Models
GitHub
Discord
Turbo
Sign in
Download
Models
Download
GitHub
Discord
Sign in
andiariffin
/
llama-3.3-nemotron-super-v1.5-bf16
:49b
2
Downloads
Updated
1 month ago
High efficiency model with leading accuracy for reasoning, tool calling, chat, and instruction following.
High efficiency model with leading accuracy for reasoning, tool calling, chat, and instruction following.
Cancel
49b
llama-3.3-nemotron-super-v1.5-bf16:49b
...
/
model
dc616736040a · 100GB
Metadata
general.architecture
deci
deci
general.file_type
BF16
BF16
deci.attention.head_count
[64, 64, 64, 64, 64, ...]
[64, 64, 64, 64, 64, ...]
deci.attention.head_count_kv
[8, 8, 8, 8, 8, ...]
[8, 8, 8, 8, 8, ...]
deci.attention.key_length
128
128
deci.attention.layer_norm_rms_epsilon
1e-05
1e-05
deci.attention.value_length
128
128
deci.block_count
80
80
deci.context_length
131072
131072
deci.embedding_length
8192
8192
deci.feed_forward_length
[14336, 28672, 28672, 28672, 28672, ...]
[14336, 28672, 28672, 28672, 28672, ...]
deci.rope.dimension_count
128
128
deci.rope.freq_base
500000
500000
deci.vocab_size
128256
128256
tokenizer.ggml.add_bos_token
true
true
tokenizer.ggml.add_sep_token
false
false
tokenizer.ggml.bos_token_id
128000
128000
tokenizer.ggml.eos_token_id
128009
128009
tokenizer.ggml.merges
[Ġ Ġ, Ġ ĠĠĠ, ĠĠ ĠĠ, ĠĠĠ Ġ, i n, ...]
[Ġ Ġ, Ġ ĠĠĠ, ĠĠ ĠĠ, ĠĠĠ Ġ, i n, ...]
tokenizer.ggml.model
gpt2
gpt2
tokenizer.ggml.padding_token_id
128009
128009
tokenizer.ggml.pre
llama-bpe
llama-bpe
tokenizer.ggml.token_type
[1, 1, 1, 1, 1, ...]
[1, 1, 1, 1, 1, ...]
tokenizer.ggml.tokens
[!, ", #, $, %, ...]
[!, ", #, $, %, ...]
Tensor
Name
Type
Shape
token_embd.weight
BF16
BF16
[8192, 128256]
blk.0
blk.0.attn_k.weight
BF16
BF16
[8192, 1024]
blk.0.attn_norm.weight
F32
F32
[8192]
blk.0.attn_output.weight
BF16
BF16
[8192, 8192]
blk.0.attn_q.weight
BF16
BF16
[8192, 8192]
blk.0.attn_v.weight
BF16
BF16
[8192, 1024]
blk.0.ffn_down.weight
BF16
BF16
[14336, 8192]
blk.0.ffn_gate.weight
BF16
BF16
[8192, 14336]
blk.0.ffn_norm.weight
F32
F32
[8192]
blk.0.ffn_up.weight
BF16
BF16
[8192, 14336]
blk.1
blk.1.attn_k.weight
BF16
BF16
[8192, 1024]
blk.1.attn_norm.weight
F32
F32
[8192]
blk.1.attn_output.weight
BF16
BF16
[8192, 8192]
blk.1.attn_q.weight
BF16
BF16
[8192, 8192]
blk.1.attn_v.weight
BF16
BF16
[8192, 1024]
blk.1.ffn_down.weight
BF16
BF16
[28672, 8192]
blk.1.ffn_gate.weight
BF16
BF16
[8192, 28672]
blk.1.ffn_norm.weight
F32
F32
[8192]
blk.1.ffn_up.weight
BF16
BF16
[8192, 28672]
blk.2
blk.2.attn_k.weight
BF16
BF16
[8192, 1024]
blk.2.attn_norm.weight
F32
F32
[8192]
blk.2.attn_output.weight
BF16
BF16
[8192, 8192]
blk.2.attn_q.weight
BF16
BF16
[8192, 8192]
blk.2.attn_v.weight
BF16
BF16
[8192, 1024]
blk.2.ffn_down.weight
BF16
BF16
[28672, 8192]
blk.2.ffn_gate.weight
BF16
BF16
[8192, 28672]
blk.2.ffn_norm.weight
F32
F32
[8192]
blk.2.ffn_up.weight
BF16
BF16
[8192, 28672]
blk.3
blk.3.attn_k.weight
BF16
BF16
[8192, 1024]
blk.3.attn_norm.weight
F32
F32
[8192]
blk.3.attn_output.weight
BF16
BF16
[8192, 8192]
blk.3.attn_q.weight
BF16
BF16
[8192, 8192]
blk.3.attn_v.weight
BF16
BF16
[8192, 1024]
blk.3.ffn_down.weight
BF16
BF16
[28672, 8192]
blk.3.ffn_gate.weight
BF16
BF16
[8192, 28672]
blk.3.ffn_norm.weight
F32
F32
[8192]
blk.3.ffn_up.weight
BF16
BF16
[8192, 28672]
blk.4
blk.4.attn_k.weight
BF16
BF16
[8192, 1024]
blk.4.attn_norm.weight
F32
F32
[8192]
blk.4.attn_output.weight
BF16
BF16
[8192, 8192]
