This comment has been hidden
alkinun PRO
AtAndDev
AI & ML interests
LLMs, Alignment, Merging, Unsloth, DPO, SFT, ORPO, SPIN..
Recent Activity
upvoted
a
paper
about 21 hours ago
Every Sample Matters: Leveraging Mixture-of-Experts and High-Quality
Data for Efficient and Accurate Code LLM
liked
a dataset
2 days ago
inclusionAI/Ling-Coder-SFT
replied to
MonsterMMORPG's
post
2 days ago
SUPIR is Still Unchallanged Image Upscaler — Supports GPUs starting from RTX 1000 series to RTX 5000 series
App Download Link
You can download SUPIR app from here : https://www.patreon.com/posts/99176057
CHECK BELOW SCREENSHOTS
It has 1-click installers for Windows (only Python 3.10.11 and Git should be sufficient), RunPod (official Pytorch 2.2.0 template) and Massed Compute template Creator > SECourses
App Info
SUPIR: Scaling Up to Excellence: Practicing Model Scaling for Photo-Realistic Image Restoration In the Wild 1 click installer scripts.
SUPIR Sampler and Text CFG Comparison : https://imgsli.com/MjU2ODQz/2/1
Gemini 2.5 Pro prompt to get image description for free :
describe this image for sdxl. write single line prompt to regenerate it exactly same. make the prompt extremely detailed
https://aistudio.google.com/prompts/new_chat
Use Default preset for highest loyalty and Replicate preset for adding more details
Human upscale from 1024x1024 to 3072x3072 (3x upscale and total 9x resolution) with face restore comparison
https://imgsli.com/NDEzMDYx
Owl upscale from 1024x1024 to 3072x3072 (3x upscale and total 9x resolution)
https://imgsli.com/NDEzMDYy
Video Tutorials
Tutorials are older but hopefully a newer one will be made and they should be still useful
Complete Guide to SUPIR Enhancing and Upscaling Images Like in Sci-Fi Movies on Your PC
How To Install SUPIR On RunPod and Massed Compute
How To Install & Use SUPIR : SOTA Image Upscaler On RunPod — 1 Click Easy Install & Run
6 September 2025 Update V91
Libraries upgraded to Torch 2.8, CUDA 12.9, xFormers 0.0.33, Flash Attention 2.8.3
You don’t need to have CUDA or anything else installed and it should work with Python 3.10.11 and Git installed
When compiling libraries, I added support for all GPUs starting from RTX 1000 to 5000 series including other GPUs like A100, H100, B200, L40, etc
Compiled for TORCH_CUDA_ARCH_LIST=6.1;7.5;8.0;8.6;8.9;9.0;10.0;12.0
Organizations
replied to
MonsterMMORPG's
post
2 days ago
replied to
MonsterMMORPG's
post
3 days ago
This comment has been hidden
reacted to
merve's
post with 👍❤️
15 days ago
Post
5942
first vision language model built off
openai/gpt-oss-20b just dropped! 🔥
InternVL3.5 comes with 32 models 🤯 pre-trained, fine-tuned, aligned in various sizes OpenGVLab/internvl35-68ac87bd52ebe953485927fb
comes with gpt-oss or Qwen3 for LLM part ⤵️
InternVL3.5 comes with 32 models 🤯 pre-trained, fine-tuned, aligned in various sizes OpenGVLab/internvl35-68ac87bd52ebe953485927fb
comes with gpt-oss or Qwen3 for LLM part ⤵️
- 1 reply
Also, I do not think someone will achive AGI as we dont know what AGI is. I think we will just do incremental perf insreases, not an "unlock" that creates AGI.
In my pov, it should be open, if I can achieve AGI, someday someone will too. So theres no need to slow things down like eu. Just let things happen, accelerate and decentralize.
reacted to
appvoid's
post with 🔥
24 days ago
Post
3510
suppose someone is working on a reasoning model, which ends up unlocking achievements that lead to agi, should it be open source?
keep in mind everybody will have access to it: scientists, governments, terrorists, average people, etc...
keep in mind everybody will have access to it: scientists, governments, terrorists, average people, etc...
