olmo2:13b-1124-instruct-fp16/model

olmo2

OLMo 2 is a new family of 7B and 13B models trained on up to 5T tokens. These models are on par with or better than equivalently sized fully open models, and competitive with open-weight models such as Llama 3.1 on English academic benchmarks.

7b 13b

4,364 Pulls Updated 6 days ago

9 Tags

olmo2:13b-1124-instruct-fp16 ... /

model

dc55b66a2732 · 27GB

Metadata

general.architecture

olmo2
general.base_model.0.name

OLMo 2 1124 13B Instruct RLVR2
general.base_model.0.organization

Allenai
general.base_model.0.repo_url

https://huggingface.co/allenai/OLMo-2-1124-13B-Instruct-RLVR2
general.base_model.count

1
general.basename

OLMo-2-1124
general.dataset.0.name

RLVR MATH
general.dataset.0.organization

Allenai
general.dataset.0.repo_url

https://huggingface.co/allenai/RLVR-MATH
general.dataset.count

1
general.file_type

1
general.finetune

Instruct
general.languages

[en]
general.license

apache-2.0
general.name

OLMo 2 1124 13B Instruct
general.quantization_version

2
general.size_label

13B
general.tags

[text-generation]
general.type

model
olmo2.attention.head_count

40
olmo2.attention.head_count_kv

40
olmo2.attention.layer_norm_rms_epsilon

1e-06
olmo2.block_count

40
olmo2.context_length

4096
olmo2.embedding_length

5120
olmo2.feed_forward_length

13824
olmo2.rope.freq_base

500000
tokenizer.ggml.bos_token_id

100257
tokenizer.ggml.eos_token_id

100257
tokenizer.ggml.merges

[Ġ Ġ, ĠĠ ĠĠ, i n, Ġ t, ĠĠĠĠ ĠĠĠĠ, ...]
tokenizer.ggml.model

gpt2
tokenizer.ggml.padding_token_id

100277
tokenizer.ggml.pre

dbrx
tokenizer.ggml.token_type

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

[!, ", #, $, %, ...]
tokenizer.ggml.unknown_token_id

100257

Tensor

Name

Type

Shape
token_embd.weight

F16

[5120, 100352]

blk.0

blk.0.attn_k.weight

F16

[5120, 5120]
blk.0.attn_k_norm.weight

F32

[5120]
blk.0.attn_output.weight

F16

[5120, 5120]
blk.0.attn_q.weight

F16

[5120, 5120]
blk.0.attn_q_norm.weight

F32

[5120]
blk.0.attn_v.weight

F16

[5120, 5120]
blk.0.ffn_down.weight

F16

[13824, 5120]
blk.0.ffn_gate.weight

F16

[5120, 13824]
blk.0.ffn_up.weight

F16

[5120, 13824]
blk.0.post_attention_norm.weight

F32

[5120]
blk.0.post_ffw_norm.weight

F32

[5120]

blk.1

blk.1.attn_k.weight

F16

[5120, 5120]
blk.1.attn_k_norm.weight

F32

[5120]
blk.1.attn_output.weight

F16

[5120, 5120]
blk.1.attn_q.weight

F16

[5120, 5120]
blk.1.attn_q_norm.weight

F32

[5120]
blk.1.attn_v.weight

F16

[5120, 5120]
blk.1.ffn_down.weight

F16

[13824, 5120]
blk.1.ffn_gate.weight

F16

[5120, 13824]
blk.1.ffn_up.weight

F16

[5120, 13824]
blk.1.post_attention_norm.weight

F32

[5120]
blk.1.post_ffw_norm.weight

F32

[5120]

blk.2

blk.2.attn_k.weight

F16

[5120, 5120]
blk.2.attn_k_norm.weight

F32

[5120]
blk.2.attn_output.weight

F16

[5120, 5120]
blk.2.attn_q.weight

F16

[5120, 5120]
blk.2.attn_q_norm.weight

F32

[5120]
blk.2.attn_v.weight

F16

[5120, 5120]
blk.2.ffn_down.weight

F16

[13824, 5120]
blk.2.ffn_gate.weight

F16

[5120, 13824]
blk.2.ffn_up.weight

F16

[5120, 13824]
blk.2.post_attention_norm.weight

F32

[5120]
blk.2.post_ffw_norm.weight

F32

[5120]

