docker: add jiwer, langid, easydict, protobuf to baked notebook deps

Continuation of the notebook-dep prebaking: the in-image notebook runner
neutralises pip cells, so declared deps must be prebaked. evaluate's WER
metric imports jiwer (Whisper), DeepSeek-R1 GRPO's reward uses langid,
some vision trust_remote_code files need easydict, and sentencepiece
tokenizer conversion needs protobuf. All pure-Python; torch pin intact.

docker/Dockerfile

@@ -272,15 +272,20 @@ RUN set -eux \
 # matplotlib rides along for the notebook crowd: plotting is table stakes in
 # a Jupyter image, and several model repos' trust_remote_code modeling files
 # (e.g. DeepSeek-OCR) import it unconditionally.
-# soundfile (TTS notebooks read/write audio; bundles libsndfile in its wheel),
-# evaluate (Whisper notebook's WER metric), and tensorboard (default
-# TrainingArguments report_to backend) are declared by notebook install cells
-# that the in-image runner neutralises, so bake them here. All pure-Python or
-# self-contained wheels; none names torch, so the cu128 pin is undisturbed.
+# These are declared by notebook install cells that the in-image runner
+# neutralises (deps are meant to be prebaked), so bake them here. All
+# pure-Python or self-contained wheels; none names torch, so the cu128 pin
+# is undisturbed:
+#   soundfile  TTS notebooks read/write audio (bundles libsndfile in its wheel)
+#   evaluate + jiwer  Whisper notebook's WER metric (evaluate.load("wer") -> jiwer)
+#   tensorboard  default TrainingArguments report_to backend
+#   langid  DeepSeek-R1 GRPO reward's language-id check
+#   easydict  some vision trust_remote_code modeling files
+#   protobuf  slow->fast tokenizer conversion for sentencepiece models
 RUN ${VENV}/bin/uv pip install \
         --python ${VENV}/bin/python \
         jupyterlab notebook ipywidgets matplotlib \
-        soundfile evaluate tensorboard
+        soundfile evaluate jiwer tensorboard langid easydict protobuf
 
 # Audio decode out of the box: the TTS/STT notebooks feed datasets' Audio
 # features, which decode through torchcodec. Three traps, all defended: