pin sdk 6.16 stack (mirrors her)

README.md

@@ -4,6 +4,7 @@ emoji: 🪐
 colorFrom: purple
 colorTo: yellow
 sdk: gradio
+sdk_version: 6.16.0
 app_file: app.py
 pinned: false
 license: mit

requirements.txt

@@ -19,7 +19,12 @@ llama-cpp-python>=0.3.8
 # mamba kernels: source builds fail under pip build isolation (setup.py imports
 # torch) — use the prebuilt cu12/torch2.8/cp310 wheels.
 # ---------------------------------------------------------------------------
+# Stack mirrors the org-blessed "her" entry (proven on zero-a10g): sdk_version
+# 6.16.0 in README + transformers v4 line (hub<1.0 world). 4.56 ≥ the 4.55.4
+# the 30B-A3B MoE config requires.
 transformers>=4.56,<5
 accelerate>=0.34.0
+einops
+sentencepiece
 causal_conv1d @ https://github.com/Dao-AILab/causal-conv1d/releases/download/v1.6.2.post1/causal_conv1d-1.6.2.post1+cu12torch2.8cxx11abiTRUE-cp310-cp310-linux_x86_64.whl
 mamba_ssm @ https://github.com/state-spaces/mamba/releases/download/v2.3.2.post1/mamba_ssm-2.3.2.post1+cu12torch2.8cxx11abiTRUE-cp310-cp310-linux_x86_64.whl