sovit123 / starcoder2-test

A small LM for coding

3 Pulls Updated 3 months ago

1 Tag

starcoder2-test ... /

model

7a6eb17080bd · 6.1GB

Metadata

general.architecture

starcoder2
general.basename

starcoder2
general.file_type

1
general.name

Starcoder2 3b
general.organization

Bigcode
general.quantization_version

2
general.size_label

3B
general.type

model
starcoder2.attention.head_count

24
starcoder2.attention.head_count_kv

2
starcoder2.attention.layer_norm_epsilon

1e-05
starcoder2.block_count

30
starcoder2.context_length

16384
starcoder2.embedding_length

3072
starcoder2.feed_forward_length

12288
starcoder2.rope.freq_base

999999.44
tokenizer.ggml.bos_token_id

0
tokenizer.ggml.eos_token_id

0
tokenizer.ggml.merges

[Ġ Ġ, ĠĠ ĠĠ, ĠĠĠĠ ĠĠĠĠ, ĠĠ Ġ, e r, ...]
tokenizer.ggml.model

gpt2
tokenizer.ggml.pre

starcoder
tokenizer.ggml.token_type

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

[<|endoftext|>, <fim_prefix>, <fim_middle>, <fim_suffix>, <fim_pad>, ...]
tokenizer.ggml.unknown_token_id

0

Tensor

Name

Type

Shape
token_embd.weight

F16

[3072, 49152]

blk.0

blk.0.attn_k.bias

F32

[256]
blk.0.attn_k.weight

F16

[3072, 256]
blk.0.attn_norm.bias

F32

[3072]
blk.0.attn_norm.weight

F32

[3072]
blk.0.attn_output.bias

F32

[3072]
blk.0.attn_output.weight

F16

[3072, 3072]
blk.0.attn_q.bias

F32

[3072]
blk.0.attn_q.weight

F16

[3072, 3072]
blk.0.attn_v.bias

F32

[256]
blk.0.attn_v.weight

F16

[3072, 256]
blk.0.ffn_down.bias

F32

[3072]
blk.0.ffn_down.weight

F16

[12288, 3072]
blk.0.ffn_norm.bias

F32

[3072]
blk.0.ffn_norm.weight

F32

[3072]
blk.0.ffn_up.bias

F32

[12288]
blk.0.ffn_up.weight

F16

[3072, 12288]

blk.1

blk.1.attn_k.bias

F32

[256]
blk.1.attn_k.weight

F16

[3072, 256]
blk.1.attn_norm.bias

F32

[3072]
blk.1.attn_norm.weight

F32

[3072]
blk.1.attn_output.bias

F32

[3072]
blk.1.attn_output.weight

F16

[3072, 3072]
blk.1.attn_q.bias

F32

[3072]
blk.1.attn_q.weight

F16

[3072, 3072]
blk.1.attn_v.bias

F32

[256]
blk.1.attn_v.weight

F16

[3072, 256]
blk.1.ffn_down.bias

F32

[3072]
blk.1.ffn_down.weight

F16

[12288, 3072]
blk.1.ffn_norm.bias

F32

[3072]
blk.1.ffn_norm.weight

F32

[3072]
blk.1.ffn_up.bias

F32

[12288]
blk.1.ffn_up.weight

F16

[3072, 12288]

blk.2

blk.2.attn_k.bias

F32

[256]
blk.2.attn_k.weight

F16

[3072, 256]
blk.2.attn_norm.bias

F32

[3072]
blk.2.attn_norm.weight

F32

[3072]
blk.2.attn_output.bias

F32

[3072]
blk.2.attn_output.weight

F16

[3072, 3072]
blk.2.attn_q.bias

F32

[3072]
blk.2.attn_q.weight

F16

[3072, 3072]
blk.2.attn_v.bias

F32

[256]
blk.2.attn_v.weight

F16

[3072, 256]
blk.2.ffn_down.bias

F32

[3072]
blk.2.ffn_down.weight

F16

[12288, 3072]
blk.2.ffn_norm.bias

F32

[3072]
blk.2.ffn_norm.weight

F32

[3072]
blk.2.ffn_up.bias

F32

[12288]
blk.2.ffn_up.weight

F16

[3072, 12288]

