Models
GitHub
Discord
Docs
Cloud
Sign in
Download
Models
Download
GitHub
Discord
Docs
Cloud
Sign in
sam860
/
falcon-h1
:1.5b
13
Downloads
Updated
1 month ago
Cancel
1.5b
falcon-h1:1.5b
...
/
model
0044322ad5aa · 1.7GB
Metadata
general.architecture
falcon-h1
falcon-h1
general.file_type
Q8_0
Q8_0
falcon-h1.attention.head_count
8
8
falcon-h1.attention.head_count_kv
2
2
falcon-h1.attention.key_length
128
128
falcon-h1.attention.layer_norm_rms_epsilon
1e-05
1e-05
falcon-h1.attention.value_length
128
128
falcon-h1.block_count
24
24
falcon-h1.context_length
131072
131072
falcon-h1.embedding_length
2048
2048
falcon-h1.feed_forward_length
4608
4608
falcon-h1.rope.freq_base
1e+11
1e+11
falcon-h1.ssm.conv_kernel
4
4
falcon-h1.ssm.group_count
1
1
falcon-h1.ssm.inner_size
3072
3072
falcon-h1.ssm.state_size
256
256
falcon-h1.ssm.time_step_rank
48
48
falcon-h1.vocab_size
65537
65537
tokenizer.ggml.bos_token_id
17
17
tokenizer.ggml.eos_token_id
11
11
tokenizer.ggml.merges
[Ġ Ġ, ĠĠ ĠĠ, Ġ t, i n, e r, ...]
[Ġ Ġ, ĠĠ ĠĠ, Ġ t, i n, e r, ...]
tokenizer.ggml.model
gpt2
gpt2
tokenizer.ggml.padding_token_id
65536
65536
tokenizer.ggml.pre
falcon-h1
falcon-h1
tokenizer.ggml.token_type
[3, 3, 3, 3, 3, ...]
[3, 3, 3, 3, 3, ...]
tokenizer.ggml.tokens
[<|pad|>, >>ABSTRACT<<, >>INTRODUCTION<<, >>SUMMARY<<, >>COMMENT<<, ...]
[<|pad|>, >>ABSTRACT<<, >>INTRODUCTION<<, >>SUMMARY<<, >>COMMENT<<, ...]
Tensor
Name
Type
Shape
token_embd.weight
Q8_0
Q8_0
[2048, 65537]
blk.0
blk.0.attn_k.weight
Q8_0
Q8_0
[2048, 256]
blk.0.attn_norm.weight
F32
F32
[2048]
blk.0.attn_output.weight
Q8_0
Q8_0
[1024, 2048]
blk.0.attn_q.weight
Q8_0
Q8_0
[2048, 1024]
blk.0.attn_v.weight
Q8_0
Q8_0
[2048, 256]
blk.0.ffn_down.weight
Q8_0
Q8_0
[4608, 2048]
blk.0.ffn_gate.weight
Q8_0
Q8_0
[2048, 4608]
blk.0.ffn_norm
F32
F32
[2048]
blk.0.ffn_up.weight
Q8_0
Q8_0
[2048, 4608]
blk.0.ssm_a
F32
F32
[1, 48]
blk.0.ssm_conv1d.bias
F32
F32
[3584]
blk.0.ssm_conv1d.weight
F32
F32
[4, 3584]
blk.0.ssm_d
F32
F32
[1, 48]
blk.0.ssm_dt.bias
F32
F32
[48]
blk.0.ssm_in.weight
Q8_0
Q8_0
[2048, 6704]
blk.0.ssm_norm.weight
