Models
Docs
Pricing
Sign in
Download
Models
Download
Docs
Pricing
Sign in
frob
/
mimo-v2.6-pro
:1023b
2
Downloads
Updated
18 hours ago
Cancel
vision
tools
thinking
1023b
mimo-v2.6-pro:1023b
...
/
model
ec5f4ed9c5f2 · 578GB
Metadata
general.architecture
mimo2
mimo2
general.file_type
(!unknown_file_type 38!)
(!unknown_file_type 38!)
mimo2.attention.head_count
128
128
mimo2.attention.head_count_kv
[8, 8, 8, 8, 8, ...]
[8, 8, 8, 8, 8, ...]
mimo2.attention.key_length
192
192
mimo2.attention.layer_norm_rms_epsilon
1e-05
1e-05
mimo2.attention.sliding_window
128
128
mimo2.attention.sliding_window_pattern
[0, 1, 1, 1, 1, ...]
[0, 1, 1, 1, 1, ...]
mimo2.attention.value_length
128
128
mimo2.attention.value_scale
0.612
0.612
mimo2.block_count
73
73
mimo2.context_length
1048576
1048576
mimo2.embedding_length
6144
6144
mimo2.expert_count
384
384
mimo2.expert_feed_forward_length
2048
2048
mimo2.expert_gating_func
2
2
mimo2.expert_group_count
1
1
mimo2.expert_group_used_count
1
1
mimo2.expert_used_count
8
8
mimo2.feed_forward_length
16384
16384
mimo2.nextn_predict_layers
3
3
mimo2.rope.dimension_count
64
64
mimo2.rope.freq_base
1e+07
1e+07
mimo2.rope.freq_base_swa
10000
10000
tokenizer.ggml.add_bos_token
false
false
tokenizer.ggml.eos_token_id
151645
151645
tokenizer.ggml.merges
[Ġ Ġ, ĠĠ ĠĠ, i n, Ġ t, ĠĠĠĠ ĠĠĠĠ, ...]
[Ġ Ġ, ĠĠ ĠĠ, i n, Ġ t, ĠĠĠĠ ĠĠĠĠ, ...]
tokenizer.ggml.model
gpt2
gpt2
tokenizer.ggml.padding_token_id
151643
151643
tokenizer.ggml.pre
qwen2
qwen2
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
[6144, 152576]
blk.0
blk.0.attn_norm.weight
F32
F32
[6144]
blk.0.attn_output.weight
BF16
BF16
[16384, 6144]
blk.0.attn_qkv.weight
BF16
BF16
[6144, 27136]
blk.0.ffn_down.weight
BF16
BF16
[16384, 6144]
blk.0.ffn_gate.weight
BF16
BF16
[6144, 16384]
blk.0.ffn_norm.weight
F32
F32
[6144]
blk.0.ffn_up.weight
BF16
BF16
[6144, 16384]
blk.1
blk.1.attn_norm.weight
F32
F32
[6144]
blk.1.attn_output.weight
BF16
BF16
[16384, 6144]
blk.1.attn_qkv.weight
BF16
BF16
[6144, 27136]
blk.1.attn_sinks.weight
F32
F32
[128]
blk.1.exp_probs_b.bias
F32
F32
[384]
blk.1.ffn_down_exps.weight
MXFP4
MXFP4
[2048, 6144, 384]
blk.1.ffn_gate_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.1.ffn_gate_inp.weight
F32
F32
[6144, 384]
blk.1.ffn_norm.weight
F32
F32
[6144]
blk.1.ffn_up_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.2
blk.2.attn_norm.weight
F32
F32
[6144]
blk.2.attn_output.weight
BF16
BF16
[16384, 6144]
blk.2.attn_qkv.weight
BF16
BF16
[6144, 27136]
blk.2.attn_sinks.weight
F32
F32
[128]
blk.2.exp_probs_b.bias
F32
F32
[384]
blk.2.ffn_down_exps.weight
MXFP4
MXFP4
[2048, 6144, 384]
blk.2.ffn_gate_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.2.ffn_gate_inp.weight
F32
F32
[6144, 384]
blk.2.ffn_norm.weight
F32
F32
[6144]
blk.2.ffn_up_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.3
blk.3.attn_norm.weight
F32
F32
[6144]
blk.3.attn_output.weight
BF16
BF16
[16384, 6144]
blk.3.attn_qkv.weight
BF16
BF16
[6144, 27136]
blk.3.attn_sinks.weight
F32
F32
[128]
blk.3.exp_probs_b.bias
F32
F32
[384]
blk.3.ffn_down_exps.weight
MXFP4
MXFP4
[2048, 6144, 384]
blk.3.ffn_gate_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.3.ffn_gate_inp.weight
F32
F32
[6144, 384]
blk.3.ffn_norm.weight
F32
F32
[6144]
blk.3.ffn_up_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.4
blk.4.attn_norm.weight
F32
F32
[6144]
blk.4.attn_output.weight
BF16
BF16
[16384, 6144]
blk.4.attn_qkv.weight
BF16
BF16
[6144, 27136]
blk.4.attn_sinks.weight
F32
F32
[128]
blk.4.exp_probs_b.bias
F32
F32
[384]
blk.4.ffn_down_exps.weight
MXFP4
MXFP4
[2048, 6144, 384]
blk.4.ffn_gate_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.4.ffn_gate_inp.weight
F32
F32
[6144, 384]
blk.4.ffn_norm.weight
F32
F32
[6144]
blk.4.ffn_up_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.5
blk.5.attn_norm.weight
F32
F32
[6144]
blk.5.attn_output.weight
BF16
BF16
[16384, 6144]
blk.5.attn_qkv.weight
BF16
BF16
[6144, 27136]
blk.5.attn_sinks.weight
F32
F32
[128]
blk.5.exp_probs_b.bias
F32
F32
[384]
blk.5.ffn_down_exps.weight
MXFP4
MXFP4
[2048, 6144, 384]
blk.5.ffn_gate_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.5.ffn_gate_inp.weight
F32
F32
[6144, 384]
blk.5.ffn_norm.weight
F32
F32
[6144]
blk.5.ffn_up_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.6
blk.6.attn_norm.weight
F32
F32
[6144]
blk.6.attn_output.weight
BF16
BF16
[16384, 6144]
blk.6.attn_qkv.weight
BF16
BF16
[6144, 27136]
blk.6.attn_sinks.weight
F32
F32
[128]
blk.6.exp_probs_b.bias
F32
F32
[384]
blk.6.ffn_down_exps.weight
MXFP4
MXFP4
[2048, 6144, 384]
blk.6.ffn_gate_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.6.ffn_gate_inp.weight
F32
F32
[6144, 384]
