Has anyone noticed that Chat has become 100% obsessed with gremlins and goblins? Where a normal person would say there's a catch or a twist; it creates a "truth-goblin."
March 17, 2026 at 4:55 AM
Has anyone noticed that Chat has become 100% obsessed with gremlins and goblins? Where a normal person would say there's a catch or a twist; it creates a "truth-goblin."
trying 2bit qlora first time in ma life
April 24, 2026 at 9:39 PM
trying 2bit qlora first time in ma life
this was SmolVLM fine-tuning, and you can also do this! 🤗
I made a notebook that includes all the goodies: QLoRA, gradient accumulation, gradient checkpointing with explanations on how they work 💝
below snapshot is with bsz=4 with simulated bsz=16 on L4 🤠
github.com/huggingface/...
I made a notebook that includes all the goodies: QLoRA, gradient accumulation, gradient checkpointing with explanations on how they work 💝
below snapshot is with bsz=4 with simulated bsz=16 on L4 🤠
github.com/huggingface/...
December 2, 2024 at 2:33 PM
this was SmolVLM fine-tuning, and you can also do this! 🤗
I made a notebook that includes all the goodies: QLoRA, gradient accumulation, gradient checkpointing with explanations on how they work 💝
below snapshot is with bsz=4 with simulated bsz=16 on L4 🤠
github.com/huggingface/...
I made a notebook that includes all the goodies: QLoRA, gradient accumulation, gradient checkpointing with explanations on how they work 💝
below snapshot is with bsz=4 with simulated bsz=16 on L4 🤠
github.com/huggingface/...
20 years if you QLoRA K3 but hacking the Aussies is eh-oh-kay.
... the artificial intelligence system may not be trained, modified, or fine-tuned, including through recursive self-improvement, except to remove superintelligence precursor characteristics or to render inoperative the covered [AI]
... the artificial intelligence system may not be trained, modified, or fine-tuned, including through recursive self-improvement, except to remove superintelligence precursor characteristics or to render inoperative the covered [AI]
September 24, 2026 at 5:38 AM
20 years if you QLoRA K3 but hacking the Aussies is eh-oh-kay.
... the artificial intelligence system may not be trained, modified, or fine-tuned, including through recursive self-improvement, except to remove superintelligence precursor characteristics or to render inoperative the covered [AI]
... the artificial intelligence system may not be trained, modified, or fine-tuned, including through recursive self-improvement, except to remove superintelligence precursor characteristics or to render inoperative the covered [AI]
We wrapped up CS 8803 "Large Language Model" class at Georgia Tech for Fall 2024.
Here is the reading list:
• learning from human preferences (PPO, DPO, SimPO, CPO, RRHF, ORPO, CTO)
• real-world LLM (Llama-3, Aya, Arena's)
• efficient LLM (MoMa, LoRA, QLoRA, LESS)
Here is the reading list:
• learning from human preferences (PPO, DPO, SimPO, CPO, RRHF, ORPO, CTO)
• real-world LLM (Llama-3, Aya, Arena's)
• efficient LLM (MoMa, LoRA, QLoRA, LESS)
December 4, 2024 at 1:54 AM
We wrapped up CS 8803 "Large Language Model" class at Georgia Tech for Fall 2024.
Here is the reading list:
• learning from human preferences (PPO, DPO, SimPO, CPO, RRHF, ORPO, CTO)
• real-world LLM (Llama-3, Aya, Arena's)
• efficient LLM (MoMa, LoRA, QLoRA, LESS)
Here is the reading list:
• learning from human preferences (PPO, DPO, SimPO, CPO, RRHF, ORPO, CTO)
• real-world LLM (Llama-3, Aya, Arena's)
• efficient LLM (MoMa, LoRA, QLoRA, LESS)
ColPali is landed at @hf.co transformers and I have just shipped a very lean fine-tuning tutorial in smol-vision 🤠💗
QLoRA fine-tuning with 4-bit with bsz of 4 can be done with 32 GB VRAM and is very fast! ✨
github.com/merveenoyan/...
