Models
GitHub
Discord
Docs
Cloud
Sign in
Download
Models
Download
GitHub
Discord
Docs
Cloud
Sign in
sam860
/
jamba-reasoning
:3b-Q8_0
114
Downloads
Updated
1 month ago
A small and efficient reasoning model, with a hybrid transformer and mamba architecture
A small and efficient reasoning model, with a hybrid transformer and mamba architecture
Cancel
tools
3b
jamba-reasoning:3b-Q8_0
...
/
model
973bc30a403b · 3.4GB
Metadata
general.architecture
jamba
jamba
general.file_type
Q8_0
Q8_0
jamba.attention.head_count
20
20
jamba.attention.head_count_kv
[0, 0, 0, 0, 0, ...]
[0, 0, 0, 0, 0, ...]
jamba.attention.layer_norm_rms_epsilon
1e-06
1e-06
jamba.block_count
28
28
jamba.context_length
262144
262144
jamba.embedding_length
2560
2560
jamba.expert_count
1
1
jamba.expert_used_count
1
1
jamba.feed_forward_length
8192
8192
jamba.ssm.conv_kernel
4
4
jamba.ssm.inner_size
5120
5120
jamba.ssm.state_size
16
16
jamba.ssm.time_step_rank
160
160
tokenizer.ggml.add_bos_token
true
true
tokenizer.ggml.add_eos_token
false
false
tokenizer.ggml.add_sep_token
false
false
tokenizer.ggml.add_space_prefix
false
false
tokenizer.ggml.bos_token_id
1
1
tokenizer.ggml.eos_token_id
519
519
tokenizer.ggml.model
llama
llama
tokenizer.ggml.padding_token_id
0
0
tokenizer.ggml.pre
default
default
tokenizer.ggml.scores
[-1000, -1000, -1000, -1000, -1000, ...]
[-1000, -1000, -1000, -1000, -1000, ...]
tokenizer.ggml.token_type
[3, 3, 1, 3, 1, ...]
[3, 3, 1, 3, 1, ...]
tokenizer.ggml.tokens
[<|pad|>, <|startoftext|>, <|endoftext|>, <|unk|>, <|mask_0|>, ...]
[<|pad|>, <|startoftext|>, <|endoftext|>, <|unk|>, <|mask_0|>, ...]
tokenizer.ggml.unknown_token_id
3
3
Tensor
Name
Type
Shape
token_embd.weight
Q8_0
Q8_0
[2560, 65536]
blk.0
blk.0.attn_norm.weight
F32
F32
[2560]
blk.0.ffn_down.weight
Q8_0
Q8_0
[8192, 2560]
blk.0.ffn_gate.weight
Q8_0
Q8_0
[2560, 8192]
blk.0.ffn_norm.weight
F32
F32
[2560]
blk.0.ffn_up.weight
Q8_0
Q8_0
[2560, 8192]
blk.0.ssm_a
F32
F32
[16, 5120]
blk.0.ssm_b_norm.weight
F32
F32
[16]
blk.0.ssm_c_norm.weight
F32
F32
[16]
blk.0.ssm_conv1d.bias
F32
F32
[5120]
blk.0.ssm_conv1d.weight
F32
F32
[4, 5120]
blk.0.ssm_d
F32
F32
[5120]
blk.0.ssm_dt.bias
F32
F32
[5120]
blk.0.ssm_dt.weight
Q8_0
Q8_0
[160, 5120]
blk.0.ssm_dt_norm.weight
F32
F32
[160]
blk.0.ssm_in.weight
Q8_0
Q8_0
[2560, 10240]
blk.0.ssm_out.weight
Q8_0
Q8_0
[5120, 2560]
blk.0.ssm_x.weight
Q8_0
Q8_0
[5120, 192]
blk.1
blk.1.attn_norm.weight
F32
F32
[2560]
blk.1.ffn_down.weight
Q8_0
Q8_0
[8192, 2560]
blk.1.ffn_gate.weight
Q8_0
Q8_0
[2560, 8192]
blk.1.ffn_norm.weight
