train: state which tokenizer files shipped with the export

train/train_ounce100m.py

@@ -330,6 +330,17 @@ def main():
 
     out = os.path.join(args.out, "final")
     tr.save_model(out)
+    # The tokenizer travels with the run so `final/` is loadable by AutoTokenizer on its own (§8). If it
+    # is not in the mix root, say so in the summary instead of skipping quietly and being surprised later.
+    shipped = []
+    import shutil as _sh
+    for fn in ("tokenizer.json", "tokenizer_config.json", "special_tokens_map.json", "vocab.json",
+               "merges.txt"):
+        src = os.path.join(args.root, fn)
+        if os.path.exists(src):
+            _sh.copy(src, out)
+            shipped.append(fn)
+    say(f"tokenizer files shipped with the export: {shipped or 'NONE -- Phase 5 must add them'}")
     SD.save_cursor(out, SD.cursor_dict(args.seq_len, samples, args.data_seed, fingerprint))
     json.dump({"step": step, "tokens_consumed": consumed, "seconds": round(elapsed, 1),
                "params": pc, "fingerprint": fingerprint, "final_loss": final_loss,