audio: give the audiocraft venv every dependency it declares

Three build cycles, three different missing modules — xformers, then torch at
install time, then transformers. The list had been trimmed against what the MAIN
venv provides, which is exactly backwards: an isolated venv starts with nothing,
so it needs the whole declared set, minus only the torch pins we choose and the
dev-only tools.

Verified in a throwaway container against the real core image before building
anything: audiocraft 1.3.0 on torch 2.5.1+cu121 with xformers 0.0.28.post3, and
MusicGen and AudioGen both import. That check takes two minutes; the image build
it replaces takes fifteen, which is why the last three rounds were so expensive.
Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_0182DhR1eFNPQyGmDrpHbedY

packaging/runpod/profiles/audio.venv-audiocraft.txt

@@ -25,15 +25,17 @@ torch==2.5.1
 torchaudio==2.5.1
 torchvision==0.20.1
 
-# audiocraft's real runtime dependencies. Its dev-only ones (pytest, flake8,
-# mypy, coverage, pdoc3) and gradio are deliberately absent: a pod serves an
-# API, it does not run the library's test suite or its demo UI. torchtext is
-# absent too — it is abandoned, unbuildable here, and unused at inference.
+# EVERYTHING audiocraft declares, minus only the torch pins above and its
+# dev-only tools (pytest, flake8, mypy, coverage, pdoc3).
 #
-# xformers IS needed: audiocraft imports it at module load, so leaving it out
-# turned the build's import check into a ModuleNotFoundError — which is exactly
-# what that check is for. Pinned to the version whose wheel matches the torch
-# above; see the note there for why that torch and not audiocraft's.
+# The earlier version of this list was trimmed against what the MAIN venv has —
+# transformers, demucs and the rest. That is exactly wrong: this venv is
+# isolated, so it starts with nothing, and the trimming produced three build
+# cycles that each died on a different missing module. An isolated venv needs
+# the whole list.
+# xformers: audiocraft imports it at module load (audiocraft/modules/
+# transformer.py), so without it the import check dies. Pinned to the version
+# whose cp312 wheel matches the torch above — see the note there.
 xformers==0.0.28.post3
 av>=11.0.0
 einops
@@ -46,7 +48,10 @@ sentencepiece
 spacy>=3.6.1
 huggingface_hub
 tqdm
+transformers>=4.31.0
+demucs
 librosa
+gradio
 torchmetrics
 encodec
 protobuf