blk.3

blk.3.attn_k.weight

F16

[5120, 5120]
blk.3.attn_k_norm.weight

F32

[5120]
blk.3.attn_output.weight

F16

[5120, 5120]
blk.3.attn_q.weight

F16

[5120, 5120]
blk.3.attn_q_norm.weight

F32

[5120]
blk.3.attn_v.weight

F16

[5120, 5120]
blk.3.ffn_down.weight

F16

[13824, 5120]
blk.3.ffn_gate.weight

F16

[5120, 13824]
blk.3.ffn_up.weight

F16

[5120, 13824]
blk.3.post_attention_norm.weight

F32

[5120]
blk.3.post_ffw_norm.weight

F32

[5120]

blk.4

blk.4.attn_k.weight

F16

[5120, 5120]
blk.4.attn_k_norm.weight

F32

[5120]
blk.4.attn_output.weight

F16

[5120, 5120]
blk.4.attn_q.weight

F16

[5120, 5120]
blk.4.attn_q_norm.weight

F32

[5120]
blk.4.attn_v.weight

F16

[5120, 5120]
blk.4.ffn_down.weight

F16

[13824, 5120]
blk.4.ffn_gate.weight

F16

[5120, 13824]
blk.4.ffn_up.weight

F16

[5120, 13824]
blk.4.post_attention_norm.weight

F32

[5120]
blk.4.post_ffw_norm.weight

F32

[5120]

blk.5

blk.5.attn_k.weight

F16

[5120, 5120]
blk.5.attn_k_norm.weight

F32

[5120]
blk.5.attn_output.weight

F16

[5120, 5120]
blk.5.attn_q.weight

F16

[5120, 5120]
blk.5.attn_q_norm.weight

F32

[5120]
blk.5.attn_v.weight

F16

[5120, 5120]
blk.5.ffn_down.weight

F16

[13824, 5120]
blk.5.ffn_gate.weight

F16

[5120, 13824]
blk.5.ffn_up.weight

F16

[5120, 13824]
blk.5.post_attention_norm.weight

F32

[5120]
blk.5.post_ffw_norm.weight

F32

[5120]

blk.6

blk.6.attn_k.weight

F16

[5120, 5120]
blk.6.attn_k_norm.weight

F32

[5120]
blk.6.attn_output.weight

F16

[5120, 5120]
blk.6.attn_q.weight

F16

[5120, 5120]
blk.6.attn_q_norm.weight

F32

[5120]
blk.6.attn_v.weight

F16

[5120, 5120]
blk.6.ffn_down.weight

F16

[13824, 5120]
blk.6.ffn_gate.weight

F16

[5120, 13824]
blk.6.ffn_up.weight

F16

[5120, 13824]
blk.6.post_attention_norm.weight

F32

[5120]
blk.6.post_ffw_norm.weight

F32

[5120]

blk.7

blk.7.attn_k.weight

F16

[5120, 5120]
blk.7.attn_k_norm.weight

F32

[5120]
blk.7.attn_output.weight

F16

[5120, 5120]
blk.7.attn_q.weight

F16

[5120, 5120]
blk.7.attn_q_norm.weight

F32

[5120]
blk.7.attn_v.weight

F16

[5120, 5120]
blk.7.ffn_down.weight

F16

[13824, 5120]
blk.7.ffn_gate.weight

F16

[5120, 13824]
blk.7.ffn_up.weight

F16

[5120, 13824]
blk.7.post_attention_norm.weight

F32

[5120]
blk.7.post_ffw_norm.weight

F32

[5120]

blk.8

blk.8.attn_k.weight

F16

[5120, 5120]
blk.8.attn_k_norm.weight

F32

[5120]
blk.8.attn_output.weight

F16

[5120, 5120]
blk.8.attn_q.weight

F16

[5120, 5120]
blk.8.attn_q_norm.weight

F32

[5120]
blk.8.attn_v.weight

F16

[5120, 5120]
blk.8.ffn_down.weight

F16

[13824, 5120]
blk.8.ffn_gate.weight

F16

[5120, 13824]
blk.8.ffn_up.weight

F16

[5120, 13824]
blk.8.post_attention_norm.weight

F32

[5120]
blk.8.post_ffw_norm.weight

F32

[5120]

