovation, Complete Teacher Forcing (CTF), conditions masked frames on fully visible frames. Taming Teacher Forcing for Masked Autoregressive Video Generation (2501.12389)
6. Go-with-the-Flow: Motion-Controllable Video Diffusion Models Using Real-Time Warped Noise (2501.08331) proposes motion control, allowing users to guide how objects or the camera move in generated videos. Its noise warping algorithm replaces random noise in videos with structured noise based on motion info.
7. Video Depth Anything model estimates depth consistently in super-long videos (several minutes or more) without sacrificing quality or speed. Video Depth Anything: Consistent Depth Estimation for Super-Long Videos (2501.12375)
AI community is making great strides toward achieving the full potential of multimodality in video generation and understanding. Last week studies showed that working with videos is now one of the main focuses for improving AI models. Another highlight of the week is that open source, once again, proves its value. For those who were impressed by DeepSeek-R1, we’re with you!
Today, we’re combining these two key focuses and bringing you a list of open-source methods for better video generation and understanding:
1. VideoLLaMA 3 model: Excels in various video and image tasks thanks to vision-centric training approach. VideoLLaMA 3: Frontier Multimodal Foundation Models for Image and Video Understanding (2501.13106)
2. FILMAGENT framework assigns roles to multiple AI agents, like a director, screenwriter, actor, and cinematographer, to automate the filmmaking process in 3D virtual environments. FilmAgent: A Multi-Agent Framework for End-to-End Film Automation in Virtual 3D Spaces (2501.12909)
3. Improving Video Generation with Human Feedback (2501.13918) proposes a new VideoReward Model and approach that uses human feedback to refine video generation models.
4. DiffuEraser video inpainting model, based on stable diffusion, is designed to fill in missing areas with detailed, realistic content and to ensure consistent structures across frames. DiffuEraser: A Diffusion Model for Video Inpainting (2501.10018)
5. MAGI is a hybrid video gen model that combines masked and casual modeling. Its key innovation, Complete Teacher Forcing (CTF), conditions masked frames on fully visible frames. Taming Teacher Forcing for Masked Autoregressive Video Generation (2501.12389)
6. Go-with-the-Flow: Motion-Controllable Video Diffusion Models Using Real-Time Warped Noise (2501.08331) proposes motion control, allowing users to guide how objects or the camera move in generated videos. Its noise warping algorithm replaces random noise in videos with structured noise based on motion info.
7. Video Depth Anything model estimates depth consistently in super-long videos (several minutes or more) without sacrificing quality or speed. Video Depth Anything: Consistent Depth Estimation for Super-Long Videos (2501.12375)
- 1 reply
reacted to
KnutJaegersberg's
post with 👀
about 2 months ago
Post
1887
Evolution and The Knightian Blindspot of Machine Learning
The paper discusses machine learning's limitations in addressing Knightian Uncertainty (KU), highlighting the fragility of models like reinforcement learning (RL) in unpredictable, open-world environments. KU refers to uncertainty that can't be quantified or predicted, a challenge that RL fails to handle due to its reliance on fixed data distributions and limited formalisms.
### Key Approaches:
1. **Artificial Life (ALife):** Simulating diverse, evolving systems to generate adaptability, mimicking biological evolution's robustness to unpredictable environments.
2. **Open-Endedness:** Creating AI systems capable of continuous innovation and adaptation, drawing inspiration from human creativity and scientific discovery.
3. **Revising RL Formalisms:** Modifying reinforcement learning (RL) models to handle dynamic, open-world environments by integrating more flexible assumptions and evolutionary strategies.
These approaches aim to address ML’s limitations in real-world uncertainty and move toward more adaptive, general intelligence.
https://arxiv.org/abs/2501.13075
The paper discusses machine learning's limitations in addressing Knightian Uncertainty (KU), highlighting the fragility of models like reinforcement learning (RL) in unpredictable, open-world environments. KU refers to uncertainty that can't be quantified or predicted, a challenge that RL fails to handle due to its reliance on fixed data distributions and limited formalisms.