blk.4.attn_q.weight
BF16
BF16
[8192, 8192]
blk.4.attn_v.weight
BF16
BF16
[8192, 1024]
blk.4.ffn_down.weight
BF16
BF16
[28672, 8192]
blk.4.ffn_gate.weight
BF16
BF16
[8192, 28672]
blk.4.ffn_norm.weight
F32
F32
[8192]
blk.4.ffn_up.weight
BF16
BF16
[8192, 28672]
blk.5
blk.5.attn_k.weight
BF16
BF16
[8192, 1024]
blk.5.attn_norm.weight
F32
F32
[8192]
blk.5.attn_output.weight
BF16
BF16
[8192, 8192]
blk.5.attn_q.weight
BF16
BF16
[8192, 8192]
blk.5.attn_v.weight
BF16
BF16
[8192, 1024]
blk.5.ffn_down.weight
BF16
BF16
[28672, 8192]
blk.5.ffn_gate.weight
BF16
BF16
[8192, 28672]
blk.5.ffn_norm.weight
F32
F32
[8192]
blk.5.ffn_up.weight
BF16
BF16
[8192, 28672]
blk.6
blk.6.ffn_down.weight
BF16
BF16
[14336, 8192]
blk.6.ffn_gate.weight
BF16
BF16
[8192, 14336]
blk.6.ffn_norm.weight
F32
F32
[8192]
blk.6.ffn_up.weight
BF16
BF16
[8192, 14336]
blk.7
blk.7.ffn_down.weight
BF16
BF16
[14336, 8192]
blk.7.ffn_gate.weight
BF16
BF16
[8192, 14336]
blk.7.ffn_norm.weight
F32
F32
[8192]
blk.7.ffn_up.weight
BF16
BF16
[8192, 14336]
blk.8
blk.8.attn_k.weight
BF16
BF16
[8192, 1024]
blk.8.attn_norm.weight
F32
F32
[8192]
blk.8.attn_output.weight
BF16
BF16
[8192, 8192]
blk.8.attn_q.weight
BF16
BF16
[8192, 8192]
blk.8.attn_v.weight
BF16
BF16
[8192, 1024]
blk.8.ffn_down.weight
BF16
BF16
[28672, 8192]
blk.8.ffn_gate.weight
BF16
BF16
[8192, 28672]
blk.8.ffn_norm.weight
F32
F32
[8192]
blk.8.ffn_up.weight
BF16
BF16
[8192, 28672]
blk.9
blk.9.attn_k.weight
BF16
BF16
[8192, 1024]
blk.9.attn_norm.weight
F32
F32
[8192]
blk.9.attn_output.weight
BF16
BF16
[8192, 8192]
blk.9.attn_q.weight
BF16
BF16
[8192, 8192]
blk.9.attn_v.weight
BF16
BF16
[8192, 1024]
blk.9.ffn_down.weight
BF16
BF16
[28672, 8192]
blk.9.ffn_gate.weight
BF16
BF16
[8192, 28672]
blk.9.ffn_norm.weight
F32
F32
[8192]
blk.9.ffn_up.weight
BF16
BF16
[8192, 28672]
blk.10
blk.10.attn_k.weight
BF16
BF16
[8192, 1024]
blk.10.attn_norm.weight
F32
F32
[8192]
blk.10.attn_output.weight
BF16
BF16
[8192, 8192]
blk.10.attn_q.weight
BF16
BF16
[8192, 8192]
blk.10.attn_v.weight
BF16
BF16
[8192, 1024]
blk.10.ffn_down.weight
BF16
BF16
[28672, 8192]
blk.10.ffn_gate.weight
BF16
BF16
[8192, 28672]
blk.10.ffn_norm.weight
F32
F32
[8192]
blk.10.ffn_up.weight
BF16
BF16
[8192, 28672]
blk.11
blk.11.ffn_down.weight
BF16
BF16
[17920, 8192]
blk.11.ffn_gate.weight
BF16
BF16
[8192, 17920]
blk.11.ffn_norm.weight
F32
F32
[8192]
blk.11.ffn_up.weight
BF16
BF16
[8192, 17920]
blk.12
blk.12.attn_k.weight
BF16
BF16
[8192, 1024]
blk.12.attn_norm.weight
F32
F32
[8192]
blk.12.attn_output.weight
BF16
BF16
[8192, 8192]
blk.12.attn_q.weight
BF16
BF16
[8192, 8192]
blk.12.attn_v.weight
BF16
BF16
[8192, 1024]
blk.12.ffn_down.weight
BF16
BF16
[28672, 8192]
blk.12.ffn_gate.weight
BF16
BF16
[8192, 28672]
blk.12.ffn_norm.weight
F32
F32
[8192]
blk.12.ffn_up.weight
BF16
BF16
[8192, 28672]
blk.13
blk.13.attn_k.weight
BF16
BF16
[8192, 1024]
blk.13.attn_norm.weight
F32
F32
[8192]
blk.13.attn_output.weight
BF16
BF16
[8192, 8192]
blk.13.attn_q.weight
BF16
BF16
[8192, 8192]
blk.13.attn_v.weight
BF16
BF16
[8192, 1024]
blk.13.ffn_down.weight
BF16
BF16
[28672, 8192]
blk.13.ffn_gate.weight
BF16
BF16
[8192, 28672]
blk.13.ffn_norm.weight
F32
F32
[8192]
blk.13.ffn_up.weight
BF16
BF16
[8192, 28672]
blk.14
blk.14.attn_k.weight
BF16
BF16
[8192, 1024]
blk.14.attn_norm.weight
F32
F32
[8192]
blk.14.attn_output.weight
BF16
BF16
[8192, 8192]
blk.14.attn_q.weight
BF16
BF16
[8192, 8192]
blk.14.attn_v.weight
