Rabe3 commited on
Commit
767c2df
·
verified ·
1 Parent(s): c85bb84

Upload README.md with huggingface_hub

Browse files
Files changed (1) hide show
  1. README.md +21 -0
README.md CHANGED
@@ -147,3 +147,24 @@ MIT (following the base model). Built on `microsoft/bitnet-b1.58-2B-4T-bf16`.
147
 
148
  ## Acknowledgements
149
  LLM2Vec (BehnamGhader et al., 2024) · BitNet b1.58 (Microsoft, 2025) · E5 · MedCPT · `llama.cpp` (ggml-org).
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
147
 
148
  ## Acknowledgements
149
  LLM2Vec (BehnamGhader et al., 2024) · BitNet b1.58 (Microsoft, 2025) · E5 · MedCPT · `llama.cpp` (ggml-org).
150
+
151
+ ---
152
+
153
+ ## Full-precision (bf16) weights
154
+
155
+ The bf16 **master weights** are also provided (in the `bf16/` subfolder) for GPU inference with 🤗 Transformers
156
+ or further fine-tuning. Note: BitNet keeps a *ternary* forward pass even from these bf16 weights (online
157
+ quantization) — bf16 is the storage/master-weight format used for training.
158
+
159
+ ```python
160
+ import torch
161
+ from transformers import AutoModel, AutoTokenizer
162
+
163
+ tok = AutoTokenizer.from_pretrained("Rabe3/1-bit-embedding-general", subfolder="bf16")
164
+ model = AutoModel.from_pretrained("Rabe3/1-bit-embedding-general", subfolder="bf16",
165
+ torch_dtype=torch.bfloat16).cuda().eval()
166
+ # NOTE: this is a decoder patched to BIDIRECTIONAL attention for embeddings; use mean pooling
167
+ # over the last hidden state and L2-normalize. See the repo scripts for the exact embedder.
168
+ ```
169
+
170
+ **Files:** `bf16/model.safetensors` (~4.8 GB, bf16), `bf16/config.json` (BitNet, ternary-online), tokenizer.