Update README.md

README.md

@@ -2,29 +2,3 @@
 library_name: transformers
 pipeline_tag: fill-mask
 ---
-
-# MyBERT (RoPE + Pre-LN, ~21M params)
-
-Custom BERT-style encoder trained with MLM on packed BookCorpus.
-Trust remote code is required because the model uses RoPE.
-
-```python
-from transformers import AutoTokenizer, AutoModelForMaskedLM
-import torch, torch.nn.functional as F
-
-tok = AutoTokenizer.from_pretrained("USERNAME/REPO")
-mdl = AutoModelForMaskedLM.from_pretrained("USERNAME/REPO", trust_remote_code=True).eval()
-
-text = f"the capital of france is {tok.mask_token}."
-enc  = tok(text, return_tensors="pt")
-with torch.no_grad():
-    logits = mdl(**enc).logits
-mask_pos = (enc["input_ids"][0] == tok.mask_token_id).nonzero()[0, 0]
-probs = F.softmax(logits[0, mask_pos], dim=-1)
-for p, i in zip(*[t.tolist() for t in probs.topk(5)]):
-    print(f"{p:.4f}  {tok.decode([i])!r}")
-```
-
-> **Note:** This is a small model trained for limited compute. It does not have
-> strong factual knowledge and is best used as a base for fine-tuning on a
-> downstream task.