blk.9

blk.9.attn_k.weight

F16

[5120, 5120]
blk.9.attn_k_norm.weight

F32

[5120]
blk.9.attn_output.weight

F16

[5120, 5120]
blk.9.attn_q.weight

F16

[5120, 5120]
blk.9.attn_q_norm.weight

F32

[5120]
blk.9.attn_v.weight

F16

[5120, 5120]
blk.9.ffn_down.weight

F16

[13824, 5120]
blk.9.ffn_gate.weight

F16

[5120, 13824]
blk.9.ffn_up.weight

F16

[5120, 13824]
blk.9.post_attention_norm.weight

F32

[5120]
blk.9.post_ffw_norm.weight

F32

[5120]

blk.10

blk.10.attn_k.weight

F16

[5120, 5120]
blk.10.attn_k_norm.weight

F32

[5120]
blk.10.attn_output.weight

F16

[5120, 5120]
blk.10.attn_q.weight

F16

[5120, 5120]
blk.10.attn_q_norm.weight

F32

[5120]
blk.10.attn_v.weight

F16

[5120, 5120]
blk.10.ffn_down.weight

F16

[13824, 5120]
blk.10.ffn_gate.weight

F16

[5120, 13824]
blk.10.ffn_up.weight

F16

[5120, 13824]
blk.10.post_attention_norm.weight

F32

[5120]
blk.10.post_ffw_norm.weight

F32

[5120]

blk.11

blk.11.attn_k.weight

F16

[5120, 5120]
blk.11.attn_k_norm.weight

F32

[5120]
blk.11.attn_output.weight

F16

[5120, 5120]
blk.11.attn_q.weight

F16

[5120, 5120]
blk.11.attn_q_norm.weight

F32

[5120]
blk.11.attn_v.weight

F16

[5120, 5120]
blk.11.ffn_down.weight

F16

[13824, 5120]
blk.11.ffn_gate.weight

F16

[5120, 13824]
blk.11.ffn_up.weight

F16

[5120, 13824]
blk.11.post_attention_norm.weight

F32

[5120]
blk.11.post_ffw_norm.weight

F32

[5120]

blk.12

blk.12.attn_k.weight

F16

[5120, 5120]
blk.12.attn_k_norm.weight

F32

[5120]
blk.12.attn_output.weight

F16

[5120, 5120]
blk.12.attn_q.weight

F16

[5120, 5120]
blk.12.attn_q_norm.weight

F32

[5120]
blk.12.attn_v.weight

F16

[5120, 5120]
blk.12.ffn_down.weight

F16

[13824, 5120]
blk.12.ffn_gate.weight

F16

[5120, 13824]
blk.12.ffn_up.weight

F16

[5120, 13824]
blk.12.post_attention_norm.weight

F32

[5120]
blk.12.post_ffw_norm.weight

F32

[5120]

blk.13

blk.13.attn_k.weight

F16

[5120, 5120]
blk.13.attn_k_norm.weight

F32

[5120]
blk.13.attn_output.weight

F16

[5120, 5120]
blk.13.attn_q.weight

F16

[5120, 5120]
blk.13.attn_q_norm.weight

F32

[5120]
blk.13.attn_v.weight

F16

[5120, 5120]
blk.13.ffn_down.weight

F16

[13824, 5120]
blk.13.ffn_gate.weight

F16

[5120, 13824]
blk.13.ffn_up.weight

F16

[5120, 13824]
blk.13.post_attention_norm.weight

F32

[5120]
blk.13.post_ffw_norm.weight

F32

[5120]

blk.14

blk.14.attn_k.weight

F16

[5120, 5120]
blk.14.attn_k_norm.weight

F32

[5120]
blk.14.attn_output.weight

F16

[5120, 5120]
blk.14.attn_q.weight

F16

[5120, 5120]
blk.14.attn_q_norm.weight

F32

[5120]
blk.14.attn_v.weight

F16

[5120, 5120]
blk.14.ffn_down.weight

F16

[13824, 5120]
blk.14.ffn_gate.weight

F16

[5120, 13824]
blk.14.ffn_up.weight

F16

[5120, 13824]
blk.14.post_attention_norm.weight

F32

[5120]
blk.14.post_ffw_norm.weight

F32

[5120]