BF16
BF16
[8192, 1024]
blk.14.ffn_down.weight
BF16
BF16
[28672, 8192]
blk.14.ffn_gate.weight
BF16
BF16
[8192, 28672]
blk.14.ffn_norm.weight
F32
F32
[8192]
blk.14.ffn_up.weight
BF16
BF16
[8192, 28672]
blk.15
blk.15.attn_k.weight
BF16
BF16
[8192, 1024]
blk.15.attn_norm.weight
F32
F32
[8192]
blk.15.attn_output.weight
BF16
BF16
[8192, 8192]
blk.15.attn_q.weight
BF16
BF16
[8192, 8192]
blk.15.attn_v.weight
BF16
BF16
[8192, 1024]
blk.15.ffn_down.weight
BF16
BF16
[28672, 8192]
blk.15.ffn_gate.weight
BF16
BF16
[8192, 28672]
blk.15.ffn_norm.weight
F32
F32
[8192]
blk.15.ffn_up.weight
BF16
BF16
[8192, 28672]
blk.16
blk.16.attn_k.weight
BF16
BF16
[8192, 1024]
blk.16.attn_norm.weight
F32
F32
[8192]
blk.16.attn_output.weight
BF16
BF16
[8192, 8192]
blk.16.attn_q.weight
BF16
BF16
[8192, 8192]
blk.16.attn_v.weight
BF16
BF16
[8192, 1024]
blk.16.ffn_down.weight
BF16
BF16
[28672, 8192]
blk.16.ffn_gate.weight
BF16
BF16
[8192, 28672]
blk.16.ffn_norm.weight
F32
F32
[8192]
blk.16.ffn_up.weight
BF16
BF16
[8192, 28672]
blk.17
blk.17.attn_k.weight
BF16
BF16
[8192, 1024]
blk.17.attn_norm.weight
F32
F32
[8192]
blk.17.attn_output.weight
BF16
BF16
[8192, 8192]
blk.17.attn_q.weight
BF16
BF16
[8192, 8192]
blk.17.attn_v.weight
BF16
BF16
[8192, 1024]
blk.17.ffn_down.weight
BF16
BF16
[28672, 8192]
blk.17.ffn_gate.weight
BF16
BF16
[8192, 28672]
blk.17.ffn_norm.weight
F32
F32
[8192]
blk.17.ffn_up.weight
BF16
BF16
[8192, 28672]
blk.18
blk.18.attn_k.weight
BF16
BF16
[8192, 1024]
blk.18.attn_norm.weight
F32
F32
[8192]
blk.18.attn_output.weight
BF16
BF16
[8192, 8192]
blk.18.attn_q.weight
BF16
BF16
[8192, 8192]
blk.18.attn_v.weight
BF16
BF16
[8192, 1024]
blk.18.ffn_down.weight
BF16
BF16
[28672, 8192]
blk.18.ffn_gate.weight
BF16
BF16
[8192, 28672]
blk.18.ffn_norm.weight
F32
F32
[8192]
blk.18.ffn_up.weight
BF16
BF16
[8192, 28672]
blk.19
blk.19.attn_k.weight
BF16
BF16
[8192, 1024]
blk.19.attn_norm.weight
F32
F32
[8192]
blk.19.attn_output.weight
BF16
BF16
[8192, 8192]
blk.19.attn_q.weight
BF16
BF16
[8192, 8192]
blk.19.attn_v.weight
BF16
BF16
[8192, 1024]
blk.19.ffn_down.weight
BF16
BF16
[28672, 8192]
blk.19.ffn_gate.weight
BF16
BF16
[8192, 28672]
blk.19.ffn_norm.weight
F32
F32
[8192]
blk.19.ffn_up.weight
BF16
BF16
[8192, 28672]
blk.20
blk.20.attn_k.weight
BF16
BF16
[8192, 1024]
blk.20.attn_norm.weight
F32
F32
[8192]
blk.20.attn_output.weight
BF16
BF16
[8192, 8192]
blk.20.attn_q.weight
BF16
BF16
[8192, 8192]
blk.20.attn_v.weight
BF16
BF16
[8192, 1024]
blk.20.ffn_down.weight
BF16
BF16
[28672, 8192]
blk.20.ffn_gate.weight
BF16
BF16
[8192, 28672]
blk.20.ffn_norm.weight
F32
F32
[8192]
blk.20.ffn_up.weight
BF16
BF16
[8192, 28672]
blk.21
blk.21.attn_k.weight
BF16
BF16
[8192, 1024]
blk.21.attn_norm.weight
F32
F32
[8192]
blk.21.attn_output.weight
BF16
BF16
[8192, 8192]
blk.21.attn_q.weight
BF16
BF16
[8192, 8192]
blk.21.attn_v.weight
BF16
BF16
[8192, 1024]
blk.21.ffn_down.weight
BF16
BF16
[28672, 8192]
blk.21.ffn_gate.weight
BF16
BF16
[8192, 28672]
blk.21.ffn_norm.weight
F32
F32
[8192]
blk.21.ffn_up.weight
BF16
BF16
[8192, 28672]
blk.22
blk.22.attn_k.weight
BF16
BF16
[8192, 1024]
blk.22.attn_norm.weight
F32
F32
[8192]
blk.22.attn_output.weight
BF16
BF16
[8192, 8192]
blk.22.attn_q.weight
BF16
BF16
[8192, 8192]
blk.22.attn_v.weight
BF16
BF16
[8192, 1024]
blk.22.ffn_down.weight
BF16
BF16
[28672, 8192]
blk.22.ffn_gate.weight
BF16
BF16
[8192, 28672]
blk.22.ffn_norm.weight
F32
F32
[8192]