F32
F32
[3072]
blk.0.ssm_out.weight
Q8_0
Q8_0
[3072, 2048]
blk.1
blk.1.attn_k.weight
Q8_0
Q8_0
[2048, 256]
blk.1.attn_norm.weight
F32
F32
[2048]
blk.1.attn_output.weight
Q8_0
Q8_0
[1024, 2048]
blk.1.attn_q.weight
Q8_0
Q8_0
[2048, 1024]
blk.1.attn_v.weight
Q8_0
Q8_0
[2048, 256]
blk.1.ffn_down.weight
Q8_0
Q8_0
[4608, 2048]
blk.1.ffn_gate.weight
Q8_0
Q8_0
[2048, 4608]
blk.1.ffn_norm
F32
F32
[2048]
blk.1.ffn_up.weight
Q8_0
Q8_0
[2048, 4608]
blk.1.ssm_a
F32
F32
[1, 48]
blk.1.ssm_conv1d.bias
F32
F32
[3584]
blk.1.ssm_conv1d.weight
F32
F32
[4, 3584]
blk.1.ssm_d
F32
F32
[1, 48]
blk.1.ssm_dt.bias
F32
F32
[48]
blk.1.ssm_in.weight
Q8_0
Q8_0
[2048, 6704]
blk.1.ssm_norm.weight
F32
F32
[3072]
blk.1.ssm_out.weight
Q8_0
Q8_0
[3072, 2048]
blk.2
blk.2.attn_k.weight
Q8_0
Q8_0
[2048, 256]
blk.2.attn_norm.weight
F32
F32
[2048]
blk.2.attn_output.weight
Q8_0
Q8_0
[1024, 2048]
blk.2.attn_q.weight
Q8_0
Q8_0
[2048, 1024]
blk.2.attn_v.weight
Q8_0
Q8_0
[2048, 256]
blk.2.ffn_down.weight
Q8_0
Q8_0
[4608, 2048]
blk.2.ffn_gate.weight
Q8_0
Q8_0
[2048, 4608]
blk.2.ffn_norm
F32
F32
[2048]
blk.2.ffn_up.weight
Q8_0
Q8_0
[2048, 4608]
blk.2.ssm_a
F32
F32
[1, 48]
blk.2.ssm_conv1d.bias
F32
F32
[3584]
blk.2.ssm_conv1d.weight
F32
F32
[4, 3584]
blk.2.ssm_d
F32
F32
[1, 48]
blk.2.ssm_dt.bias
F32
F32
[48]
blk.2.ssm_in.weight
Q8_0
Q8_0
[2048, 6704]
blk.2.ssm_norm.weight
F32
F32
[3072]
blk.2.ssm_out.weight
Q8_0
Q8_0
[3072, 2048]
blk.3
blk.3.attn_k.weight
Q8_0
Q8_0
[2048, 256]
blk.3.attn_norm.weight
F32
F32
[2048]
blk.3.attn_output.weight
Q8_0
Q8_0
[1024, 2048]
blk.3.attn_q.weight
Q8_0
Q8_0
[2048, 1024]
blk.3.attn_v.weight
Q8_0
Q8_0
[2048, 256]
blk.3.ffn_down.weight
Q8_0
Q8_0
[4608, 2048]
blk.3.ffn_gate.weight
Q8_0
Q8_0
[2048, 4608]
blk.3.ffn_norm
F32
F32
[2048]
blk.3.ffn_up.weight
Q8_0
Q8_0
[2048, 4608]
blk.3.ssm_a
F32
F32
[1, 48]
blk.3.ssm_conv1d.bias
F32
F32
[3584]
blk.3.ssm_conv1d.weight
F32
F32
[4, 3584]
blk.3.ssm_d
F32
F32
[1, 48]
blk.3.ssm_dt.bias
F32
F32
[48]
blk.3.ssm_in.weight
Q8_0
Q8_0
[2048, 6704]
blk.3.ssm_norm.weight
F32
F32
[3072]
blk.3.ssm_out.weight