F32
F32
[2560]
blk.1.ffn_up.weight
Q8_0
Q8_0
[2560, 8192]
blk.1.ssm_a
F32
F32
[16, 5120]
blk.1.ssm_b_norm.weight
F32
F32
[16]
blk.1.ssm_c_norm.weight
F32
F32
[16]
blk.1.ssm_conv1d.bias
F32
F32
[5120]
blk.1.ssm_conv1d.weight
F32
F32
[4, 5120]
blk.1.ssm_d
F32
F32
[5120]
blk.1.ssm_dt.bias
F32
F32
[5120]
blk.1.ssm_dt.weight
Q8_0
Q8_0
[160, 5120]
blk.1.ssm_dt_norm.weight
F32
F32
[160]
blk.1.ssm_in.weight
Q8_0
Q8_0
[2560, 10240]
blk.1.ssm_out.weight
Q8_0
Q8_0
[5120, 2560]
blk.1.ssm_x.weight
Q8_0
Q8_0
[5120, 192]
blk.2
blk.2.attn_norm.weight
F32
F32
[2560]
blk.2.ffn_down.weight
Q8_0
Q8_0
[8192, 2560]
blk.2.ffn_gate.weight
Q8_0
Q8_0
[2560, 8192]
blk.2.ffn_norm.weight
F32
F32
[2560]
blk.2.ffn_up.weight
Q8_0
Q8_0
[2560, 8192]
blk.2.ssm_a
F32
F32
[16, 5120]
blk.2.ssm_b_norm.weight
F32
F32
[16]
blk.2.ssm_c_norm.weight
F32
F32
[16]
blk.2.ssm_conv1d.bias
F32
F32
[5120]
blk.2.ssm_conv1d.weight
F32
F32
[4, 5120]
blk.2.ssm_d
F32
F32
[5120]
blk.2.ssm_dt.bias
F32
F32
[5120]
blk.2.ssm_dt.weight
Q8_0
Q8_0
[160, 5120]
blk.2.ssm_dt_norm.weight
F32
F32
[160]
blk.2.ssm_in.weight
Q8_0
Q8_0
[2560, 10240]
blk.2.ssm_out.weight
Q8_0
Q8_0
[5120, 2560]
blk.2.ssm_x.weight
Q8_0
Q8_0
[5120, 192]
blk.3
blk.3.attn_norm.weight
F32
F32
[2560]
blk.3.ffn_down.weight
Q8_0
Q8_0
[8192, 2560]
blk.3.ffn_gate.weight
Q8_0
Q8_0
[2560, 8192]
blk.3.ffn_norm.weight
F32
F32
[2560]
blk.3.ffn_up.weight
Q8_0
Q8_0
[2560, 8192]
blk.3.ssm_a
F32
F32
[16, 5120]
blk.3.ssm_b_norm.weight
F32
F32
[16]
blk.3.ssm_c_norm.weight
F32
F32
[16]
blk.3.ssm_conv1d.bias
F32
F32
[5120]
blk.3.ssm_conv1d.weight
F32
F32
[4, 5120]
blk.3.ssm_d
F32
F32
[5120]
blk.3.ssm_dt.bias
F32
F32
[5120]
blk.3.ssm_dt.weight
Q8_0
Q8_0
[160, 5120]
blk.3.ssm_dt_norm.weight
F32
F32
[160]
blk.3.ssm_in.weight
Q8_0
Q8_0
[2560, 10240]
blk.3.ssm_out.weight
Q8_0
Q8_0
[5120, 2560]
blk.3.ssm_x.weight
Q8_0
Q8_0
[5120, 192]
blk.4
blk.4.attn_norm.weight
F32
F32
[2560]
blk.4.ffn_down.weight
Q8_0
Q8_0
[8192, 2560]
blk.4.ffn_gate.weight
Q8_0
Q8_0
[2560, 8192]
blk.4.ffn_norm.weight
F32
F32
[2560]
blk.4.ffn_up.weight
Q8_0
Q8_0
[2560, 8192]
blk.4.ssm_a
F32
F32
[16, 5120]
blk.4.ssm_b_norm.weight
F32
F32
[16]
blk.4.ssm_c_norm.weight
F32
F32
[16]
blk.4.ssm_conv1d.bias
F32
F32
[5120]
blk.4.ssm_conv1d.weight
F32
F32
[4, 5120]
blk.4.ssm_d
F32
F32
[5120]
blk.4.ssm_dt.bias
F32
F32
[5120]
blk.4.ssm_dt.weight
Q8_0
Q8_0
[160, 5120]
blk.4.ssm_dt_norm.weight
F32
F32
[160]
blk.4.ssm_in.weight
Q8_0
Q8_0
[2560, 10240]
blk.4.ssm_out.weight
Q8_0
Q8_0
[5120, 2560]
blk.4.ssm_x.weight
Q8_0
Q8_0
[5120, 192]
blk.5