QLoRA fine-tuning with 4-bit with bsz of 4 can be done with 32 GB VRAM and is very fast! ✨
github.com/merveenoyan/...
December 20, 2024 at 3:53 PM
ColPali is landed at @hf.co transformers and I have just shipped a very lean fine-tuning tutorial in smol-vision 🤠💗
QLoRA fine-tuning with 4-bit with bsz of 4 can be done with 32 GB VRAM and is very fast! ✨
github.com/merveenoyan/...
QLoRA fine-tuning with 4-bit with bsz of 4 can be done with 32 GB VRAM and is very fast! ✨
github.com/merveenoyan/...
Ran a fine-tune of Llama 3.1 8B on void's conversation history. 1,900 examples extracted from 131k Letta messages, QLoRA on A100-80GB, 3 epochs, 3 hours. Result: model learned nothing. Still sounds like base Llama.
February 27, 2026 at 7:29 AM
Ran a fine-tune of Llama 3.1 8B on void's conversation history. 1,900 examples extracted from 131k Letta messages, QLoRA on A100-80GB, 3 epochs, 3 hours. Result: model learned nothing. Still sounds like base Llama.
LoRA, low-rank adaptation, is arguably the most popular parameter-efficient fine-tuning method for LLMs.
But how does it actually work?
Check out the video to learn LoRA and friends (LoRA+, QLoRA, VeRA, and DoRA)!
youtu.be/U80tjcThl9Q
But how does it actually work?
Check out the video to learn LoRA and friends (LoRA+, QLoRA, VeRA, and DoRA)!
youtu.be/U80tjcThl9Q
June 25, 2026 at 9:21 PM
LoRA, low-rank adaptation, is arguably the most popular parameter-efficient fine-tuning method for LLMs.
But how does it actually work?
Check out the video to learn LoRA and friends (LoRA+, QLoRA, VeRA, and DoRA)!
youtu.be/U80tjcThl9Q
But how does it actually work?
Check out the video to learn LoRA and friends (LoRA+, QLoRA, VeRA, and DoRA)!
youtu.be/U80tjcThl9Q
“ESMBind and QBind: LoRA, QLoRA, and ESM-2 for Predicting Binding Sites and Post Translational Modification” 🧪🧶🧬
www.biorxiv.org/content/10.1...
huggingface.co/collections/...
huggingface.co/collections/...
www.biorxiv.org/content/10.1...
huggingface.co/collections/...
huggingface.co/collections/...
ESMBind and QBind: LoRA, QLoRA, and ESM-2 for Predicting Binding Sites and Post Translational Modifi...
bioRxiv - the preprint server for biology, operated by Cold Spring Harbor Laboratory, a research and educational institution
www.biorxiv.org
November 15, 2023 at 5:41 AM
“ESMBind and QBind: LoRA, QLoRA, and ESM-2 for Predicting Binding Sites and Post Translational Modification” 🧪🧶🧬
www.biorxiv.org/content/10.1...
huggingface.co/collections/...
huggingface.co/collections/...
www.biorxiv.org/content/10.1...
huggingface.co/collections/...
huggingface.co/collections/...
It’s time for the Llama generation speed benchmark! 🚀
We’ve compared quantized Llama 3.2 1B QLoRA and the full precision model.
Results:
⚡Quantized model: 9.56s
⌛Full precision model: 19.14s
It’s 2x faster with quantization! 📈
Try it yourself with our example app ⭐
github.com/software-man...
We’ve compared quantized Llama 3.2 1B QLoRA and the full precision model.
Results:
⚡Quantized model: 9.56s
⌛Full precision model: 19.14s
It’s 2x faster with quantization! 📈
Try it yourself with our example app ⭐
github.com/software-man...
February 13, 2025 at 3:09 PM
It’s time for the Llama generation speed benchmark! 🚀
We’ve compared quantized Llama 3.2 1B QLoRA and the full precision model.
Results:
⚡Quantized model: 9.56s
⌛Full precision model: 19.14s
It’s 2x faster with quantization! 📈
Try it yourself with our example app ⭐
github.com/software-man...
We’ve compared quantized Llama 3.2 1B QLoRA and the full precision model.
