jmorgan / gpt2

42 Pulls Updated 2 months ago

gpt2:latest ... /

model

abbfaae58dbb · 328MB

Metadata

general.architecture

gpt2
general.file_type

F16
gpt2.attention.head_count

12
gpt2.attention.layer_norm_epsilon

1e-05
gpt2.block_count

12
gpt2.context_length

1024
gpt2.embedding_length

768
gpt2.feed_forward_length

3072
tokenizer.ggml.bos_token_id

50256
tokenizer.ggml.eos_token_id

50256
tokenizer.ggml.merges

[Ġ t Ġ a h e i n r e ...]
tokenizer.ggml.model

gpt2
tokenizer.ggml.token_type

[1 1 1 1 1 ...]
tokenizer.ggml.tokens

[! " # $ % ...]

Tensors

Name

Type

Shape
position_embd.weight

F16

[768 1024]
token_embd.weight

F16

[768 50257]

blk.0

blk.0.attn_qkv.bias

F32

[2304]
blk.0.attn_qkv.weight

F16

[768 2304]
blk.0.attn_output.bias

F32

[768]
blk.0.attn_output.weight

F16

[768 768]
blk.0.attn_norm.bias

F32

[768]
blk.0.attn_norm.weight

F32

[768]
blk.0.ffn_norm.bias

F32

[768]
blk.0.ffn_norm.weight

F32

[768]
blk.0.ffn_up.bias

F32

[3072]
blk.0.ffn_up.weight

F16

[768 3072]
blk.0.ffn_down.bias

F32

[768]
blk.0.ffn_down.weight

F16

[3072 768]

blk.1

blk.1.attn_qkv.bias

F32

[2304]
blk.1.attn_qkv.weight

F16

[768 2304]
blk.1.attn_output.bias

F32

[768]
blk.1.attn_output.weight

F16

[768 768]
blk.1.attn_norm.bias

F32

[768]
blk.1.attn_norm.weight

F32

[768]
blk.1.ffn_norm.bias

F32

[768]
blk.1.ffn_norm.weight

F32

[768]
blk.1.ffn_up.bias

F32

[3072]
blk.1.ffn_up.weight

F16

[768 3072]
blk.1.ffn_down.bias

F32

[768]
blk.1.ffn_down.weight

F16

[3072 768]

blk.2

blk.2.attn_qkv.bias

F32

[2304]
blk.2.attn_qkv.weight

F16

[768 2304]
blk.2.attn_output.bias

F32

[768]
blk.2.attn_output.weight

F16

[768 768]
blk.2.attn_norm.bias

F32

[768]
blk.2.attn_norm.weight

F32

[768]
blk.2.ffn_norm.bias

F32

[768]
blk.2.ffn_norm.weight

F32

[768]
blk.2.ffn_up.bias

F32

[3072]
blk.2.ffn_up.weight

F16

[768 3072]
blk.2.ffn_down.bias

F32

[768]
blk.2.ffn_down.weight

F16

[3072 768]

blk.3

blk.3.attn_qkv.bias

F32

[2304]
blk.3.attn_qkv.weight

F16

[768 2304]
blk.3.attn_output.bias

F32

[768]
blk.3.attn_output.weight

F16

[768 768]
blk.3.attn_norm.bias

F32

[768]
blk.3.attn_norm.weight

F32

[768]
blk.3.ffn_norm.bias

F32

[768]
blk.3.ffn_norm.weight

F32

[768]
blk.3.ffn_up.bias

F32

[3072]
blk.3.ffn_up.weight

F16

[768 3072]
blk.3.ffn_down.bias

F32

[768]
blk.3.ffn_down.weight

F16

[3072 768]

blk.4

blk.4.attn_qkv.bias

F32

[2304]
blk.4.attn_qkv.weight

F16

[768 2304]
blk.4.attn_output.bias

F32

[768]
blk.4.attn_output.weight

F16

[768 768]
blk.4.attn_norm.bias

F32

[768]
blk.4.attn_norm.weight

F32

[768]
blk.4.ffn_norm.bias

F32

[768]
blk.4.ffn_norm.weight

F32

[768]
blk.4.ffn_up.bias

F32

[3072]
blk.4.ffn_up.weight

F16

[768 3072]
blk.4.ffn_down.bias

F32

[768]
blk.4.ffn_down.weight

F16

[3072 768]