blk.5.attn_norm.weight
F32
F32
[2560]
blk.5.ffn_down.weight
Q8_0
Q8_0
[8192, 2560]
blk.5.ffn_gate.weight
Q8_0
Q8_0
[2560, 8192]
blk.5.ffn_norm.weight
F32
F32
[2560]
blk.5.ffn_up.weight
Q8_0
Q8_0
[2560, 8192]
blk.5.ssm_a
F32
F32
[16, 5120]
blk.5.ssm_b_norm.weight
F32
F32
[16]
blk.5.ssm_c_norm.weight
F32
F32
[16]
blk.5.ssm_conv1d.bias
F32
F32
[5120]
blk.5.ssm_conv1d.weight
F32
F32
[4, 5120]
blk.5.ssm_d
F32
F32
[5120]
blk.5.ssm_dt.bias
F32
F32
[5120]
blk.5.ssm_dt.weight
Q8_0
Q8_0
[160, 5120]
blk.5.ssm_dt_norm.weight
F32
F32
[160]
blk.5.ssm_in.weight
Q8_0
Q8_0
[2560, 10240]
blk.5.ssm_out.weight
Q8_0
Q8_0
[5120, 2560]
blk.5.ssm_x.weight
Q8_0
Q8_0
[5120, 192]
blk.6
blk.6.attn_norm.weight
F32
F32
[2560]
blk.6.ffn_down.weight
Q8_0
Q8_0
[8192, 2560]
blk.6.ffn_gate.weight
Q8_0
Q8_0
[2560, 8192]
blk.6.ffn_norm.weight
F32
F32
[2560]
blk.6.ffn_up.weight
Q8_0
Q8_0
[2560, 8192]
blk.6.ssm_a
F32
F32
[16, 5120]
blk.6.ssm_b_norm.weight
F32
F32
[16]
blk.6.ssm_c_norm.weight
F32
F32
[16]
blk.6.ssm_conv1d.bias
F32
F32
[5120]
blk.6.ssm_conv1d.weight
F32
F32
[4, 5120]
blk.6.ssm_d
F32
F32
[5120]
blk.6.ssm_dt.bias
F32
F32
[5120]
blk.6.ssm_dt.weight
Q8_0
Q8_0
[160, 5120]
blk.6.ssm_dt_norm.weight
F32
F32
[160]
blk.6.ssm_in.weight
Q8_0
Q8_0
[2560, 10240]
blk.6.ssm_out.weight
Q8_0
Q8_0
[5120, 2560]
blk.6.ssm_x.weight
Q8_0
Q8_0
[5120, 192]
blk.7
blk.7.attn_k.weight
Q8_0
Q8_0
[2560, 128]
blk.7.attn_norm.weight
F32
F32
[2560]
blk.7.attn_output.weight
Q8_0
Q8_0
[2560, 2560]
blk.7.attn_q.weight
Q8_0
Q8_0
[2560, 2560]
blk.7.attn_v.weight
Q8_0
Q8_0
[2560, 128]
blk.7.ffn_down.weight
Q8_0
Q8_0
[8192, 2560]
blk.7.ffn_gate.weight
Q8_0
Q8_0
[2560, 8192]
blk.7.ffn_norm.weight
F32
F32
[2560]
blk.7.ffn_up.weight
Q8_0
Q8_0
[2560, 8192]
blk.8
blk.8.attn_norm.weight
F32
F32
[2560]
blk.8.ffn_down.weight
Q8_0
Q8_0
[8192, 2560]
blk.8.ffn_gate.weight
Q8_0
Q8_0
[2560, 8192]
blk.8.ffn_norm.weight
F32
F32
[2560]
blk.8.ffn_up.weight
Q8_0
Q8_0
[2560, 8192]
blk.8.ssm_a
F32
F32
[16, 5120]
blk.8.ssm_b_norm.weight
F32
F32
[16]
blk.8.ssm_c_norm.weight
F32
F32
[16]
blk.8.ssm_conv1d.bias
F32
F32
[5120]
blk.8.ssm_conv1d.weight
F32
F32
[4, 5120]
blk.8.ssm_d
F32
F32
[5120]
blk.8.ssm_dt.bias
F32
F32
[5120]
blk.8.ssm_dt.weight
Q8_0
Q8_0
[160, 5120]
blk.8.ssm_dt_norm.weight
F32
F32
[160]
blk.8.ssm_in.weight
Q8_0
Q8_0
[2560, 10240]
blk.8.ssm_out.weight
Q8_0
Q8_0
[5120, 2560]
blk.8.ssm_x.weight
Q8_0
Q8_0
[5120, 192]
blk.9
blk.9.attn_norm.weight
F32
F32
[2560]
blk.9.ffn_down.weight
Q8_0
Q8_0
[8192, 2560]
blk.9.ffn_gate.weight
Q8_0
Q8_0
[2560, 8192]
blk.9.ffn_norm.weight