blk.22.ffn_up.weight
BF16
BF16
[8192, 28672]
blk.23
blk.23.attn_k.weight
BF16
BF16
[8192, 1024]
blk.23.attn_norm.weight
F32
F32
[8192]
blk.23.attn_output.weight
BF16
BF16
[8192, 8192]
blk.23.attn_q.weight
BF16
BF16
[8192, 8192]
blk.23.attn_v.weight
BF16
BF16
[8192, 1024]
blk.23.ffn_down.weight
BF16
BF16
[28672, 8192]
blk.23.ffn_gate.weight
BF16
BF16
[8192, 28672]
blk.23.ffn_norm.weight
F32
F32
[8192]
blk.23.ffn_up.weight
BF16
BF16
[8192, 28672]
blk.24
blk.24.attn_k.weight
BF16
BF16
[8192, 1024]
blk.24.attn_norm.weight
F32
F32
[8192]
blk.24.attn_output.weight
BF16
BF16
[8192, 8192]
blk.24.attn_q.weight
BF16
BF16
[8192, 8192]
blk.24.attn_v.weight
BF16
BF16
[8192, 1024]
blk.24.ffn_down.weight
BF16
BF16
[28672, 8192]
blk.24.ffn_gate.weight
BF16
BF16
[8192, 28672]
blk.24.ffn_norm.weight
F32
F32
[8192]
blk.24.ffn_up.weight
BF16
BF16
[8192, 28672]
blk.25
blk.25.attn_k.weight
BF16
BF16
[8192, 1024]
blk.25.attn_norm.weight
F32
F32
[8192]
blk.25.attn_output.weight
BF16
BF16
[8192, 8192]
blk.25.attn_q.weight
BF16
BF16
[8192, 8192]
blk.25.attn_v.weight
BF16
BF16
[8192, 1024]
blk.25.ffn_down.weight
BF16
BF16
[28672, 8192]
blk.25.ffn_gate.weight
BF16
BF16
[8192, 28672]
blk.25.ffn_norm.weight
F32
F32
[8192]
blk.25.ffn_up.weight
BF16
BF16
[8192, 28672]
blk.26
blk.26.attn_k.weight
BF16
BF16
[8192, 1024]
blk.26.attn_norm.weight
F32
F32
[8192]
blk.26.attn_output.weight
BF16
BF16
[8192, 8192]
blk.26.attn_q.weight
BF16
BF16
[8192, 8192]
blk.26.attn_v.weight
BF16
BF16
[8192, 1024]
blk.26.ffn_down.weight
BF16
BF16
[28672, 8192]
blk.26.ffn_gate.weight
BF16
BF16
[8192, 28672]
blk.26.ffn_norm.weight
F32
F32
[8192]
blk.26.ffn_up.weight
BF16
BF16
[8192, 28672]
blk.27
blk.27.attn_k.weight
BF16
BF16
[8192, 1024]
blk.27.attn_norm.weight
F32
F32
[8192]
blk.27.attn_output.weight
BF16
BF16
[8192, 8192]
blk.27.attn_q.weight
BF16
BF16
[8192, 8192]
blk.27.attn_v.weight
BF16
BF16
[8192, 1024]
blk.27.ffn_down.weight
BF16
BF16
[28672, 8192]
blk.27.ffn_gate.weight
BF16
BF16
[8192, 28672]
blk.27.ffn_norm.weight
F32
F32
[8192]
blk.27.ffn_up.weight
BF16
BF16
[8192, 28672]
blk.28
blk.28.attn_k.weight
BF16
BF16
[8192, 1024]
blk.28.attn_norm.weight
F32
F32
[8192]
blk.28.attn_output.weight
BF16
BF16
[8192, 8192]
blk.28.attn_q.weight
BF16
BF16
[8192, 8192]
blk.28.attn_v.weight
BF16
BF16
[8192, 1024]
blk.28.ffn_down.weight
BF16
BF16
[28672, 8192]
blk.28.ffn_gate.weight
BF16
BF16
[8192, 28672]
blk.28.ffn_norm.weight
F32
F32
[8192]
blk.28.ffn_up.weight
BF16
BF16
[8192, 28672]
blk.29
blk.29.attn_k.weight
BF16
BF16
[8192, 1024]
blk.29.attn_norm.weight
F32
F32
[8192]
blk.29.attn_output.weight
BF16
BF16
[8192, 8192]
blk.29.attn_q.weight
BF16
BF16
[8192, 8192]
blk.29.attn_v.weight
BF16
BF16
[8192, 1024]
blk.29.ffn_down.weight
BF16
BF16
[28672, 8192]
blk.29.ffn_gate.weight
BF16
BF16
[8192, 28672]
blk.29.ffn_norm.weight
F32
F32
[8192]
blk.29.ffn_up.weight
BF16
BF16
[8192, 28672]
blk.30
blk.30.attn_k.weight
BF16
BF16
[8192, 1024]
blk.30.attn_norm.weight
F32
F32
[8192]
blk.30.attn_output.weight
BF16
BF16
[8192, 8192]
blk.30.attn_q.weight
BF16
BF16
[8192, 8192]
blk.30.attn_v.weight
BF16
BF16
[8192, 1024]
blk.30.ffn_down.weight
BF16
BF16
[28672, 8192]
blk.30.ffn_gate.weight
BF16
BF16
[8192, 28672]
blk.30.ffn_norm.weight
F32
F32
[8192]
blk.30.ffn_up.weight
BF16
BF16
[8192, 28672]
blk.31
blk.31.attn_k.weight
BF16
BF16
[8192, 1024]
blk.31.attn_norm.weight
F32
F32
[8192]
blk.31.attn_output.weight