blk.6.ffn_norm.weight
F32
F32
[6144]
blk.6.ffn_up_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.7
blk.7.attn_norm.weight
F32
F32
[6144]
blk.7.attn_output.weight
BF16
BF16
[16384, 6144]
blk.7.attn_qkv.weight
BF16
BF16
[6144, 27136]
blk.7.exp_probs_b.bias
F32
F32
[384]
blk.7.ffn_down_exps.weight
MXFP4
MXFP4
[2048, 6144, 384]
blk.7.ffn_gate_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.7.ffn_gate_inp.weight
F32
F32
[6144, 384]
blk.7.ffn_norm.weight
F32
F32
[6144]
blk.7.ffn_up_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.8
blk.8.attn_norm.weight
F32
F32
[6144]
blk.8.attn_output.weight
BF16
BF16
[16384, 6144]
blk.8.attn_qkv.weight
BF16
BF16
[6144, 27136]
blk.8.attn_sinks.weight
F32
F32
[128]
blk.8.exp_probs_b.bias
F32
F32
[384]
blk.8.ffn_down_exps.weight
MXFP4
MXFP4
[2048, 6144, 384]
blk.8.ffn_gate_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.8.ffn_gate_inp.weight
F32
F32
[6144, 384]
blk.8.ffn_norm.weight
F32
F32
[6144]
blk.8.ffn_up_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.9
blk.9.attn_norm.weight
F32
F32
[6144]
blk.9.attn_output.weight
BF16
BF16
[16384, 6144]
blk.9.attn_qkv.weight
BF16
BF16
[6144, 27136]
blk.9.attn_sinks.weight
F32
F32
[128]
blk.9.exp_probs_b.bias
F32
F32
[384]
blk.9.ffn_down_exps.weight
MXFP4
MXFP4
[2048, 6144, 384]
blk.9.ffn_gate_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.9.ffn_gate_inp.weight
F32
F32
[6144, 384]
blk.9.ffn_norm.weight
F32
F32
[6144]
blk.9.ffn_up_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.10
blk.10.attn_norm.weight
F32
F32
[6144]
blk.10.attn_output.weight
BF16
BF16
[16384, 6144]
blk.10.attn_qkv.weight
BF16
BF16
[6144, 27136]
blk.10.attn_sinks.weight
F32
F32
[128]
blk.10.exp_probs_b.bias
F32
F32
[384]
blk.10.ffn_down_exps.weight
MXFP4
MXFP4
[2048, 6144, 384]
blk.10.ffn_gate_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.10.ffn_gate_inp.weight
F32
F32
[6144, 384]
blk.10.ffn_norm.weight
F32
F32
[6144]
blk.10.ffn_up_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.11
blk.11.attn_norm.weight
F32
F32
[6144]
blk.11.attn_output.weight
BF16
BF16
[16384, 6144]
blk.11.attn_qkv.weight
BF16
BF16
[6144, 27136]
blk.11.attn_sinks.weight
F32
F32
[128]
blk.11.exp_probs_b.bias
F32
F32
[384]
blk.11.ffn_down_exps.weight
MXFP4
MXFP4
[2048, 6144, 384]
blk.11.ffn_gate_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.11.ffn_gate_inp.weight
F32
F32
[6144, 384]
blk.11.ffn_norm.weight
F32
F32
[6144]
blk.11.ffn_up_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.12
blk.12.attn_norm.weight
F32
F32
[6144]
blk.12.attn_output.weight
BF16
BF16
[16384, 6144]
blk.12.attn_qkv.weight
BF16
BF16
[6144, 27136]
blk.12.attn_sinks.weight
F32
F32
[128]
blk.12.exp_probs_b.bias
F32
F32
[384]
blk.12.ffn_down_exps.weight
MXFP4
MXFP4
[2048, 6144, 384]
blk.12.ffn_gate_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.12.ffn_gate_inp.weight
F32
F32
[6144, 384]
blk.12.ffn_norm.weight
F32
F32
[6144]
blk.12.ffn_up_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.13
blk.13.attn_norm.weight
F32
F32
[6144]
blk.13.attn_output.weight
BF16
BF16
[16384, 6144]
blk.13.attn_qkv.weight
BF16
BF16
[6144, 27136]
blk.13.attn_sinks.weight
F32
F32
[128]
blk.13.exp_probs_b.bias
F32
F32
[384]
blk.13.ffn_down_exps.weight
MXFP4
MXFP4
[2048, 6144, 384]
blk.13.ffn_gate_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.13.ffn_gate_inp.weight
F32
F32
[6144, 384]
blk.13.ffn_norm.weight
F32
F32
[6144]
blk.13.ffn_up_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.14
blk.14.attn_norm.weight
F32
F32
[6144]
blk.14.attn_output.weight
BF16
BF16
[16384, 6144]
blk.14.attn_qkv.weight
BF16
BF16
[6144, 27136]
blk.14.attn_sinks.weight
F32
F32
[128]
blk.14.exp_probs_b.bias
F32
F32
[384]
blk.14.ffn_down_exps.weight
MXFP4
MXFP4
[2048, 6144, 384]
blk.14.ffn_gate_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.14.ffn_gate_inp.weight
F32
F32
[6144, 384]
blk.14.ffn_norm.weight
F32
F32
[6144]
blk.14.ffn_up_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.15
blk.15.attn_norm.weight
F32
F32
[6144]
blk.15.attn_output.weight
BF16
BF16
[16384, 6144]
blk.15.attn_qkv.weight
BF16
BF16
[6144, 27136]
blk.15.exp_probs_b.bias
F32
F32
[384]
blk.15.ffn_down_exps.weight
MXFP4
MXFP4
[2048, 6144, 384]
blk.15.ffn_gate_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.15.ffn_gate_inp.weight
F32
F32
[6144, 384]
blk.15.ffn_norm.weight
F32
F32
[6144]
blk.15.ffn_up_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.16
blk.16.attn_norm.weight
F32
F32
[6144]
blk.16.attn_output.weight
BF16
BF16
[16384, 6144]
blk.16.attn_qkv.weight
BF16
BF16
[6144, 27136]
blk.16.attn_sinks.weight
F32
F32
[128]
blk.16.exp_probs_b.bias