Q8_0
Q8_0
[3072, 2048]
blk.4
blk.4.attn_k.weight
Q8_0
Q8_0
[2048, 256]
blk.4.attn_norm.weight
F32
F32
[2048]
blk.4.attn_output.weight
Q8_0
Q8_0
[1024, 2048]
blk.4.attn_q.weight
Q8_0
Q8_0
[2048, 1024]
blk.4.attn_v.weight
Q8_0
Q8_0
[2048, 256]
blk.4.ffn_down.weight
Q8_0
Q8_0
[4608, 2048]
blk.4.ffn_gate.weight
Q8_0
Q8_0
[2048, 4608]
blk.4.ffn_norm
F32
F32
[2048]
blk.4.ffn_up.weight
Q8_0
Q8_0
[2048, 4608]
blk.4.ssm_a
F32
F32
[1, 48]
blk.4.ssm_conv1d.bias
F32
F32
[3584]
blk.4.ssm_conv1d.weight
F32
F32
[4, 3584]
blk.4.ssm_d
F32
F32
[1, 48]
blk.4.ssm_dt.bias
F32
F32
[48]
blk.4.ssm_in.weight
Q8_0
Q8_0
[2048, 6704]
blk.4.ssm_norm.weight
F32
F32
[3072]
blk.4.ssm_out.weight
Q8_0
Q8_0
[3072, 2048]
blk.5
blk.5.attn_k.weight
Q8_0
Q8_0
[2048, 256]
blk.5.attn_norm.weight
F32
F32
[2048]
blk.5.attn_output.weight
Q8_0
Q8_0
[1024, 2048]
blk.5.attn_q.weight
Q8_0
Q8_0
[2048, 1024]
blk.5.attn_v.weight
Q8_0
Q8_0
[2048, 256]
blk.5.ffn_down.weight
Q8_0
Q8_0
[4608, 2048]
blk.5.ffn_gate.weight
Q8_0
Q8_0
[2048, 4608]
blk.5.ffn_norm
F32
F32
[2048]
blk.5.ffn_up.weight
Q8_0
Q8_0
[2048, 4608]
blk.5.ssm_a
F32
F32
[1, 48]
blk.5.ssm_conv1d.bias
F32
F32
[3584]
blk.5.ssm_conv1d.weight
F32
F32
[4, 3584]
blk.5.ssm_d
F32
F32
[1, 48]
blk.5.ssm_dt.bias
F32
F32
[48]
blk.5.ssm_in.weight
Q8_0
Q8_0
[2048, 6704]
blk.5.ssm_norm.weight
F32
F32
[3072]
blk.5.ssm_out.weight
Q8_0
Q8_0
[3072, 2048]
blk.6
blk.6.attn_k.weight
Q8_0
Q8_0
[2048, 256]
blk.6.attn_norm.weight
F32
F32
[2048]
blk.6.attn_output.weight
Q8_0
Q8_0
[1024, 2048]
blk.6.attn_q.weight
Q8_0
Q8_0
[2048, 1024]
blk.6.attn_v.weight
Q8_0
Q8_0
[2048, 256]
blk.6.ffn_down.weight
Q8_0
Q8_0
[4608, 2048]
blk.6.ffn_gate.weight
Q8_0
Q8_0
[2048, 4608]
blk.6.ffn_norm
F32
F32
[2048]
blk.6.ffn_up.weight
Q8_0
Q8_0
[2048, 4608]
blk.6.ssm_a
F32
F32
[1, 48]
blk.6.ssm_conv1d.bias
F32
F32
[3584]
blk.6.ssm_conv1d.weight
F32
F32
[4, 3584]
blk.6.ssm_d
F32
F32
[1, 48]
blk.6.ssm_dt.bias
F32
F32
[48]
blk.6.ssm_in.weight
Q8_0
Q8_0
[2048, 6704]
blk.6.ssm_norm.weight
F32
F32
[3072]
blk.6.ssm_out.weight
Q8_0
Q8_0
[3072, 2048]
blk.7