-
- 11 replies
reacted to
prithivMLmods's
post with 👀👍
29 days ago
Post
4391
On the verge of releasing Poseidon-Reasoning-5M, a dataset built to excel in general thought processes, mathematics, and science across a diverse mixture of domains, I’m also dropping the Gargantua-R1-Compact dataset, a collection of over six million high-quality reasoning QA pair traces. 🤗🚀
✦ Gargantua-R1-Compact : prithivMLmods/Gargantua-R1-Compact
Additionally, I’m adding the mini version of Gargantua — the Gargantua-R1-Wee : prithivMLmods/Gargantua-R1-Wee
The composition spans 73.93% core mathematical reasoning involving problems, proofs, and computational challenges, 12.11% across diverse scientific domains such as physics, chemistry, biology, and interdisciplinary topics, 11.35% in competitive coding covering algorithms and data structures, 1.37% in academic science focusing on research-level methodology, 0.95% in creative and analytical reasoning through logic puzzles and problem-solving tasks, 0.25% in specialized technical areas like MLOps, LLMs, diffusion models, and CUDA, and 0.06% involving data from graphs and charts converted into structured JSON formats. Designed with both rich contextual depth and formal structural clarity, Gargantua-R1-Compact is an optimal resource for advancing research in symbolic reasoning, interpretability, and high-precision question answering in mathematical domains.
✦ Collection : prithivMLmods/gargantua-r1-mod-6896bfd7834e82b89ad2b38b
To know more about it, visit the dataset card of the respective dataset. !!
✦ Gargantua-R1-Compact : prithivMLmods/Gargantua-R1-Compact
from datasets import load_dataset
dataset = load_dataset("prithivMLmods/Gargantua-R1-Compact", split="train")Additionally, I’m adding the mini version of Gargantua — the Gargantua-R1-Wee : prithivMLmods/Gargantua-R1-Wee
from datasets import load_dataset
dataset = load_dataset("prithivMLmods/Gargantua-R1-Wee", split="train")The composition spans 73.93% core mathematical reasoning involving problems, proofs, and computational challenges, 12.11% across diverse scientific domains such as physics, chemistry, biology, and interdisciplinary topics, 11.35% in competitive coding covering algorithms and data structures, 1.37% in academic science focusing on research-level methodology, 0.95% in creative and analytical reasoning through logic puzzles and problem-solving tasks, 0.25% in specialized technical areas like MLOps, LLMs, diffusion models, and CUDA, and 0.06% involving data from graphs and charts converted into structured JSON formats. Designed with both rich contextual depth and formal structural clarity, Gargantua-R1-Compact is an optimal resource for advancing research in symbolic reasoning, interpretability, and high-precision question answering in mathematical domains.
✦ Collection : prithivMLmods/gargantua-r1-mod-6896bfd7834e82b89ad2b38b
To know more about it, visit the dataset card of the respective dataset. !!
reacted to
fdaudens's
post with 👍🚀
30 days ago
Post
3376
OpenAI’s GPT-OSS has sparked ~400 new models on Hugging Face and racked up 5M downloads in less than a week, already outpacing DeepSeek R1’s first-week numbers.
For comparison: when R1 launched, I tracked 550 derivatives (across 8 base models) in a week, with ~3M downloads. GPT-OSS is ahead on adoption and engagement.
It’s also the most-liked release of any major LLM this summer. The 20B and 120B versions quickly shot past Kimi K2, GLM 4.5, and others in likes.
Most-downloaded GPT-OSS models include LM Studio and Unsloth AI versions:
1️⃣ openai/gpt-oss-20b - 2.0M
2️⃣ lmstudio-community/gpt-oss-20b-MLX-8bit - 750K
3️⃣ openai/gpt-oss-120b - 430K
4️⃣ unsloth/gpt-oss-20b-GGUF - 380K
5️⃣ lmstudio-community/gpt-oss-20b-GGUF - 330K
The 20B version is clearly finding its audience, showing the power of smaller, faster, more memory- and energy-efficient models. (These numbers don’t include calls to the models via inference providers, so the real usage is likely even bigger, especially for the 120B version)
Open-weight models let anyone build on top. Empower the builders, and innovation takes off. 🚀
For comparison: when R1 launched, I tracked 550 derivatives (across 8 base models) in a week, with ~3M downloads. GPT-OSS is ahead on adoption and engagement.
It’s also the most-liked release of any major LLM this summer. The 20B and 120B versions quickly shot past Kimi K2, GLM 4.5, and others in likes.