BF16
BF16
[8192, 8192]
blk.31.attn_q.weight
BF16
BF16
[8192, 8192]
blk.31.attn_v.weight
BF16
BF16
[8192, 1024]
blk.31.ffn_down.weight
BF16
BF16
[28672, 8192]
blk.31.ffn_gate.weight
BF16
BF16
[8192, 28672]
blk.31.ffn_norm.weight
F32
F32
[8192]
blk.31.ffn_up.weight
BF16
BF16
[8192, 28672]
blk.32
blk.32.attn_k.weight
BF16
BF16
[8192, 1024]
blk.32.attn_norm.weight
F32
F32
[8192]
blk.32.attn_output.weight
BF16
BF16
[8192, 8192]
blk.32.attn_q.weight
BF16
BF16
[8192, 8192]
blk.32.attn_v.weight
BF16
BF16
[8192, 1024]
blk.32.ffn_down.weight
BF16
BF16
[28672, 8192]
blk.32.ffn_gate.weight
BF16
BF16
[8192, 28672]
blk.32.ffn_norm.weight
F32
F32
[8192]
blk.32.ffn_up.weight
BF16
BF16
[8192, 28672]
blk.33
blk.33.attn_k.weight
BF16
BF16
[8192, 1024]
blk.33.attn_norm.weight
F32
F32
[8192]
blk.33.attn_output.weight
BF16
BF16
[8192, 8192]
blk.33.attn_q.weight
BF16
BF16
[8192, 8192]
blk.33.attn_v.weight
BF16
BF16
[8192, 1024]
blk.33.ffn_down.weight
BF16
BF16
[28672, 8192]
blk.33.ffn_gate.weight
BF16
BF16
[8192, 28672]
blk.33.ffn_norm.weight
F32
F32
[8192]
blk.33.ffn_up.weight
BF16
BF16
[8192, 28672]
blk.34
blk.34.attn_k.weight
BF16
BF16
[8192, 1024]
blk.34.attn_norm.weight
F32
F32
[8192]
blk.34.attn_output.weight
BF16
BF16
[8192, 8192]
blk.34.attn_q.weight
BF16
BF16
[8192, 8192]
blk.34.attn_v.weight
BF16
BF16
[8192, 1024]
blk.34.ffn_down.weight
BF16
BF16
[28672, 8192]
blk.34.ffn_gate.weight
BF16
BF16
[8192, 28672]
blk.34.ffn_norm.weight
F32
F32
[8192]
blk.34.ffn_up.weight
BF16
BF16
[8192, 28672]
blk.35
blk.35.attn_k.weight
BF16
BF16
[8192, 1024]
blk.35.attn_norm.weight
F32
F32
[8192]
blk.35.attn_output.weight
BF16
BF16
[8192, 8192]
blk.35.attn_q.weight
BF16
BF16
[8192, 8192]
blk.35.attn_v.weight
BF16
BF16
[8192, 1024]
blk.35.ffn_down.weight
BF16
BF16
[28672, 8192]
blk.35.ffn_gate.weight
BF16
BF16
[8192, 28672]
blk.35.ffn_norm.weight
F32
F32
[8192]
blk.35.ffn_up.weight
BF16
BF16
[8192, 28672]
blk.36
blk.36.attn_k.weight
BF16
BF16
[8192, 1024]
blk.36.attn_norm.weight
F32
F32
[8192]
blk.36.attn_output.weight
BF16
BF16
[8192, 8192]
blk.36.attn_q.weight
BF16
BF16
[8192, 8192]
blk.36.attn_v.weight
BF16
BF16
[8192, 1024]
blk.36.ffn_down.weight
BF16
BF16
[28672, 8192]
blk.36.ffn_gate.weight
BF16
BF16
[8192, 28672]
blk.36.ffn_norm.weight
F32
F32
[8192]
blk.36.ffn_up.weight
BF16
BF16
[8192, 28672]
blk.37
blk.37.attn_k.weight
BF16
BF16
[8192, 1024]
blk.37.attn_norm.weight
F32
F32
[8192]
blk.37.attn_output.weight
BF16
BF16
[8192, 8192]
blk.37.attn_q.weight
BF16
BF16
[8192, 8192]
blk.37.attn_v.weight
BF16
BF16
[8192, 1024]
blk.37.ffn_down.weight
BF16
BF16
[28672, 8192]
blk.37.ffn_gate.weight
BF16
BF16
[8192, 28672]
blk.37.ffn_norm.weight
F32
F32
[8192]
blk.37.ffn_up.weight
BF16
BF16
[8192, 28672]
blk.38
blk.38.attn_k.weight
BF16
BF16
[8192, 1024]
blk.38.attn_norm.weight
F32
F32
[8192]
blk.38.attn_output.weight
BF16
BF16
[8192, 8192]
blk.38.attn_q.weight
BF16
BF16
[8192, 8192]
blk.38.attn_v.weight
BF16
BF16
[8192, 1024]
blk.38.ffn_down.weight
BF16
BF16
[28672, 8192]
blk.38.ffn_gate.weight
BF16
BF16
[8192, 28672]
blk.38.ffn_norm.weight
F32
F32
[8192]
blk.38.ffn_up.weight
BF16
BF16
[8192, 28672]
blk.39
blk.39.attn_k.weight
BF16
BF16
[8192, 1024]
blk.39.attn_norm.weight
F32
F32
[8192]
blk.39.attn_output.weight
BF16
BF16
[8192, 8192]
blk.39.attn_q.weight
BF16
BF16
[8192, 8192]
blk.39.attn_v.weight
BF16
BF16
[8192, 1024]
blk.39.ffn_down.weight
BF16
BF16
[28672, 8192]