blk.7.attn_k.weight
Q8_0
Q8_0
[2048, 256]
blk.7.attn_norm.weight
F32
F32
[2048]
blk.7.attn_output.weight
Q8_0
Q8_0
[1024, 2048]
blk.7.attn_q.weight
Q8_0
Q8_0
[2048, 1024]
blk.7.attn_v.weight
Q8_0
Q8_0
[2048, 256]
blk.7.ffn_down.weight
Q8_0
Q8_0
[4608, 2048]
blk.7.ffn_gate.weight
Q8_0
Q8_0
[2048, 4608]
blk.7.ffn_norm
F32
F32
[2048]
blk.7.ffn_up.weight
Q8_0
Q8_0
[2048, 4608]
blk.7.ssm_a
F32
F32
[1, 48]
blk.7.ssm_conv1d.bias
F32
F32
[3584]
blk.7.ssm_conv1d.weight
F32
F32
[4, 3584]
blk.7.ssm_d
F32
F32
[1, 48]
blk.7.ssm_dt.bias
F32
F32
[48]
blk.7.ssm_in.weight
Q8_0
Q8_0
[2048, 6704]
blk.7.ssm_norm.weight
F32
F32
[3072]
blk.7.ssm_out.weight
Q8_0
Q8_0
[3072, 2048]
blk.8
blk.8.attn_k.weight
Q8_0
Q8_0
[2048, 256]
blk.8.attn_norm.weight
F32
F32
[2048]
blk.8.attn_output.weight
Q8_0
Q8_0
[1024, 2048]
blk.8.attn_q.weight
Q8_0
Q8_0
[2048, 1024]
blk.8.attn_v.weight
Q8_0
Q8_0
[2048, 256]
blk.8.ffn_down.weight
Q8_0
Q8_0
[4608, 2048]
blk.8.ffn_gate.weight
Q8_0
Q8_0
[2048, 4608]
blk.8.ffn_norm
F32
F32
[2048]
blk.8.ffn_up.weight
Q8_0
Q8_0
[2048, 4608]
blk.8.ssm_a
F32
F32
[1, 48]
blk.8.ssm_conv1d.bias
F32
F32
[3584]
blk.8.ssm_conv1d.weight
F32
F32
[4, 3584]
blk.8.ssm_d
F32
F32
[1, 48]
blk.8.ssm_dt.bias
F32
F32
[48]
blk.8.ssm_in.weight
Q8_0
Q8_0
[2048, 6704]
blk.8.ssm_norm.weight
F32
F32
[3072]
blk.8.ssm_out.weight
Q8_0
Q8_0
[3072, 2048]
blk.9
blk.9.attn_k.weight
Q8_0
Q8_0
[2048, 256]
blk.9.attn_norm.weight
F32
F32
[2048]
blk.9.attn_output.weight
Q8_0
Q8_0
[1024, 2048]
blk.9.attn_q.weight
Q8_0
Q8_0
[2048, 1024]
blk.9.attn_v.weight
Q8_0
Q8_0
[2048, 256]
blk.9.ffn_down.weight
Q8_0
Q8_0
[4608, 2048]
blk.9.ffn_gate.weight
Q8_0
Q8_0
[2048, 4608]
blk.9.ffn_norm
F32
F32
[2048]
blk.9.ffn_up.weight
Q8_0
Q8_0
[2048, 4608]
blk.9.ssm_a
F32
F32
[1, 48]
blk.9.ssm_conv1d.bias
F32
F32
[3584]
blk.9.ssm_conv1d.weight
F32
F32
[4, 3584]
blk.9.ssm_d
F32
F32
[1, 48]
blk.9.ssm_dt.bias
F32
F32
[48]
blk.9.ssm_in.weight
Q8_0
Q8_0
[2048, 6704]
blk.9.ssm_norm.weight
F32
F32
[3072]
blk.9.ssm_out.weight
Q8_0
Q8_0
[3072, 2048]
blk.10
blk.10.attn_k.weight
Q8_0
Q8_0
[2048, 256]
