qllama/bge-small-zh-v1.5:q4_k_m/model

qllama/

bge-small-zh-v1.5:q4_k_m

655 Downloads Updated 6 months ago

quantize https://huggingface.co/BAAI/bge-small-zh-v1.5 to f16 / q8_0 (latest) / q4_k_m

embedding

bge-small-zh-v1.5:q4_k_m ... /

model

f1b50deed303 · 18MB

Metadata

general.architecture

bert

bert
general.file_type

Q4_K_M

Q4_K_M
bert.attention.causal

false

false
bert.attention.head_count

8

8
bert.attention.layer_norm_epsilon

1e-12

1e-12
bert.block_count

4

4
bert.context_length

512

512
bert.embedding_length

512

512
bert.feed_forward_length

2048

2048
bert.pooling_type

CLS

CLS
tokenizer.ggml.cls_token_id

101

101
tokenizer.ggml.mask_token_id

103

103
tokenizer.ggml.model

bert

bert
tokenizer.ggml.padding_token_id

0

0
tokenizer.ggml.seperator_token_id

102

102
tokenizer.ggml.token_type

[3, 1, 1, 1, 1, ...]

[3, 1, 1, 1, 1, ...]
tokenizer.ggml.token_type_count

2

2
tokenizer.ggml.tokens

[[PAD], [unused1], [unused2], [unused3], [unused4], ...]

[[PAD], [unused1], [unused2], [unused3], [unused4], ...]
tokenizer.ggml.unknown_token_id

100

100

Tensor

Name

Type

Shape
token_embd.weight

Q6_K

Q6_K

[512, 21128]

blk.0

blk.0.attn_k.bias

F32

F32

[512]
blk.0.attn_k.weight

Q4_K

Q4_K

[512, 512]
blk.0.attn_output.bias

F32

F32

[512]
blk.0.attn_output.weight

Q4_K

Q4_K

[512, 512]
blk.0.attn_output_norm.bias

F32

F32

[512]
blk.0.attn_output_norm.weight

F32

F32

[512]
blk.0.attn_q.bias

F32

F32

[512]
blk.0.attn_q.weight

Q4_K

Q4_K

[512, 512]
blk.0.attn_v.bias

F32

F32

[512]
blk.0.attn_v.weight

Q4_K

Q4_K

[512, 512]
blk.0.ffn_down.bias

F32

F32

[512]
blk.0.ffn_down.weight

Q4_K

Q4_K

[2048, 512]
blk.0.ffn_up.bias

F32

F32

[2048]
blk.0.ffn_up.weight

Q4_K

Q4_K

[512, 2048]
blk.0.layer_output_norm.bias

F32

F32

[512]
blk.0.layer_output_norm.weight

F32

F32

[512]

blk.1

blk.1.attn_k.bias

F32

F32

[512]
blk.1.attn_k.weight

Q4_K

Q4_K

[512, 512]
blk.1.attn_output.bias

F32

F32

[512]
blk.1.attn_output.weight

Q4_K

Q4_K

[512, 512]
blk.1.attn_output_norm.bias

F32

F32

[512]
blk.1.attn_output_norm.weight

F32

F32

[512]
blk.1.attn_q.bias

F32

F32

[512]
blk.1.attn_q.weight

Q4_K

Q4_K

[512, 512]
blk.1.attn_v.bias

F32

F32

[512]
blk.1.attn_v.weight

Q4_K

Q4_K

[512, 512]
blk.1.ffn_down.bias

F32

F32

[512]
blk.1.ffn_down.weight

Q4_K

Q4_K

[2048, 512]
blk.1.ffn_up.bias

F32

F32

[2048]
blk.1.ffn_up.weight

Q4_K

Q4_K

[512, 2048]
blk.1.layer_output_norm.bias

F32

F32

[512]
blk.1.layer_output_norm.weight

F32

F32

[512]

blk.2

blk.2.attn_k.bias

F32

F32

[512]
blk.2.attn_k.weight

Q4_K

Q4_K

[512, 512]
blk.2.attn_output.bias

F32

F32

[512]
blk.2.attn_output.weight

Q4_K

Q4_K

[512, 512]
blk.2.attn_output_norm.bias

F32

F32

[512]
blk.2.attn_output_norm.weight

F32

F32

[512]
blk.2.attn_q.bias

F32

F32

[512]
blk.2.attn_q.weight

Q4_K

Q4_K

[512, 512]
blk.2.attn_v.bias

F32

F32

[512]
blk.2.attn_v.weight

Q6_K

Q6_K

[512, 512]
blk.2.ffn_down.bias

F32

F32

[512]
blk.2.ffn_down.weight

Q6_K

Q6_K

[2048, 512]
blk.2.ffn_up.bias

F32

F32

[2048]
blk.2.ffn_up.weight

Q4_K

Q4_K

[512, 2048]
blk.2.layer_output_norm.bias

F32

F32

[512]
blk.2.layer_output_norm.weight

F32

F32

[512]

blk.3

blk.3.attn_k.bias

F32

F32

[512]
blk.3.attn_k.weight

Q4_K

Q4_K

[512, 512]
blk.3.attn_output.bias

F32

F32

[512]
blk.3.attn_output.weight

Q4_K

Q4_K

[512, 512]
blk.3.attn_output_norm.bias

F32

F32

[512]
blk.3.attn_output_norm.weight

F32

F32

[512]
blk.3.attn_q.bias

F32

F32

[512]
blk.3.attn_q.weight

Q4_K

Q4_K

[512, 512]
blk.3.attn_v.bias

F32

F32

[512]
blk.3.attn_v.weight

Q6_K

Q6_K

[512, 512]
blk.3.ffn_down.bias

F32

F32

[512]
blk.3.ffn_down.weight

Q6_K

Q6_K

[2048, 512]
blk.3.ffn_up.bias

F32

F32

[2048]
blk.3.ffn_up.weight

Q4_K

Q4_K

[512, 2048]
blk.3.layer_output_norm.bias

F32

F32

[512]
blk.3.layer_output_norm.weight

F32

F32

[512]

position_embd.weight

F16

F16

[512, 512]
token_embd_norm.bias

F32

F32

[512]
token_embd_norm.weight

F32

F32

[512]
token_types.weight

F32

F32

[512, 2]