blk.39.ffn_gate.weight
BF16
BF16
[8192, 28672]
blk.39.ffn_norm.weight
F32
F32
[8192]
blk.39.ffn_up.weight
BF16
BF16
[8192, 28672]
blk.40
blk.40.attn_k.weight
BF16
BF16
[8192, 1024]
blk.40.attn_norm.weight
F32
F32
[8192]
blk.40.attn_output.weight
BF16
BF16
[8192, 8192]
blk.40.attn_q.weight
BF16
BF16
[8192, 8192]
blk.40.attn_v.weight
BF16
BF16
[8192, 1024]
blk.40.ffn_down.weight
BF16
BF16
[28672, 8192]
blk.40.ffn_gate.weight
BF16
BF16
[8192, 28672]
blk.40.ffn_norm.weight
F32
F32
[8192]
blk.40.ffn_up.weight
BF16
BF16
[8192, 28672]
blk.41
blk.41.attn_k.weight
BF16
BF16
[8192, 1024]
blk.41.attn_norm.weight
F32
F32
[8192]
blk.41.attn_output.weight
BF16
BF16
[8192, 8192]
blk.41.attn_q.weight
BF16
BF16
[8192, 8192]
blk.41.attn_v.weight
BF16
BF16
[8192, 1024]
blk.41.ffn_down.weight
BF16
BF16
[28672, 8192]
blk.41.ffn_gate.weight
BF16
BF16
[8192, 28672]
blk.41.ffn_norm.weight
F32
F32
[8192]
blk.41.ffn_up.weight
BF16
BF16
[8192, 28672]
blk.42
blk.42.ffn_down.weight
BF16
BF16
[7168, 8192]
blk.42.ffn_gate.weight
BF16
BF16
[8192, 7168]
blk.42.ffn_norm.weight
F32
F32
[8192]
blk.42.ffn_up.weight
BF16
BF16
[8192, 7168]
blk.43
blk.43.ffn_down.weight
BF16
BF16
[14336, 8192]
blk.43.ffn_gate.weight
BF16
BF16
[8192, 14336]
blk.43.ffn_norm.weight
F32
F32
[8192]
blk.43.ffn_up.weight
BF16
BF16
[8192, 14336]
blk.44
blk.44.ffn_down.weight
BF16
BF16
[14336, 8192]
blk.44.ffn_gate.weight
BF16
BF16
[8192, 14336]
blk.44.ffn_norm.weight
F32
F32
[8192]
blk.44.ffn_up.weight
BF16
BF16
[8192, 14336]
blk.45
blk.45.ffn_down.weight
BF16
BF16
[7168, 8192]
blk.45.ffn_gate.weight
BF16
BF16
[8192, 7168]
blk.45.ffn_norm.weight
F32
F32
[8192]
blk.45.ffn_up.weight
BF16
BF16
[8192, 7168]
blk.46
blk.46.ffn_down.weight
BF16
BF16
[28672, 8192]
blk.46.ffn_gate.weight
BF16
BF16
[8192, 28672]
blk.46.ffn_norm.weight
F32
F32
[8192]
blk.46.ffn_up.weight
BF16
BF16
[8192, 28672]
blk.47
blk.47.ffn_down.weight
BF16
BF16
[7168, 8192]
blk.47.ffn_gate.weight
BF16
BF16
[8192, 7168]
blk.47.ffn_norm.weight
F32
F32
[8192]
blk.47.ffn_up.weight
BF16
BF16
[8192, 7168]
blk.48
blk.48.ffn_down.weight
BF16
BF16
[14336, 8192]
blk.48.ffn_gate.weight
BF16
BF16
[8192, 14336]
blk.48.ffn_norm.weight
F32
F32
[8192]
blk.48.ffn_up.weight
BF16
BF16
[8192, 14336]
blk.49
blk.49.ffn_down.weight
BF16
BF16
[7168, 8192]
blk.49.ffn_gate.weight
BF16
BF16
[8192, 7168]
blk.49.ffn_norm.weight
F32
F32
[8192]
blk.49.ffn_up.weight
BF16
BF16
[8192, 7168]
blk.50
blk.50.ffn_down.weight
BF16
BF16
[7168, 8192]
blk.50.ffn_gate.weight
BF16
BF16
[8192, 7168]
blk.50.ffn_norm.weight
F32
F32
[8192]
blk.50.ffn_up.weight
BF16
BF16
[8192, 7168]
blk.51
blk.51.ffn_down.weight
BF16
BF16
[7168, 8192]
blk.51.ffn_gate.weight
BF16
BF16
[8192, 7168]
blk.51.ffn_norm.weight
F32
F32
[8192]
blk.51.ffn_up.weight
BF16
BF16
[8192, 7168]
blk.52
blk.52.attn_k.weight
BF16
BF16
[8192, 1024]
blk.52.attn_norm.weight
F32
F32
[8192]
blk.52.attn_output.weight
BF16
BF16
[8192, 8192]
blk.52.attn_q.weight
BF16
BF16
[8192, 8192]
blk.52.attn_v.weight
BF16
BF16
[8192, 1024]
blk.52.ffn_down.weight
BF16
BF16
[28672, 8192]
blk.52.ffn_gate.weight
BF16
BF16
[8192, 28672]
blk.52.ffn_norm.weight
F32
F32
[8192]
blk.52.ffn_up.weight
BF16
BF16
[8192, 28672]
blk.53
blk.53.ffn_down.weight
BF16
BF16
[7168, 8192]
blk.53.ffn_gate.weight
BF16
BF16
[8192, 7168]
blk.53.ffn_norm.weight
F32
F32
[8192]
blk.53.ffn_up.weight
BF16
BF16
[8192, 7168]
blk.54
blk.54.ffn_down.weight
BF16
BF16
