jina-embeddings-v2-base-zh is a Chinese/English bilingual text embedding model supporting 8192 sequence length.
embedding
160m
51 Pulls Updated 2 weeks ago
ae1783c29cd6 · 322MB
-
general.architecturejina-bert-v2
-
general.basenamejina-embeddings
-
general.file_type1
-
general.languages[en, zh]
-
general.licenseapache-2.0
-
general.nameJina Bert Implementation
-
general.organizationJinaai
-
general.quantization_version2
-
general.size_label160M
-
general.tags[sentence-transformers, feature-extraction, sentence-similarity, mteb, transformers, ...]
-
general.typemodel
-
general.versionv2
-
jina-bert-v2.attention.causalfalse
-
jina-bert-v2.attention.head_count12
-
jina-bert-v2.attention.layer_norm_epsilon1e-12
-
jina-bert-v2.block_count12
-
jina-bert-v2.context_length8192
-
jina-bert-v2.embedding_length768
-
jina-bert-v2.feed_forward_length3072
-
jina-bert-v2.pooling_type1
-
tokenizer.ggml.add_bos_tokentrue
-
tokenizer.ggml.add_eos_tokentrue
-
tokenizer.ggml.bos_token_id0
-
tokenizer.ggml.cls_token_id0
-
tokenizer.ggml.eos_token_id2
-
tokenizer.ggml.mask_token_id4
-
tokenizer.ggml.merges[t h, i n, a n, e r, th e, ...]
-
tokenizer.ggml.modelgpt2
-
tokenizer.ggml.padding_token_id1
-
tokenizer.ggml.prejina-v1-en
-
tokenizer.ggml.seperator_token_id2
-
tokenizer.ggml.token_type[3, 3, 3, 3, 3, ...]
-
tokenizer.ggml.token_type_count2
-
tokenizer.ggml.tokens[<s>, <pad>, </s>, <unk>, <mask>, ...]
-
tokenizer.ggml.unknown_token_id3
-
NameTypeShape
-
token_embd.weightF16[768, 61056]
-
blk.0.attn_k.biasF32[768]
-
blk.0.attn_k.weightF16[768, 768]
-
blk.0.attn_output.biasF32[768]
-
blk.0.attn_output.weightF16[768, 768]
-
blk.0.attn_output_norm.biasF32[768]
-
blk.0.attn_output_norm.weightF32[768]
-
blk.0.attn_q.biasF32[768]
-
blk.0.attn_q.weightF16[768, 768]
-
blk.0.attn_v.biasF32[768]
-
blk.0.attn_v.weightF16[768, 768]
-
blk.0.ffn_down.biasF32[768]
-
blk.0.ffn_down.weightF16[3072, 768]
-
blk.0.ffn_gate.weightF16[768, 3072]
-
blk.0.ffn_up.weightF16[768, 3072]
-
blk.0.layer_output_norm.biasF32[768]
-
blk.0.layer_output_norm.weightF32[768]
-
blk.1.attn_k.biasF32[768]
-
blk.1.attn_k.weightF16[768, 768]
-
blk.1.attn_output.biasF32[768]
-
blk.1.attn_output.weightF16[768, 768]
-
blk.1.attn_output_norm.biasF32[768]
-
blk.1.attn_output_norm.weightF32[768]
-
blk.1.attn_q.biasF32[768]
-
blk.1.attn_q.weightF16[768, 768]
-
blk.1.attn_v.biasF32[768]
-
blk.1.attn_v.weightF16[768, 768]
-
blk.1.ffn_down.biasF32[768]
-
blk.1.ffn_down.weightF16[3072, 768]
-
blk.1.ffn_gate.weightF16[768, 3072]
-
blk.1.ffn_up.weightF16[768, 3072]
-
blk.1.layer_output_norm.biasF32[768]
-
blk.1.layer_output_norm.weightF32[768]
-
blk.2.attn_k.biasF32[768]
-
blk.2.attn_k.weightF16[768, 768]
-
blk.2.attn_output.biasF32[768]
-
blk.2.attn_output.weightF16[768, 768]
-
blk.2.attn_output_norm.biasF32[768]
-
blk.2.attn_output_norm.weightF32[768]