F32
F32
[384]
blk.16.ffn_down_exps.weight
MXFP4
MXFP4
[2048, 6144, 384]
blk.16.ffn_gate_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.16.ffn_gate_inp.weight
F32
F32
[6144, 384]
blk.16.ffn_norm.weight
F32
F32
[6144]
blk.16.ffn_up_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.17
blk.17.attn_norm.weight
F32
F32
[6144]
blk.17.attn_output.weight
BF16
BF16
[16384, 6144]
blk.17.attn_qkv.weight
BF16
BF16
[6144, 27136]
blk.17.attn_sinks.weight
F32
F32
[128]
blk.17.exp_probs_b.bias
F32
F32
[384]
blk.17.ffn_down_exps.weight
MXFP4
MXFP4
[2048, 6144, 384]
blk.17.ffn_gate_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.17.ffn_gate_inp.weight
F32
F32
[6144, 384]
blk.17.ffn_norm.weight
F32
F32
[6144]
blk.17.ffn_up_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.18
blk.18.attn_norm.weight
F32
F32
[6144]
blk.18.attn_output.weight
BF16
BF16
[16384, 6144]
blk.18.attn_qkv.weight
BF16
BF16
[6144, 27136]
blk.18.attn_sinks.weight
F32
F32
[128]
blk.18.exp_probs_b.bias
F32
F32
[384]
blk.18.ffn_down_exps.weight
MXFP4
MXFP4
[2048, 6144, 384]
blk.18.ffn_gate_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.18.ffn_gate_inp.weight
F32
F32
[6144, 384]
blk.18.ffn_norm.weight
F32
F32
[6144]
blk.18.ffn_up_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.19
blk.19.attn_norm.weight
F32
F32
[6144]
blk.19.attn_output.weight
BF16
BF16
[16384, 6144]
blk.19.attn_qkv.weight
BF16
BF16
[6144, 27136]
blk.19.attn_sinks.weight
F32
F32
[128]
blk.19.exp_probs_b.bias
F32
F32
[384]
blk.19.ffn_down_exps.weight
MXFP4
MXFP4
[2048, 6144, 384]
blk.19.ffn_gate_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.19.ffn_gate_inp.weight
F32
F32
[6144, 384]
blk.19.ffn_norm.weight
F32
F32
[6144]
blk.19.ffn_up_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.20
blk.20.attn_norm.weight
F32
F32
[6144]
blk.20.attn_output.weight
BF16
BF16
[16384, 6144]
blk.20.attn_qkv.weight
BF16
BF16
[6144, 27136]
blk.20.attn_sinks.weight
F32
F32
[128]
blk.20.exp_probs_b.bias
F32
F32
[384]
blk.20.ffn_down_exps.weight
MXFP4
MXFP4
[2048, 6144, 384]
blk.20.ffn_gate_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.20.ffn_gate_inp.weight
F32
F32
[6144, 384]
blk.20.ffn_norm.weight
F32
F32
[6144]
blk.20.ffn_up_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.21
blk.21.attn_norm.weight
F32
F32
[6144]
blk.21.attn_output.weight
BF16
BF16
[16384, 6144]
blk.21.attn_qkv.weight
BF16
BF16
[6144, 27136]
blk.21.attn_sinks.weight
F32
F32
[128]
blk.21.exp_probs_b.bias
F32
F32
[384]
blk.21.ffn_down_exps.weight
MXFP4
MXFP4
[2048, 6144, 384]
blk.21.ffn_gate_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.21.ffn_gate_inp.weight
F32
F32
[6144, 384]
blk.21.ffn_norm.weight
F32
F32
[6144]
blk.21.ffn_up_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.22
blk.22.attn_norm.weight
F32
F32
[6144]
blk.22.attn_output.weight
BF16
BF16
[16384, 6144]
blk.22.attn_qkv.weight
BF16
BF16
[6144, 27136]
blk.22.attn_sinks.weight
F32
F32
[128]
blk.22.exp_probs_b.bias
F32
F32
[384]
blk.22.ffn_down_exps.weight
MXFP4
MXFP4
[2048, 6144, 384]
blk.22.ffn_gate_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.22.ffn_gate_inp.weight
F32
F32
[6144, 384]
blk.22.ffn_norm.weight
F32
F32
[6144]
blk.22.ffn_up_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.23
blk.23.attn_norm.weight
F32
F32
[6144]
blk.23.attn_output.weight
BF16
BF16
[16384, 6144]
blk.23.attn_qkv.weight
BF16
BF16
[6144, 27136]
blk.23.exp_probs_b.bias
F32
F32
[384]
blk.23.ffn_down_exps.weight
MXFP4
MXFP4
[2048, 6144, 384]
blk.23.ffn_gate_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.23.ffn_gate_inp.weight
F32
F32
[6144, 384]
blk.23.ffn_norm.weight
F32
F32
[6144]
blk.23.ffn_up_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.24
blk.24.attn_norm.weight
F32
F32
[6144]
blk.24.attn_output.weight
BF16
BF16
[16384, 6144]
blk.24.attn_qkv.weight
BF16
BF16
[6144, 27136]
blk.24.attn_sinks.weight
F32
F32
[128]
blk.24.exp_probs_b.bias
F32
F32
[384]
blk.24.ffn_down_exps.weight
MXFP4
MXFP4
[2048, 6144, 384]
blk.24.ffn_gate_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.24.ffn_gate_inp.weight
F32
F32
[6144, 384]
blk.24.ffn_norm.weight
F32
F32
[6144]
blk.24.ffn_up_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.25
blk.25.attn_norm.weight
F32
F32
[6144]
blk.25.attn_output.weight
BF16
BF16
[16384, 6144]
blk.25.attn_qkv.weight
BF16
BF16
[6144, 27136]
blk.25.attn_sinks.weight
F32
F32
[128]
blk.25.exp_probs_b.bias
F32
F32
[384]
blk.25.ffn_down_exps.weight
MXFP4
MXFP4
[2048, 6144, 384]
blk.25.ffn_gate_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.25.ffn_gate_inp.weight
F32
F32
[6144, 384]
blk.25.ffn_norm.weight
F32
F32
[6144]