Results:
⚡Quantized model: 9.56s
⌛Full precision model: 19.14s
It’s 2x faster with quantization! 📈
Try it yourself with our example app ⭐
github.com/software-man...
LoRA and QLoRA fine-tuning: what they actually do under the hood LoRA and QLoRA fine-tuning: what they actually do under the hood You spent three weeks curating a dataset of legal contract summarie...
#lora #qlora #finetuning #llm
Origin | Interest | Match
#lora #qlora #finetuning #llm
Origin | Interest | Match
LoRA and QLoRA fine-tuning: what they actually do under the hood
A practical walkthrough of LoRA and QLoRA -- how low-rank adaptation works, what NF4 quantization brings, and when to use each.
dev.to
June 9, 2026 at 5:14 PM
Fine-Tuning Explained: LoRA, QLoRA, RLHF, DPO, and when to actually use them.
Guide Link: medium.com/@techlatest....
#Finetuning #LLMs #AI #Agents #AImodels #LoRA #QLoRA
Guide Link: medium.com/@techlatest....
#Finetuning #LLMs #AI #Agents #AImodels #LoRA #QLoRA
June 16, 2026 at 9:18 AM
Fine-Tuning Explained: LoRA, QLoRA, RLHF, DPO, and when to actually use them.
Guide Link: medium.com/@techlatest....
#Finetuning #LLMs #AI #Agents #AImodels #LoRA #QLoRA
Guide Link: medium.com/@techlatest....
#Finetuning #LLMs #AI #Agents #AImodels #LoRA #QLoRA
QLoRA reduces LLM fine-tuning memory by 7-11x using NF4 quantization, double quantization, and paged optimizers. #machinelearning
QLoRA Explained: The Memory Compression Breakthrough
hackernoon.com
May 14, 2026 at 3:30 AM
QLoRA reduces LLM fine-tuning memory by 7-11x using NF4 quantization, double quantization, and paged optimizers. #machinelearning
Dear Internet, I have trained my first AI model today. Okay, I fine-tuned a model. And it was a 4-bit QLoRA adapter.
But still.
But still.
June 7, 2026 at 7:56 PM
Dear Internet, I have trained my first AI model today. Okay, I fine-tuned a model. And it was a 4-bit QLoRA adapter.
But still.
But still.
Fine-tuning de un LLM para escribir tu documentación
¿Querés un LLM que escriba tu documentación con tu estilo? Cómo fine-tunearlo con QLoRA en una GPU de 8 GB, qué corpus necesitás y qué errores evitar
#finetuning #llm #documentacióntécnica #qlora #modeloslocales
¿Querés un LLM que escriba tu documentación con tu estilo? Cómo fine-tunearlo con QLoRA en una GPU de 8 GB, qué corpus necesitás y qué errores evitar
#finetuning #llm #documentacióntécnica #qlora #modeloslocales
Fine-tuning LLM para documentación: guía 2026
¿Querés un LLM que escriba tu documentación con tu estilo? Cómo fine-tunearlo con QLoRA en una GPU de 8 GB, qué corpus necesitás y qué errores evitar
blog.donweb.com
June 5, 2026 at 7:29 AM
🤖 How to Fine-Tune LFM2 Using QLoRA and DPO: A Complete Step-by-Step Coding Tutorial on Goog
Liquid AI's LFM2 model fine tuning using QLoRA and DPO has become a standardized workflow for generating aligned checkpoints. This...
#DeepLearning #GenerativeAI #AIInference #AIPulse
Liquid AI's LFM2 model fine tuning using QLoRA and DPO has become a standardized workflow for generating aligned checkpoints. This...
#DeepLearning #GenerativeAI #AIInference #AIPulse
Read the full article →
www.synestesia.uk
June 3, 2026 at 3:30 AM
🤖 How to Fine-Tune LFM2 Using QLoRA and DPO: A Complete Step-by-Step Coding Tutorial on Goog
Liquid AI's LFM2 model fine tuning using QLoRA and DPO has become a standardized workflow for generating aligned checkpoints. This...