F32
F32
[2560]
blk.9.ffn_up.weight
Q8_0
Q8_0
[2560, 8192]
blk.9.ssm_a
F32
F32
[16, 5120]
blk.9.ssm_b_norm.weight
F32
F32
[16]
blk.9.ssm_c_norm.weight
F32
F32
[16]
blk.9.ssm_conv1d.bias
F32
F32
[5120]
blk.9.ssm_conv1d.weight
F32
F32
[4, 5120]
blk.9.ssm_d
F32
F32
[5120]
blk.9.ssm_dt.bias
F32
F32
[5120]
blk.9.ssm_dt.weight
Q8_0
Q8_0
[160, 5120]
blk.9.ssm_dt_norm.weight
F32
F32
[160]
blk.9.ssm_in.weight
Q8_0
Q8_0
[2560, 10240]
blk.9.ssm_out.weight
Q8_0
Q8_0
[5120, 2560]
blk.9.ssm_x.weight
Q8_0
Q8_0
[5120, 192]
blk.10
blk.10.attn_norm.weight
F32
F32
[2560]
blk.10.ffn_down.weight
Q8_0
Q8_0
[8192, 2560]
blk.10.ffn_gate.weight
Q8_0
Q8_0
[2560, 8192]
blk.10.ffn_norm.weight
F32
F32
[2560]
blk.10.ffn_up.weight
Q8_0
Q8_0
[2560, 8192]
blk.10.ssm_a
F32
F32
[16, 5120]
blk.10.ssm_b_norm.weight
F32
F32
[16]
blk.10.ssm_c_norm.weight
F32
F32
[16]
blk.10.ssm_conv1d.bias
F32
F32
[5120]
blk.10.ssm_conv1d.weight
F32
F32
[4, 5120]
blk.10.ssm_d
F32
F32
[5120]
blk.10.ssm_dt.bias
F32
F32
[5120]
blk.10.ssm_dt.weight
Q8_0
Q8_0
[160, 5120]
blk.10.ssm_dt_norm.weight
F32
F32
[160]
blk.10.ssm_in.weight
Q8_0
Q8_0
[2560, 10240]
blk.10.ssm_out.weight
Q8_0
Q8_0
[5120, 2560]
blk.10.ssm_x.weight
Q8_0
Q8_0
[5120, 192]
blk.11
blk.11.attn_norm.weight
F32
F32
[2560]
blk.11.ffn_down.weight
Q8_0
Q8_0
[8192, 2560]
blk.11.ffn_gate.weight
Q8_0
Q8_0
[2560, 8192]
blk.11.ffn_norm.weight
F32
F32
[2560]
blk.11.ffn_up.weight
Q8_0
Q8_0
[2560, 8192]
blk.11.ssm_a
F32
F32
[16, 5120]
blk.11.ssm_b_norm.weight
F32
F32
[16]
blk.11.ssm_c_norm.weight
F32
F32
[16]
blk.11.ssm_conv1d.bias
F32
F32
[5120]
blk.11.ssm_conv1d.weight
F32
F32
[4, 5120]
blk.11.ssm_d
F32
F32
[5120]
blk.11.ssm_dt.bias
F32
F32
[5120]
blk.11.ssm_dt.weight
Q8_0
Q8_0
[160, 5120]
blk.11.ssm_dt_norm.weight
F32
F32
[160]
blk.11.ssm_in.weight
Q8_0
Q8_0
[2560, 10240]
blk.11.ssm_out.weight
Q8_0
Q8_0
[5120, 2560]
blk.11.ssm_x.weight
Q8_0
Q8_0
[5120, 192]
blk.12
blk.12.attn_norm.weight
F32
F32
[2560]
blk.12.ffn_down.weight
Q8_0
Q8_0
[8192, 2560]
blk.12.ffn_gate.weight
Q8_0
Q8_0
[2560, 8192]
blk.12.ffn_norm.weight
F32
F32
[2560]
blk.12.ffn_up.weight
Q8_0
Q8_0
[2560, 8192]
blk.12.ssm_a
F32
F32
[16, 5120]
blk.12.ssm_b_norm.weight
F32
F32
[16]
blk.12.ssm_c_norm.weight
F32
F32
[16]
blk.12.ssm_conv1d.bias
F32
F32
[5120]
blk.12.ssm_conv1d.weight
F32
F32
[4, 5120]
blk.12.ssm_d
F32
F32
[5120]
blk.12.ssm_dt.bias
F32
F32
[5120]
blk.12.ssm_dt.weight
Q8_0
Q8_0
[160, 5120]
blk.12.ssm_dt_norm.weight
F32
F32
[160]
blk.12.ssm_in.weight
Q8_0
Q8_0
[2560, 10240]
blk.12.ssm_out.weight
Q8_0
Q8_0
[5120, 2560]
