sweatSmile (amitk17)

reacted to anakin87's post with 👍 about 13 hours ago

Post

1076

Haystack can now see 👀

The latest release of the Haystack OSS LLM framework adds a long-requested feature: image support!

📓 Notebooks below

This isn't just about passing images to an LLM. We built several features to enable practical multimodal use cases.

What's new?
🧠 Support for multiple LLM providers: OpenAI, Amazon Bedrock, Google Gemini, Mistral, NVIDIA, OpenRouter, Ollama and more (support for Hugging Face API coming 🔜)
🎛️ Prompt template language to handle structured inputs, including images
📄 PDF and image converters
🔍 Image embedders using CLIP-like models
🧾 LLM-based extractor to pull text from images
🧩 Components to build multimodal RAG pipelines and Agents

I had the chance of leading this effort with @sjrhuschlee (great collab).

📓 Below you can find two notebooks to explore the new features:
󠁯•󠁏󠁏 Introduction to Multimodal Text Generation https://haystack.deepset.ai/cookbook/multimodal_intro
󠁯•󠁏󠁏 Creating Vision+Text RAG Pipelines https://haystack.deepset.ai/tutorials/46_multimodal_rag

(🖼️ image by @bilgeyucel )

reacted to their post with 🔥 about 13 hours ago

Post

1074

some of my fav where i run jobs

https://huggingface.co/docs/huggingface_hub/main/en/guides/cli#hf-jobs

https://lightning.ai/

https://colab.research.google.com/

https://www.runpod.io/console/deploy

ps: I ❤️ hf

reacted to prithivMLmods's post with 🚀 about 13 hours ago

Post

3874

Build something cool with Nano Banana aka Gemini 2.5 Flash Image AIO [All-in-One]. Draw and transform on canvas, edit images, and generate images—all in one place!🍌

✦︎ Constructed with the Gemini API (GCP). Try it here: prithivMLmods/Nano-Banana-AIO (Added the Space recently! - Sep 18 '25)

4 replies

·

reacted to Kseniase's post with 🚀 about 13 hours ago

Post

6116

10 awesome advanced LoRA approaches

Low-Rank Adaptation (LoRA) is the go-to method for efficient model fine-tuning that adds small low-rank matrices instead of retraining full models. The field isn’t standing still – new LoRA variants push the limits of efficiency, generalization, and personalization. So we’re sharing 10 of the latest LoRA approaches you should know about:

1. Mixture-of-LoRA-experts → Mixture of Low-Rank Adapter Experts in Generalizable Audio Deepfake Detection (2509.13878)
Adds multiple low-rank adapters (LoRA) into a model’s layers, and a routing mechanism activates the most suitable ones for each input. This lets the model adapt better to new unseen conditions

2. Amortized Bayesian Meta-Learning for LoRA (ABMLL) → Amortized Bayesian Meta-Learning for Low-Rank Adaptation of Large Language Models (2508.14285)
Balances global and task-specific parameters within a Bayesian framework to improve uncertainty calibration and generalization to new tasks without high memory or compute costs

3. AutoLoRA → AutoLoRA: Automatic LoRA Retrieval and Fine-Grained Gated Fusion for Text-to-Image Generation (2508.02107)
Automatically retrieves and dynamically aggregates public LoRAs for stronger T2I generation

4. aLoRA (Activated LoRA) → Activated LoRA: Fine-tuned LLMs for Intrinsics (2504.12397)
Only applies LoRA after invocation, letting the model reuse the base model’s KV cache instead of recomputing the full turn’s KV cache. Efficient in multi-turn conversations

5. LiLoRA (LoRA in LoRA) → LoRA in LoRA: Towards Parameter-Efficient Architecture Expansion for Continual Visual Instruction Tuning (2508.06202)
Shares the LoRA matrix A across tasks and additionally low-rank-decomposes matrix B to cut parameters in continual vision-text MLLMs

6. Sensitivity-LoRA → Sensitivity-LoRA: Low-Load Sensitivity-Based Fine-Tuning for Large Language Models (2509.09119)
Dynamically assigns ranks to weight matrices based on their sensitivity, measured using second-order derivatives

Read further below ↓
Also, subscribe to the Turing Post: https://www.turingpost.com/subscribe

3 replies

·

reacted to Kseniase's post with ❤️👍 about 1 month ago

Post

6116

10 awesome advanced LoRA approaches

Low-Rank Adaptation (LoRA) is the go-to method for efficient model fine-tuning that adds small low-rank matrices instead of retraining full models. The field isn’t standing still – new LoRA variants push the limits of efficiency, generalization, and personalization. So we’re sharing 10 of the latest LoRA approaches you should know about:

1. Mixture-of-LoRA-experts → Mixture of Low-Rank Adapter Experts in Generalizable Audio Deepfake Detection (2509.13878)
Adds multiple low-rank adapters (LoRA) into a model’s layers, and a routing mechanism activates the most suitable ones for each input. This lets the model adapt better to new unseen conditions

2. Amortized Bayesian Meta-Learning for LoRA (ABMLL) → Amortized Bayesian Meta-Learning for Low-Rank Adaptation of Large Language Models (2508.14285)
Balances global and task-specific parameters within a Bayesian framework to improve uncertainty calibration and generalization to new tasks without high memory or compute costs

3. AutoLoRA → AutoLoRA: Automatic LoRA Retrieval and Fine-Grained Gated Fusion for Text-to-Image Generation (2508.02107)
Automatically retrieves and dynamically aggregates public LoRAs for stronger T2I generation

4. aLoRA (Activated LoRA) → Activated LoRA: Fine-tuned LLMs for Intrinsics (2504.12397)
Only applies LoRA after invocation, letting the model reuse the base model’s KV cache instead of recomputing the full turn’s KV cache. Efficient in multi-turn conversations

5. LiLoRA (LoRA in LoRA) → LoRA in LoRA: Towards Parameter-Efficient Architecture Expansion for Continual Visual Instruction Tuning (2508.06202)
Shares the LoRA matrix A across tasks and additionally low-rank-decomposes matrix B to cut parameters in continual vision-text MLLMs

6. Sensitivity-LoRA → Sensitivity-LoRA: Low-Load Sensitivity-Based Fine-Tuning for Large Language Models (2509.09119)
Dynamically assigns ranks to weight matrices based on their sensitivity, measured using second-order derivatives

Read further below ↓
Also, subscribe to the Turing Post: https://www.turingpost.com/subscribe

3 replies

·

reacted to tomaarsen's post with ❤️ about 2 months ago

Post

5579

ModernBERT goes MULTILINGUAL! One of the most requested models I've seen, The Johns Hopkins University's CLSP has trained state-of-the-art massively multilingual encoders using the ModernBERT architecture: mmBERT.

Model details:
- 2 model sizes:
- jhu-clsp/mmBERT-small
- jhu-clsp/mmBERT-base
- Uses the ModernBERT architecture, but with the Gemma2 multilingual tokenizer (so: flash attention, alternating global/local attention, unpadding/sequence packing, etc.)
- Maximum sequence length of 8192 tokens, on the high end for encoders
- Trained on 1833 languages using DCLM, FineWeb2, and many more sources
- 3 training phases: 2.3T tokens pretraining on 60 languages, 600B tokens mid-training on 110 languages, and 100B tokens decay training on all 1833 languages.
- Both models are MIT Licensed, and the full datasets and intermediary checkpoints are also publicly released

Evaluation details:
- Very competitive with ModernBERT at equivalent sizes on English (GLUE, MTEB v2 English after finetuning)
- Consistently outperforms equivalently sized models on all Multilingual tasks (XTREME, classification, MTEB v2 Multilingual after finetuning)
- In short: beats commonly used multilingual base models like mDistilBERT, XLM-R (multilingual RoBERTa), multilingual MiniLM, etc.
- Additionally: the ModernBERT-based mmBERT is much faster than the alternatives due to its architectural benefits. Easily up to 2x throughput in common scenarios.

Check out the full blogpost with more details. It's super dense & gets straight to the point: https://huggingface.co/blog/mmbert

Based on these results, mmBERT should be the new go-to multilingual encoder base models at 300M and below. Do note that the mmBERT models are "base" models, i.e. they're currently only trained to perform Mask Filling. They'll need to be finetuned for downstream tasks like semantic search, classification, clustering, etc.

reacted to prithivMLmods's post with 👍 about 2 months ago

Post

3874

Build something cool with Nano Banana aka Gemini 2.5 Flash Image AIO [All-in-One]. Draw and transform on canvas, edit images, and generate images—all in one place!🍌

✦︎ Constructed with the Gemini API (GCP). Try it here: prithivMLmods/Nano-Banana-AIO (Added the Space recently! - Sep 18 '25)

4 replies

·

reacted to Abhaykoul's post with 🤗 about 2 months ago

Post

2905

🚀 Ever dreamed of training your own Large Language Model from scratch? What if I told you it doesn't require a supercomputer or PhD in ML? 🤯