blk.25.ffn_up_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.26
blk.26.attn_norm.weight
F32
F32
[6144]
blk.26.attn_output.weight
BF16
BF16
[16384, 6144]
blk.26.attn_qkv.weight
BF16
BF16
[6144, 27136]
blk.26.attn_sinks.weight
F32
F32
[128]
blk.26.exp_probs_b.bias
F32
F32
[384]
blk.26.ffn_down_exps.weight
MXFP4
MXFP4
[2048, 6144, 384]
blk.26.ffn_gate_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.26.ffn_gate_inp.weight
F32
F32
[6144, 384]
blk.26.ffn_norm.weight
F32
F32
[6144]
blk.26.ffn_up_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.27
blk.27.attn_norm.weight
F32
F32
[6144]
blk.27.attn_output.weight
BF16
BF16
[16384, 6144]
blk.27.attn_qkv.weight
BF16
BF16
[6144, 27136]
blk.27.attn_sinks.weight
F32
F32
[128]
blk.27.exp_probs_b.bias
F32
F32
[384]
blk.27.ffn_down_exps.weight
MXFP4
MXFP4
[2048, 6144, 384]
blk.27.ffn_gate_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.27.ffn_gate_inp.weight
F32
F32
[6144, 384]
blk.27.ffn_norm.weight
F32
F32
[6144]
blk.27.ffn_up_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.28
blk.28.attn_norm.weight
F32
F32
[6144]
blk.28.attn_output.weight
BF16
BF16
[16384, 6144]
blk.28.attn_qkv.weight
BF16
BF16
[6144, 27136]
blk.28.attn_sinks.weight
F32
F32
[128]
blk.28.exp_probs_b.bias
F32
F32
[384]
blk.28.ffn_down_exps.weight
MXFP4
MXFP4
[2048, 6144, 384]
blk.28.ffn_gate_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.28.ffn_gate_inp.weight
F32
F32
[6144, 384]
blk.28.ffn_norm.weight
F32
F32
[6144]
blk.28.ffn_up_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.29
blk.29.attn_norm.weight
F32
F32
[6144]
blk.29.attn_output.weight
BF16
BF16
[16384, 6144]
blk.29.attn_qkv.weight
BF16
BF16
[6144, 27136]
blk.29.attn_sinks.weight
F32
F32
[128]
blk.29.exp_probs_b.bias
F32
F32
[384]
blk.29.ffn_down_exps.weight
MXFP4
MXFP4
[2048, 6144, 384]
blk.29.ffn_gate_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.29.ffn_gate_inp.weight
F32
F32
[6144, 384]
blk.29.ffn_norm.weight
F32
F32
[6144]
blk.29.ffn_up_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.30
blk.30.attn_norm.weight
F32
F32
[6144]
blk.30.attn_output.weight
BF16
BF16
[16384, 6144]
blk.30.attn_qkv.weight
BF16
BF16
[6144, 27136]
blk.30.attn_sinks.weight
F32
F32
[128]
blk.30.exp_probs_b.bias
F32
F32
[384]
blk.30.ffn_down_exps.weight
MXFP4
MXFP4
[2048, 6144, 384]
blk.30.ffn_gate_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.30.ffn_gate_inp.weight
F32
F32
[6144, 384]
blk.30.ffn_norm.weight
F32
F32
[6144]
blk.30.ffn_up_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.31
blk.31.attn_norm.weight
F32
F32
[6144]
blk.31.attn_output.weight
BF16
BF16
[16384, 6144]
blk.31.attn_qkv.weight
BF16
BF16
[6144, 27136]
blk.31.exp_probs_b.bias
F32
F32
[384]
blk.31.ffn_down_exps.weight
MXFP4
MXFP4
[2048, 6144, 384]
blk.31.ffn_gate_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.31.ffn_gate_inp.weight
F32
F32
[6144, 384]
blk.31.ffn_norm.weight
F32
F32
[6144]
blk.31.ffn_up_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.32
blk.32.attn_norm.weight
F32
F32
[6144]
blk.32.attn_output.weight
BF16
BF16
[16384, 6144]
blk.32.attn_qkv.weight
BF16
BF16
[6144, 27136]
blk.32.attn_sinks.weight
F32
F32
[128]
blk.32.exp_probs_b.bias
F32
F32
[384]
blk.32.ffn_down_exps.weight
MXFP4
MXFP4
[2048, 6144, 384]
blk.32.ffn_gate_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.32.ffn_gate_inp.weight
F32
F32
[6144, 384]
blk.32.ffn_norm.weight
F32
F32
[6144]
blk.32.ffn_up_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.33
blk.33.attn_norm.weight
F32
F32
[6144]
blk.33.attn_output.weight
BF16
BF16
[16384, 6144]
blk.33.attn_qkv.weight
BF16
BF16
[6144, 27136]
blk.33.attn_sinks.weight
F32
F32
[128]
blk.33.exp_probs_b.bias
F32
F32
[384]
blk.33.ffn_down_exps.weight
MXFP4
MXFP4
[2048, 6144, 384]
blk.33.ffn_gate_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.33.ffn_gate_inp.weight
F32
F32
[6144, 384]
blk.33.ffn_norm.weight
F32
F32
[6144]
blk.33.ffn_up_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.34
blk.34.attn_norm.weight
F32
F32
[6144]
blk.34.attn_output.weight
BF16
BF16
[16384, 6144]
blk.34.attn_qkv.weight
BF16
BF16
[6144, 27136]
blk.34.attn_sinks.weight
F32
F32
[128]
blk.34.exp_probs_b.bias
F32
F32
[384]
blk.34.ffn_down_exps.weight
MXFP4
MXFP4
[2048, 6144, 384]
blk.34.ffn_gate_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.34.ffn_gate_inp.weight
F32
F32
[6144, 384]
blk.34.ffn_norm.weight
F32
F32
[6144]
blk.34.ffn_up_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.35
blk.35.attn_norm.weight
F32
F32
[6144]
blk.35.attn_output.weight
BF16
BF16
[16384, 6144]
blk.35.attn_qkv.weight
BF16
BF16
[6144, 27136]
blk.35.attn_sinks.weight
F32
F32
[128]