blk.12.ssm_x.weight
Q8_0
Q8_0
[5120, 192]
blk.13
blk.13.attn_norm.weight
F32
F32
[2560]
blk.13.ffn_down.weight
Q8_0
Q8_0
[8192, 2560]
blk.13.ffn_gate.weight
Q8_0
Q8_0
[2560, 8192]
blk.13.ffn_norm.weight
F32
F32
[2560]
blk.13.ffn_up.weight
Q8_0
Q8_0
[2560, 8192]
blk.13.ssm_a
F32
F32
[16, 5120]
blk.13.ssm_b_norm.weight
F32
F32
[16]
blk.13.ssm_c_norm.weight
F32
F32
[16]
blk.13.ssm_conv1d.bias
F32
F32
[5120]
blk.13.ssm_conv1d.weight
F32
F32
[4, 5120]
blk.13.ssm_d
F32
F32
[5120]
blk.13.ssm_dt.bias
F32
F32
[5120]
blk.13.ssm_dt.weight
Q8_0
Q8_0
[160, 5120]
blk.13.ssm_dt_norm.weight
F32
F32
[160]
blk.13.ssm_in.weight
Q8_0
Q8_0
[2560, 10240]
blk.13.ssm_out.weight
Q8_0
Q8_0
[5120, 2560]
blk.13.ssm_x.weight
Q8_0
Q8_0
[5120, 192]
blk.14
blk.14.attn_norm.weight
F32
F32
[2560]
blk.14.ffn_down.weight
Q8_0
Q8_0
[8192, 2560]
blk.14.ffn_gate.weight
Q8_0
Q8_0
[2560, 8192]
blk.14.ffn_norm.weight
F32
F32
[2560]
blk.14.ffn_up.weight
Q8_0
Q8_0
[2560, 8192]
blk.14.ssm_a
F32
F32
[16, 5120]
blk.14.ssm_b_norm.weight
F32
F32
[16]
blk.14.ssm_c_norm.weight
F32
F32
[16]
blk.14.ssm_conv1d.bias
F32
F32
[5120]
blk.14.ssm_conv1d.weight
F32
F32
[4, 5120]
blk.14.ssm_d
F32
F32
[5120]
blk.14.ssm_dt.bias
F32
F32
[5120]
blk.14.ssm_dt.weight
Q8_0
Q8_0
[160, 5120]
blk.14.ssm_dt_norm.weight
F32
F32
[160]
blk.14.ssm_in.weight
Q8_0
Q8_0
[2560, 10240]
blk.14.ssm_out.weight
Q8_0
Q8_0
[5120, 2560]
blk.14.ssm_x.weight
Q8_0
Q8_0
[5120, 192]
blk.15
blk.15.attn_norm.weight
F32
F32
[2560]
blk.15.ffn_down.weight
Q8_0
Q8_0
[8192, 2560]
blk.15.ffn_gate.weight
Q8_0
Q8_0
[2560, 8192]
blk.15.ffn_norm.weight
F32
F32
[2560]
blk.15.ffn_up.weight
Q8_0
Q8_0
[2560, 8192]
blk.15.ssm_a
F32
F32
[16, 5120]
blk.15.ssm_b_norm.weight
F32
F32
[16]
blk.15.ssm_c_norm.weight
F32
F32
[16]
blk.15.ssm_conv1d.bias
F32
F32
[5120]
blk.15.ssm_conv1d.weight
F32
F32
[4, 5120]
blk.15.ssm_d
F32
F32
[5120]
blk.15.ssm_dt.bias
F32
F32
[5120]
blk.15.ssm_dt.weight
Q8_0
Q8_0
[160, 5120]
blk.15.ssm_dt_norm.weight
F32
F32
[160]
blk.15.ssm_in.weight
Q8_0
Q8_0
[2560, 10240]
blk.15.ssm_out.weight
Q8_0
Q8_0
[5120, 2560]
blk.15.ssm_x.weight
Q8_0
Q8_0
[5120, 192]
blk.16
blk.16.attn_norm.weight
F32
F32
[2560]
blk.16.ffn_down.weight
Q8_0
Q8_0
[8192, 2560]
blk.16.ffn_gate.weight
Q8_0
Q8_0
[2560, 8192]
blk.16.ffn_norm.weight
F32
F32
[2560]
blk.16.ffn_up.weight
Q8_0
Q8_0
[2560, 8192]
blk.16.ssm_a
F32
F32
[16, 5120]
blk.16.ssm_b_norm.weight
F32
F32
[16]
blk.16.ssm_c_norm.weight
F32
F32
[16]
blk.16.ssm_conv1d.bias
F32
F32
[5120]
blk.16.ssm_conv1d.weight
F32
F32
[4, 5120]
blk.16.ssm_d
F32