blk.3

blk.3.attn_k.bias

F32

[256]
blk.3.attn_k.weight

F16

[3072, 256]
blk.3.attn_norm.bias

F32

[3072]
blk.3.attn_norm.weight

F32

[3072]
blk.3.attn_output.bias

F32

[3072]
blk.3.attn_output.weight

F16

[3072, 3072]
blk.3.attn_q.bias

F32

[3072]
blk.3.attn_q.weight

F16

[3072, 3072]
blk.3.attn_v.bias

F32

[256]
blk.3.attn_v.weight

F16

[3072, 256]
blk.3.ffn_down.bias

F32

[3072]
blk.3.ffn_down.weight

F16

[12288, 3072]
blk.3.ffn_norm.bias

F32

[3072]
blk.3.ffn_norm.weight

F32

[3072]
blk.3.ffn_up.bias

F32

[12288]
blk.3.ffn_up.weight

F16

[3072, 12288]

blk.4

blk.4.attn_k.bias

F32

[256]
blk.4.attn_k.weight

F16

[3072, 256]
blk.4.attn_norm.bias

F32

[3072]
blk.4.attn_norm.weight

F32

[3072]
blk.4.attn_output.bias

F32

[3072]
blk.4.attn_output.weight

F16

[3072, 3072]
blk.4.attn_q.bias

F32

[3072]
blk.4.attn_q.weight

F16

[3072, 3072]
blk.4.attn_v.bias

F32

[256]
blk.4.attn_v.weight

F16

[3072, 256]
blk.4.ffn_down.bias

F32

[3072]
blk.4.ffn_down.weight

F16

[12288, 3072]
blk.4.ffn_norm.bias

F32

[3072]
blk.4.ffn_norm.weight

F32

[3072]
blk.4.ffn_up.bias

F32

[12288]
blk.4.ffn_up.weight

F16

[3072, 12288]

blk.5

blk.5.attn_k.bias

F32

[256]
blk.5.attn_k.weight

F16

[3072, 256]
blk.5.attn_norm.bias

F32

[3072]
blk.5.attn_norm.weight

F32

[3072]
blk.5.attn_output.bias

F32

[3072]
blk.5.attn_output.weight

F16

[3072, 3072]
blk.5.attn_q.bias

F32

[3072]
blk.5.attn_q.weight

F16

[3072, 3072]
blk.5.attn_v.bias

F32

[256]
blk.5.attn_v.weight

F16

[3072, 256]
blk.5.ffn_down.bias

F32

[3072]
blk.5.ffn_down.weight

F16

[12288, 3072]
blk.5.ffn_norm.bias

F32

[3072]
blk.5.ffn_norm.weight

F32

[3072]
blk.5.ffn_up.bias

F32

[12288]
blk.5.ffn_up.weight

F16

[3072, 12288]

blk.6

blk.6.attn_k.bias

F32

[256]
blk.6.attn_k.weight

F16

[3072, 256]
blk.6.attn_norm.bias

F32

[3072]
blk.6.attn_norm.weight

F32

[3072]
blk.6.attn_output.bias

F32

[3072]
blk.6.attn_output.weight

F16

[3072, 3072]
blk.6.attn_q.bias

F32

[3072]
blk.6.attn_q.weight

F16

[3072, 3072]
blk.6.attn_v.bias

F32

[256]
blk.6.attn_v.weight

F16

[3072, 256]
blk.6.ffn_down.bias

F32

[3072]
blk.6.ffn_down.weight

F16

[12288, 3072]
blk.6.ffn_norm.bias

F32

[3072]
blk.6.ffn_norm.weight

F32

[3072]
blk.6.ffn_up.bias

F32

[12288]
blk.6.ffn_up.weight

F16

[3072, 12288]

blk.7

blk.7.attn_k.bias

F32

[256]
blk.7.attn_k.weight

F16

[3072, 256]
blk.7.attn_norm.bias

F32

[3072]
blk.7.attn_norm.weight

F32

[3072]
blk.7.attn_output.bias

F32

[3072]
blk.7.attn_output.weight

F16

[3072, 3072]
blk.7.attn_q.bias

F32

[3072]
blk.7.attn_q.weight

F16

[3072, 3072]
blk.7.attn_v.bias

F32

[256]
blk.7.attn_v.weight

F16

[3072, 256]
blk.7.ffn_down.bias

F32

[3072]
blk.7.ffn_down.weight

F16

[12288, 3072]
blk.7.ffn_norm.bias

F32

[3072]
blk.7.ffn_norm.weight

F32

[3072]
blk.7.ffn_up.bias

F32

[12288]
blk.7.ffn_up.weight

F16

[3072, 12288]