blk.15

blk.15.attn_k.weight

F16

[5120, 5120]
blk.15.attn_k_norm.weight

F32

[5120]
blk.15.attn_output.weight

F16

[5120, 5120]
blk.15.attn_q.weight

F16

[5120, 5120]
blk.15.attn_q_norm.weight

F32

[5120]
blk.15.attn_v.weight

F16

[5120, 5120]
blk.15.ffn_down.weight

F16

[13824, 5120]
blk.15.ffn_gate.weight

F16

[5120, 13824]
blk.15.ffn_up.weight

F16

[5120, 13824]
blk.15.post_attention_norm.weight

F32

[5120]
blk.15.post_ffw_norm.weight

F32

[5120]

blk.16

blk.16.attn_k.weight

F16

[5120, 5120]
blk.16.attn_k_norm.weight

F32

[5120]
blk.16.attn_output.weight

F16

[5120, 5120]
blk.16.attn_q.weight

F16

[5120, 5120]
blk.16.attn_q_norm.weight

F32

[5120]
blk.16.attn_v.weight

F16

[5120, 5120]
blk.16.ffn_down.weight

F16

[13824, 5120]
blk.16.ffn_gate.weight

F16

[5120, 13824]
blk.16.ffn_up.weight

F16

[5120, 13824]
blk.16.post_attention_norm.weight

F32

[5120]
blk.16.post_ffw_norm.weight

F32

[5120]

blk.17

blk.17.attn_k.weight

F16

[5120, 5120]
blk.17.attn_k_norm.weight

F32

[5120]
blk.17.attn_output.weight

F16

[5120, 5120]
blk.17.attn_q.weight

F16

[5120, 5120]
blk.17.attn_q_norm.weight

F32

[5120]
blk.17.attn_v.weight

F16

[5120, 5120]
blk.17.ffn_down.weight

F16

[13824, 5120]
blk.17.ffn_gate.weight

F16

[5120, 13824]
blk.17.ffn_up.weight

F16

[5120, 13824]
blk.17.post_attention_norm.weight

F32

[5120]
blk.17.post_ffw_norm.weight

F32

[5120]

blk.18

blk.18.attn_k.weight

F16

[5120, 5120]
blk.18.attn_k_norm.weight

F32

[5120]
blk.18.attn_output.weight

F16

[5120, 5120]
blk.18.attn_q.weight

F16

[5120, 5120]
blk.18.attn_q_norm.weight

F32

[5120]
blk.18.attn_v.weight

F16

[5120, 5120]
blk.18.ffn_down.weight

F16

[13824, 5120]
blk.18.ffn_gate.weight

F16

[5120, 13824]
blk.18.ffn_up.weight

F16

[5120, 13824]
blk.18.post_attention_norm.weight

F32

[5120]
blk.18.post_ffw_norm.weight

F32

[5120]

blk.19

blk.19.attn_k.weight

F16

[5120, 5120]
blk.19.attn_k_norm.weight

F32

[5120]
blk.19.attn_output.weight

F16

[5120, 5120]
blk.19.attn_q.weight

F16

[5120, 5120]
blk.19.attn_q_norm.weight

F32

[5120]
blk.19.attn_v.weight

F16

[5120, 5120]
blk.19.ffn_down.weight

F16

[13824, 5120]
blk.19.ffn_gate.weight

F16

[5120, 13824]
blk.19.ffn_up.weight

F16

[5120, 13824]
blk.19.post_attention_norm.weight

F32

[5120]
blk.19.post_ffw_norm.weight

F32

[5120]

blk.20

blk.20.attn_k.weight

F16

[5120, 5120]
blk.20.attn_k_norm.weight

F32

[5120]
blk.20.attn_output.weight

F16

[5120, 5120]
blk.20.attn_q.weight

F16

[5120, 5120]
blk.20.attn_q_norm.weight

F32

[5120]
blk.20.attn_v.weight

F16

[5120, 5120]
blk.20.ffn_down.weight

F16

[13824, 5120]
blk.20.ffn_gate.weight

F16

[5120, 13824]
blk.20.ffn_up.weight

F16

[5120, 13824]
blk.20.post_attention_norm.weight

F32

[5120]
blk.20.post_ffw_norm.weight

F32

[5120]

blk.21

blk.21.attn_k.weight

F16

[5120, 5120]
blk.21.attn_k_norm.weight

F32

[5120]
blk.21.attn_output.weight

F16

[5120, 5120]
blk.21.attn_q.weight

F16

[5120, 5120]
blk.21.attn_q_norm.weight

F32

[5120]
blk.21.attn_v.weight

F16

[5120, 5120]
blk.21.ffn_down.weight

F16

[13824, 5120]
blk.21.ffn_gate.weight

F16

[5120, 13824]
blk.21.ffn_up.weight

F16

[5120, 13824]
blk.21.post_attention_norm.weight

F32

[5120]
blk.21.post_ffw_norm.weight

F32

[5120]