### Key Approaches:
1. **Artificial Life (ALife):** Simulating diverse, evolving systems to generate adaptability, mimicking biological evolution's robustness to unpredictable environments.
2. **Open-Endedness:** Creating AI systems capable of continuous innovation and adaptation, drawing inspiration from human creativity and scientific discovery.
3. **Revising RL Formalisms:** Modifying reinforcement learning (RL) models to handle dynamic, open-world environments by integrating more flexible assumptions and evolutionary strategies.
These approaches aim to address ML’s limitations in real-world uncertainty and move toward more adaptive, general intelligence.
https://arxiv.org/abs/2501.13075
reacted to
fantos's
post with 🔥
about 2 months ago
Post
4280
🚀 HuggingFace Spaces Ranking Tracker - Your Complete AI Trend Analytics!
Introducing the Spaces Ranking Tracker, a comprehensive analytics dashboard that tracks and analyzes every AI application in the HuggingFace ecosystem.
✨ Key Features:
• Real-time tracking of daily ranking changes over 30 days
• Detailed analysis of top 100 trending spaces
• User-based integrated score visualization
• One-click access to space details
• Interactive rank change graphs
📊 Dashboard Components:
1. Main Dashboard
- Daily rank trend graphs
- Top 20 creators' combined score chart
- Detailed space information cards
- Real-time trending score updates
2. Space Detailed Analysis
- Creation date, current rank, and trending score
- 30-day ranking history
- Direct space access
- Custom color coding for intuitive rank display
🎯 How to Use:
• Monitor latest AI community trends
• Track your project's performance
• Discover popular AI demos
• Analyze competing projects
• Follow AI ecosystem dynamics
3. Interactive Features
- Custom filtering options
- Sorting by various metrics
- Detailed performance statistics
- Comprehensive trending scores
- Historical data tracking
Stay on top of every movement in the HuggingFace ecosystem with daily ranking updates! 👉 Try it now!
🔗 Access Dashboard: fantos/Ranking-Tracker
#HuggingFace #AI #DataVisualization #TrendAnalysis #AITrends
Introducing the Spaces Ranking Tracker, a comprehensive analytics dashboard that tracks and analyzes every AI application in the HuggingFace ecosystem.
✨ Key Features:
• Real-time tracking of daily ranking changes over 30 days
• Detailed analysis of top 100 trending spaces
• User-based integrated score visualization
• One-click access to space details
• Interactive rank change graphs
📊 Dashboard Components:
1. Main Dashboard
- Daily rank trend graphs
- Top 20 creators' combined score chart
- Detailed space information cards
- Real-time trending score updates
2. Space Detailed Analysis
- Creation date, current rank, and trending score
- 30-day ranking history
- Direct space access
- Custom color coding for intuitive rank display
🎯 How to Use:
• Monitor latest AI community trends
• Track your project's performance
• Discover popular AI demos
• Analyze competing projects
• Follow AI ecosystem dynamics
3. Interactive Features
- Custom filtering options
- Sorting by various metrics
- Detailed performance statistics
- Comprehensive trending scores
- Historical data tracking
Stay on top of every movement in the HuggingFace ecosystem with daily ranking updates! 👉 Try it now!