blk.35.exp_probs_b.bias
F32
F32
[384]
blk.35.ffn_down_exps.weight
MXFP4
MXFP4
[2048, 6144, 384]
blk.35.ffn_gate_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.35.ffn_gate_inp.weight
F32
F32
[6144, 384]
blk.35.ffn_norm.weight
F32
F32
[6144]
blk.35.ffn_up_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.36
blk.36.attn_norm.weight
F32
F32
[6144]
blk.36.attn_output.weight
BF16
BF16
[16384, 6144]
blk.36.attn_qkv.weight
BF16
BF16
[6144, 27136]
blk.36.attn_sinks.weight
F32
F32
[128]
blk.36.exp_probs_b.bias
F32
F32
[384]
blk.36.ffn_down_exps.weight
MXFP4
MXFP4
[2048, 6144, 384]
blk.36.ffn_gate_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.36.ffn_gate_inp.weight
F32
F32
[6144, 384]
blk.36.ffn_norm.weight
F32
F32
[6144]
blk.36.ffn_up_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.37
blk.37.attn_norm.weight
F32
F32
[6144]
blk.37.attn_output.weight
BF16
BF16
[16384, 6144]
blk.37.attn_qkv.weight
BF16
BF16
[6144, 27136]
blk.37.attn_sinks.weight
F32
F32
[128]
blk.37.exp_probs_b.bias
F32
F32
[384]
blk.37.ffn_down_exps.weight
MXFP4
MXFP4
[2048, 6144, 384]
blk.37.ffn_gate_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.37.ffn_gate_inp.weight
F32
F32
[6144, 384]
blk.37.ffn_norm.weight
F32
F32
[6144]
blk.37.ffn_up_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.38
blk.38.attn_norm.weight
F32
F32
[6144]
blk.38.attn_output.weight
BF16
BF16
[16384, 6144]
blk.38.attn_qkv.weight
BF16
BF16
[6144, 27136]
blk.38.attn_sinks.weight
F32
F32
[128]
blk.38.exp_probs_b.bias
F32
F32
[384]
blk.38.ffn_down_exps.weight
MXFP4
MXFP4
[2048, 6144, 384]
blk.38.ffn_gate_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.38.ffn_gate_inp.weight
F32
F32
[6144, 384]
blk.38.ffn_norm.weight
F32
F32
[6144]
blk.38.ffn_up_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.39
blk.39.attn_norm.weight
F32
F32
[6144]
blk.39.attn_output.weight
BF16
BF16
[16384, 6144]
blk.39.attn_qkv.weight
BF16
BF16
[6144, 27136]
blk.39.exp_probs_b.bias
F32
F32
[384]
blk.39.ffn_down_exps.weight
MXFP4
MXFP4
[2048, 6144, 384]
blk.39.ffn_gate_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.39.ffn_gate_inp.weight
F32
F32
[6144, 384]
blk.39.ffn_norm.weight
F32
F32
[6144]
blk.39.ffn_up_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.40
blk.40.attn_norm.weight
F32
F32
[6144]
blk.40.attn_output.weight
BF16
BF16
[16384, 6144]
blk.40.attn_qkv.weight
BF16
BF16
[6144, 27136]
blk.40.attn_sinks.weight
F32
F32
[128]
blk.40.exp_probs_b.bias
F32
F32
[384]
blk.40.ffn_down_exps.weight
MXFP4
MXFP4
[2048, 6144, 384]
blk.40.ffn_gate_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.40.ffn_gate_inp.weight
F32
F32
[6144, 384]
blk.40.ffn_norm.weight
F32
F32
[6144]
blk.40.ffn_up_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.41
blk.41.attn_norm.weight
F32
F32
[6144]
blk.41.attn_output.weight
BF16
BF16
[16384, 6144]
blk.41.attn_qkv.weight
BF16
BF16
[6144, 27136]
blk.41.attn_sinks.weight
F32
F32
[128]
blk.41.exp_probs_b.bias
F32
F32
[384]
blk.41.ffn_down_exps.weight
MXFP4
MXFP4
[2048, 6144, 384]
blk.41.ffn_gate_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.41.ffn_gate_inp.weight
F32
F32
[6144, 384]
blk.41.ffn_norm.weight
F32
F32
[6144]
blk.41.ffn_up_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.42
blk.42.attn_norm.weight
F32
F32
[6144]
blk.42.attn_output.weight
BF16
BF16
[16384, 6144]
blk.42.attn_qkv.weight
BF16
BF16
[6144, 27136]
blk.42.attn_sinks.weight
F32
F32
[128]
blk.42.exp_probs_b.bias
F32
F32
[384]
blk.42.ffn_down_exps.weight
MXFP4
MXFP4
[2048, 6144, 384]
blk.42.ffn_gate_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.42.ffn_gate_inp.weight
F32
F32
[6144, 384]
blk.42.ffn_norm.weight
F32
F32
[6144]
blk.42.ffn_up_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.43
blk.43.attn_norm.weight
F32
F32
[6144]
blk.43.attn_output.weight
BF16
BF16
[16384, 6144]
blk.43.attn_qkv.weight
BF16
BF16
[6144, 27136]
blk.43.attn_sinks.weight
F32
F32
[128]
blk.43.exp_probs_b.bias
F32
F32
[384]
blk.43.ffn_down_exps.weight
MXFP4
MXFP4
[2048, 6144, 384]
blk.43.ffn_gate_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.43.ffn_gate_inp.weight
F32
F32
[6144, 384]
blk.43.ffn_norm.weight
F32
F32
[6144]
blk.43.ffn_up_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.44
blk.44.attn_norm.weight
F32
F32
[6144]
blk.44.attn_output.weight
BF16
BF16
[16384, 6144]
blk.44.attn_qkv.weight
BF16
BF16
[6144, 27136]
blk.44.attn_sinks.weight
F32
F32
[128]
blk.44.exp_probs_b.bias
F32
F32
[384]
blk.44.ffn_down_exps.weight
MXFP4
MXFP4
[2048, 6144, 384]
blk.44.ffn_gate_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.44.ffn_gate_inp.weight
F32
F32
[6144, 384]
blk.44.ffn_norm.weight
F32
F32
[6144]