blk.22

blk.22.attn_k.weight

F16

[5120, 5120]
blk.22.attn_k_norm.weight

F32

[5120]
blk.22.attn_output.weight

F16

[5120, 5120]
blk.22.attn_q.weight

F16

[5120, 5120]
blk.22.attn_q_norm.weight

F32

[5120]
blk.22.attn_v.weight

F16

[5120, 5120]
blk.22.ffn_down.weight

F16

[13824, 5120]
blk.22.ffn_gate.weight

F16

[5120, 13824]
blk.22.ffn_up.weight

F16

[5120, 13824]
blk.22.post_attention_norm.weight

F32

[5120]
blk.22.post_ffw_norm.weight

F32

[5120]

blk.23

blk.23.attn_k.weight

F16

[5120, 5120]
blk.23.attn_k_norm.weight

F32

[5120]
blk.23.attn_output.weight

F16

[5120, 5120]
blk.23.attn_q.weight

F16

[5120, 5120]
blk.23.attn_q_norm.weight

F32

[5120]
blk.23.attn_v.weight

F16

[5120, 5120]
blk.23.ffn_down.weight

F16

[13824, 5120]
blk.23.ffn_gate.weight

F16

[5120, 13824]
blk.23.ffn_up.weight

F16

[5120, 13824]
blk.23.post_attention_norm.weight

F32

[5120]
blk.23.post_ffw_norm.weight

F32

[5120]

blk.24

blk.24.attn_k.weight

F16

[5120, 5120]
blk.24.attn_k_norm.weight

F32

[5120]
blk.24.attn_output.weight

F16

[5120, 5120]
blk.24.attn_q.weight

F16

[5120, 5120]
blk.24.attn_q_norm.weight

F32

[5120]
blk.24.attn_v.weight

F16

[5120, 5120]
blk.24.ffn_down.weight

F16

[13824, 5120]
blk.24.ffn_gate.weight

F16

[5120, 13824]
blk.24.ffn_up.weight

F16

[5120, 13824]
blk.24.post_attention_norm.weight

F32

[5120]
blk.24.post_ffw_norm.weight

F32

[5120]

blk.25

blk.25.attn_k.weight

F16

[5120, 5120]
blk.25.attn_k_norm.weight

F32

[5120]
blk.25.attn_output.weight

F16

[5120, 5120]
blk.25.attn_q.weight

F16

[5120, 5120]
blk.25.attn_q_norm.weight

F32

[5120]
blk.25.attn_v.weight

F16

[5120, 5120]
blk.25.ffn_down.weight

F16

[13824, 5120]
blk.25.ffn_gate.weight

F16

[5120, 13824]
blk.25.ffn_up.weight

F16

[5120, 13824]
blk.25.post_attention_norm.weight

F32

[5120]
blk.25.post_ffw_norm.weight

F32

[5120]

blk.26

blk.26.attn_k.weight

F16

[5120, 5120]
blk.26.attn_k_norm.weight

F32

[5120]
blk.26.attn_output.weight

F16

[5120, 5120]
blk.26.attn_q.weight

F16

[5120, 5120]
blk.26.attn_q_norm.weight

F32

[5120]
blk.26.attn_v.weight

F16

[5120, 5120]
blk.26.ffn_down.weight

F16

[13824, 5120]
blk.26.ffn_gate.weight

F16

[5120, 13824]
blk.26.ffn_up.weight

F16

[5120, 13824]
blk.26.post_attention_norm.weight

F32

[5120]
blk.26.post_ffw_norm.weight

F32

[5120]

blk.27

blk.27.attn_k.weight

F16

[5120, 5120]
blk.27.attn_k_norm.weight

F32

[5120]
blk.27.attn_output.weight

F16

[5120, 5120]
blk.27.attn_q.weight

F16

[5120, 5120]
blk.27.attn_q_norm.weight

F32

[5120]
blk.27.attn_v.weight

F16

[5120, 5120]
blk.27.ffn_down.weight

F16

[13824, 5120]
blk.27.ffn_gate.weight

F16

[5120, 13824]
blk.27.ffn_up.weight

F16

[5120, 13824]
blk.27.post_attention_norm.weight

F32

[5120]
blk.27.post_ffw_norm.weight

F32

[5120]