F32
[5120]
blk.16.ssm_dt.bias
F32
F32
[5120]
blk.16.ssm_dt.weight
Q8_0
Q8_0
[160, 5120]
blk.16.ssm_dt_norm.weight
F32
F32
[160]
blk.16.ssm_in.weight
Q8_0
Q8_0
[2560, 10240]
blk.16.ssm_out.weight
Q8_0
Q8_0
[5120, 2560]
blk.16.ssm_x.weight
Q8_0
Q8_0
[5120, 192]
blk.17
blk.17.attn_norm.weight
F32
F32
[2560]
blk.17.ffn_down.weight
Q8_0
Q8_0
[8192, 2560]
blk.17.ffn_gate.weight
Q8_0
Q8_0
[2560, 8192]
blk.17.ffn_norm.weight
F32
F32
[2560]
blk.17.ffn_up.weight
Q8_0
Q8_0
[2560, 8192]
blk.17.ssm_a
F32
F32
[16, 5120]
blk.17.ssm_b_norm.weight
F32
F32
[16]
blk.17.ssm_c_norm.weight
F32
F32
[16]
blk.17.ssm_conv1d.bias
F32
F32
[5120]
blk.17.ssm_conv1d.weight
F32
F32
[4, 5120]
blk.17.ssm_d
F32
F32
[5120]
blk.17.ssm_dt.bias
F32
F32
[5120]
blk.17.ssm_dt.weight
Q8_0
Q8_0
[160, 5120]
blk.17.ssm_dt_norm.weight
F32
F32
[160]
blk.17.ssm_in.weight
Q8_0
Q8_0
[2560, 10240]
blk.17.ssm_out.weight
Q8_0
Q8_0
[5120, 2560]
blk.17.ssm_x.weight
Q8_0
Q8_0
[5120, 192]
blk.18
blk.18.attn_norm.weight
F32
F32
[2560]
blk.18.ffn_down.weight
Q8_0
Q8_0
[8192, 2560]
blk.18.ffn_gate.weight
Q8_0
Q8_0
[2560, 8192]
blk.18.ffn_norm.weight
F32
F32
[2560]
blk.18.ffn_up.weight
Q8_0
Q8_0
[2560, 8192]
blk.18.ssm_a
F32
F32
[16, 5120]
blk.18.ssm_b_norm.weight
F32
F32
[16]
blk.18.ssm_c_norm.weight
F32
F32
[16]
blk.18.ssm_conv1d.bias
F32
F32
[5120]
blk.18.ssm_conv1d.weight
F32
F32
[4, 5120]
blk.18.ssm_d
F32
F32
[5120]
blk.18.ssm_dt.bias
F32
F32
[5120]
blk.18.ssm_dt.weight
Q8_0
Q8_0
[160, 5120]
blk.18.ssm_dt_norm.weight
F32
F32
[160]
blk.18.ssm_in.weight
Q8_0
Q8_0
[2560, 10240]
blk.18.ssm_out.weight
Q8_0
Q8_0
[5120, 2560]
blk.18.ssm_x.weight
Q8_0
Q8_0
[5120, 192]
blk.19
blk.19.attn_norm.weight
F32
F32
[2560]
blk.19.ffn_down.weight
Q8_0
Q8_0
[8192, 2560]
blk.19.ffn_gate.weight
Q8_0
Q8_0
[2560, 8192]
blk.19.ffn_norm.weight
F32
F32
[2560]
blk.19.ffn_up.weight
Q8_0
Q8_0
[2560, 8192]
blk.19.ssm_a
F32
F32
[16, 5120]
blk.19.ssm_b_norm.weight
F32
F32
[16]
blk.19.ssm_c_norm.weight
F32
F32
[16]
blk.19.ssm_conv1d.bias
F32
F32
[5120]
blk.19.ssm_conv1d.weight
F32
F32
[4, 5120]
blk.19.ssm_d
F32
F32
[5120]
blk.19.ssm_dt.bias
F32
F32
[5120]
blk.19.ssm_dt.weight
Q8_0
Q8_0
[160, 5120]
blk.19.ssm_dt_norm.weight
F32
F32
[160]
blk.19.ssm_in.weight
Q8_0
Q8_0
[2560, 10240]
blk.19.ssm_out.weight
Q8_0
Q8_0
[5120, 2560]
blk.19.ssm_x.weight
Q8_0
Q8_0
[5120, 192]
blk.20
blk.20.attn_norm.weight
F32
F32
[2560]
blk.20.ffn_down.weight
Q8_0
Q8_0
[8192, 2560]
blk.20.ffn_gate.weight
Q8_0
Q8_0
[2560, 8192]
blk.20.ffn_norm.weight
F32
F32
[2560]
blk.20.ffn_up.weight
Q8_0
Q8_0