Most-downloaded GPT-OSS models include LM Studio and Unsloth AI versions:
1️⃣ openai/gpt-oss-20b - 2.0M
2️⃣ lmstudio-community/gpt-oss-20b-MLX-8bit - 750K
3️⃣ openai/gpt-oss-120b - 430K
4️⃣ unsloth/gpt-oss-20b-GGUF - 380K
5️⃣ lmstudio-community/gpt-oss-20b-GGUF - 330K
The 20B version is clearly finding its audience, showing the power of smaller, faster, more memory- and energy-efficient models. (These numbers don’t include calls to the models via inference providers, so the real usage is likely even bigger, especially for the 120B version)
Open-weight models let anyone build on top. Empower the builders, and innovation takes off. 🚀
- 1 reply
reacted to
ovi054's
post with 🔥
30 days ago
Post
3723
WAN 2.2 Text to Image ⚡
ovi054/wan2-2-text-to-image
We all know that WAN 2.2 A14B is a video model. But It turns out this video model can also produce great image results with incredible prompt adherence! The image output is sharp, detailed, and sticks to the prompt better than most.
👉 Try it now: ovi054/wan2-2-text-to-image
ovi054/wan2-2-text-to-image
We all know that WAN 2.2 A14B is a video model. But It turns out this video model can also produce great image results with incredible prompt adherence! The image output is sharp, detailed, and sticks to the prompt better than most.
👉 Try it now: ovi054/wan2-2-text-to-image
- 1 reply
sad
reacted to
FlameF0X's
post with 👀😔
about 1 month ago
Post
3849
I am very sad to say that the budget in creating of SnowflakeCore-G1 1b and 7b MoE models ran out and I can't pre-train them anymore.
-
- 4 replies
reacted to
sweatSmile's
post with ❤️🚀
about 1 month ago
Post
2787
Teaching a 7B Model to Be Just the Right Amount of Snark
Ever wondered if a language model could get sarcasm? I fine-tuned Mistral-7B using LoRA and 4-bit quantisation—on just ~720 hand-picked sarcastic prompt–response pairs from Reddit, Twitter, and real-life conversations.
The challenge? Keeping it sarcastic but still helpful.
LoRA rank 16 to avoid overfitting
4-bit NF4 quantization to fit on limited GPU memory
10 carefully monitored epochs so it didn’t turn into a full-time comedian
Result: a model that understands “Oh great, another meeting” exactly as you mean it.
Read the full journey, tech details, and lessons learned on my blog:
Fine-Tuning Mistral-7B for Sarcasm with LoRA and 4-Bit Quantisation
Try the model here on Hugging Face: sweatSmile/Mistral-7B-Instruct-v0.1-Sarcasm.
Ever wondered if a language model could get sarcasm? I fine-tuned Mistral-7B using LoRA and 4-bit quantisation—on just ~720 hand-picked sarcastic prompt–response pairs from Reddit, Twitter, and real-life conversations.
The challenge? Keeping it sarcastic but still helpful.
LoRA rank 16 to avoid overfitting
4-bit NF4 quantization to fit on limited GPU memory
10 carefully monitored epochs so it didn’t turn into a full-time comedian
Result: a model that understands “Oh great, another meeting” exactly as you mean it.
Read the full journey, tech details, and lessons learned on my blog:
Fine-Tuning Mistral-7B for Sarcasm with LoRA and 4-Bit Quantisation
Try the model here on Hugging Face: sweatSmile/Mistral-7B-Instruct-v0.1-Sarcasm.
Interesting! Did you have a chance to try k2, glm 4.5 or sonnet 4?
reacted to
dhruv3006's
post with 👀
about 1 month ago
Post
1977
GPT 5 for Computer Use agents.
Same tasks, same grounding model we just swapped GPT 4o with GPT 5 as the thinking model.
Left = 4o, right = 5.
Watch GPT 5 pull away.
Reasoning model: OpenAI GPT-5
Grounding model: Salesforce GTA1-7B
Action space: CUA Cloud Instances (macOS/Linux/Windows)
The task is: "Navigate to {random_url} and play the game until you reach a score of 5/5”....each task is set up by having claude generate a random app from a predefined list of prompts (multiple choice trivia, form filling, or color matching)"
Try it yourself here : https://github.com/trycua/cua
Docs : https://docs.trycua.com/docs/agent-sdk/supported-agents/composed-agents
Same tasks, same grounding model we just swapped GPT 4o with GPT 5 as the thinking model.
Left = 4o, right = 5.
Watch GPT 5 pull away.
Reasoning model: OpenAI GPT-5
Grounding model: Salesforce GTA1-7B
Action space: CUA Cloud Instances (macOS/Linux/Windows)
The task is: "Navigate to {random_url} and play the game until you reach a score of 5/5”....each task is set up by having claude generate a random app from a predefined list of prompts (multiple choice trivia, form filling, or color matching)"
Try it yourself here : https://github.com/trycua/cua
Docs : https://docs.trycua.com/docs/agent-sdk/supported-agents/composed-agents
-
- 2 replies