Add NOTICE attributing upstream Qwen Team
Browse files
NOTICE
ADDED
|
@@ -0,0 +1,17 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
Qwen3-Reranker-4B (OpenVINO IR conversion)
|
| 2 |
+
|
| 3 |
+
This OpenVINO IR conversion is a derivative of Qwen/Qwen3-Reranker-4B
|
| 4 |
+
(https://huggingface.co/Qwen/Qwen3-Reranker-4B), redistributed under the
|
| 5 |
+
Apache License, Version 2.0.
|
| 6 |
+
|
| 7 |
+
Original work:
|
| 8 |
+
Copyright 2024-2025 Alibaba Cloud (Qwen Team).
|
| 9 |
+
Licensed under the Apache License, Version 2.0.
|
| 10 |
+
|
| 11 |
+
Modifications by the redistributor of this OpenVINO conversion:
|
| 12 |
+
- Exported model weights to OpenVINO IR via optimum-intel.
|
| 13 |
+
- Applied NNCF INT8 (asymmetric, weight-only) quantization.
|
| 14 |
+
- Generated OpenVINO Tokenizer / Detokenizer IR via openvino-tokenizers.
|
| 15 |
+
|
| 16 |
+
No model weights were retrained. Tokenizer vocabulary, chat template, and
|
| 17 |
+
generation config are preserved verbatim from the upstream release.
|