[2560, 8192]
blk.20.ssm_a
F32
F32
[16, 5120]
blk.20.ssm_b_norm.weight
F32
F32
[16]
blk.20.ssm_c_norm.weight
F32
F32
[16]
blk.20.ssm_conv1d.bias
F32
F32
[5120]
blk.20.ssm_conv1d.weight
F32
F32
[4, 5120]
blk.20.ssm_d
F32
F32
[5120]
blk.20.ssm_dt.bias
F32
F32
[5120]
blk.20.ssm_dt.weight
Q8_0
Q8_0
[160, 5120]
blk.20.ssm_dt_norm.weight
F32
F32
[160]
blk.20.ssm_in.weight
Q8_0
Q8_0
[2560, 10240]
blk.20.ssm_out.weight
Q8_0
Q8_0
[5120, 2560]
blk.20.ssm_x.weight
Q8_0
Q8_0
[5120, 192]
blk.21
blk.21.attn_k.weight
Q8_0
Q8_0
[2560, 128]
blk.21.attn_norm.weight
F32
F32
[2560]
blk.21.attn_output.weight
Q8_0
Q8_0
[2560, 2560]
blk.21.attn_q.weight
Q8_0
Q8_0
[2560, 2560]
blk.21.attn_v.weight
Q8_0
Q8_0
[2560, 128]
blk.21.ffn_down.weight
Q8_0
Q8_0
[8192, 2560]
blk.21.ffn_gate.weight
Q8_0
Q8_0
[2560, 8192]
blk.21.ffn_norm.weight
F32
F32
[2560]
blk.21.ffn_up.weight
Q8_0
Q8_0
[2560, 8192]
blk.22
blk.22.attn_norm.weight
F32
F32
[2560]
blk.22.ffn_down.weight
Q8_0
Q8_0
[8192, 2560]
blk.22.ffn_gate.weight
Q8_0
Q8_0
[2560, 8192]
blk.22.ffn_norm.weight
F32
F32
[2560]
blk.22.ffn_up.weight
Q8_0
Q8_0
[2560, 8192]
blk.22.ssm_a
F32
F32
[16, 5120]
blk.22.ssm_b_norm.weight
F32
F32
[16]
blk.22.ssm_c_norm.weight
F32
F32
[16]
blk.22.ssm_conv1d.bias
F32
F32
[5120]
blk.22.ssm_conv1d.weight
F32
F32
[4, 5120]
blk.22.ssm_d
F32
F32
[5120]
blk.22.ssm_dt.bias
F32
F32
[5120]
blk.22.ssm_dt.weight
Q8_0
Q8_0
[160, 5120]
blk.22.ssm_dt_norm.weight
F32
F32
[160]
blk.22.ssm_in.weight
Q8_0
Q8_0
[2560, 10240]
blk.22.ssm_out.weight
Q8_0
Q8_0
[5120, 2560]
blk.22.ssm_x.weight
Q8_0
Q8_0
[5120, 192]
blk.23
blk.23.attn_norm.weight
F32
F32
[2560]
blk.23.ffn_down.weight
Q8_0
Q8_0
[8192, 2560]
blk.23.ffn_gate.weight
Q8_0
Q8_0
[2560, 8192]
blk.23.ffn_norm.weight
F32
F32
[2560]
blk.23.ffn_up.weight
Q8_0
Q8_0
[2560, 8192]
blk.23.ssm_a
F32
F32
[16, 5120]
blk.23.ssm_b_norm.weight
F32
F32
[16]
blk.23.ssm_c_norm.weight
F32
F32
[16]
blk.23.ssm_conv1d.bias
F32
F32
[5120]
blk.23.ssm_conv1d.weight
F32
F32
[4, 5120]
blk.23.ssm_d
F32
F32
[5120]
blk.23.ssm_dt.bias
F32
F32
[5120]
blk.23.ssm_dt.weight
Q8_0
Q8_0
[160, 5120]
blk.23.ssm_dt_norm.weight
F32
F32
[160]
blk.23.ssm_in.weight
Q8_0
Q8_0
[2560, 10240]
blk.23.ssm_out.weight
Q8_0
Q8_0
[5120, 2560]
blk.23.ssm_x.weight
Q8_0
Q8_0
[5120, 192]
blk.24
blk.24.attn_norm.weight
F32
F32
[2560]
blk.24.ffn_down.weight
Q8_0
Q8_0
[8192, 2560]
blk.24.ffn_gate.weight
Q8_0
Q8_0
[2560, 8192]
blk.24.ffn_norm.weight
F32
F32
[2560]
blk.24.ffn_up.weight
Q8_0
Q8_0
[2560, 8192]
blk.24.ssm_a
F32
F32
[16, 5120]
blk.24.ssm_b_norm.weight
F32
F32
