Upload README.md with huggingface_hub
Browse files
README.md
ADDED
|
@@ -0,0 +1,9 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
# Streamalign-SLM (R16)
|
| 2 |
+
|
| 3 |
+
StreamAlign speech language model (Llama-3.2-1B backbone, R16 hier
|
| 4 |
+
durfirst-durreg, text-KL recipe). Scores speech via the paired R16 RVQ
|
| 5 |
+
teacher `Streamalign-R16`.
|
| 6 |
+
|
| 7 |
+
- Original checkpoint: `R16_step180k_lossWtext5_kl05_210k/step_00185000.pt`
|
| 8 |
+
- Speech tokenizer / teacher: `Streamalign-R16`
|
| 9 |
+
- Reported metrics: SALMon 69.1 (4-task, no pitch), StoryCloze 72.1 (sSC 59.0 / tSC 85.1)
|