blk.44.ffn_up_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.45
blk.45.attn_norm.weight
F32
F32
[6144]
blk.45.attn_output.weight
BF16
BF16
[16384, 6144]
blk.45.attn_qkv.weight
BF16
BF16
[6144, 27136]
blk.45.attn_sinks.weight
F32
F32
[128]
blk.45.exp_probs_b.bias
F32
F32
[384]
blk.45.ffn_down_exps.weight
MXFP4
MXFP4
[2048, 6144, 384]
blk.45.ffn_gate_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.45.ffn_gate_inp.weight
F32
F32
[6144, 384]
blk.45.ffn_norm.weight
F32
F32
[6144]
blk.45.ffn_up_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.46
blk.46.attn_norm.weight
F32
F32
[6144]
blk.46.attn_output.weight
BF16
BF16
[16384, 6144]
blk.46.attn_qkv.weight
BF16
BF16
[6144, 27136]
blk.46.attn_sinks.weight
F32
F32
[128]
blk.46.exp_probs_b.bias
F32
F32
[384]
blk.46.ffn_down_exps.weight
MXFP4
MXFP4
[2048, 6144, 384]
blk.46.ffn_gate_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.46.ffn_gate_inp.weight
F32
F32
[6144, 384]
blk.46.ffn_norm.weight
F32
F32
[6144]
blk.46.ffn_up_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.47
blk.47.attn_norm.weight
F32
F32
[6144]
blk.47.attn_output.weight
BF16
BF16
[16384, 6144]
blk.47.attn_qkv.weight
BF16
BF16
[6144, 27136]
blk.47.exp_probs_b.bias
F32
F32
[384]
blk.47.ffn_down_exps.weight
MXFP4
MXFP4
[2048, 6144, 384]
blk.47.ffn_gate_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.47.ffn_gate_inp.weight
F32
F32
[6144, 384]
blk.47.ffn_norm.weight
F32
F32
[6144]
blk.47.ffn_up_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.48
blk.48.attn_norm.weight
F32
F32
[6144]
blk.48.attn_output.weight
BF16
BF16
[16384, 6144]
blk.48.attn_qkv.weight
BF16
BF16
[6144, 27136]
blk.48.attn_sinks.weight
F32
F32
[128]
blk.48.exp_probs_b.bias
F32
F32
[384]
blk.48.ffn_down_exps.weight
MXFP4
MXFP4
[2048, 6144, 384]
blk.48.ffn_gate_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.48.ffn_gate_inp.weight
F32
F32
[6144, 384]
blk.48.ffn_norm.weight
F32
F32
[6144]
blk.48.ffn_up_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.49
blk.49.attn_norm.weight
F32
F32
[6144]
blk.49.attn_output.weight
BF16
BF16
[16384, 6144]
blk.49.attn_qkv.weight
BF16
BF16
[6144, 27136]
blk.49.attn_sinks.weight
F32
F32
[128]
blk.49.exp_probs_b.bias
F32
F32
[384]
blk.49.ffn_down_exps.weight
MXFP4
MXFP4
[2048, 6144, 384]
blk.49.ffn_gate_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.49.ffn_gate_inp.weight
F32
F32
[6144, 384]
blk.49.ffn_norm.weight
F32
F32
[6144]
blk.49.ffn_up_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.50
blk.50.attn_norm.weight
F32
F32
[6144]
blk.50.attn_output.weight
BF16
BF16
[16384, 6144]
blk.50.attn_qkv.weight
BF16
BF16
[6144, 27136]
blk.50.attn_sinks.weight
F32
F32
[128]
blk.50.exp_probs_b.bias
F32
F32
[384]
blk.50.ffn_down_exps.weight
MXFP4
MXFP4
[2048, 6144, 384]
blk.50.ffn_gate_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.50.ffn_gate_inp.weight
F32
F32
[6144, 384]
blk.50.ffn_norm.weight
F32
F32
[6144]
blk.50.ffn_up_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.51
blk.51.attn_norm.weight
F32
F32
[6144]
blk.51.attn_output.weight
BF16
BF16
[16384, 6144]
blk.51.attn_qkv.weight
BF16
BF16
[6144, 27136]
blk.51.attn_sinks.weight
F32
F32
[128]
blk.51.exp_probs_b.bias
F32
F32
[384]
blk.51.ffn_down_exps.weight
MXFP4
MXFP4
[2048, 6144, 384]
blk.51.ffn_gate_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.51.ffn_gate_inp.weight
F32
F32
[6144, 384]
blk.51.ffn_norm.weight
F32
F32
[6144]
blk.51.ffn_up_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.52
blk.52.attn_norm.weight
F32
F32
[6144]
blk.52.attn_output.weight
BF16
BF16
[16384, 6144]
blk.52.attn_qkv.weight
BF16
BF16
[6144, 27136]
blk.52.attn_sinks.weight
F32
F32
[128]
blk.52.exp_probs_b.bias
F32
F32
[384]
blk.52.ffn_down_exps.weight
MXFP4
MXFP4
[2048, 6144, 384]
blk.52.ffn_gate_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.52.ffn_gate_inp.weight
F32
F32
[6144, 384]
blk.52.ffn_norm.weight
F32
F32
[6144]
blk.52.ffn_up_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.53
blk.53.attn_norm.weight
F32
F32
[6144]
blk.53.attn_output.weight
BF16
BF16
[16384, 6144]
blk.53.attn_qkv.weight
BF16
BF16
[6144, 27136]
blk.53.attn_sinks.weight
F32
F32
[128]
blk.53.exp_probs_b.bias
F32
F32
[384]
blk.53.ffn_down_exps.weight
MXFP4
MXFP4
[2048, 6144, 384]
blk.53.ffn_gate_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.53.ffn_gate_inp.weight
F32
F32
[6144, 384]
blk.53.ffn_norm.weight
F32
F32
[6144]
blk.53.ffn_up_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.54
blk.54.attn_norm.weight
F32
F32
[6144]
blk.54.attn_output.weight
BF16
BF16
[16384, 6144]
blk.54.attn_qkv.weight
BF16
BF16
[6144, 27136]
blk.54.attn_sinks.weight
F32
F32
[128]