[5632, 8192]
blk.54.ffn_gate.weight
BF16
BF16
[8192, 5632]
blk.54.ffn_norm.weight
F32
F32
[8192]
blk.54.ffn_up.weight
BF16
BF16
[8192, 5632]
blk.55
blk.55.ffn_down.weight
BF16
BF16
[5632, 8192]
blk.55.ffn_gate.weight
BF16
BF16
[8192, 5632]
blk.55.ffn_norm.weight
F32
F32
[8192]
blk.55.ffn_up.weight
BF16
BF16
[8192, 5632]
blk.56
blk.56.ffn_down.weight
BF16
BF16
[7168, 8192]
blk.56.ffn_gate.weight
BF16
BF16
[8192, 7168]
blk.56.ffn_norm.weight
F32
F32
[8192]
blk.56.ffn_up.weight
BF16
BF16
[8192, 7168]
blk.57
blk.57.ffn_down.weight
BF16
BF16
[5632, 8192]
blk.57.ffn_gate.weight
BF16
BF16
[8192, 5632]
blk.57.ffn_norm.weight
F32
F32
[8192]
blk.57.ffn_up.weight
BF16
BF16
[8192, 5632]
blk.58
blk.58.ffn_down.weight
BF16
BF16
[5632, 8192]
blk.58.ffn_gate.weight
BF16
BF16
[8192, 5632]
blk.58.ffn_norm.weight
F32
F32
[8192]
blk.58.ffn_up.weight
BF16
BF16
[8192, 5632]
blk.59
blk.59.ffn_down.weight
BF16
BF16
[5632, 8192]
blk.59.ffn_gate.weight
BF16
BF16
[8192, 5632]
blk.59.ffn_norm.weight
F32
F32
[8192]
blk.59.ffn_up.weight
BF16
BF16
[8192, 5632]
blk.60
blk.60.ffn_down.weight
BF16
BF16
[7168, 8192]
blk.60.ffn_gate.weight
BF16
BF16
[8192, 7168]
blk.60.ffn_norm.weight
F32
F32
[8192]
blk.60.ffn_up.weight
BF16
BF16
[8192, 7168]
blk.61
blk.61.ffn_down.weight
BF16
BF16
[7168, 8192]
blk.61.ffn_gate.weight
BF16
BF16
[8192, 7168]
blk.61.ffn_norm.weight
F32
F32
[8192]
blk.61.ffn_up.weight
BF16
BF16
[8192, 7168]
blk.62
blk.62.ffn_down.weight
BF16
BF16
[2816, 8192]
blk.62.ffn_gate.weight
BF16
BF16
[8192, 2816]
blk.62.ffn_norm.weight
F32
F32
[8192]
blk.62.ffn_up.weight
BF16
BF16
[8192, 2816]
blk.63
blk.63.ffn_down.weight
BF16
BF16
[2816, 8192]
blk.63.ffn_gate.weight
BF16
BF16
[8192, 2816]
blk.63.ffn_norm.weight
F32
F32
[8192]
blk.63.ffn_up.weight
BF16
BF16
[8192, 2816]
blk.64
blk.64.ffn_down.weight
BF16
BF16
[5632, 8192]
blk.64.ffn_gate.weight
BF16
BF16
[8192, 5632]
blk.64.ffn_norm.weight
F32
F32
[8192]
blk.64.ffn_up.weight
BF16
BF16
[8192, 5632]
blk.65
blk.65.ffn_down.weight
BF16
BF16
[5632, 8192]
blk.65.ffn_gate.weight
BF16
BF16
[8192, 5632]
blk.65.ffn_norm.weight
F32
F32
[8192]
blk.65.ffn_up.weight
BF16
BF16
[8192, 5632]
blk.66
blk.66.ffn_down.weight
BF16
BF16
[2816, 8192]
blk.66.ffn_gate.weight
BF16
BF16
[8192, 2816]
blk.66.ffn_norm.weight
F32
F32
[8192]
blk.66.ffn_up.weight
BF16
BF16
[8192, 2816]
blk.67
blk.67.ffn_down.weight
BF16
BF16
[2816, 8192]
blk.67.ffn_gate.weight
BF16
BF16
[8192, 2816]
blk.67.ffn_norm.weight
F32
F32
[8192]
blk.67.ffn_up.weight
BF16
BF16
[8192, 2816]
blk.68
blk.68.ffn_down.weight
BF16
BF16
[5632, 8192]
blk.68.ffn_gate.weight
BF16
BF16
[8192, 5632]
blk.68.ffn_norm.weight
F32
F32
[8192]
blk.68.ffn_up.weight
BF16
BF16
[8192, 5632]
blk.69
blk.69.ffn_down.weight
BF16
BF16
[2816, 8192]
blk.69.ffn_gate.weight
BF16
BF16
[8192, 2816]
blk.69.ffn_norm.weight
F32
F32
[8192]
blk.69.ffn_up.weight
BF16
BF16
[8192, 2816]
blk.70
blk.70.ffn_down.weight
BF16
BF16
[2816, 8192]
blk.70.ffn_gate.weight
BF16
BF16
[8192, 2816]
blk.70.ffn_norm.weight
F32
F32
[8192]
blk.70.ffn_up.weight
BF16
BF16
[8192, 2816]
blk.71
blk.71.attn_k.weight
BF16
BF16
[8192, 1024]
blk.71.attn_norm.weight
F32
F32
[8192]
blk.71.attn_output.weight
BF16
BF16
[8192, 8192]
blk.71.attn_q.weight
BF16
BF16
[8192, 8192]
blk.71.attn_v.weight
BF16
BF16
[8192, 1024]
blk.71.ffn_down.weight
BF16
BF16
[28672, 8192]
blk.71.ffn_gate.weight
BF16
BF16
[8192, 28672]
blk.71.ffn_norm.weight
F32
F32
