Repository associated with paper titled "CLP: Finetuning Vision-Language-Action Models Requires Fewer Layers Than You Think"
-
Updated
Jul 6, 2026 - Jupyter Notebook
Repository associated with paper titled "CLP: Finetuning Vision-Language-Action Models Requires Fewer Layers Than You Think"
Caphlon — Lightest Qwen 3.8 27B GGUF model family (en hafif Qwen 3.8 27B modeli): Caphlon-2 (4.52 GiB PTQ1_SB via Ternary-Bonsai-2, 80.0% GPQA-Diamond, 87.4% Halu-270) & Caphlon-1 (6.87 GiB IQ2_S, 96.7% HumanEval). Runs on 6 GB VRAM / 8 GB RAM.
[NIPS 2016] Learning Structured Sparsity in Deep Neural Networks
Similarity-based block and sublayer dropping for ViT, DeiT, DINOv2, DINOv3 and SwinV2, evaluated on 13 datasets, with speed and energy benchmarks, statistical tests and automated model selection.
Official Pytorch implementation of "Filter Pruning by Image Channel Reduction in Pre-Trained Convolutional Neural Networks".
Intelligent layer pruning toolkit for LLMs featuring iterative optimization, self-healing algorithms, and comprehensive benchmarking.
Code for the papers: "Efficient Speech Translation through Model Compression and Knowledge Distillation" and "Iterative Layer Pruning for Efficient Translation Inference"
TALE: training-free task-aware layer pruning for improving LLM accuracy and reducing inference cost
Investigates whether transformer layer redundancy in small multilingual LLMs (Qwen, LLaMA, Gemma) is language-dependent. Tests if English-calibrated layer pruning silently harms Hindi and Marathi performance, and whether a mixed calibration set fixes it.
Based on yolov3 and channel and layer pruning
To associate your repository with the layer-pruning topic, visit your repo's landing page and select "manage topics."