Hugging Face's logo

Collections

Discover the best community collections!

Collections trending this week
Qwen3.8-27B DSpark Drafters — FP8_BLOCK (dummy layer sweep)
Dummy FP8_BLOCK DSpark drafters (5/15/30 layers) cycled from the 5-layer base. For size/latency experiments only.
Don't Stop LCCP!
Model checkpoints trained with varying LCCP token budgets, paper: “Don't Stop Long-Context Continual Pre-Training Too Early: An Empirical Scaling Law”
NAMAA at KnowledgeGraphEval 2026 Shared Task
NAMAA at KnowledgeGraphEval 2026 (ArabicNLP 2026): AdaptNER 1st (75.96), RE 2nd (93.48).
FLUX.3 Image Klein for Apple Silicon · MLX
Local FLUX.3 image generation and editing on Apple Silicon. Five MLX variants, runtime, benchmarks and original image comparisons.
Qwen3.8-27B DSpark Drafters — BF16 (dummy layer sweep)
Dummy BF16 DSpark drafters (5/15/30 layers) cycled from the 5-layer base. For size/latency experiments only.
NAMAA at KnowledgeGraphEval 2026 Shared Task
NAMAA at KnowledgeGraphEval 2026 (ArabicNLP 2026): AdaptNER 1st (75.96), RE 2nd (93.48).
FLUX.3 Image Klein for Apple Silicon · MLX
Local FLUX.3 image generation and editing on Apple Silicon. Five MLX variants, runtime, benchmarks and original image comparisons.
Qwen3.8-27B DSpark Drafters — FP8_BLOCK (dummy layer sweep)
Dummy FP8_BLOCK DSpark drafters (5/15/30 layers) cycled from the 5-layer base. For size/latency experiments only.
Qwen3.8-27B DSpark Drafters — BF16 (dummy layer sweep)
Dummy BF16 DSpark drafters (5/15/30 layers) cycled from the 5-layer base. For size/latency experiments only.
Don't Stop LCCP!
Model checkpoints trained with varying LCCP token budgets, paper: “Don't Stop Long-Context Continual Pre-Training Too Early: An Empirical Scaling Law”