[8192]
blk.71.ffn_up.weight
BF16
BF16
[8192, 28672]
blk.72
blk.72.attn_k.weight
BF16
BF16
[8192, 1024]
blk.72.attn_norm.weight
F32
F32
[8192]
blk.72.attn_output.weight
BF16
BF16
[8192, 8192]
blk.72.attn_q.weight
BF16
BF16
[8192, 8192]
blk.72.attn_v.weight
BF16
BF16
[8192, 1024]
blk.72.ffn_down.weight
BF16
BF16
[28672, 8192]
blk.72.ffn_gate.weight
BF16
BF16
[8192, 28672]
blk.72.ffn_norm.weight
F32
F32
[8192]
blk.72.ffn_up.weight
BF16
BF16
[8192, 28672]
blk.73
blk.73.attn_k.weight
BF16
BF16
[8192, 1024]
blk.73.attn_norm.weight
F32
F32
[8192]
blk.73.attn_output.weight
BF16
BF16
[8192, 8192]
blk.73.attn_q.weight
BF16
BF16
[8192, 8192]
blk.73.attn_v.weight
BF16
BF16
[8192, 1024]
blk.73.ffn_down.weight
BF16
BF16
[28672, 8192]
blk.73.ffn_gate.weight
BF16
BF16
[8192, 28672]
blk.73.ffn_norm.weight
F32
F32
[8192]
blk.73.ffn_up.weight
BF16
BF16
[8192, 28672]
blk.74
blk.74.attn_k.weight
BF16
BF16
[8192, 1024]
blk.74.attn_norm.weight
F32
F32
[8192]
blk.74.attn_output.weight
BF16
BF16
[8192, 8192]
blk.74.attn_q.weight
BF16
BF16
[8192, 8192]
blk.74.attn_v.weight
BF16
BF16
[8192, 1024]
blk.74.ffn_down.weight
BF16
BF16
[28672, 8192]
blk.74.ffn_gate.weight
BF16
BF16
[8192, 28672]
blk.74.ffn_norm.weight
F32
F32
[8192]
blk.74.ffn_up.weight
BF16
BF16
[8192, 28672]
blk.75
blk.75.attn_k.weight
BF16
BF16
[8192, 1024]
blk.75.attn_norm.weight
F32
F32
[8192]
blk.75.attn_output.weight
BF16
BF16
[8192, 8192]
blk.75.attn_q.weight
BF16
BF16
[8192, 8192]
blk.75.attn_v.weight
BF16
BF16
[8192, 1024]
blk.75.ffn_down.weight
BF16
BF16
[28672, 8192]
blk.75.ffn_gate.weight
BF16
BF16
[8192, 28672]
blk.75.ffn_norm.weight
F32
F32
[8192]
blk.75.ffn_up.weight
BF16
BF16
[8192, 28672]
blk.76
blk.76.attn_k.weight
BF16
BF16
[8192, 1024]
blk.76.attn_norm.weight
F32
F32
[8192]
blk.76.attn_output.weight
BF16
BF16
[8192, 8192]
blk.76.attn_q.weight
BF16
BF16
[8192, 8192]
blk.76.attn_v.weight
BF16
BF16
[8192, 1024]
blk.76.ffn_down.weight
BF16
BF16
[28672, 8192]
blk.76.ffn_gate.weight
BF16
BF16
[8192, 28672]
blk.76.ffn_norm.weight
F32
F32
[8192]
blk.76.ffn_up.weight
BF16
BF16
[8192, 28672]
blk.77
blk.77.attn_k.weight
BF16
BF16
[8192, 1024]
blk.77.attn_norm.weight
F32
F32
[8192]
blk.77.attn_output.weight
BF16
BF16
[8192, 8192]
blk.77.attn_q.weight
BF16
BF16
[8192, 8192]
blk.77.attn_v.weight
BF16
BF16
[8192, 1024]
blk.77.ffn_down.weight
BF16
BF16
[28672, 8192]
blk.77.ffn_gate.weight
BF16
BF16
[8192, 28672]
blk.77.ffn_norm.weight
F32
F32
[8192]
blk.77.ffn_up.weight
BF16
BF16
[8192, 28672]
blk.78
blk.78.attn_k.weight
BF16
BF16
[8192, 1024]
blk.78.attn_norm.weight
F32
F32
[8192]
blk.78.attn_output.weight
BF16
BF16
[8192, 8192]
blk.78.attn_q.weight
BF16
BF16
[8192, 8192]
blk.78.attn_v.weight
BF16
BF16
[8192, 1024]
blk.78.ffn_down.weight
BF16
BF16
[28672, 8192]
blk.78.ffn_gate.weight
BF16
BF16
[8192, 28672]
blk.78.ffn_norm.weight
F32
F32
[8192]
blk.78.ffn_up.weight
BF16
BF16
[8192, 28672]
blk.79
blk.79.attn_k.weight
BF16
BF16
[8192, 1024]
blk.79.attn_norm.weight
F32
F32
[8192]
blk.79.attn_output.weight
BF16
BF16
[8192, 8192]
blk.79.attn_q.weight
BF16
BF16
[8192, 8192]
blk.79.attn_v.weight
BF16
BF16
[8192, 1024]
blk.79.ffn_down.weight
BF16
BF16
[28672, 8192]
blk.79.ffn_gate.weight
BF16
BF16
[8192, 28672]
blk.79.ffn_norm.weight
F32
F32
[8192]
blk.79.ffn_up.weight
BF16
BF16
[8192, 28672]
output.weight
BF16
BF16
[8192, 128256]
rope_freqs.weight
F32
F32
[64]
output_norm.weight
F32
F32
[8192]