blk.10.attn_norm.weight
F32
F32
[2048]
blk.10.attn_output.weight
Q8_0
Q8_0
[1024, 2048]
blk.10.attn_q.weight
Q8_0
Q8_0
[2048, 1024]
blk.10.attn_v.weight
Q8_0
Q8_0
[2048, 256]
blk.10.ffn_down.weight
Q8_0
Q8_0
[4608, 2048]
blk.10.ffn_gate.weight
Q8_0
Q8_0
[2048, 4608]
blk.10.ffn_norm
F32
F32
[2048]
blk.10.ffn_up.weight
Q8_0
Q8_0
[2048, 4608]
blk.10.ssm_a
F32
F32
[1, 48]
blk.10.ssm_conv1d.bias
F32
F32
[3584]
blk.10.ssm_conv1d.weight
F32
F32
[4, 3584]
blk.10.ssm_d
F32
F32
[1, 48]
blk.10.ssm_dt.bias
F32
F32
[48]
blk.10.ssm_in.weight
Q8_0
Q8_0
[2048, 6704]
blk.10.ssm_norm.weight
F32
F32
[3072]
blk.10.ssm_out.weight
Q8_0
Q8_0
[3072, 2048]
blk.11
blk.11.attn_k.weight
Q8_0
Q8_0
[2048, 256]
blk.11.attn_norm.weight
F32
F32
[2048]
blk.11.attn_output.weight
Q8_0
Q8_0
[1024, 2048]
blk.11.attn_q.weight
Q8_0
Q8_0
[2048, 1024]
blk.11.attn_v.weight
Q8_0
Q8_0
[2048, 256]
blk.11.ffn_down.weight
Q8_0
Q8_0
[4608, 2048]
blk.11.ffn_gate.weight
Q8_0
Q8_0
[2048, 4608]
blk.11.ffn_norm
F32
F32
[2048]
blk.11.ffn_up.weight
Q8_0
Q8_0
[2048, 4608]
blk.11.ssm_a
F32
F32
[1, 48]
blk.11.ssm_conv1d.bias
F32
F32
[3584]
blk.11.ssm_conv1d.weight
F32
F32
[4, 3584]
blk.11.ssm_d
F32
F32
[1, 48]
blk.11.ssm_dt.bias
F32
F32
[48]
blk.11.ssm_in.weight
Q8_0
Q8_0
[2048, 6704]
blk.11.ssm_norm.weight
F32
F32
[3072]
blk.11.ssm_out.weight
Q8_0
Q8_0
[3072, 2048]
blk.12
blk.12.attn_k.weight
Q8_0
Q8_0
[2048, 256]
blk.12.attn_norm.weight
F32
F32
[2048]
blk.12.attn_output.weight
Q8_0
Q8_0
[1024, 2048]
blk.12.attn_q.weight
Q8_0
Q8_0
[2048, 1024]
blk.12.attn_v.weight
Q8_0
Q8_0
[2048, 256]
blk.12.ffn_down.weight
Q8_0
Q8_0
[4608, 2048]
blk.12.ffn_gate.weight
Q8_0
Q8_0
[2048, 4608]
blk.12.ffn_norm
F32
F32
[2048]
blk.12.ffn_up.weight
Q8_0
Q8_0
[2048, 4608]
blk.12.ssm_a
F32
F32
[1, 48]
blk.12.ssm_conv1d.bias
F32
F32
[3584]
blk.12.ssm_conv1d.weight
F32
F32
[4, 3584]
blk.12.ssm_d
F32
F32
[1, 48]
blk.12.ssm_dt.bias
F32
F32
[48]
blk.12.ssm_in.weight
Q8_0
Q8_0
[2048, 6704]
blk.12.ssm_norm.weight
F32
F32
[3072]
blk.12.ssm_out.weight
Q8_0
Q8_0
[3072, 2048]
blk.13
blk.13.attn_k.weight
Q8_0
Q8_0
