Oleg Silkin's contributions
Article
Unsloth and Training Hub: Lightning-fast LoRA and QLoRA fine-tuning
Learn how to fine-tune large language models in enterprise environments with Training Hub, an open source library for LLM post-training. Discover the benefits of LoRA and QLoRA using Unsloth, including reduced VRAM requirements and faster training times.
Article
Sculpting subspaces: How we solved continual learning in LLMs
Discover how the adaptive SVD approach enables LLMs to continually learn and adapt without forgetting previously acquired knowledge.
Article
Lessons on reproducing R1-like reasoning in small LLMs
Learn about an efficient inference scaling method that can improve your model's reasoning ability and performance at runtime while saving on compute costs.
Article
On reasoning versus inference-time scaling
Progress in small LLM reasoning: Our Qwen-32B model, using particle filtering, now surpasses o1-preview on Math500.
Article
Granite, LIMO, and small LLM reasoning
On reproducing R1-like reasoning in small LLMs: LIMO dataset ineffective for Llama/Granite; synthetic data generation shows promise but fine-tuning is tricky.
Article
How particle filtering makes small LLMs think big
An update on reproducing R1-like reasoning in small LLMs: Granite models show big gains with particle filtering, outperforming GPT-4o on benchmarks.
Article
GPU enablement on MicroShift
MicroShift is a low-footprint alternative to OpenShift. Learn how to enable it to take advantage of GPU computing power.
Unsloth and Training Hub: Lightning-fast LoRA and QLoRA fine-tuning
Learn how to fine-tune large language models in enterprise environments with Training Hub, an open source library for LLM post-training. Discover the benefits of LoRA and QLoRA using Unsloth, including reduced VRAM requirements and faster training times.
Sculpting subspaces: How we solved continual learning in LLMs
Discover how the adaptive SVD approach enables LLMs to continually learn and adapt without forgetting previously acquired knowledge.
Lessons on reproducing R1-like reasoning in small LLMs
Learn about an efficient inference scaling method that can improve your model's reasoning ability and performance at runtime while saving on compute costs.
On reasoning versus inference-time scaling
Progress in small LLM reasoning: Our Qwen-32B model, using particle filtering, now surpasses o1-preview on Math500.
Granite, LIMO, and small LLM reasoning
On reproducing R1-like reasoning in small LLMs: LIMO dataset ineffective for Llama/Granite; synthetic data generation shows promise but fine-tuning is tricky.
How particle filtering makes small LLMs think big
An update on reproducing R1-like reasoning in small LLMs: Granite models show big gains with particle filtering, outperforming GPT-4o on benchmarks.
GPU enablement on MicroShift
MicroShift is a low-footprint alternative to OpenShift. Learn how to enable it to take advantage of GPU computing power.
