snowflake-arctic-embed:137m/model

snowflake-arctic-embed:137m

838.4K Downloads Updated 1 year ago

A suite of text embedding models by Snowflake, optimized for performance.

embedding 22m 33m 110m 137m 335m

snowflake-arctic-embed:137m ... /

model

4c5716ded514 · 274MB

Metadata

general.architecture

nomic-bert

nomic-bert
general.file_type

F16

F16
nomic-bert.attention.causal

false

false
nomic-bert.attention.head_count

12

12
nomic-bert.attention.layer_norm_epsilon

1e-12

1e-12
nomic-bert.block_count

12

12
nomic-bert.context_length

2048

2048
nomic-bert.embedding_length

768

768
nomic-bert.feed_forward_length

3072

3072
nomic-bert.pooling_type

CLS

CLS
nomic-bert.rope.freq_base

1000

1000
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], [unused0], [unused1], [unused2], [unused3], ...]

[[PAD], [unused0], [unused1], [unused2], [unused3], ...]
tokenizer.ggml.unknown_token_id

100

100

Tensor

Name

Type

Shape
token_embd.weight

F16

F16

[768, 30528]

blk.0

blk.0.attn_output.weight

F16

F16

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

F32

F32

[768]
blk.0.attn_output_norm.weight

F32

F32

[768]
blk.0.attn_qkv.weight

F16

F16

[768, 2304]
blk.0.ffn_down.weight

F16

F16

[3072, 768]
blk.0.ffn_gate.weight

F16

F16

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

F16

F16

[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_output.weight

F16

F16

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

F32

F32

[768]
blk.1.attn_output_norm.weight

F32

F32

[768]
blk.1.attn_qkv.weight

F16

F16

[768, 2304]
blk.1.ffn_down.weight

F16

F16

[3072, 768]
blk.1.ffn_gate.weight

F16

F16

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

F16

F16

[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_output.weight

F16

F16

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

F32

F32

[768]
blk.2.attn_output_norm.weight

F32

F32

[768]
blk.2.attn_qkv.weight

F16

F16

[768, 2304]
blk.2.ffn_down.weight

F16

F16

[3072, 768]
blk.2.ffn_gate.weight

F16

F16

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

F16

F16

[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_output.weight

F16

F16

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

F32

F32

[768]
blk.3.attn_output_norm.weight

F32

F32

[768]
blk.3.attn_qkv.weight

F16

F16

[768, 2304]
blk.3.ffn_down.weight

F16

F16

[3072, 768]
blk.3.ffn_gate.weight

F16

F16

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

F16

F16

[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_output.weight

F16

F16

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

F32

F32

[768]
blk.4.attn_output_norm.weight

F32

F32

[768]
blk.4.attn_qkv.weight

F16

F16

[768, 2304]
blk.4.ffn_down.weight

F16

F16

[3072, 768]
blk.4.ffn_gate.weight

F16

F16

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

F16

F16

[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_output.weight

F16

F16

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

F32

F32

[768]
blk.5.attn_output_norm.weight

F32

F32

[768]
blk.5.attn_qkv.weight

F16

F16

[768, 2304]
blk.5.ffn_down.weight

F16

F16

[3072, 768]
blk.5.ffn_gate.weight

F16

F16

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

F16

F16

[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_output.weight

F16

F16

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

F32

F32

[768]
blk.6.attn_output_norm.weight

F32

F32

[768]
blk.6.attn_qkv.weight

F16

F16

[768, 2304]
blk.6.ffn_down.weight

F16

F16

[3072, 768]
blk.6.ffn_gate.weight

F16

F16

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

F16

F16

[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_output.weight

F16

F16

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

F32

F32

[768]
blk.7.attn_output_norm.weight

F32

F32

[768]
blk.7.attn_qkv.weight

F16

F16

[768, 2304]
blk.7.ffn_down.weight

F16

F16

[3072, 768]
blk.7.ffn_gate.weight

F16

F16

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

F16

F16

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

F32

F32

[768]
blk.7.layer_output_norm.weight

F32

F32

[768]

blk.8

blk.8.attn_output.weight

F16

F16

[768, 768]
blk.8.attn_output_norm.bias

F32

F32

[768]
blk.8.attn_output_norm.weight

F32

F32

[768]
blk.8.attn_qkv.weight

F16

F16

[768, 2304]
blk.8.ffn_down.weight

F16

F16

[3072, 768]
blk.8.ffn_gate.weight

F16

F16

[768, 3072]
blk.8.ffn_up.weight

F16

F16

[768, 3072]
blk.8.layer_output_norm.bias

F32

F32

[768]
blk.8.layer_output_norm.weight

F32

F32

[768]

blk.9

blk.9.attn_output.weight

F16

F16

[768, 768]
blk.9.attn_output_norm.bias

F32

F32

[768]
blk.9.attn_output_norm.weight

F32

F32

[768]
blk.9.attn_qkv.weight

F16

F16

[768, 2304]
blk.9.ffn_down.weight

F16

F16

[3072, 768]
blk.9.ffn_gate.weight

F16

F16

[768, 3072]
blk.9.ffn_up.weight

F16

F16

[768, 3072]
blk.9.layer_output_norm.bias

F32

F32

[768]
blk.9.layer_output_norm.weight

F32

F32

[768]

blk.10

blk.10.attn_output.weight

F16

F16

[768, 768]
blk.10.attn_output_norm.bias

F32

F32

[768]
blk.10.attn_output_norm.weight

F32

F32

[768]
blk.10.attn_qkv.weight

F16

F16

[768, 2304]
blk.10.ffn_down.weight

F16

F16

[3072, 768]
blk.10.ffn_gate.weight

F16

F16

[768, 3072]
blk.10.ffn_up.weight

F16

F16

[768, 3072]
blk.10.layer_output_norm.bias

F32

F32

[768]
blk.10.layer_output_norm.weight

F32

F32

[768]

blk.11

blk.11.attn_output.weight

F16

F16

[768, 768]
blk.11.attn_output_norm.bias

F32

F32

[768]
blk.11.attn_output_norm.weight

F32

F32

[768]
blk.11.attn_qkv.weight

F16

F16

[768, 2304]
blk.11.ffn_down.weight

F16

F16

[3072, 768]
blk.11.ffn_gate.weight

F16

F16

[768, 3072]
blk.11.ffn_up.weight

F16

F16

[768, 3072]
blk.11.layer_output_norm.bias

F32

F32

[768]
blk.11.layer_output_norm.weight

F32

F32

[768]

token_embd_norm.bias

F32

F32

[768]
token_embd_norm.weight

F32

F32

[768]
token_types.weight

F32

F32

[768, 2]