Upload folder using huggingface_hub

README.md

@@ -1,3 +1,28 @@
+license mit
+language
+- vi
+- ty
+tags
+- translation
+- transformer
+library_name pytorch
 ---
-license: apache-2.0
----
+
+# Vi → Tày Transformer (custom)
+
+- Kiến trúc Encoder–Decoder Transformer (PyTorch thuần), beam search.
+- Tokenizer SentencePiece (file `spm.model`).
+- Trained for Vietnamese → Tày.
+
+## Cách load
+```python
+import json, torch, sentencepiece as spm
+from model import ModelConfig, Seq2SeqTransformer, PAD, BOS, EOS, LANG2ID  # (bạn định nghĩa trong code của mình)
+
+cfg = json.load(open(config.json,r,encoding=utf-8))
+cfg = ModelConfig(cfg)
+model = Seq2SeqTransformer(cfg)
+model.load_state_dict(torch.load(pytorch_model.bin, map_location=cpu))
+model.eval()
+
+sp = spm.SentencePieceProcessor(model_file=spm.model)

config.json

@@ -0,0 +1,16 @@
+{
+  "vocab_size": 32000,
+  "d_model": 384,
+  "num_heads": 6,
+  "d_ff": 1536,
+  "num_encoder_layers": 6,
+  "num_decoder_layers": 6,
+  "max_pos": 1024,
+  "emb_dropout": 0.1,
+  "attn_pdrop": 0.1,
+  "resid_pdrop": 0.1,
+  "layerdrop": 0.1,
+  "pad_token_id": 0,
+  "tie_embeddings": true,
+  "num_langs": 2
+}

pytorch_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8c77e4768b976b2ac7642626bc8d3589b3e1558aaa41baf8c8c9c355e3df8832
+size 166757503

spm.model

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bd2fa575aee9f09b9d95917125e4f1fe483a6db55527abd8a61bdf7fae54b68a
+size 765927