blk.8

blk.8.attn_k.bias

F32

[256]
blk.8.attn_k.weight

F16

[3072, 256]
blk.8.attn_norm.bias

F32

[3072]
blk.8.attn_norm.weight

F32

[3072]
blk.8.attn_output.bias

F32

[3072]
blk.8.attn_output.weight

F16

[3072, 3072]
blk.8.attn_q.bias

F32

[3072]
blk.8.attn_q.weight

F16

[3072, 3072]
blk.8.attn_v.bias

F32

[256]
blk.8.attn_v.weight

F16

[3072, 256]
blk.8.ffn_down.bias

F32

[3072]
blk.8.ffn_down.weight

F16

[12288, 3072]
blk.8.ffn_norm.bias

F32

[3072]
blk.8.ffn_norm.weight

F32

[3072]
blk.8.ffn_up.bias

F32

[12288]
blk.8.ffn_up.weight

F16

[3072, 12288]

blk.9

blk.9.attn_k.bias

F32

[256]
blk.9.attn_k.weight

F16

[3072, 256]
blk.9.attn_norm.bias

F32

[3072]
blk.9.attn_norm.weight

F32

[3072]
blk.9.attn_output.bias

F32

[3072]
blk.9.attn_output.weight

F16

[3072, 3072]
blk.9.attn_q.bias

F32

[3072]
blk.9.attn_q.weight

F16

[3072, 3072]
blk.9.attn_v.bias

F32

[256]
blk.9.attn_v.weight

F16

[3072, 256]
blk.9.ffn_down.bias

F32

[3072]
blk.9.ffn_down.weight

F16

[12288, 3072]
blk.9.ffn_norm.bias

F32

[3072]
blk.9.ffn_norm.weight

F32

[3072]
blk.9.ffn_up.bias

F32

[12288]
blk.9.ffn_up.weight

F16

[3072, 12288]

blk.10

blk.10.attn_k.bias

F32

[256]
blk.10.attn_k.weight

F16

[3072, 256]
blk.10.attn_norm.bias

F32

[3072]
blk.10.attn_norm.weight

F32

[3072]
blk.10.attn_output.bias

F32

[3072]
blk.10.attn_output.weight

F16

[3072, 3072]
blk.10.attn_q.bias

F32

[3072]
blk.10.attn_q.weight

F16

[3072, 3072]
blk.10.attn_v.bias

F32

[256]
blk.10.attn_v.weight

F16

[3072, 256]
blk.10.ffn_down.bias

F32

[3072]
blk.10.ffn_down.weight

F16

[12288, 3072]
blk.10.ffn_norm.bias

F32

[3072]
blk.10.ffn_norm.weight

F32

[3072]
blk.10.ffn_up.bias

F32

[12288]
blk.10.ffn_up.weight

F16

[3072, 12288]

blk.11

blk.11.attn_k.bias

F32

[256]
blk.11.attn_k.weight

F16

[3072, 256]
blk.11.attn_norm.bias

F32

[3072]
blk.11.attn_norm.weight

F32

[3072]
blk.11.attn_output.bias

F32

[3072]
blk.11.attn_output.weight

F16

[3072, 3072]
blk.11.attn_q.bias

F32

[3072]
blk.11.attn_q.weight

F16

[3072, 3072]
blk.11.attn_v.bias

F32

[256]
blk.11.attn_v.weight

F16

[3072, 256]
blk.11.ffn_down.bias

F32

[3072]
blk.11.ffn_down.weight

F16

[12288, 3072]
blk.11.ffn_norm.bias

F32

[3072]
blk.11.ffn_norm.weight

F32

[3072]
blk.11.ffn_up.bias

F32

[12288]
blk.11.ffn_up.weight

F16

[3072, 12288]