blk.5

blk.5.attn_qkv.bias

F32

[2304]
blk.5.attn_qkv.weight

F16

[768 2304]
blk.5.attn_output.bias

F32

[768]
blk.5.attn_output.weight

F16

[768 768]
blk.5.attn_norm.bias

F32

[768]
blk.5.attn_norm.weight

F32

[768]
blk.5.ffn_norm.bias

F32

[768]
blk.5.ffn_norm.weight

F32

[768]
blk.5.ffn_up.bias

F32

[3072]
blk.5.ffn_up.weight

F16

[768 3072]
blk.5.ffn_down.bias

F32

[768]
blk.5.ffn_down.weight

F16

[3072 768]

blk.6

blk.6.attn_qkv.bias

F32

[2304]
blk.6.attn_qkv.weight

F16

[768 2304]
blk.6.attn_output.bias

F32

[768]
blk.6.attn_output.weight

F16

[768 768]
blk.6.attn_norm.bias

F32

[768]
blk.6.attn_norm.weight

F32

[768]
blk.6.ffn_norm.bias

F32

[768]
blk.6.ffn_norm.weight

F32

[768]
blk.6.ffn_up.bias

F32

[3072]
blk.6.ffn_up.weight

F16

[768 3072]
blk.6.ffn_down.bias

F32

[768]
blk.6.ffn_down.weight

F16

[3072 768]

blk.7

blk.7.attn_qkv.bias

F32

[2304]
blk.7.attn_qkv.weight

F16

[768 2304]
blk.7.attn_output.bias

F32

[768]
blk.7.attn_output.weight

F16

[768 768]
blk.7.attn_norm.bias

F32

[768]
blk.7.attn_norm.weight

F32

[768]
blk.7.ffn_norm.bias

F32

[768]
blk.7.ffn_norm.weight

F32

[768]
blk.7.ffn_up.bias

F32

[3072]
blk.7.ffn_up.weight

F16

[768 3072]
blk.7.ffn_down.bias

F32

[768]
blk.7.ffn_down.weight

F16

[3072 768]

blk.8

blk.8.attn_qkv.bias

F32

[2304]
blk.8.attn_qkv.weight

F16

[768 2304]
blk.8.attn_output.bias

F32

[768]
blk.8.attn_output.weight

F16

[768 768]
blk.8.attn_norm.bias

F32

[768]
blk.8.attn_norm.weight

F32

[768]
blk.8.ffn_norm.bias

F32

[768]
blk.8.ffn_norm.weight

F32

[768]
blk.8.ffn_up.bias

F32

[3072]
blk.8.ffn_up.weight

F16

[768 3072]
blk.8.ffn_down.bias

F32

[768]
blk.8.ffn_down.weight

F16

[3072 768]

blk.9

blk.9.attn_qkv.bias

F32

[2304]
blk.9.attn_qkv.weight

F16

[768 2304]
blk.9.attn_output.bias

F32

[768]
blk.9.attn_output.weight

F16

[768 768]
blk.9.attn_norm.bias

F32

[768]
blk.9.attn_norm.weight

F32

[768]
blk.9.ffn_norm.bias

F32

[768]
blk.9.ffn_norm.weight

F32

[768]
blk.9.ffn_up.bias

F32

[3072]
blk.9.ffn_up.weight

F16

[768 3072]
blk.9.ffn_down.bias

F32

[768]
blk.9.ffn_down.weight

F16

[3072 768]

blk.10

blk.10.attn_qkv.bias

F32

[2304]
blk.10.attn_qkv.weight

F16

[768 2304]
blk.10.attn_output.bias

F32

[768]
blk.10.attn_output.weight

F16

[768 768]
blk.10.attn_norm.bias

F32

[768]
blk.10.attn_norm.weight

F32

[768]
blk.10.ffn_norm.bias

F32

[768]
blk.10.ffn_norm.weight

F32

[768]
blk.10.ffn_up.bias

F32

[3072]
blk.10.ffn_up.weight

F16

[768 3072]
blk.10.ffn_down.bias

F32

[768]
blk.10.ffn_down.weight

F16

[3072 768]

blk.11

blk.11.attn_qkv.bias

F32

[2304]
blk.11.attn_qkv.weight

F16

[768 2304]
blk.11.attn_output.bias

F32

[768]
blk.11.attn_output.weight

F16

[768 768]
blk.11.attn_norm.bias

F32

[768]
blk.11.attn_norm.weight

F32

[768]
blk.11.ffn_norm.bias

F32

[768]
blk.11.ffn_norm.weight

F32

[768]
blk.11.ffn_up.bias

F32

[3072]
blk.11.ffn_up.weight

F16

[768 3072]
blk.11.ffn_down.bias

F32

[768]
blk.11.ffn_down.weight

F16

[3072 768]

output_norm.bias

F32

[768]
output_norm.weight

F32

[768]
output.weight

F16

[768 50257]