shaw/dmeta-embedding-zh-small-q4/model

shaw/

dmeta-embedding-zh-small-q4:latest

885 Downloads Updated 1 year ago

The Q4_K_M quantized version of dmeta-embedding-zh-small.

embedding

dmeta-embedding-zh-small-q4:latest ... /

model

327df418e257 · 52MB

Metadata

general.architecture

bert

bert
general.file_type

Q4_K_M

Q4_K_M
bert.attention.causal

false

false
bert.attention.head_count

12

12
bert.attention.layer_norm_epsilon

1e-12

1e-12
bert.block_count

8

8
bert.context_length

1024

1024
bert.embedding_length

768

768
bert.feed_forward_length

3072

3072
bert.pooling_type

CLS

CLS
tokenizer.ggml.bos_token_id

0

0
tokenizer.ggml.cls_token_id

101

101
tokenizer.ggml.eos_token_id

2

2
tokenizer.ggml.mask_token_id

103

103
tokenizer.ggml.model

bert

bert
tokenizer.ggml.padding_token_id

0

0
tokenizer.ggml.scores

[-1000, -1000, -1000, -1000, -1000, ...]

[-1000, -1000, -1000, -1000, -1000, ...]
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

[768, 21128]

blk.0

blk.0.attn_k.bias

F32

F32

[768]
blk.0.attn_k.weight

Q4_K

Q4_K

[768, 768]
blk.0.attn_output.bias

F32

F32

[768]
blk.0.attn_output.weight

Q4_K

Q4_K

[768, 768]
blk.0.attn_output_norm.bias

F32

F32

[768]
blk.0.attn_output_norm.weight

F32

F32

[768]
blk.0.attn_q.bias

F32

F32

[768]
blk.0.attn_q.weight

Q4_K

Q4_K

[768, 768]
blk.0.attn_v.bias

F32

F32

[768]
blk.0.attn_v.weight

Q6_K

Q6_K

[768, 768]
blk.0.ffn_down.bias

F32

F32

[768]
blk.0.ffn_down.weight

Q6_K

Q6_K

[3072, 768]
blk.0.ffn_up.bias

F32

F32

[3072]
blk.0.ffn_up.weight

Q4_K

Q4_K

[768, 3072]
blk.0.layer_output_norm.bias

F32

F32

[768]
blk.0.layer_output_norm.weight

F32

F32

[768]

blk.1

blk.1.attn_k.bias

F32

F32

[768]
blk.1.attn_k.weight

Q4_K

Q4_K

[768, 768]
blk.1.attn_output.bias

F32

F32

[768]
blk.1.attn_output.weight

Q4_K

Q4_K

[768, 768]
blk.1.attn_output_norm.bias

F32

F32

[768]
blk.1.attn_output_norm.weight

F32

F32

[768]
blk.1.attn_q.bias

F32

F32

[768]
blk.1.attn_q.weight

Q4_K

Q4_K

[768, 768]
blk.1.attn_v.bias

F32

F32

[768]
blk.1.attn_v.weight

Q4_K

Q4_K

[768, 768]
blk.1.ffn_down.bias

F32

F32

[768]
blk.1.ffn_down.weight

Q6_K

Q6_K

[3072, 768]
blk.1.ffn_up.bias

F32

F32

[3072]
blk.1.ffn_up.weight

Q4_K

Q4_K

[768, 3072]
blk.1.layer_output_norm.bias

F32

F32

[768]
blk.1.layer_output_norm.weight

F32

F32

[768]

blk.2

blk.2.attn_k.bias

F32

F32

[768]
blk.2.attn_k.weight

Q4_K

Q4_K

[768, 768]
blk.2.attn_output.bias

F32

F32

[768]
blk.2.attn_output.weight

Q4_K

Q4_K

[768, 768]
blk.2.attn_output_norm.bias

F32

F32

[768]
blk.2.attn_output_norm.weight

F32

F32

[768]
blk.2.attn_q.bias

F32

F32

[768]
blk.2.attn_q.weight

Q4_K

Q4_K

[768, 768]
blk.2.attn_v.bias

F32

F32

[768]
blk.2.attn_v.weight

Q4_K

Q4_K

[768, 768]
blk.2.ffn_down.bias

F32

F32

[768]
blk.2.ffn_down.weight

Q4_K

Q4_K

[3072, 768]
blk.2.ffn_up.bias

F32

F32

[3072]
blk.2.ffn_up.weight

Q4_K

Q4_K

[768, 3072]
blk.2.layer_output_norm.bias

F32

F32

[768]
blk.2.layer_output_norm.weight

F32

F32

[768]

blk.3

blk.3.attn_k.bias

F32

F32

[768]
blk.3.attn_k.weight

Q4_K

Q4_K

[768, 768]
blk.3.attn_output.bias

F32

F32

[768]
blk.3.attn_output.weight

Q4_K

Q4_K

[768, 768]
blk.3.attn_output_norm.bias

F32

F32

[768]
blk.3.attn_output_norm.weight

F32

F32

[768]
blk.3.attn_q.bias

F32

F32

[768]
blk.3.attn_q.weight

Q4_K

Q4_K

[768, 768]
blk.3.attn_v.bias

F32

F32

[768]
blk.3.attn_v.weight

Q6_K

Q6_K

[768, 768]
blk.3.ffn_down.bias

F32

F32

[768]
blk.3.ffn_down.weight

Q4_K

Q4_K

[3072, 768]
blk.3.ffn_up.bias

F32

F32

[3072]
blk.3.ffn_up.weight

Q4_K

Q4_K

[768, 3072]
blk.3.layer_output_norm.bias

F32

F32

[768]
blk.3.layer_output_norm.weight

F32

F32

[768]