blk.28

blk.28.attn_k.weight

F16

[5120, 5120]
blk.28.attn_k_norm.weight

F32

[5120]
blk.28.attn_output.weight

F16

[5120, 5120]
blk.28.attn_q.weight

F16

[5120, 5120]
blk.28.attn_q_norm.weight

F32

[5120]
blk.28.attn_v.weight

F16

[5120, 5120]
blk.28.ffn_down.weight

F16

[13824, 5120]
blk.28.ffn_gate.weight

F16

[5120, 13824]
blk.28.ffn_up.weight

F16

[5120, 13824]
blk.28.post_attention_norm.weight

F32

[5120]
blk.28.post_ffw_norm.weight

F32

[5120]

blk.29

blk.29.attn_k.weight

F16

[5120, 5120]
blk.29.attn_k_norm.weight

F32

[5120]
blk.29.attn_output.weight

F16

[5120, 5120]
blk.29.attn_q.weight

F16

[5120, 5120]
blk.29.attn_q_norm.weight

F32

[5120]
blk.29.attn_v.weight

F16

[5120, 5120]
blk.29.ffn_down.weight

F16

[13824, 5120]
blk.29.ffn_gate.weight

F16

[5120, 13824]
blk.29.ffn_up.weight

F16

[5120, 13824]
blk.29.post_attention_norm.weight

F32

[5120]
blk.29.post_ffw_norm.weight

F32

[5120]

blk.30

blk.30.attn_k.weight

F16

[5120, 5120]
blk.30.attn_k_norm.weight

F32

[5120]
blk.30.attn_output.weight

F16

[5120, 5120]
blk.30.attn_q.weight

F16

[5120, 5120]
blk.30.attn_q_norm.weight

F32

[5120]
blk.30.attn_v.weight

F16

[5120, 5120]
blk.30.ffn_down.weight

F16

[13824, 5120]
blk.30.ffn_gate.weight

F16

[5120, 13824]
blk.30.ffn_up.weight

F16

[5120, 13824]
blk.30.post_attention_norm.weight

F32

[5120]
blk.30.post_ffw_norm.weight

F32

[5120]

blk.31

blk.31.attn_k.weight

F16

[5120, 5120]
blk.31.attn_k_norm.weight

F32

[5120]
blk.31.attn_output.weight

F16

[5120, 5120]
blk.31.attn_q.weight

F16

[5120, 5120]
blk.31.attn_q_norm.weight

F32

[5120]
blk.31.attn_v.weight

F16

[5120, 5120]
blk.31.ffn_down.weight

F16

[13824, 5120]
blk.31.ffn_gate.weight

F16

[5120, 13824]
blk.31.ffn_up.weight

F16

[5120, 13824]
blk.31.post_attention_norm.weight

F32

[5120]
blk.31.post_ffw_norm.weight

F32

[5120]

blk.32

blk.32.attn_k.weight

F16

[5120, 5120]
blk.32.attn_k_norm.weight

F32

[5120]
blk.32.attn_output.weight

F16

[5120, 5120]
blk.32.attn_q.weight

F16

[5120, 5120]
blk.32.attn_q_norm.weight

F32

[5120]
blk.32.attn_v.weight

F16

[5120, 5120]
blk.32.ffn_down.weight

F16

[13824, 5120]
blk.32.ffn_gate.weight

F16

[5120, 13824]
blk.32.ffn_up.weight

F16

[5120, 13824]
blk.32.post_attention_norm.weight

F32

[5120]
blk.32.post_ffw_norm.weight

F32

[5120]

blk.33

blk.33.attn_k.weight

F16

[5120, 5120]
blk.33.attn_k_norm.weight

F32

[5120]
blk.33.attn_output.weight

F16

[5120, 5120]
blk.33.attn_q.weight

F16

[5120, 5120]
blk.33.attn_q_norm.weight

F32

[5120]
blk.33.attn_v.weight

F16

[5120, 5120]
blk.33.ffn_down.weight

F16

[13824, 5120]
blk.33.ffn_gate.weight

F16

[5120, 13824]
blk.33.ffn_up.weight

F16

[5120, 13824]
blk.33.post_attention_norm.weight

F32

[5120]
blk.33.post_ffw_norm.weight

F32

[5120]