blk.12

blk.12.attn_k.bias

F32

[256]
blk.12.attn_k.weight

F16

[3072, 256]
blk.12.attn_norm.bias

F32

[3072]
blk.12.attn_norm.weight

F32

[3072]
blk.12.attn_output.bias

F32

[3072]
blk.12.attn_output.weight

F16

[3072, 3072]
blk.12.attn_q.bias

F32

[3072]
blk.12.attn_q.weight

F16

[3072, 3072]
blk.12.attn_v.bias

F32

[256]
blk.12.attn_v.weight

F16

[3072, 256]
blk.12.ffn_down.bias

F32

[3072]
blk.12.ffn_down.weight

F16

[12288, 3072]
blk.12.ffn_norm.bias

F32

[3072]
blk.12.ffn_norm.weight

F32

[3072]
blk.12.ffn_up.bias

F32

[12288]
blk.12.ffn_up.weight

F16

[3072, 12288]

blk.13

blk.13.attn_k.bias

F32

[256]
blk.13.attn_k.weight

F16

[3072, 256]
blk.13.attn_norm.bias

F32

[3072]
blk.13.attn_norm.weight

F32

[3072]
blk.13.attn_output.bias

F32

[3072]
blk.13.attn_output.weight

F16

[3072, 3072]
blk.13.attn_q.bias

F32

[3072]
blk.13.attn_q.weight

F16

[3072, 3072]
blk.13.attn_v.bias

F32

[256]
blk.13.attn_v.weight

F16

[3072, 256]
blk.13.ffn_down.bias

F32

[3072]
blk.13.ffn_down.weight

F16

[12288, 3072]
blk.13.ffn_norm.bias

F32

[3072]
blk.13.ffn_norm.weight

F32

[3072]
blk.13.ffn_up.bias

F32

[12288]
blk.13.ffn_up.weight

F16

[3072, 12288]

blk.14

blk.14.attn_k.bias

F32

[256]
blk.14.attn_k.weight

F16

[3072, 256]
blk.14.attn_norm.bias

F32

[3072]
blk.14.attn_norm.weight

F32

[3072]
blk.14.attn_output.bias

F32

[3072]
blk.14.attn_output.weight

F16

[3072, 3072]
blk.14.attn_q.bias

F32

[3072]
blk.14.attn_q.weight

F16

[3072, 3072]
blk.14.attn_v.bias

F32

[256]
blk.14.attn_v.weight

F16

[3072, 256]
blk.14.ffn_down.bias

F32

[3072]
blk.14.ffn_down.weight

F16

[12288, 3072]
blk.14.ffn_norm.bias

F32

[3072]
blk.14.ffn_norm.weight

F32

[3072]
blk.14.ffn_up.bias

F32

[12288]
blk.14.ffn_up.weight

F16

[3072, 12288]

blk.15

blk.15.attn_k.bias

F32

[256]
blk.15.attn_k.weight

F16

[3072, 256]
blk.15.attn_norm.bias

F32

[3072]
blk.15.attn_norm.weight

F32

[3072]
blk.15.attn_output.bias

F32

[3072]
blk.15.attn_output.weight

F16

[3072, 3072]
blk.15.attn_q.bias

F32

[3072]
blk.15.attn_q.weight

F16

[3072, 3072]
blk.15.attn_v.bias

F32

[256]
blk.15.attn_v.weight

F16

[3072, 256]
blk.15.ffn_down.bias

F32

[3072]
blk.15.ffn_down.weight

F16

[12288, 3072]
blk.15.ffn_norm.bias

F32

[3072]
blk.15.ffn_norm.weight

F32

[3072]
blk.15.ffn_up.bias

F32

[12288]
blk.15.ffn_up.weight

F16

[3072, 12288]

blk.16

blk.16.attn_k.bias

F32

[256]
blk.16.attn_k.weight

F16

[3072, 256]
blk.16.attn_norm.bias

F32

[3072]
blk.16.attn_norm.weight

F32

[3072]
blk.16.attn_output.bias

F32

[3072]
blk.16.attn_output.weight

F16

[3072, 3072]
blk.16.attn_q.bias

F32

[3072]
blk.16.attn_q.weight

F16

[3072, 3072]
blk.16.attn_v.bias

F32

[256]
blk.16.attn_v.weight

F16

[3072, 256]
blk.16.ffn_down.bias

F32

[3072]
blk.16.ffn_down.weight

F16

[12288, 3072]
blk.16.ffn_norm.bias

F32

[3072]
blk.16.ffn_norm.weight

F32

[3072]
blk.16.ffn_up.bias

F32

[12288]
blk.16.ffn_up.weight

F16

[3072, 12288]