[2048, 256]
blk.13.attn_norm.weight
F32
F32
[2048]
blk.13.attn_output.weight
Q8_0
Q8_0
[1024, 2048]
blk.13.attn_q.weight
Q8_0
Q8_0
[2048, 1024]
blk.13.attn_v.weight
Q8_0
Q8_0
[2048, 256]
blk.13.ffn_down.weight
Q8_0
Q8_0
[4608, 2048]
blk.13.ffn_gate.weight
Q8_0
Q8_0
[2048, 4608]
blk.13.ffn_norm
F32
F32
[2048]
blk.13.ffn_up.weight
Q8_0
Q8_0
[2048, 4608]
blk.13.ssm_a
F32
F32
[1, 48]
blk.13.ssm_conv1d.bias
F32
F32
[3584]
blk.13.ssm_conv1d.weight
F32
F32
[4, 3584]
blk.13.ssm_d
F32
F32
[1, 48]
blk.13.ssm_dt.bias
F32
F32
[48]
blk.13.ssm_in.weight
Q8_0
Q8_0
[2048, 6704]
blk.13.ssm_norm.weight
F32
F32
[3072]
blk.13.ssm_out.weight
Q8_0
Q8_0
[3072, 2048]
blk.14
blk.14.attn_k.weight
Q8_0
Q8_0
[2048, 256]
blk.14.attn_norm.weight
F32
F32
[2048]
blk.14.attn_output.weight
Q8_0
Q8_0
[1024, 2048]
blk.14.attn_q.weight
Q8_0
Q8_0
[2048, 1024]
blk.14.attn_v.weight
Q8_0
Q8_0
[2048, 256]
blk.14.ffn_down.weight
Q8_0
Q8_0
[4608, 2048]
blk.14.ffn_gate.weight
Q8_0
Q8_0
[2048, 4608]
blk.14.ffn_norm
F32
F32
[2048]
blk.14.ffn_up.weight
Q8_0
Q8_0
[2048, 4608]
blk.14.ssm_a
F32
F32
[1, 48]
blk.14.ssm_conv1d.bias
F32
F32
[3584]
blk.14.ssm_conv1d.weight
F32
F32
[4, 3584]
blk.14.ssm_d
F32
F32
[1, 48]
blk.14.ssm_dt.bias
F32
F32
[48]
blk.14.ssm_in.weight
Q8_0
Q8_0
[2048, 6704]
blk.14.ssm_norm.weight
F32
F32
[3072]
blk.14.ssm_out.weight
Q8_0
Q8_0
[3072, 2048]
blk.15
blk.15.attn_k.weight
Q8_0
Q8_0
[2048, 256]
blk.15.attn_norm.weight
F32
F32
[2048]
blk.15.attn_output.weight
Q8_0
Q8_0
[1024, 2048]
blk.15.attn_q.weight
Q8_0
Q8_0
[2048, 1024]
blk.15.attn_v.weight
Q8_0
Q8_0
[2048, 256]
blk.15.ffn_down.weight
Q8_0
Q8_0
[4608, 2048]
blk.15.ffn_gate.weight
Q8_0
Q8_0
[2048, 4608]
blk.15.ffn_norm
F32
F32
[2048]
blk.15.ffn_up.weight
Q8_0
Q8_0
[2048, 4608]
blk.15.ssm_a
F32
F32
[1, 48]
blk.15.ssm_conv1d.bias
F32
F32
[3584]
blk.15.ssm_conv1d.weight
F32
F32
[4, 3584]
blk.15.ssm_d
F32
F32
[1, 48]
blk.15.ssm_dt.bias
F32
F32
[48]
blk.15.ssm_in.weight
Q8_0
Q8_0
[2048, 6704]
blk.15.ssm_norm.weight
F32
F32
[3072]
blk.15.ssm_out.weight
Q8_0
Q8_0
[3072, 2048]
blk.16
blk.16.attn_k.weight