-
blk.2.attn_q.biasF32[768]
-
blk.2.attn_q.weightF16[768, 768]
-
blk.2.attn_v.biasF32[768]
-
blk.2.attn_v.weightF16[768, 768]
-
blk.2.ffn_down.biasF32[768]
-
blk.2.ffn_down.weightF16[3072, 768]
-
blk.2.ffn_gate.weightF16[768, 3072]
-
blk.2.ffn_up.weightF16[768, 3072]
-
blk.2.layer_output_norm.biasF32[768]
-
blk.2.layer_output_norm.weightF32[768]
-
blk.3.attn_k.biasF32[768]
-
blk.3.attn_k.weightF16[768, 768]
-
blk.3.attn_output.biasF32[768]
-
blk.3.attn_output.weightF16[768, 768]
-
blk.3.attn_output_norm.biasF32[768]
-
blk.3.attn_output_norm.weightF32[768]
-
blk.3.attn_q.biasF32[768]
-
blk.3.attn_q.weightF16[768, 768]
-
blk.3.attn_v.biasF32[768]
-
blk.3.attn_v.weightF16[768, 768]
-
blk.3.ffn_down.biasF32[768]
-
blk.3.ffn_down.weightF16[3072, 768]
-
blk.3.ffn_gate.weightF16[768, 3072]
-
blk.3.ffn_up.weightF16[768, 3072]
-
blk.3.layer_output_norm.biasF32[768]
-
blk.3.layer_output_norm.weightF32[768]
-
blk.4.attn_k.biasF32[768]
-
blk.4.attn_k.weightF16[768, 768]
-
blk.4.attn_output.biasF32[768]
-
blk.4.attn_output.weightF16[768, 768]
-
blk.4.attn_output_norm.biasF32[768]
-
blk.4.attn_output_norm.weightF32[768]
-
blk.4.attn_q.biasF32[768]
-
blk.4.attn_q.weightF16[768, 768]
-
blk.4.attn_v.biasF32[768]
-
blk.4.attn_v.weightF16[768, 768]
-
blk.4.ffn_down.biasF32[768]
-
blk.4.ffn_down.weightF16[3072, 768]
-
blk.4.ffn_gate.weightF16[768, 3072]
-
blk.4.ffn_up.weightF16[768, 3072]
-
blk.4.layer_output_norm.biasF32[768]
-
blk.4.layer_output_norm.weightF32[768]
-
blk.5.attn_k.biasF32[768]
-
blk.5.attn_k.weightF16[768, 768]
-
blk.5.attn_output.biasF32[768]
-
blk.5.attn_output.weightF16[768, 768]
-
blk.5.attn_output_norm.biasF32[768]
-
blk.5.attn_output_norm.weightF32[768]
-
blk.5.attn_q.biasF32[768]
-
blk.5.attn_q.weightF16[768, 768]
-
blk.5.attn_v.biasF32[768]
-
blk.5.attn_v.weightF16[768, 768]
-
blk.5.ffn_down.biasF32[768]
-
blk.5.ffn_down.weightF16[3072, 768]
-
blk.5.ffn_gate.weightF16[768, 3072]
-
blk.5.ffn_up.weightF16[768, 3072]
-
blk.5.layer_output_norm.biasF32[768]
-
blk.5.layer_output_norm.weightF32[768]
-
blk.6.attn_k.biasF32[768]
-
blk.6.attn_k.weightF16[768, 768]
-
blk.6.attn_output.biasF32[768]
-
blk.6.attn_output.weightF16[768, 768]
-
blk.6.attn_output_norm.biasF32[768]
-
blk.6.attn_output_norm.weightF32[768]
-
blk.6.attn_q.biasF32[768]
-
blk.6.attn_q.weightF16[768, 768]
-
blk.6.attn_v.biasF32[768]
-
blk.6.attn_v.weightF16[768, 768]
-
blk.6.ffn_down.biasF32[768]
-
blk.6.ffn_down.weightF16[3072, 768]
-
blk.6.ffn_gate.weightF16[768, 3072]
-
blk.6.ffn_up.weightF16[768, 3072]
-
blk.6.layer_output_norm.biasF32[768]
-
blk.6.layer_output_norm.weightF32[768]
-
blk.7.attn_k.biasF32[768]
-
blk.7.attn_k.weightF16[768, 768]
-
blk.7.attn_output.biasF32[768]
-
blk.7.attn_output.weightF16[768, 768]
-
blk.7.attn_output_norm.biasF32[768]
-
blk.7.attn_output_norm.weightF32[768]