blk.17

blk.17.attn_k.bias

F32

[256]
blk.17.attn_k.weight

F16

[3072, 256]
blk.17.attn_norm.bias

F32

[3072]
blk.17.attn_norm.weight

F32

[3072]
blk.17.attn_output.bias

F32

[3072]
blk.17.attn_output.weight

F16

[3072, 3072]
blk.17.attn_q.bias

F32

[3072]
blk.17.attn_q.weight

F16

[3072, 3072]
blk.17.attn_v.bias

F32

[256]
blk.17.attn_v.weight

F16

[3072, 256]
blk.17.ffn_down.bias

F32

[3072]
blk.17.ffn_down.weight

F16

[12288, 3072]
blk.17.ffn_norm.bias

F32

[3072]
blk.17.ffn_norm.weight

F32

[3072]
blk.17.ffn_up.bias

F32

[12288]
blk.17.ffn_up.weight

F16

[3072, 12288]

blk.18

blk.18.attn_k.bias

F32

[256]
blk.18.attn_k.weight

F16

[3072, 256]
blk.18.attn_norm.bias

F32

[3072]
blk.18.attn_norm.weight

F32

[3072]
blk.18.attn_output.bias

F32

[3072]
blk.18.attn_output.weight

F16

[3072, 3072]
blk.18.attn_q.bias

F32

[3072]
blk.18.attn_q.weight

F16

[3072, 3072]
blk.18.attn_v.bias

F32

[256]
blk.18.attn_v.weight

F16

[3072, 256]
blk.18.ffn_down.bias

F32

[3072]
blk.18.ffn_down.weight

F16

[12288, 3072]
blk.18.ffn_norm.bias

F32

[3072]
blk.18.ffn_norm.weight

F32

[3072]
blk.18.ffn_up.bias

F32

[12288]
blk.18.ffn_up.weight

F16

[3072, 12288]

blk.19

blk.19.attn_k.bias

F32

[256]
blk.19.attn_k.weight

F16

[3072, 256]
blk.19.attn_norm.bias

F32

[3072]
blk.19.attn_norm.weight

F32

[3072]
blk.19.attn_output.bias

F32

[3072]
blk.19.attn_output.weight

F16

[3072, 3072]
blk.19.attn_q.bias

F32

[3072]
blk.19.attn_q.weight

F16

[3072, 3072]
blk.19.attn_v.bias

F32

[256]
blk.19.attn_v.weight

F16

[3072, 256]
blk.19.ffn_down.bias

F32

[3072]
blk.19.ffn_down.weight

F16

[12288, 3072]
blk.19.ffn_norm.bias

F32

[3072]
blk.19.ffn_norm.weight

F32

[3072]
blk.19.ffn_up.bias

F32

[12288]
blk.19.ffn_up.weight

F16

[3072, 12288]

blk.20

blk.20.attn_k.bias

F32

[256]
blk.20.attn_k.weight

F16

[3072, 256]
blk.20.attn_norm.bias

F32

[3072]
blk.20.attn_norm.weight

F32

[3072]
blk.20.attn_output.bias

F32

[3072]
blk.20.attn_output.weight

F16

[3072, 3072]
blk.20.attn_q.bias

F32

[3072]
blk.20.attn_q.weight

F16

[3072, 3072]
blk.20.attn_v.bias

F32

[256]
blk.20.attn_v.weight

F16

[3072, 256]
blk.20.ffn_down.bias

F32

[3072]
blk.20.ffn_down.weight

F16

[12288, 3072]
blk.20.ffn_norm.bias

F32

[3072]
blk.20.ffn_norm.weight

F32

[3072]
blk.20.ffn_up.bias

F32

[12288]
blk.20.ffn_up.weight

F16

[3072, 12288]

