Upload fp32.sft

Updated weights (the prior ones were "post-fix trained" with an out of date tokenizer)

models/ckpt/ar+nar-tts+stt-llama-8/fp32.sft

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:3e64d7e69932ffa1fbc2c8ae786f65ea5eebf7ee5ac3c6c92be801ac1388509d
-size 456274426
+oid sha256:fa86f2361aca832b24b8a09c7113de4d0c17a7c069de8f3a544b1f8717d9faa0
+size 456274378