-
blk.7.attn_q.biasF32[768]
-
blk.7.attn_q.weightF16[768, 768]
-
blk.7.attn_v.biasF32[768]
-
blk.7.attn_v.weightF16[768, 768]
-
blk.7.ffn_down.biasF32[768]
-
blk.7.ffn_down.weightF16[3072, 768]
-
blk.7.ffn_gate.weightF16[768, 3072]
-
blk.7.ffn_up.weightF16[768, 3072]
-
blk.7.layer_output_norm.biasF32[768]
-
blk.7.layer_output_norm.weightF32[768]
-
blk.8.attn_k.biasF32[768]
-
blk.8.attn_k.weightF16[768, 768]
-
blk.8.attn_output.biasF32[768]
-
blk.8.attn_output.weightF16[768, 768]
-
blk.8.attn_output_norm.biasF32[768]
-
blk.8.attn_output_norm.weightF32[768]
-
blk.8.attn_q.biasF32[768]
-
blk.8.attn_q.weightF16[768, 768]
-
blk.8.attn_v.biasF32[768]
-
blk.8.attn_v.weightF16[768, 768]
-
blk.8.ffn_down.biasF32[768]
-
blk.8.ffn_down.weightF16[3072, 768]
-
blk.8.ffn_gate.weightF16[768, 3072]
-
blk.8.ffn_up.weightF16[768, 3072]
-
blk.8.layer_output_norm.biasF32[768]
-
blk.8.layer_output_norm.weightF32[768]
-
blk.9.attn_k.biasF32[768]
-
blk.9.attn_k.weightF16[768, 768]
-
blk.9.attn_output.biasF32[768]
-
blk.9.attn_output.weightF16[768, 768]
-
blk.9.attn_output_norm.biasF32[768]
-
blk.9.attn_output_norm.weightF32[768]
-
blk.9.attn_q.biasF32[768]
-
blk.9.attn_q.weightF16[768, 768]
-
blk.9.attn_v.biasF32[768]
-
blk.9.attn_v.weightF16[768, 768]
-
blk.9.ffn_down.biasF32[768]
-
blk.9.ffn_down.weightF16[3072, 768]
-
blk.9.ffn_gate.weightF16[768, 3072]
-
blk.9.ffn_up.weightF16[768, 3072]
-
blk.9.layer_output_norm.biasF32[768]
-
blk.9.layer_output_norm.weightF32[768]
-
blk.10.attn_k.biasF32[768]
-
blk.10.attn_k.weightF16[768, 768]
-
blk.10.attn_output.biasF32[768]
-
blk.10.attn_output.weightF16[768, 768]
-
blk.10.attn_output_norm.biasF32[768]
-
blk.10.attn_output_norm.weightF32[768]
-
blk.10.attn_q.biasF32[768]
-
blk.10.attn_q.weightF16[768, 768]
-
blk.10.attn_v.biasF32[768]
-
blk.10.attn_v.weightF16[768, 768]
-
blk.10.ffn_down.biasF32[768]
-
blk.10.ffn_down.weightF16[3072, 768]
-
blk.10.ffn_gate.weightF16[768, 3072]
-
blk.10.ffn_up.weightF16[768, 3072]
-
blk.10.layer_output_norm.biasF32[768]
-
blk.10.layer_output_norm.weightF32[768]
-
blk.11.attn_k.biasF32[768]
-
blk.11.attn_k.weightF16[768, 768]
-
blk.11.attn_output.biasF32[768]
-
blk.11.attn_output.weightF16[768, 768]
-
blk.11.attn_output_norm.biasF32[768]
-
blk.11.attn_output_norm.weightF32[768]
-
blk.11.attn_q.biasF32[768]
-
blk.11.attn_q.weightF16[768, 768]
-
blk.11.attn_v.biasF32[768]
-
blk.11.attn_v.weightF16[768, 768]
-
blk.11.ffn_down.biasF32[768]
-
blk.11.ffn_down.weightF16[3072, 768]
-
blk.11.ffn_gate.weightF16[768, 3072]
-
blk.11.ffn_up.weightF16[768, 3072]
-
blk.11.layer_output_norm.biasF32[768]
-
blk.11.layer_output_norm.weightF32[768]
-
token_embd_norm.biasF32[768]
-
token_embd_norm.weightF32[768]
-
token_types.weightF32[768, 2]
Metadata
Tensor
blk.0
blk.1
blk.2
blk.3
blk.4
blk.5
blk.6
blk.7
blk.8
blk.9
blk.10
blk.11