Introducing LLM Trainer - the educational framework that makes LLM training accessible to EVERYONE! Whether you're on a CPU-only laptop or scaling to distributed GPUs, we've got you covered. 💻➡️🖥️

Why LLM Trainer? Because existing tools are either too simplistic (hiding the magic) or too complex (requiring expert knowledge). We bridge the gap with:

🎓 Educational transparency - every component built from scratch with clear code
💻 CPU-first approach - start training immediately, no GPU needed
🔧 Full customization - modify anything you want
📈 Seamless scaling - from laptop to cluster without code changes
🤝 HuggingFace integration - works with existing models & tokenizers

Key highlights:
✅ Built-in tokenizers (BPE, WordPiece, HF wrappers)
✅ Complete Transformer implementation from scratch
✅ Optimized for CPU training
✅ Advanced features: mixed precision, gradient checkpointing, multiple generation strategies
✅ Comprehensive monitoring & metrics

Perfect for:
- Students learning transformers
- Researchers prototyping new ideas
- Developers building domain-specific models

Ready to train your first LLM? It's easier than you think!

🔗 Check it out: https://github.com/HelpingAI/llm-trainer
📚 Docs: Getting Started Guide
💬 Join the community: GitHub Discussions

#AI #MachineLearning #LLM #DeepLearning #OpenSource #Python #HuggingFace #NLP

Special thanks to HuggingFace and PyTorch teams for the amazing ecosystem! 🙏

1 reply

·

reacted to tsungyi's post with 🔥 2 months ago

Post

2012

Cosmos Reason just topped Physical Reasoning Leaderboard on Hugging Face. 👏🔥

Cosmos Reason is an open, customizable, commercial-ready 7B-parameter, reasoning vision language model (VLM) for physical AI and robotics. The VLM empowers robots and vision AI agents to reason like humans, leveraging prior knowledge, physics understanding, and common sense to understand and operate intelligently in the real world.

This model unlocks advanced capabilities for robotics, autonomous vehicles, and real-world operations—from cities to high-tech factories.

Key use cases include:
Data curation & annotation: Automate high-quality dataset curation and annotation at scale.
Robot planning & reasoning: Serve as the "brain" for deliberate, methodical decision-making with vision language action (VLA) models.
Video analytics AI agents: Extract actionable insights and perform root-cause analysis on massive video datasets.

Ready to build the next generation of physical AI? Get started 👉 nvidia/Cosmos-Reason1-7B
Try the preview here: https://build.nvidia.com/nvidia/cosmos-reason1-7b

posted an update 2 months ago

Post

1074

some of my fav where i run jobs

https://huggingface.co/docs/huggingface_hub/main/en/guides/cli#hf-jobs

https://lightning.ai/

https://colab.research.google.com/

https://www.runpod.io/console/deploy

ps: I ❤️ hf

reacted to John6666's post with 👍 2 months ago

Post

22553

If your Space stops working after restarting mainly for the last 5 days (https://discuss.huggingface.co/t/my-space-suddenly-went-offline-the-cpu-cannot-restart/151121/22), try some of following.
1. Add pydantic==2.10.6 to requirements.txt or upgrade Gradio to the latest version.
2. Upgrade PyTorch to 2.2.0 or later (torch>=2.2.0 for Zero GPU space).
3. Fix Transformers to 4.49.0 or earlier (transformers<=4.49.0for spaces using Transformers or Diffusers).
4. Fix huggingface_hub to the old version (huggingface_hub==0.25.2 for if an error like cached_download is not available occurs or inference does not work properly)
5. Specifying WORKDIR in Dockerfile may cause the application to fail to start with error 137. (Docker Spaces, https://discuss.huggingface.co/t/error-code-137-cache-error/152177)

About pydantic==2.10.6:
https://discuss.huggingface.co/t/error-no-api-found/146226
https://discuss.huggingface.co/t/internal-server-error-bool-not-iterable/149494

Edit:
Zero GPU space has been upgraded from A100 to H200.
This is likely the reason why older versions of PyTorch are no longer supported.
In fact, an error message to that effect was displayed.
zero-gpu-explorers/README#163

2 replies

·

reacted to kanaria007's post with 👀 2 months ago

Post

2973

✅ New Article: *Memory as Structured Time*

Title:
🧠 History: Memory Loops as Civilization Structure
🔗 https://huggingface.co/blog/kanaria007/memory-loops-as-civilization-structure