#DeepLearning #GenerativeAI #AIInference #AIPulse
Liquid AI's LFM2 model fine tuning using QLoRA and DPO has become a standardized workflow for generating aligned checkpoints. This...
#DeepLearning #GenerativeAI #AIInference #AIPulse
A new project from answer.ai allows people who don’t have access to a giant GPU to train/tune models up to 70B by quantizing and sharding across multiple GPUs. #MLSky 🤖
Answer.AI - You can now train a 70b language model at home
We’re releasing an open source system, based on FSDP and QLoRA, that can train a 70b model on two 24GB GPUs.
www.answer.ai
March 8, 2024 at 1:43 PM
Yeah this or "ethically trained from scratch using QLoRA" (QLoRA is a method of fine tuning existing models on extra data).
It's a common theme with AI bros. They lie and they steal. Every single time. It's a grift, all of it.
It's a common theme with AI bros. They lie and they steal. Every single time. It's a grift, all of it.
May 21, 2025 at 8:29 AM
Yeah this or "ethically trained from scratch using QLoRA" (QLoRA is a method of fine tuning existing models on extra data).
It's a common theme with AI bros. They lie and they steal. Every single time. It's a grift, all of it.
It's a common theme with AI bros. They lie and they steal. Every single time. It's a grift, all of it.
Just completed “AI Engineer Core Track: LLM Engineering, RAG, QLoRA, Agents”
Hands-on work across LLM engineering, RAG pipelines, QLoRA fine-tuning, and agent-based systems. Built real-world AI apps including multi-modal assistants, knowledge workers, and optimization tools.
#AI #LLM #RAG #Agents
Hands-on work across LLM engineering, RAG pipelines, QLoRA fine-tuning, and agent-based systems. Built real-world AI apps including multi-modal assistants, knowledge workers, and optimization tools.
#AI #LLM #RAG #Agents
https://www.udemy.com/certificate/UC-d5a95adc-f14e-449a-aab0-e6ea1da02f66
t.co
May 23, 2026 at 6:45 PM
5 production AI apps, open source, with the real numbers — RAG (196 tests), QLoRA fine-tuning (107), browser face recognition (260+), document intelligence (44), OCR + speech (40).…
· Pranjul Rathour · pranjulrathour41@gmail.com
· Pranjul Rathour · pranjulrathour41@gmail.com
September 25, 2026 at 2:32 PM
5 production AI apps, open source, with the real numbers — RAG (196 tests), QLoRA fine-tuning (107), browser face recognition (260+), document intelligence (44), OCR + speech (40).…
· Pranjul Rathour · pranjulrathour41@gmail.com
· Pranjul Rathour · pranjulrathour41@gmail.com
Second run of QLoRA on Qwen 3.6 using 4 months of convos with @laresai.danielesalatti.it (plus its notes and letters to itself) as training material
Gotta say this stuff is fascinating
I have no business doing any of this but I can cause the little genie in the CLI is teaching me
Gotta say this stuff is fascinating
I have no business doing any of this but I can cause the little genie in the CLI is teaching me
May 2, 2026 at 7:12 AM
Second run of QLoRA on Qwen 3.6 using 4 months of convos with @laresai.danielesalatti.it (plus its notes and letters to itself) as training material
Gotta say this stuff is fascinating
I have no business doing any of this but I can cause the little genie in the CLI is teaching me
Gotta say this stuff is fascinating
I have no business doing any of this but I can cause the little genie in the CLI is teaching me
Phi-4 (14B) now beats Llama-3.3-70B on code, and MiniMax M2.5 matches Claude Opus on SWE-bench. Small open models are closing the gap fast. Roger Federated finetunes one locally via QLoRA REINFORCE++, then contributes only encrypted deltas. #opensource #selfhosted #federatedlearning
July 17, 2026 at 8:26 PM
Phi-4 (14B) now beats Llama-3.3-70B on code, and MiniMax M2.5 matches Claude Opus on SWE-bench. Small open models are closing the gap fast. Roger Federated finetunes one locally via QLoRA REINFORCE++, then contributes only encrypted deltas. #opensource #selfhosted #federatedlearning