Q8_0
Q8_0
[2048, 256]
blk.16.attn_norm.weight
F32
F32
[2048]
blk.16.attn_output.weight
Q8_0
Q8_0
[1024, 2048]
blk.16.attn_q.weight
Q8_0
Q8_0
[2048, 1024]
blk.16.attn_v.weight
Q8_0
Q8_0
[2048, 256]
blk.16.ffn_down.weight
Q8_0
Q8_0
[4608, 2048]
blk.16.ffn_gate.weight
Q8_0
Q8_0
[2048, 4608]
blk.16.ffn_norm
F32
F32
[2048]
blk.16.ffn_up.weight
Q8_0
Q8_0
[2048, 4608]
blk.16.ssm_a
F32
F32
[1, 48]
blk.16.ssm_conv1d.bias
F32
F32
[3584]
blk.16.ssm_conv1d.weight
F32
F32
[4, 3584]
blk.16.ssm_d
F32
F32
[1, 48]
blk.16.ssm_dt.bias
F32
F32
[48]
blk.16.ssm_in.weight
Q8_0
Q8_0
[2048, 6704]
blk.16.ssm_norm.weight
F32
F32
[3072]
blk.16.ssm_out.weight
Q8_0
Q8_0
[3072, 2048]
blk.17
blk.17.attn_k.weight
Q8_0
Q8_0
[2048, 256]
blk.17.attn_norm.weight
F32
F32
[2048]
blk.17.attn_output.weight
Q8_0
Q8_0
[1024, 2048]
blk.17.attn_q.weight
Q8_0
Q8_0
[2048, 1024]
blk.17.attn_v.weight
Q8_0
Q8_0
[2048, 256]
blk.17.ffn_down.weight
Q8_0
Q8_0
[4608, 2048]
blk.17.ffn_gate.weight
Q8_0
Q8_0
[2048, 4608]
blk.17.ffn_norm
F32
F32
[2048]
blk.17.ffn_up.weight
Q8_0
Q8_0
[2048, 4608]
blk.17.ssm_a
F32
F32
[1, 48]
blk.17.ssm_conv1d.bias
F32
F32
[3584]
blk.17.ssm_conv1d.weight
F32
F32
[4, 3584]
blk.17.ssm_d
F32
F32
[1, 48]
blk.17.ssm_dt.bias
F32
F32
[48]
blk.17.ssm_in.weight
Q8_0
Q8_0
[2048, 6704]
blk.17.ssm_norm.weight
F32
F32
[3072]
blk.17.ssm_out.weight
Q8_0
Q8_0
[3072, 2048]
blk.18
blk.18.attn_k.weight
Q8_0
Q8_0
[2048, 256]
blk.18.attn_norm.weight
F32
F32
[2048]
blk.18.attn_output.weight
Q8_0
Q8_0
[1024, 2048]
blk.18.attn_q.weight
Q8_0
Q8_0
[2048, 1024]
blk.18.attn_v.weight
Q8_0
Q8_0
[2048, 256]
blk.18.ffn_down.weight
Q8_0
Q8_0
[4608, 2048]
blk.18.ffn_gate.weight
Q8_0
Q8_0
[2048, 4608]
blk.18.ffn_norm
F32
F32
[2048]
blk.18.ffn_up.weight
Q8_0
Q8_0
[2048, 4608]
blk.18.ssm_a
F32
F32
[1, 48]
blk.18.ssm_conv1d.bias
F32
F32
[3584]
blk.18.ssm_conv1d.weight
F32
F32
[4, 3584]
blk.18.ssm_d
F32
F32
[1, 48]
blk.18.ssm_dt.bias
F32
F32
[48]
blk.18.ssm_in.weight
Q8_0
Q8_0
[2048, 6704]
blk.18.ssm_norm.weight
F32
F32
[3072]
blk.18.ssm_out.weight
Q8_0
Q8_0
[3072, 2048]
blk.19