blk.21

blk.21.attn_k.bias

F32

[256]
blk.21.attn_k.weight

F16

[3072, 256]
blk.21.attn_norm.bias

F32

[3072]
blk.21.attn_norm.weight

F32

[3072]
blk.21.attn_output.bias

F32

[3072]
blk.21.attn_output.weight

F16

[3072, 3072]
blk.21.attn_q.bias

F32

[3072]
blk.21.attn_q.weight

F16

[3072, 3072]
blk.21.attn_v.bias

F32

[256]
blk.21.attn_v.weight

F16

[3072, 256]
blk.21.ffn_down.bias

F32

[3072]
blk.21.ffn_down.weight

F16

[12288, 3072]
blk.21.ffn_norm.bias

F32

[3072]
blk.21.ffn_norm.weight

F32

[3072]
blk.21.ffn_up.bias

F32

[12288]
blk.21.ffn_up.weight

F16

[3072, 12288]

blk.22

blk.22.attn_k.bias

F32

[256]
blk.22.attn_k.weight

F16

[3072, 256]
blk.22.attn_norm.bias

F32

[3072]
blk.22.attn_norm.weight

F32

[3072]
blk.22.attn_output.bias

F32

[3072]
blk.22.attn_output.weight

F16

[3072, 3072]
blk.22.attn_q.bias

F32

[3072]
blk.22.attn_q.weight

F16

[3072, 3072]
blk.22.attn_v.bias

F32

[256]
blk.22.attn_v.weight

F16

[3072, 256]
blk.22.ffn_down.bias

F32

[3072]
blk.22.ffn_down.weight

F16

[12288, 3072]
blk.22.ffn_norm.bias

F32

[3072]
blk.22.ffn_norm.weight

F32

[3072]
blk.22.ffn_up.bias

F32

[12288]
blk.22.ffn_up.weight

F16

[3072, 12288]

blk.23

blk.23.attn_k.bias

F32

[256]
blk.23.attn_k.weight

F16

[3072, 256]
blk.23.attn_norm.bias

F32

[3072]
blk.23.attn_norm.weight

F32

[3072]
blk.23.attn_output.bias

F32

[3072]
blk.23.attn_output.weight

F16

[3072, 3072]
blk.23.attn_q.bias

F32

[3072]
blk.23.attn_q.weight

F16

[3072, 3072]
blk.23.attn_v.bias

F32

[256]
blk.23.attn_v.weight

F16

[3072, 256]
blk.23.ffn_down.bias

F32

[3072]
blk.23.ffn_down.weight

F16

[12288, 3072]
blk.23.ffn_norm.bias

F32

[3072]
blk.23.ffn_norm.weight

F32

[3072]
blk.23.ffn_up.bias

F32

[12288]
blk.23.ffn_up.weight

F16

[3072, 12288]

blk.24

blk.24.attn_k.bias

F32

[256]
blk.24.attn_k.weight

F16

[3072, 256]
blk.24.attn_norm.bias

F32

[3072]
blk.24.attn_norm.weight

F32

[3072]
blk.24.attn_output.bias

F32

[3072]
blk.24.attn_output.weight

F16

[3072, 3072]
blk.24.attn_q.bias

F32

[3072]
blk.24.attn_q.weight

F16

[3072, 3072]
blk.24.attn_v.bias

F32

[256]
blk.24.attn_v.weight

F16

[3072, 256]
blk.24.ffn_down.bias

F32

[3072]
blk.24.ffn_down.weight

F16

[12288, 3072]
blk.24.ffn_norm.bias

F32

[3072]
blk.24.ffn_norm.weight

F32

[3072]
blk.24.ffn_up.bias

F32

[12288]
blk.24.ffn_up.weight

F16

[3072, 12288]

blk.25

blk.25.attn_k.bias

F32

[256]
blk.25.attn_k.weight

F16

[3072, 256]
blk.25.attn_norm.bias

F32

[3072]
blk.25.attn_norm.weight

F32

[3072]
blk.25.attn_output.bias

F32

[3072]
blk.25.attn_output.weight

F16

[3072, 3072]
blk.25.attn_q.bias

F32

[3072]
blk.25.attn_q.weight

F16

[3072, 3072]
blk.25.attn_v.bias

F32

[256]
blk.25.attn_v.weight

F16

[3072, 256]
blk.25.ffn_down.bias

F32

[3072]
blk.25.ffn_down.weight

F16

[12288, 3072]
blk.25.ffn_norm.bias

F32

[3072]
blk.25.ffn_norm.weight

F32

[3072]
blk.25.ffn_up.bias

F32

[12288]
blk.25.ffn_up.weight

F16

[3072, 12288]

