Sentence Similarity
Transformers
Safetensors
Vietnamese
sai_embedding
feature-extraction
embeddings
retrieval
llm2vec
matryoshka
custom-code
custom_code
Instructions to use thongbuind/SAI-Embedding_100M with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Transformers
How to use thongbuind/SAI-Embedding_100M with Transformers:
# pip install -U transformers accelerate # Load model directly from transformers import AutoModel model = AutoModel.from_pretrained("thongbuind/SAI-Embedding_100M", trust_remote_code=True, device_map="auto") - Notebooks
- Google Colab
- Kaggle
Download TransformerModel.py from thongbuind/SAI-Embedding_100M: direct link, hf CLI and curl.
- Browser
- Download file 3.98 kB
-
https://huggingface.co/thongbuind/SAI-Embedding_100M/resolve/main/TransformerModel.py
- Command line
-
hf download hf://thongbuind/SAI-Embedding_100M/TransformerModel.py
-
curl -L -o TransformerModel.py https://huggingface.co/thongbuind/SAI-Embedding_100M/resolve/main/TransformerModel.py
3.98 kB
| import torch | |
| import torch.nn as nn | |
| from .DecoderBlock import DecoderBlock | |
| from .RotaryPositionalEmbedding import RotaryPositionalEmbedding | |
| class TransformerModel(nn.Module): | |
| def __init__(self, vocab_size: int, d_model: int, num_heads: int, num_kv_heads: int, num_layers: int, ff_dim: int, max_seq_len: int, dropout: float, pad_token_id: int = 0): | |
| super().__init__() | |
| self.d_model = d_model | |
| self.num_heads = num_heads | |
| self.num_kv_heads = num_kv_heads | |
| self.num_layers = num_layers | |
| self.pad_token_id = pad_token_id | |
| self.max_seq_len = max_seq_len | |
| self.embed = nn.Embedding(vocab_size, d_model) | |
| d_k = d_model // num_heads | |
| self.rope = RotaryPositionalEmbedding(d_k, max_seq_len) | |
| self.blocks = nn.ModuleList([ | |
| DecoderBlock(d_model, num_heads, num_kv_heads, ff_dim, dropout) | |
| for _ in range(num_layers) | |
| ]) | |
| self.norm = nn.RMSNorm(d_model, eps=1e-6) | |
| self.lm_head = nn.Linear(d_model, vocab_size, bias=False) | |
| self.lm_head.weight = self.embed.weight | |
| causal = torch.triu( | |
| torch.full((max_seq_len, max_seq_len), float('-inf')), diagonal=1 | |
| ) | |
| self.register_buffer("causal_mask", causal, persistent=False) | |
| self._init_weights() | |
| def _init_weights(self): | |
| std = self.d_model ** -0.5 | |
| nn.init.normal_(self.embed.weight, mean=0.0, std=std) | |
| for module in self.modules(): | |
| if isinstance(module, nn.Linear): | |
| nn.init.normal_(module.weight, mean=0.0, std=std) | |
| if module.bias is not None: | |
| nn.init.zeros_(module.bias) | |
| def _build_attn_mask(self, T: int, pad_mask, device, causal: bool = True): | |
| if not causal: | |
| # Bidirectional (embedding): chỉ chặn key là PAD, mọi token nhìn thấy nhau. | |
| pad = torch.zeros(pad_mask.shape[0], 1, 1, T, device=device) | |
| pad.masked_fill_(~pad_mask[:, None, None, :], float('-inf')) | |
| return pad | |
| attn_mask = self.causal_mask[:T, :T][None, None, :, :] | |
| if pad_mask is not None: | |
| pad = torch.zeros(pad_mask.shape[0], 1, 1, T, device=device) | |
| pad.masked_fill_(~pad_mask[:, None, None, :], float('-inf')) | |
| attn_mask = attn_mask + pad | |
| return attn_mask | |
| def forward_features(self, input_ids: torch.Tensor, attention_mask=None, has_padding: bool = True, causal: bool = True) -> torch.Tensor: | |
| """Giống forward() nhưng DỪNG TRƯỚC lm_head — dùng cho training loss chunked | |
| (tránh vật lý hóa logits full (B*T, vocab_size)). | |
| causal=False bật attention hai chiều cho embedding model (LLM2Vec); | |
| causal=True giữ đúng hành vi của LLM gốc (dùng để đo baseline zero-shot).""" | |
| pad_mask = attention_mask.bool() if attention_mask is not None \ | |
| else (input_ids != self.pad_token_id) | |
| B, T = input_ids.shape | |
| x = self.embed(input_ids) | |
| pos = torch.arange(T, device=input_ids.device) | |
| cos, sin = self.rope.get_cos_sin(pos) | |
| # Chỉ build mask (và cộng pad-bias) khi batch này thực sự có token PAD. | |
| # Nếu không có PAD, causal mask thuần == causal+pad mask về mặt toán học | |
| # (pad-bias toàn số 0) -> bỏ qua an toàn, không đổi kết quả. | |
| # has_padding được tính sẵn trên CPU trong collate_fn nên không tốn sync GPU ở đây. | |
| attn_mask = self._build_attn_mask(T, pad_mask, x.device, causal) if has_padding else None | |
| for block in self.blocks: | |
| x = block(x, cos, sin, attn_mask, causal) | |
| return self.norm(x) # (B, T, d_model) — CHƯA qua lm_head | |
| def forward(self, input_ids: torch.Tensor, attention_mask=None, has_padding: bool = True) -> torch.Tensor: | |
| x = self.forward_features(input_ids, attention_mask, has_padding) | |
| return self.lm_head(x) | |