blk.54.exp_probs_b.bias
F32
F32
[384]
blk.54.ffn_down_exps.weight
MXFP4
MXFP4
[2048, 6144, 384]
blk.54.ffn_gate_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.54.ffn_gate_inp.weight
F32
F32
[6144, 384]
blk.54.ffn_norm.weight
F32
F32
[6144]
blk.54.ffn_up_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.55
blk.55.attn_norm.weight
F32
F32
[6144]
blk.55.attn_output.weight
BF16
BF16
[16384, 6144]
blk.55.attn_qkv.weight
BF16
BF16
[6144, 27136]
blk.55.exp_probs_b.bias
F32
F32
[384]
blk.55.ffn_down_exps.weight
MXFP4
MXFP4
[2048, 6144, 384]
blk.55.ffn_gate_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.55.ffn_gate_inp.weight
F32
F32
[6144, 384]
blk.55.ffn_norm.weight
F32
F32
[6144]
blk.55.ffn_up_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.56
blk.56.attn_norm.weight
F32
F32
[6144]
blk.56.attn_output.weight
BF16
BF16
[16384, 6144]
blk.56.attn_qkv.weight
BF16
BF16
[6144, 27136]
blk.56.attn_sinks.weight
F32
F32
[128]
blk.56.exp_probs_b.bias
F32
F32
[384]
blk.56.ffn_down_exps.weight
MXFP4
MXFP4
[2048, 6144, 384]
blk.56.ffn_gate_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.56.ffn_gate_inp.weight
F32
F32
[6144, 384]
blk.56.ffn_norm.weight
F32
F32
[6144]
blk.56.ffn_up_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.57
blk.57.attn_norm.weight
F32
F32
[6144]
blk.57.attn_output.weight
BF16
BF16
[16384, 6144]
blk.57.attn_qkv.weight
BF16
BF16
[6144, 27136]
blk.57.attn_sinks.weight
F32
F32
[128]
blk.57.exp_probs_b.bias
F32
F32
[384]
blk.57.ffn_down_exps.weight
MXFP4
MXFP4
[2048, 6144, 384]
blk.57.ffn_gate_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.57.ffn_gate_inp.weight
F32
F32
[6144, 384]
blk.57.ffn_norm.weight
F32
F32
[6144]
blk.57.ffn_up_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.58
blk.58.attn_norm.weight
F32
F32
[6144]
blk.58.attn_output.weight
BF16
BF16
[16384, 6144]
blk.58.attn_qkv.weight
BF16
BF16
[6144, 27136]
blk.58.attn_sinks.weight
F32
F32
[128]
blk.58.exp_probs_b.bias
F32
F32
[384]
blk.58.ffn_down_exps.weight
MXFP4
MXFP4
[2048, 6144, 384]
blk.58.ffn_gate_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.58.ffn_gate_inp.weight
F32
F32
[6144, 384]
blk.58.ffn_norm.weight
F32
F32
[6144]
blk.58.ffn_up_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.59
blk.59.attn_norm.weight
F32
F32
[6144]
blk.59.attn_output.weight
BF16
BF16
[16384, 6144]
blk.59.attn_qkv.weight
BF16
BF16
[6144, 27136]
blk.59.attn_sinks.weight
F32
F32
[128]
blk.59.exp_probs_b.bias
F32
F32
[384]
blk.59.ffn_down_exps.weight
MXFP4
MXFP4
[2048, 6144, 384]
blk.59.ffn_gate_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.59.ffn_gate_inp.weight
F32
F32
[6144, 384]
blk.59.ffn_norm.weight
F32
F32
[6144]
blk.59.ffn_up_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.60
blk.60.attn_norm.weight
F32
F32
[6144]
blk.60.attn_output.weight
BF16
BF16
[16384, 6144]
blk.60.attn_qkv.weight
BF16
BF16
[6144, 27136]
blk.60.attn_sinks.weight
F32
F32
[128]
blk.60.exp_probs_b.bias
F32
F32
[384]
blk.60.ffn_down_exps.weight
MXFP4
MXFP4
[2048, 6144, 384]
blk.60.ffn_gate_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.60.ffn_gate_inp.weight
F32
F32
[6144, 384]
blk.60.ffn_norm.weight
F32
F32
[6144]
blk.60.ffn_up_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.61
blk.61.attn_norm.weight
F32
F32
[6144]
blk.61.attn_output.weight
BF16
BF16
[16384, 6144]
blk.61.attn_qkv.weight
BF16
BF16
[6144, 27136]
blk.61.attn_sinks.weight
F32
F32
[128]
blk.61.exp_probs_b.bias
F32
F32
[384]
blk.61.ffn_down_exps.weight
MXFP4
MXFP4
[2048, 6144, 384]
blk.61.ffn_gate_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.61.ffn_gate_inp.weight
F32
F32
[6144, 384]
blk.61.ffn_norm.weight
F32
F32
[6144]
blk.61.ffn_up_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.62
blk.62.attn_norm.weight
F32
F32
[6144]
blk.62.attn_output.weight
BF16
BF16
[16384, 6144]
blk.62.attn_qkv.weight
BF16
BF16
[6144, 27136]
blk.62.exp_probs_b.bias
F32
F32
[384]
blk.62.ffn_down_exps.weight
MXFP4
MXFP4
[2048, 6144, 384]
blk.62.ffn_gate_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.62.ffn_gate_inp.weight
F32
F32
[6144, 384]
blk.62.ffn_norm.weight
F32
F32
[6144]
blk.62.ffn_up_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.63
blk.63.attn_norm.weight
F32
F32
[6144]
blk.63.attn_output.weight
BF16
BF16
[16384, 6144]
blk.63.attn_qkv.weight
BF16
BF16
[6144, 27136]
blk.63.attn_sinks.weight
F32
F32
[128]
blk.63.exp_probs_b.bias
F32
F32
[384]
blk.63.ffn_down_exps.weight
MXFP4
MXFP4
[2048, 6144, 384]
blk.63.ffn_gate_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.63.ffn_gate_inp.weight
F32
F32
[6144, 384]
blk.63.ffn_norm.weight
F32
F32
[6144]
