Hudson/pythia-instruct:70m/model

Hudson/

pythia-instruct:70m

70 Downloads Updated 11 months ago

70m

pythia-instruct:70m ... /

model

f719f9ed21c3 · 143MB

Metadata

general.architecture

gptneox

gptneox
gptneox.attention.head_count

8

8
gptneox.attention.layer_norm_epsilon

1e-05

1e-05
gptneox.block_count

6

6
gptneox.context_length

2048

2048
gptneox.embedding_length

512

512
gptneox.feed_forward_length

2048

2048
gptneox.rope.dimension_count

16

16
gptneox.use_parallel_residual

true

true
tokenizer.ggml.bos_token_id

0

0
tokenizer.ggml.eos_token_id

0

0
tokenizer.ggml.merges

[Ġ Ġ, Ġ t, Ġ a, h e, i n, ...]

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

gpt2

gpt2
tokenizer.ggml.pre

olmo

olmo
tokenizer.ggml.token_type

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

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

[<|endoftext|>, <|padding|>, !, ", #, ...]

[<|endoftext|>, <|padding|>, !, ", #, ...]
tokenizer.ggml.unknown_token_id

0

0

Tensor

Name

Type

Shape
token_embd.weight

F16

F16

[512, 50279]

blk.0

blk.0.attn_norm.bias

F32

F32

[512]
blk.0.attn_norm.weight

F32

F32

[512]
blk.0.attn_output.bias

F32

F32

[512]
blk.0.attn_output.weight

F16

F16

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

F32

F32

[1536]
blk.0.attn_qkv.weight

F16

F16

[512, 1536]
blk.0.ffn_down.bias

F32

F32

[512]
blk.0.ffn_down.weight

F16

F16

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

F32

F32

[512]
blk.0.ffn_norm.weight

F32

F32

[512]
blk.0.ffn_up.bias

F32

F32

[2048]
blk.0.ffn_up.weight

F16

F16

[512, 2048]

blk.1

blk.1.attn_norm.bias

F32

F32

[512]
blk.1.attn_norm.weight

F32

F32

[512]
blk.1.attn_output.bias

F32

F32

[512]
blk.1.attn_output.weight

F16

F16

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

F32

F32

[1536]
blk.1.attn_qkv.weight

F16

F16

[512, 1536]
blk.1.ffn_down.bias

F32

F32

[512]
blk.1.ffn_down.weight

F16

F16

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

F32

F32

[512]
blk.1.ffn_norm.weight

F32

F32

[512]
blk.1.ffn_up.bias

F32

F32

[2048]
blk.1.ffn_up.weight

F16

F16

[512, 2048]

blk.2

blk.2.attn_norm.bias

F32

F32

[512]
blk.2.attn_norm.weight

F32

F32

[512]
blk.2.attn_output.bias

F32

F32

[512]
blk.2.attn_output.weight

F16

F16

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

F32

F32

[1536]
blk.2.attn_qkv.weight

F16

F16

[512, 1536]
blk.2.ffn_down.bias

F32

F32

[512]
blk.2.ffn_down.weight

F16

F16

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

F32

F32

[512]
blk.2.ffn_norm.weight

F32

F32

[512]
blk.2.ffn_up.bias

F32

F32

[2048]
blk.2.ffn_up.weight

F16

F16

[512, 2048]

blk.3

blk.3.attn_norm.bias

F32

F32

[512]
blk.3.attn_norm.weight

F32

F32

[512]
blk.3.attn_output.bias

F32

F32

[512]
blk.3.attn_output.weight

F16

F16

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

F32

F32

[1536]
blk.3.attn_qkv.weight

F16

F16

[512, 1536]
blk.3.ffn_down.bias

F32

F32

[512]
blk.3.ffn_down.weight

F16

F16

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

F32

F32

[512]
blk.3.ffn_norm.weight

F32

F32

[512]
blk.3.ffn_up.bias

F32

F32

[2048]
blk.3.ffn_up.weight

F16

F16

[512, 2048]

blk.4

blk.4.attn_norm.bias

F32

F32

[512]
blk.4.attn_norm.weight

F32

F32

[512]
blk.4.attn_output.bias

F32

F32

[512]
blk.4.attn_output.weight

F16

F16

[512, 512]
blk.4.attn_qkv.bias

F32

F32

[1536]
blk.4.attn_qkv.weight

F16

F16

[512, 1536]
blk.4.ffn_down.bias

F32

F32

[512]
blk.4.ffn_down.weight

F16

F16

[2048, 512]
blk.4.ffn_norm.bias

F32

F32

[512]
blk.4.ffn_norm.weight

F32

F32

[512]
blk.4.ffn_up.bias

F32

F32

[2048]
blk.4.ffn_up.weight

F16

F16

[512, 2048]

blk.5

blk.5.attn_norm.bias

F32

F32

[512]
blk.5.attn_norm.weight

F32

F32

[512]
blk.5.attn_output.bias

F32

F32

[512]
blk.5.attn_output.weight

F16

F16

[512, 512]
blk.5.attn_qkv.bias

F32

F32

[1536]
blk.5.attn_qkv.weight

F16

F16

[512, 1536]
blk.5.ffn_down.bias

F32

F32

[512]
blk.5.ffn_down.weight

F16

F16

[2048, 512]
blk.5.ffn_norm.bias

F32

F32

[512]
blk.5.ffn_norm.weight

F32

F32

[512]
blk.5.ffn_up.bias

F32

F32

[2048]
blk.5.ffn_up.weight

F16

F16

[512, 2048]

output.weight

F16

F16

[512, 50279]
output_norm.bias

F32

F32

[512]
output_norm.weight

F32

F32

[512]