Commits
User selector
Commit history
Commits on Jul 11, 2026
test(gguf): expand FuzzParse with F1/F2/F3 seeds and LoadTensors coverage (S139.3.1)
Show description for 97b7ab4committedfix(registry): reject non-https registry URLs unless explicitly opted in (OCI-2, T139.5)
Show description for 3177eaccommittedfix(registry): support out-of-band expected hash for HF pulls (HF-1, T143.10)
Show description for 7557dfbcommitted
Commits on Jul 10, 2026
fix(gguf): cap tensor dimension count at 8 (F3, T139.3)
Show description for 7ad5cc0committedfix(registry): verify blob digest on OCI pull (OCI-1, T139.4)
Show description for 026ca1fcommitted
Commits on Jul 4, 2026
fix(gguf): restore missing math import in loader_mmap.go and split_file.go
Show description for ca1eb41committedfix(gguf): check-before-multiply on tensor element-count overflow (F1, T139.1)
Show description for 285fafecommittedfix(gguf): bounds-check tensor offset before mmap slice (F2, T139.2)
Show description for ab01d5dcommitted
Commits on Apr 28, 2026
refactor(layout): move modelcache/+modeldsl/+registry/ to model/ (T124.5.5)
Show description for e06dc97committed
Commits on Apr 27, 2026
refactor(gguf): use isEmbeddingShape helper in decodeQ8Tensor
Show description for fc1bf12committed
Commits on Apr 22, 2026
feat(gguf): T99.2.2.9 H21 -- keep native Q4_K/Q5_K/Q6_K for embedding tensors
Show description for d84760bcommitted
Commits on Apr 14, 2026
fix(gguf): correct Gemma 4 AttentionKEqV and SlidingWindowPattern defaults
Show description for b1c24eecommittedfeat(gguf): add Gemma 4 edge tensor name mapping
Show description for 3932fb9committedfeat(gguf): route Gemma 4 sub-variants; add e2e integration test
Show description for c6580c0committedfix(gguf): extract Gemma 4 metadata from canonical llama.cpp keys
Show description for 8213a7ecommitted
Commits on Apr 13, 2026
feat(gguf): add Gemma 4 per-layer attention config fields
Show description for 9d108d3committed
Commits on Apr 1, 2026
perf(gguf): re-quantize all weight types to Q4_0 for maximum throughput
Show description for 4b202d2committedperf(gguf): re-quantize Q8_0 weight matrices to Q4_K for merged QKV
Show description for 023505dcommittedperf(gguf): re-quantize Q5_0/Q5_K/Q6_K to Q4_K for uniform fast GEMV
Show description for b408c30committedfix(gguf): update Q5_K/Q6_K decode tests for native storage
Show description for 21cc2bccommitted- committed
perf(gguf): keep native Q4_K storage (eliminate lossy re-quantization)
Show description for 63bdd54committedfix(gguf): match raw GGUF tensor names for embedding native decode
Show description for c9219cacommittedReapply "fix(gguf): skip Q4_0 re-quantization for embedding tensors"
Show description for 42cc834committedRevert "fix(gguf): skip Q4_0 re-quantization for embedding tensors"
Show description for 96eddefcommittedfix(gguf): skip Q4_0 re-quantization for embedding tensors
Show description for badeadbcommittedfix(gguf): upgrade embeddings to F32, keep Q6_K re-quantization for weights
Show description for b887b10committedfix(gguf): keep native Q6_K storage (don't re-quantize to Q4_0)
Show description for eee3a54committedperf(gguf): restore Q4_0 re-quantization for Q4_K/Q5_K/Q6_K/Q5_0
Show description for 4e554e0committed
Commits on Mar 31, 2026
fix(gguf): use native Q4_K storage instead of lossy re-quantization to Q4_0
Show description for 1d56d2ecommitted
Commits on Mar 29, 2026
feat(gguf): add audio encoder config fields and Voxtral tensor name mapping
Show description for ce9b18dcommittedfeat(inference): add split-GGUF support and default mmap loading
Show description for a72391acommitted
Commits on Mar 28, 2026
feat(gguf): add ScoringFunc to ModelConfig for expert gating function
Show description for 4ba0cc2committedfeat(moe): add sigmoid gating option for MoE expert routing
Show description for 69f0f5ecommittedfeat(gguf): add Nemotron-H SSM config fields and tensor name mappings
Show description for 16f2c4acommitted