OpenMOSS-Team/Qwen2.5-7B-base-mla-topk8-rank256开源了新模型

OpenMOSS OpenMOSS-Team/Qwen2.5-7B-base-mla-topk8-rank256 OpenMOSS-Team/Qwen2.5-7B-base-mla-topk8-rank256 https://huggingface.co/OpenMOSS-Team/Qwen2.5-7B-base-mla-topk8-rank256

OpenMOSS-Team/Qwen2.5-7B-base-mla-topk8-rank256 · Hugging Face Hugging Face (https://huggingface.co/) Models (https://huggingface.co/models) Datasets (https://huggingface.co/datasets) Spaces (https://huggingface.co/spaces) Buckets new (https://huggingface.co/storage) Docs (https://huggingface.co/docs) Enterprise (https://huggingface.co/enterprise) Pricing (https://huggingface.co/pricing) Website Tasks (https://huggingface.co/tasks) HuggingChat (https://huggingface.co/chat) Collections (https://huggingface.co/collections) Languages (https://huggingface.co/languages) Organizations (https://huggingface.co/organizations) Community Blog (https://huggingface.co/blog) Posts (https://huggingface.co/posts) Daily Papers (https://huggingface.co/papers) Hardware (https://huggingface.co/hardware) Learn (https://huggingface.co/learn) Discord (https://huggingface.co/join/discord) Forum (https://discuss.huggingface.co/) GitHub (https://github.com/huggingface) Solutions Team & Enterprise (https://huggingface.co/enterprise) Hugging Face PRO (https://huggingface.co/pro) Enterprise Support (https://huggingface.co/support) Inference Providers (https://huggingface.co/inference/models) Inference Endpoints (https://huggingface.co/inference-endpoints) Storage Buckets (https://huggingface.co/storage) Log In (https://huggingface.co/login) Sign Up (https://huggingface.co/join) (https://huggingface.co/OpenMOSS-Team) OpenMOSS-Team (https://huggingface.co/OpenMOSS-Team) / Qwen2.5-7B-base-mla-topk8-rank256 (https://huggingface.co/OpenMOSS-Team/Qwen2.5-7B-base-mla-topk8-rank256) like 0 Follow OpenMOSS 1.18k Safetensors (https://huggingface.co/models?library=safetensors) qwen2 (https://huggingface.co/models?other=qwen2) qwen2.5 (https://huggingface.co/models?other=qwen2.5) mla (https://huggingface.co/models?other=mla) mlafication (https://huggingface.co/models?other=mlafication) delta-weights (https://huggingface.co/models?other=delta-weights) License: apache-2.0 Model card (https://huggingface.co/OpenMOSS-Team/Qwen2.5-7B-base-mla-topk8-rank256) Files Files and versions xet (https://huggingface.co/OpenMOSS-Team/Qwen2.5-7B-base-mla-topk8-rank256/tree/main) Community (https://huggingface.co/OpenMOSS-Team/Qwen2.5-7B-base-mla-topk8-rank256/discussions) Copy to bucket new Qwen2.5-7B-base-mla-topk8-rank256 (https://huggingface.co/OpenMOSS-Team/Qwen2.5-7B-base-mla-topk8-rank256#qwen25-7b-base-mla-topk8-rank256) Variant (https://huggingface.co/OpenMOSS-Team/Qwen2.5-7B-base-mla-topk8-rank256#variant) Loading (https://huggingface.co/OpenMOSS-Team/Qwen2.5-7B-base-mla-topk8-rank256#loading) Files (https://huggingface.co/OpenMOSS-Team/Qwen2.5-7B-base-mla-topk8-rank256#files) License (https://huggingface.co/OpenMOSS-Team/Qwen2.5-7B-base-mla-topk8-rank256#license) (https://huggingface.co/OpenMOSS-Team/Qwen2.5-7B-base-mla-topk8-rank256#qwen25-7b-base-mla-topk8-rank256) Qwen2.5-7B-base-mla-topk8-rank256 This repository contains the attention-only incremental weights for an MLAfication conversion of Qwen/Qwen2.5-7B (https://huggingface.co/Qwen/Qwen2.5-7B) . It is not a standalone full-model checkpoint. (https://huggingface.co/OpenMOSS-Team/Qwen2.5-7B-base-mla-topk8-rank256#variant) Variant Field Value Base model Qwen/Qwen2.5-7B Base revision d149729398750b98c0af14eb82c78cfe92750796 MLA latent rank 256 RoPE dimensions per KV head 8 Training checkpoint step 2160 Training variant stage1+2-distill-QKV Delta tensors 336 Delta size 1.41 GiB SHA-256 7e460722e713378421a2bb42686a13d4004b93e05ddd1582d25aa471d287e88b The training run froze every parameter outside attention and also froze each attention output projection. The uploaded delta therefore contains exactly the parameters matching "attn" in name and "o_proj" not in name; embeddings, MLP, normalization outside attention, output projections, and LM head are omitted. Frozen tensors from the historical full checkpoint were verified exactly against the pinned base weights after ca