Hugging Face's logo
Stephen Genusa's picture

Stephen Genusa PRO

StephenGenusa
7 17 80
regisss's profile picture tahamajs's profile picture Wampanoag's profile picture
·

AI & ML interests

NLP, RAG/Hybrid/Graph, LLM, SLM, Optimized Quantization, Vision, Multimodality

Recent Activity

liked a model about 9 hours ago
endless-frontier/BigBang-v1
liked a model about 9 hours ago
meta-models/Muse-Glimmer-30B
reacted to Banaxi-Tech's post with 🔥 about 1 month ago
Today we are releasing BananaMind-KV1-8M-2Bit-Experimental, a KV-cache-aware trained model that stores its generation KV cache in 2-bit precision instead of the usual 16-bit precision. Result: 5.33x smaller KV cache vs FP16, with 0.0916 mean KLD against a 16-bit KV cache reference on WikiText-2. Model: https://huggingface.co/BananaMind/BananaMind-KV1-8M-2Bit-Experimental The important part: this is not just post-training KV cache quantization. Instead we take the BitNet approach. KV1 is trained with a 2-bit-aware K/V path. Instead of training a normal model and quantizing the cache afterwards, the model learns during training to operate under the low-bit KV constraint, closer in spirit to the BitNet idea of training for the low-bit regime. During generation, each K/V vector is quantized into 4 affine levels and packed into uint8 tensors, with four 2-bit values stored per byte. WikiText-2 eval vs 16-bit KV cache reference: Mean KLD: 0.0916 nats/token Mean KLD: 0.1322 bits/token Average KV cache shrink vs FP16: 5.33x Evaluated positions: 372,675 If this actually gets used in models like Qwen or Gemma, then it may be possible to run 128K or even 256K Context on a Normal Machine! Try it here: https://huggingface.co/BananaMind/BananaMind-KV1-8M-2Bit-Experimental Code: https://github.com/Banaxi-Tech/kv1
View all activity

Organizations

Social Post Explorers's profile picture