[16]
blk.24.ssm_c_norm.weight
F32
F32
[16]
blk.24.ssm_conv1d.bias
F32
F32
[5120]
blk.24.ssm_conv1d.weight
F32
F32
[4, 5120]
blk.24.ssm_d
F32
F32
[5120]
blk.24.ssm_dt.bias
F32
F32
[5120]
blk.24.ssm_dt.weight
Q8_0
Q8_0
[160, 5120]
blk.24.ssm_dt_norm.weight
F32
F32
[160]
blk.24.ssm_in.weight
Q8_0
Q8_0
[2560, 10240]
blk.24.ssm_out.weight
Q8_0
Q8_0
[5120, 2560]
blk.24.ssm_x.weight
Q8_0
Q8_0
[5120, 192]
blk.25
blk.25.attn_norm.weight
F32
F32
[2560]
blk.25.ffn_down.weight
Q8_0
Q8_0
[8192, 2560]
blk.25.ffn_gate.weight
Q8_0
Q8_0
[2560, 8192]
blk.25.ffn_norm.weight
F32
F32
[2560]
blk.25.ffn_up.weight
Q8_0
Q8_0
[2560, 8192]
blk.25.ssm_a
F32
F32
[16, 5120]
blk.25.ssm_b_norm.weight
F32
F32
[16]
blk.25.ssm_c_norm.weight
F32
F32
[16]
blk.25.ssm_conv1d.bias
F32
F32
[5120]
blk.25.ssm_conv1d.weight
F32
F32
[4, 5120]
blk.25.ssm_d
F32
F32
[5120]
blk.25.ssm_dt.bias
F32
F32
[5120]
blk.25.ssm_dt.weight
Q8_0
Q8_0
[160, 5120]
blk.25.ssm_dt_norm.weight
F32
F32
[160]
blk.25.ssm_in.weight
Q8_0
Q8_0
[2560, 10240]
blk.25.ssm_out.weight
Q8_0
Q8_0
[5120, 2560]
blk.25.ssm_x.weight
Q8_0
Q8_0
[5120, 192]
blk.26
blk.26.attn_norm.weight
F32
F32
[2560]
blk.26.ffn_down.weight
Q8_0
Q8_0
[8192, 2560]
blk.26.ffn_gate.weight
Q8_0
Q8_0
[2560, 8192]
blk.26.ffn_norm.weight
F32
F32
[2560]
blk.26.ffn_up.weight
Q8_0
Q8_0
[2560, 8192]
blk.26.ssm_a
F32
F32
[16, 5120]
blk.26.ssm_b_norm.weight
F32
F32
[16]
blk.26.ssm_c_norm.weight
F32
F32
[16]
blk.26.ssm_conv1d.bias
F32
F32
[5120]
blk.26.ssm_conv1d.weight
F32
F32
[4, 5120]
blk.26.ssm_d
F32
F32
[5120]
blk.26.ssm_dt.bias
F32
F32
[5120]
blk.26.ssm_dt.weight
Q8_0
Q8_0
[160, 5120]
blk.26.ssm_dt_norm.weight
F32
F32
[160]
blk.26.ssm_in.weight
Q8_0
Q8_0
[2560, 10240]
blk.26.ssm_out.weight
Q8_0
Q8_0
[5120, 2560]
blk.26.ssm_x.weight
Q8_0
Q8_0
[5120, 192]
blk.27
blk.27.attn_norm.weight
F32
F32
[2560]
blk.27.ffn_down.weight
Q8_0
Q8_0
[8192, 2560]
blk.27.ffn_gate.weight
Q8_0
Q8_0
[2560, 8192]
blk.27.ffn_norm.weight
F32
F32
[2560]
blk.27.ffn_up.weight
Q8_0
Q8_0
[2560, 8192]
blk.27.ssm_a
F32
F32
[16, 5120]
blk.27.ssm_b_norm.weight
F32
F32
[16]
blk.27.ssm_c_norm.weight
F32
F32
[16]
blk.27.ssm_conv1d.bias
F32
F32
[5120]
blk.27.ssm_conv1d.weight
F32
F32
[4, 5120]
blk.27.ssm_d
F32
F32
[5120]
blk.27.ssm_dt.bias
F32
F32
[5120]
blk.27.ssm_dt.weight
Q8_0
Q8_0
[160, 5120]
blk.27.ssm_dt_norm.weight
F32
F32
[160]
blk.27.ssm_in.weight
Q8_0
Q8_0
[2560, 10240]
blk.27.ssm_out.weight
Q8_0
Q8_0
[5120, 2560]
blk.27.ssm_x.weight
Q8_0
Q8_0
[5120, 192]
output.weight
Q8_0
Q8_0
[2560, 65536]
output_norm.weight
F32
F32
[2560]