blk.34

blk.34.attn_k.weight

F16

[5120, 5120]
blk.34.attn_k_norm.weight

F32

[5120]
blk.34.attn_output.weight

F16

[5120, 5120]
blk.34.attn_q.weight

F16

[5120, 5120]
blk.34.attn_q_norm.weight

F32

[5120]
blk.34.attn_v.weight

F16

[5120, 5120]
blk.34.ffn_down.weight

F16

[13824, 5120]
blk.34.ffn_gate.weight

F16

[5120, 13824]
blk.34.ffn_up.weight

F16

[5120, 13824]
blk.34.post_attention_norm.weight

F32

[5120]
blk.34.post_ffw_norm.weight

F32

[5120]

blk.35

blk.35.attn_k.weight

F16

[5120, 5120]
blk.35.attn_k_norm.weight

F32

[5120]
blk.35.attn_output.weight

F16

[5120, 5120]
blk.35.attn_q.weight

F16

[5120, 5120]
blk.35.attn_q_norm.weight

F32

[5120]
blk.35.attn_v.weight

F16

[5120, 5120]
blk.35.ffn_down.weight

F16

[13824, 5120]
blk.35.ffn_gate.weight

F16

[5120, 13824]
blk.35.ffn_up.weight

F16

[5120, 13824]
blk.35.post_attention_norm.weight

F32

[5120]
blk.35.post_ffw_norm.weight

F32

[5120]

blk.36

blk.36.attn_k.weight

F16

[5120, 5120]
blk.36.attn_k_norm.weight

F32

[5120]
blk.36.attn_output.weight

F16

[5120, 5120]
blk.36.attn_q.weight

F16

[5120, 5120]
blk.36.attn_q_norm.weight

F32

[5120]
blk.36.attn_v.weight

F16

[5120, 5120]
blk.36.ffn_down.weight

F16

[13824, 5120]
blk.36.ffn_gate.weight

F16

[5120, 13824]
blk.36.ffn_up.weight

F16

[5120, 13824]
blk.36.post_attention_norm.weight

F32

[5120]
blk.36.post_ffw_norm.weight

F32

[5120]

blk.37

blk.37.attn_k.weight

F16

[5120, 5120]
blk.37.attn_k_norm.weight

F32

[5120]
blk.37.attn_output.weight

F16

[5120, 5120]
blk.37.attn_q.weight

F16

[5120, 5120]
blk.37.attn_q_norm.weight

F32

[5120]
blk.37.attn_v.weight

F16

[5120, 5120]
blk.37.ffn_down.weight

F16

[13824, 5120]
blk.37.ffn_gate.weight

F16

[5120, 13824]
blk.37.ffn_up.weight

F16

[5120, 13824]
blk.37.post_attention_norm.weight

F32

[5120]
blk.37.post_ffw_norm.weight

F32

[5120]

blk.38

blk.38.attn_k.weight

F16

[5120, 5120]
blk.38.attn_k_norm.weight

F32

[5120]
blk.38.attn_output.weight

F16

[5120, 5120]
blk.38.attn_q.weight

F16

[5120, 5120]
blk.38.attn_q_norm.weight

F32

[5120]
blk.38.attn_v.weight

F16

[5120, 5120]
blk.38.ffn_down.weight

F16

[13824, 5120]
blk.38.ffn_gate.weight

F16

[5120, 13824]
blk.38.ffn_up.weight

F16

[5120, 13824]
blk.38.post_attention_norm.weight

F32

[5120]
blk.38.post_ffw_norm.weight

F32

[5120]

blk.39

blk.39.attn_k.weight

F16

[5120, 5120]
blk.39.attn_k_norm.weight

F32

[5120]
blk.39.attn_output.weight

F16

[5120, 5120]
blk.39.attn_q.weight

F16

[5120, 5120]
blk.39.attn_q_norm.weight

F32

[5120]
blk.39.attn_v.weight

F16

[5120, 5120]
blk.39.ffn_down.weight

F16

[13824, 5120]
blk.39.ffn_gate.weight

F16

[5120, 13824]
blk.39.ffn_up.weight

F16

[5120, 13824]
blk.39.post_attention_norm.weight

F32

[5120]
blk.39.post_ffw_norm.weight

F32

[5120]

output.weight

F16

[5120, 100352]
output_norm.weight

F32

[5120]