feat(v0.2.9): Unified LLM Interface with ModelSpec Abstraction#80

Merged

m96-chan merged 14 commits intomainfrom

Dec 16, 2025

Commits on Dec 15, 2025

refactor: organize root directory structure

m96-chan
and
claude
committed

Commits on Dec 16, 2025

feat(v0.2.9): General LLM Execution - Attention layer and E2E inference

m96-chan
and
claude
committed
feat(v0.2.9): Unified Transformer + Hybrid Attention

m96-chan
and
claude
committed
feat(llm): Add ShardedSafeTensorsFile and Qwen3 support

m96-chan
and
claude
committed
feat(llm): Add FP16 inference support for Qwen3-8B

m96-chan
and
claude
committed
feat(ops): Add native FP16/BF16 kernels for transpose_3d_021 and reshape_copy

m96-chan
and
claude
committed
refactor(llm): Add ModelSpec abstraction for model-specific differences

m96-chan
and
claude
committed
refactor(llm): convert model loaders to thin wrappers

m96-chan
and
claude
committed
refactor(llm): unify all models into CausalTransformerModel

m96-chan
and
claude
committed
test(llm): add unified interface tests for v0.2.9

m96-chan
and
claude
committed
demo(llm): update E2E demo for unified ModelSpec interface

m96-chan
and
claude
committed
docs: clarify tokenizer is experimental, update README for v0.2.9

m96-chan
and
claude
committed
docs(readme): add What's New in v0.2.9 section

m96-chan
and
claude
committed
fix(llm): resolve mypy type errors in model.py

m96-chan
and
claude
committed