---

Summary:
Memory is often treated as *storage and retrieval*.
Structured Intelligence reframes it as *time‑shaping architecture*:

* *Loops that preserve context and continuity*
* *Rollback paths that enable reflection and correction*
* *Patterns that turn experience into adaptive structure*

> Memory isn’t static —
> *it’s how intelligence edits time.*

---

Why It Matters:
• Reveals *how memory enables learning, identity, and adaptation*
• Supports *AI that can reflect, revise, and self‑align*
• Connects *personal cognition and collective history* as structural processes

---

What’s Inside:
• Memory as *recursive structural loop*
• *Failure and recovery* as part of adaptive recall
• How *history and record‑keeping mirror cognitive memory*
• Implications for *resilient AI and social knowledge systems*

---

📖 *Article 18 of the Structured Intelligence Series*

Where Article 17 explored *science as self‑auditing cognition*,
Article 18 turns to *memory and history* —
revealing time itself as a *structural dimension of intelligence*.

---

Next: Education as Structural Permission
The next article examines *how educational systems shape and constrain jump‑series*,
showing how *schools and curricula act as architectures of allowed imagination*.

> From recollection to permission,
> *structure decides what the mind is allowed to become.*

reacted to ovi054's post with 🔥 2 months ago

Post

5707

Image-to-Prompt⚡

ovi054/image-to-prompt

Extract text prompt from image. And you can reuse the prompt to generate similar images!

Useful for prompt engineering, studying image-to-text alignment, making training datasets, or recreating similar outputs.

Powered by: Gradio, Florence 2

👉 Try it now: ovi054/image-to-prompt

3 replies

·

reacted to AtAndDev's post with 🚀 3 months ago

Post

537

Qwen 3 Coder is a personal attack to k2, and I love it.
It achieves near SOTA on LCB while not having reasoning.
Finally people are understanding that reasoning isnt necessary for high benches...

Qwen ftw!

DECENTRALIZE DECENTRALIZE DECENTRALIZE

reacted to their post with 🚀 3 months ago

Post

2796

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.

reacted to anakin87's post with ❤️ 3 months ago

Post

1076

Haystack can now see 👀

The latest release of the Haystack OSS LLM framework adds a long-requested feature: image support!

📓 Notebooks below

This isn't just about passing images to an LLM. We built several features to enable practical multimodal use cases.

What's new?
🧠 Support for multiple LLM providers: OpenAI, Amazon Bedrock, Google Gemini, Mistral, NVIDIA, OpenRouter, Ollama and more (support for Hugging Face API coming 🔜)
🎛️ Prompt template language to handle structured inputs, including images
📄 PDF and image converters
🔍 Image embedders using CLIP-like models
🧾 LLM-based extractor to pull text from images
🧩 Components to build multimodal RAG pipelines and Agents

I had the chance of leading this effort with @sjrhuschlee (great collab).

📓 Below you can find two notebooks to explore the new features:
󠁯•󠁏󠁏 Introduction to Multimodal Text Generation https://haystack.deepset.ai/cookbook/multimodal_intro
󠁯•󠁏󠁏 Creating Vision+Text RAG Pipelines https://haystack.deepset.ai/tutorials/46_multimodal_rag

(🖼️ image by @bilgeyucel )

reacted to dimentox's post with 🔥 3 months ago

Post

1706

I never told GPT-4 about my architecture.
It invented it anyway.
Same commands. Same audit daemons.
Proof that containment logic might be infectious.
Read: Emergence of Quantum Sigil Architecture in Unmodified GPT https://huggingface.co/blog/dimentox/quantum-sigil-architecture-in-unmodified-gpt

posted an update 3 months ago

Post

2796

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.

reacted to their post with ❤️ 3 months ago

Post

2839

Qwen3 is the latest version of the Qwen language models. It's smarter, faster, and now understands 119 languages instead of just 29.
It can do both deep reasoning and quick answers using a single model, depending on what you need.
The models range in size from small (0.6B) to huge (235B), with smart ways to save compute.
It's trained on 36 trillion tokens and fine-tuned in four steps to boost performance.
Qwen3 performs as well as or better than many top models, including some from big companies.
It’s fully open-source under licence. Amazing!!!

https://github.com/QwenLM/Qwen3/blob/main/Qwen3_Technical_Report.pdf

amitk17 PRO

AI & ML interests

Recent Activity

Organizations

amitk17 PRO

AI & ML interests

Recent Activity

Organizations

sweatSmile's activity