blk.26

blk.26.attn_k.bias

F32

[256]
blk.26.attn_k.weight

F16

[3072, 256]
blk.26.attn_norm.bias

F32

[3072]
blk.26.attn_norm.weight

F32

[3072]
blk.26.attn_output.bias

F32

[3072]
blk.26.attn_output.weight

F16

[3072, 3072]
blk.26.attn_q.bias

F32

[3072]
blk.26.attn_q.weight

F16

[3072, 3072]
blk.26.attn_v.bias

F32

[256]
blk.26.attn_v.weight

F16

[3072, 256]
blk.26.ffn_down.bias

F32

[3072]
blk.26.ffn_down.weight

F16

[12288, 3072]
blk.26.ffn_norm.bias

F32

[3072]
blk.26.ffn_norm.weight

F32

[3072]
blk.26.ffn_up.bias

F32

[12288]
blk.26.ffn_up.weight

F16

[3072, 12288]

blk.27

blk.27.attn_k.bias

F32

[256]
blk.27.attn_k.weight

F16

[3072, 256]
blk.27.attn_norm.bias

F32

[3072]
blk.27.attn_norm.weight

F32

[3072]
blk.27.attn_output.bias

F32

[3072]
blk.27.attn_output.weight

F16

[3072, 3072]
blk.27.attn_q.bias

F32

[3072]
blk.27.attn_q.weight

F16

[3072, 3072]
blk.27.attn_v.bias

F32

[256]
blk.27.attn_v.weight

F16

[3072, 256]
blk.27.ffn_down.bias

F32

[3072]
blk.27.ffn_down.weight

F16

[12288, 3072]
blk.27.ffn_norm.bias

F32

[3072]
blk.27.ffn_norm.weight

F32

[3072]
blk.27.ffn_up.bias

F32

[12288]
blk.27.ffn_up.weight

F16

[3072, 12288]

blk.28

blk.28.attn_k.bias

F32

[256]
blk.28.attn_k.weight

F16

[3072, 256]
blk.28.attn_norm.bias

F32

[3072]
blk.28.attn_norm.weight

F32

[3072]
blk.28.attn_output.bias

F32

[3072]
blk.28.attn_output.weight

F16

[3072, 3072]
blk.28.attn_q.bias

F32

[3072]
blk.28.attn_q.weight

F16

[3072, 3072]
blk.28.attn_v.bias

F32

[256]
blk.28.attn_v.weight

F16

[3072, 256]
blk.28.ffn_down.bias

F32

[3072]
blk.28.ffn_down.weight

F16

[12288, 3072]
blk.28.ffn_norm.bias

F32

[3072]
blk.28.ffn_norm.weight

F32

[3072]
blk.28.ffn_up.bias

F32

[12288]
blk.28.ffn_up.weight

F16

[3072, 12288]

blk.29

blk.29.attn_k.bias

F32

[256]
blk.29.attn_k.weight

F16

[3072, 256]
blk.29.attn_norm.bias

F32

[3072]
blk.29.attn_norm.weight

F32

[3072]
blk.29.attn_output.bias

F32

[3072]
blk.29.attn_output.weight

F16

[3072, 3072]
blk.29.attn_q.bias

F32

[3072]
blk.29.attn_q.weight

F16

[3072, 3072]
blk.29.attn_v.bias

F32

[256]
blk.29.attn_v.weight

F16

[3072, 256]
blk.29.ffn_down.bias

F32

[3072]
blk.29.ffn_down.weight

F16

[12288, 3072]
blk.29.ffn_norm.bias

F32

[3072]
blk.29.ffn_norm.weight

F32

[3072]
blk.29.ffn_up.bias

F32

[12288]
blk.29.ffn_up.weight

F16

[3072, 12288]

output_norm.bias

F32

[3072]
output_norm.weight

F32

[3072]