blk.4

blk.4.attn_k.bias

F32

F32

[768]
blk.4.attn_k.weight

Q4_K

Q4_K

[768, 768]
blk.4.attn_output.bias

F32

F32

[768]
blk.4.attn_output.weight

Q4_K

Q4_K

[768, 768]
blk.4.attn_output_norm.bias

F32

F32

[768]
blk.4.attn_output_norm.weight

F32

F32

[768]
blk.4.attn_q.bias

F32

F32

[768]
blk.4.attn_q.weight

Q4_K

Q4_K

[768, 768]
blk.4.attn_v.bias

F32

F32

[768]
blk.4.attn_v.weight

Q4_K

Q4_K

[768, 768]
blk.4.ffn_down.bias

F32

F32

[768]
blk.4.ffn_down.weight

Q6_K

Q6_K

[3072, 768]
blk.4.ffn_up.bias

F32

F32

[3072]
blk.4.ffn_up.weight

Q4_K

Q4_K

[768, 3072]
blk.4.layer_output_norm.bias

F32

F32

[768]
blk.4.layer_output_norm.weight

F32

F32

[768]

blk.5

blk.5.attn_k.bias

F32

F32

[768]
blk.5.attn_k.weight

Q4_K

Q4_K

[768, 768]
blk.5.attn_output.bias

F32

F32

[768]
blk.5.attn_output.weight

Q4_K

Q4_K

[768, 768]
blk.5.attn_output_norm.bias

F32

F32

[768]
blk.5.attn_output_norm.weight

F32

F32

[768]
blk.5.attn_q.bias

F32

F32

[768]
blk.5.attn_q.weight

Q4_K

Q4_K

[768, 768]
blk.5.attn_v.bias

F32

F32

[768]
blk.5.attn_v.weight

Q4_K

Q4_K

[768, 768]
blk.5.ffn_down.bias

F32

F32

[768]
blk.5.ffn_down.weight

Q4_K

Q4_K

[3072, 768]
blk.5.ffn_up.bias

F32

F32

[3072]
blk.5.ffn_up.weight

Q4_K

Q4_K

[768, 3072]
blk.5.layer_output_norm.bias

F32

F32

[768]
blk.5.layer_output_norm.weight

F32

F32

[768]

blk.6

blk.6.attn_k.bias

F32

F32

[768]
blk.6.attn_k.weight

Q4_K

Q4_K

[768, 768]
blk.6.attn_output.bias

F32

F32

[768]
blk.6.attn_output.weight

Q4_K

Q4_K

[768, 768]
blk.6.attn_output_norm.bias

F32

F32

[768]
blk.6.attn_output_norm.weight

F32

F32

[768]
blk.6.attn_q.bias

F32

F32

[768]
blk.6.attn_q.weight

Q4_K

Q4_K

[768, 768]
blk.6.attn_v.bias

F32

F32

[768]
blk.6.attn_v.weight

Q6_K

Q6_K

[768, 768]
blk.6.ffn_down.bias

F32

F32

[768]
blk.6.ffn_down.weight

Q4_K

Q4_K

[3072, 768]
blk.6.ffn_up.bias

F32

F32

[3072]
blk.6.ffn_up.weight

Q4_K

Q4_K

[768, 3072]
blk.6.layer_output_norm.bias

F32

F32

[768]
blk.6.layer_output_norm.weight

F32

F32

[768]

blk.7

blk.7.attn_k.bias

F32

F32

[768]
blk.7.attn_k.weight

Q4_K

Q4_K

[768, 768]
blk.7.attn_output.bias

F32

F32

[768]
blk.7.attn_output.weight

Q4_K

Q4_K

[768, 768]
blk.7.attn_output_norm.bias

F32

F32

[768]
blk.7.attn_output_norm.weight

F32

F32

[768]
blk.7.attn_q.bias

F32

F32

[768]
blk.7.attn_q.weight

Q4_K

Q4_K

[768, 768]
blk.7.attn_v.bias

F32

F32

[768]
blk.7.attn_v.weight

Q6_K

Q6_K

[768, 768]
blk.7.ffn_down.bias

F32

F32

[768]
blk.7.ffn_down.weight

Q6_K

Q6_K

[3072, 768]
blk.7.ffn_up.bias

F32

F32

[3072]
blk.7.ffn_up.weight

Q4_K

Q4_K

[768, 3072]
blk.7.layer_output_norm.bias

F32

F32

[768]
blk.7.layer_output_norm.weight

F32

F32

[768]

position_embd.weight

F32

F32

[768, 1024]
token_embd_norm.bias

F32

F32

[768]
token_embd_norm.weight

F32

F32

[768]
token_types.weight

F32

F32

[768, 2]