Embedding models on very large sentence level datasets.
embedding
22m
33m
204.5K Pulls Updated 7 months ago
797b70c4edf8 · 46MB
-
bert.attention.causalfalse
-
bert.attention.head_count12
-
bert.attention.layer_norm_epsilon1e-12
-
bert.block_count6
-
bert.context_length512
-
bert.embedding_length384
-
bert.feed_forward_length1536
-
bert.pooling_type1
-
general.architecturebert
-
general.file_type1
-
general.nameall-MiniLM-L6-v2
-
tokenizer.ggml.bos_token_id101
-
tokenizer.ggml.cls_token_id101
-
tokenizer.ggml.eos_token_id102
-
tokenizer.ggml.mask_token_id103
-
tokenizer.ggml.modelbert
-
tokenizer.ggml.padding_token_id0
-
tokenizer.ggml.scores[-1000, -1000, -1000, -1000, -1000, ...]
-
tokenizer.ggml.seperator_token_id102
-
tokenizer.ggml.token_type[3, 1, 1, 1, 1, ...]
-
tokenizer.ggml.token_type_count2
-
tokenizer.ggml.tokens[[PAD], [unused0], [unused1], [unused2], [unused3], ...]
-
tokenizer.ggml.unknown_token_id100
-
NameTypeShape
-
token_embd.weightF16[384, 30522]
-
blk.0.attn_k.biasF32[384]
-
blk.0.attn_k.weightF16[384, 384]
-
blk.0.attn_output.biasF32[384]
-
blk.0.attn_output.weightF16[384, 384]
-
blk.0.attn_output_norm.biasF32[384]
-
blk.0.attn_output_norm.weightF32[384]
-
blk.0.attn_q.biasF32[384]
-
blk.0.attn_q.weightF16[384, 384]
-
blk.0.attn_v.biasF32[384]
-
blk.0.attn_v.weightF16[384, 384]
-
blk.0.ffn_down.biasF32[384]
-
blk.0.ffn_down.weightF16[1536, 384]
-
blk.0.ffn_up.biasF32[1536]
-
blk.0.ffn_up.weightF16[384, 1536]
-
blk.0.layer_output_norm.biasF32[384]
-
blk.0.layer_output_norm.weightF32[384]
-
blk.1.attn_k.biasF32[384]
-
blk.1.attn_k.weightF16[384, 384]
-
blk.1.attn_output.biasF32[384]
-
blk.1.attn_output.weightF16[384, 384]
-
blk.1.attn_output_norm.biasF32[384]
-
blk.1.attn_output_norm.weightF32[384]
-
blk.1.attn_q.biasF32[384]
-
blk.1.attn_q.weightF16[384, 384]
-
blk.1.attn_v.biasF32[384]
-
blk.1.attn_v.weightF16[384, 384]
-
blk.1.ffn_down.biasF32[384]
-
blk.1.ffn_down.weightF16[1536, 384]
-
blk.1.ffn_up.biasF32[1536]
-
blk.1.ffn_up.weightF16[384, 1536]
-
blk.1.layer_output_norm.biasF32[384]
-
blk.1.layer_output_norm.weightF32[384]
-
blk.2.attn_k.biasF32[384]
-
blk.2.attn_k.weightF16[384, 384]
-
blk.2.attn_output.biasF32[384]
-
blk.2.attn_output.weightF16[384, 384]
-
blk.2.attn_output_norm.biasF32[384]
-
blk.2.attn_output_norm.weightF32[384]
-
blk.2.attn_q.biasF32[384]
-
blk.2.attn_q.weightF16[384, 384]
-
blk.2.attn_v.biasF32[384]
-
blk.2.attn_v.weightF16[384, 384]
-
blk.2.ffn_down.biasF32[384]
-
blk.2.ffn_down.weightF16[1536, 384]
-
blk.2.ffn_up.biasF32[1536]
-
blk.2.ffn_up.weightF16[384, 1536]
-
blk.2.layer_output_norm.biasF32[384]
-
blk.2.layer_output_norm.weightF32[384]
-
blk.3.attn_k.biasF32[384]
-
blk.3.attn_k.weightF16[384, 384]
-
blk.3.attn_output.biasF32[384]
-
blk.3.attn_output.weightF16[384, 384]
-
blk.3.attn_output_norm.biasF32[384]
-
blk.3.attn_output_norm.weightF32[384]
-
blk.3.attn_q.biasF32[384]
-
blk.3.attn_q.weightF16[384, 384]
-
blk.3.attn_v.biasF32[384]
-
blk.3.attn_v.weightF16[384, 384]
-
blk.3.ffn_down.biasF32[384]
-
blk.3.ffn_down.weightF16[1536, 384]
-
blk.3.ffn_up.biasF32[1536]
-
blk.3.ffn_up.weightF16[384, 1536]
-
blk.3.layer_output_norm.biasF32[384]
-
blk.3.layer_output_norm.weightF32[384]
-
blk.4.attn_k.biasF32[384]
-
blk.4.attn_k.weightF16[384, 384]
-
blk.4.attn_output.biasF32[384]
-
blk.4.attn_output.weightF16[384, 384]
-
blk.4.attn_output_norm.biasF32[384]
-
blk.4.attn_output_norm.weightF32[384]
-
blk.4.attn_q.biasF32[384]
-
blk.4.attn_q.weightF16[384, 384]
-
blk.4.attn_v.biasF32[384]
-
blk.4.attn_v.weightF16[384, 384]
-
blk.4.ffn_down.biasF32[384]
-
blk.4.ffn_down.weightF16[1536, 384]
-
blk.4.ffn_up.biasF32[1536]
-
blk.4.ffn_up.weightF16[384, 1536]
-
blk.4.layer_output_norm.biasF32[384]
-
blk.4.layer_output_norm.weightF32[384]
-
blk.5.attn_k.biasF32[384]
-
blk.5.attn_k.weightF16[384, 384]
-
blk.5.attn_output.biasF32[384]
-
blk.5.attn_output.weightF16[384, 384]
-
blk.5.attn_output_norm.biasF32[384]
-
blk.5.attn_output_norm.weightF32[384]
-
blk.5.attn_q.biasF32[384]
-
blk.5.attn_q.weightF16[384, 384]
-
blk.5.attn_v.biasF32[384]
-
blk.5.attn_v.weightF16[384, 384]
-
blk.5.ffn_down.biasF32[384]
-
blk.5.ffn_down.weightF16[1536, 384]
-
blk.5.ffn_up.biasF32[1536]
-
blk.5.ffn_up.weightF16[384, 1536]
-
blk.5.layer_output_norm.biasF32[384]
-
blk.5.layer_output_norm.weightF32[384]
-
position_embd.weightF16[384, 512]
-
token_embd_norm.biasF32[384]
-
token_embd_norm.weightF32[384]
-
token_types.weightF32[384, 2]
Metadata
Tensor
blk.0
blk.1
blk.2
blk.3
blk.4
blk.5