🔗 Access Dashboard: fantos/Ranking-Tracker
#HuggingFace #AI #DataVisualization #TrendAnalysis #AITrends
-
- 1 reply
liked
a
dataset
about 2 months ago
reacted to
bartowski's
post with 👍
about 2 months ago
Post
66477
Looks like Q4_0_N_M file types are going away
Before you panic, there's a new "preferred" method which is online (I prefer the term on-the-fly) repacking, so if you download Q4_0 and your setup can benefit from repacking the weights into interleaved rows (what Q4_0_4_4 was doing), it will do that automatically and give you similar performance (minor losses I think due to using intrinsics instead of assembly, but intrinsics are more maintainable)
You can see the reference PR here:
https://github.com/ggerganov/llama.cpp/pull/10446
So if you update your llama.cpp past that point, you won't be able to run Q4_0_4_4 (unless they add backwards compatibility back), but Q4_0 should be the same speeds (though it may currently be bugged on some platforms)
As such, I'll stop making those newer model formats soon, probably end of this week unless something changes, but you should be safe to download and Q4_0 quants and use those !
Also IQ4_NL supports repacking though not in as many shapes yet, but should get a respectable speed up on ARM chips, PR for that can be found here: https://github.com/ggerganov/llama.cpp/pull/10541
Remember, these are not meant for Apple silicon since those use the GPU and don't benefit from the repacking of weights
Before you panic, there's a new "preferred" method which is online (I prefer the term on-the-fly) repacking, so if you download Q4_0 and your setup can benefit from repacking the weights into interleaved rows (what Q4_0_4_4 was doing), it will do that automatically and give you similar performance (minor losses I think due to using intrinsics instead of assembly, but intrinsics are more maintainable)
You can see the reference PR here:
https://github.com/ggerganov/llama.cpp/pull/10446
So if you update your llama.cpp past that point, you won't be able to run Q4_0_4_4 (unless they add backwards compatibility back), but Q4_0 should be the same speeds (though it may currently be bugged on some platforms)
As such, I'll stop making those newer model formats soon, probably end of this week unless something changes, but you should be safe to download and Q4_0 quants and use those !
Also IQ4_NL supports repacking though not in as many shapes yet, but should get a respectable speed up on ARM chips, PR for that can be found here: https://github.com/ggerganov/llama.cpp/pull/10541
Remember, these are not meant for Apple silicon since those use the GPU and don't benefit from the repacking of weights
- 16 replies
reacted to
burtenshaw's
post with 🚀
about 2 months ago
Post
3344
Manic few days in open source AI, with game changing development all over the place. Here's a round up of the resources:
- The science team at @huggingface reproduced and open source the seek r1. https://github.com/huggingface/open-r1
- @qwen released a series of models with 1 million token context! https://qwenlm.github.io/blog/qwen2.5-1m/
- SmolVLM got even smaller with completely new variants at 256m and 500m https://huggingface.co/blog/smolervlm
There's so much you could do with these developments. Especially combining them together into agentic applications or fine-tuning them on your use case.
- The science team at @huggingface reproduced and open source the seek r1. https://github.com/huggingface/open-r1
- @qwen released a series of models with 1 million token context! https://qwenlm.github.io/blog/qwen2.5-1m/
- SmolVLM got even smaller with completely new variants at 256m and 500m https://huggingface.co/blog/smolervlm
There's so much you could do with these developments. Especially combining them together into agentic applications or fine-tuning them on your use case.
-
- 1 reply
updated
a
collection
about 2 months ago
reacted to
haritzpuerto's
post with 👍
about 2 months ago
Post
1470
I'm excited to announce that my internship paper at Parameter Lab was accepted to Findings of #NAACL2025 🎉
TLDR: Stating an LLM was trained on a sentence might not be possible 😥 , but it is possible for large enough amounts of tokens, such as long documents or multiple documents! 🤯
Scaling Up Membership Inference: When and How Attacks Succeed on Large Language Models (2411.00154)
🔗 https://github.com/parameterlab/mia-scaling
TLDR: Stating an LLM was trained on a sentence might not be possible 😥 , but it is possible for large enough amounts of tokens, such as long documents or multiple documents! 🤯
Scaling Up Membership Inference: When and How Attacks Succeed on Large Language Models (2411.00154)
🔗 https://github.com/parameterlab/mia-scaling
updated
a
model
about 2 months ago
updated
a
collection
about 2 months ago