Added test_mla_defaults_n_kv_to_1_when_heads_absent to verify MLA path uses n_kv=1 (not n_heads) when head_count_kv is absent. |
||
|---|---|---|
| .. | ||
| backend | ||
| frontend | ||
| __init__.py | ||
| install_llama_prebuilt.py | ||
| install_python_stack.py | ||
| LICENSE.AGPL-3.0 | ||
| setup.bat | ||
| setup.ps1 | ||
| setup.sh | ||
| Unsloth_Studio_Colab.ipynb | ||