blk.63.ffn_up_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.64
blk.64.attn_norm.weight
F32
F32
[6144]
blk.64.attn_output.weight
BF16
BF16
[16384, 6144]
blk.64.attn_qkv.weight
BF16
BF16
[6144, 27136]
blk.64.attn_sinks.weight
F32
F32
[128]
blk.64.exp_probs_b.bias
F32
F32
[384]
blk.64.ffn_down_exps.weight
MXFP4
MXFP4
[2048, 6144, 384]
blk.64.ffn_gate_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.64.ffn_gate_inp.weight
F32
F32
[6144, 384]
blk.64.ffn_norm.weight
F32
F32
[6144]
blk.64.ffn_up_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.65
blk.65.attn_norm.weight
F32
F32
[6144]
blk.65.attn_output.weight
BF16
BF16
[16384, 6144]
blk.65.attn_qkv.weight
BF16
BF16
[6144, 27136]
blk.65.attn_sinks.weight
F32
F32
[128]
blk.65.exp_probs_b.bias
F32
F32
[384]
blk.65.ffn_down_exps.weight
MXFP4
MXFP4
[2048, 6144, 384]
blk.65.ffn_gate_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.65.ffn_gate_inp.weight
F32
F32
[6144, 384]
blk.65.ffn_norm.weight
F32
F32
[6144]
blk.65.ffn_up_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.66
blk.66.attn_norm.weight
F32
F32
[6144]
blk.66.attn_output.weight
BF16
BF16
[16384, 6144]
blk.66.attn_qkv.weight
BF16
BF16
[6144, 27136]
blk.66.attn_sinks.weight
F32
F32
[128]
blk.66.exp_probs_b.bias
F32
F32
[384]
blk.66.ffn_down_exps.weight
MXFP4
MXFP4
[2048, 6144, 384]
blk.66.ffn_gate_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.66.ffn_gate_inp.weight
F32
F32
[6144, 384]
blk.66.ffn_norm.weight
F32
F32
[6144]
blk.66.ffn_up_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.67
blk.67.attn_norm.weight
F32
F32
[6144]
blk.67.attn_output.weight
BF16
BF16
[16384, 6144]
blk.67.attn_qkv.weight
BF16
BF16
[6144, 27136]
blk.67.attn_sinks.weight
F32
F32
[128]
blk.67.exp_probs_b.bias
F32
F32
[384]
blk.67.ffn_down_exps.weight
MXFP4
MXFP4
[2048, 6144, 384]
blk.67.ffn_gate_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.67.ffn_gate_inp.weight
F32
F32
[6144, 384]
blk.67.ffn_norm.weight
F32
F32
[6144]
blk.67.ffn_up_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.68
blk.68.attn_norm.weight
F32
F32
[6144]
blk.68.attn_output.weight
BF16
BF16
[16384, 6144]
blk.68.attn_qkv.weight
BF16
BF16
[6144, 27136]
blk.68.attn_sinks.weight
F32
F32
[128]
blk.68.exp_probs_b.bias
F32
F32
[384]
blk.68.ffn_down_exps.weight
MXFP4
MXFP4
[2048, 6144, 384]
blk.68.ffn_gate_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.68.ffn_gate_inp.weight
F32
F32
[6144, 384]
blk.68.ffn_norm.weight
F32
F32
[6144]
blk.68.ffn_up_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.69
blk.69.attn_norm.weight
F32
F32
[6144]
blk.69.attn_output.weight
BF16
BF16
[16384, 6144]
blk.69.attn_qkv.weight
BF16
BF16
[6144, 27136]
blk.69.exp_probs_b.bias
F32
F32
[384]
blk.69.ffn_down_exps.weight
MXFP4
MXFP4
[2048, 6144, 384]
blk.69.ffn_gate_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.69.ffn_gate_inp.weight
F32
F32
[6144, 384]
blk.69.ffn_norm.weight
F32
F32
[6144]
blk.69.ffn_up_exps.weight
MXFP4
MXFP4
[6144, 2048, 384]
blk.70
blk.70.attn_norm.weight
F32
F32
[6144]
blk.70.attn_output.weight
BF16
BF16
[16384, 6144]
blk.70.attn_qkv.weight
BF16
BF16
[6144, 27136]
blk.70.attn_sinks.weight
F32
F32
[128]
blk.70.ffn_down.weight
BF16
BF16
[16384, 6144]
blk.70.ffn_gate.weight
BF16
BF16
[6144, 16384]
blk.70.ffn_norm.weight
F32
F32
[6144]
blk.70.ffn_up.weight
BF16
BF16
[6144, 16384]
blk.70.layer_output_norm.weight
F32
F32
[6144]
blk.70.nextn.eh_proj.weight
BF16
BF16
[12288, 6144]
blk.70.nextn.enorm.weight
F32
F32
[6144]
blk.70.nextn.hnorm.weight
F32
F32
[6144]
blk.71
blk.71.attn_norm.weight
F32
F32
[6144]
blk.71.attn_output.weight
BF16
BF16
[16384, 6144]
blk.71.attn_qkv.weight
BF16
BF16
[6144, 27136]
blk.71.attn_sinks.weight
F32
F32
[128]
blk.71.ffn_down.weight
BF16
BF16
[16384, 6144]
blk.71.ffn_gate.weight
BF16
BF16
[6144, 16384]
blk.71.ffn_norm.weight
F32
F32
[6144]
blk.71.ffn_up.weight
BF16
BF16
[6144, 16384]
blk.71.layer_output_norm.weight
F32
F32
[6144]
blk.71.nextn.eh_proj.weight
BF16
BF16
[12288, 6144]
blk.71.nextn.enorm.weight
F32
F32
[6144]
blk.71.nextn.hnorm.weight
F32
F32
[6144]
blk.72
blk.72.attn_norm.weight
F32
F32
[6144]
blk.72.attn_output.weight
BF16
BF16
[16384, 6144]
blk.72.attn_qkv.weight
BF16
BF16
[6144, 27136]
blk.72.attn_sinks.weight
F32
F32
[128]
blk.72.ffn_down.weight
BF16
BF16
[16384, 6144]
blk.72.ffn_gate.weight
BF16
BF16
[6144, 16384]
blk.72.ffn_norm.weight
F32
F32
[6144]
blk.72.ffn_up.weight
BF16
BF16
[6144, 16384]
blk.72.layer_output_norm.weight
F32
F32
[6144]
blk.72.nextn.eh_proj.weight
BF16
BF16
[12288, 6144]
blk.72.nextn.enorm.weight
F32
F32
[6144]
blk.72.nextn.hnorm.weight
F32
F32
[6144]
output.weight
BF16
BF16
[6144, 152576]
output_norm.weight
F32
F32
[6144]