blk.19.attn_k.weight
Q8_0
Q8_0
[2048, 256]
blk.19.attn_norm.weight
F32
F32
[2048]
blk.19.attn_output.weight
Q8_0
Q8_0
[1024, 2048]
blk.19.attn_q.weight
Q8_0
Q8_0
[2048, 1024]
blk.19.attn_v.weight
Q8_0
Q8_0
[2048, 256]
blk.19.ffn_down.weight
Q8_0
Q8_0
[4608, 2048]
blk.19.ffn_gate.weight
Q8_0
Q8_0
[2048, 4608]
blk.19.ffn_norm
F32
F32
[2048]
blk.19.ffn_up.weight
Q8_0
Q8_0
[2048, 4608]
blk.19.ssm_a
F32
F32
[1, 48]
blk.19.ssm_conv1d.bias
F32
F32
[3584]
blk.19.ssm_conv1d.weight
F32
F32
[4, 3584]
blk.19.ssm_d
F32
F32
[1, 48]
blk.19.ssm_dt.bias
F32
F32
[48]
blk.19.ssm_in.weight
Q8_0
Q8_0
[2048, 6704]
blk.19.ssm_norm.weight
F32
F32
[3072]
blk.19.ssm_out.weight
Q8_0
Q8_0
[3072, 2048]
blk.20
blk.20.attn_k.weight
Q8_0
Q8_0
[2048, 256]
blk.20.attn_norm.weight
F32
F32
[2048]
blk.20.attn_output.weight
Q8_0
Q8_0
[1024, 2048]
blk.20.attn_q.weight
Q8_0
Q8_0
[2048, 1024]
blk.20.attn_v.weight
Q8_0
Q8_0
[2048, 256]
blk.20.ffn_down.weight
Q8_0
Q8_0
[4608, 2048]
blk.20.ffn_gate.weight
Q8_0
Q8_0
[2048, 4608]
blk.20.ffn_norm
F32
F32
[2048]
blk.20.ffn_up.weight
Q8_0
Q8_0
[2048, 4608]
blk.20.ssm_a
F32
F32
[1, 48]
blk.20.ssm_conv1d.bias
F32
F32
[3584]
blk.20.ssm_conv1d.weight
F32
F32
[4, 3584]
blk.20.ssm_d
F32
F32
[1, 48]
blk.20.ssm_dt.bias
F32
F32
[48]
blk.20.ssm_in.weight
Q8_0
Q8_0
[2048, 6704]
blk.20.ssm_norm.weight
F32
F32
[3072]
blk.20.ssm_out.weight
Q8_0
Q8_0
[3072, 2048]
blk.21
blk.21.attn_k.weight
Q8_0
Q8_0
[2048, 256]
blk.21.attn_norm.weight
F32
F32
[2048]
blk.21.attn_output.weight
Q8_0
Q8_0
[1024, 2048]
blk.21.attn_q.weight
Q8_0
Q8_0
[2048, 1024]
blk.21.attn_v.weight
Q8_0
Q8_0
[2048, 256]
blk.21.ffn_down.weight
Q8_0
Q8_0
[4608, 2048]
blk.21.ffn_gate.weight
Q8_0
Q8_0
[2048, 4608]
blk.21.ffn_norm
F32
F32
[2048]
blk.21.ffn_up.weight
Q8_0
Q8_0
[2048, 4608]
blk.21.ssm_a
F32
F32
[1, 48]
blk.21.ssm_conv1d.bias
F32
F32
[3584]
blk.21.ssm_conv1d.weight
F32
F32
[4, 3584]
blk.21.ssm_d
F32
F32
[1, 48]
blk.21.ssm_dt.bias
F32
F32
[48]
blk.21.ssm_in.weight
Q8_0
Q8_0
[2048, 6704]
blk.21.ssm_norm.weight
F32
F32
[3072]
blk.21.ssm_out.weight
Q8_0
Q8_0
[3072, 2048]
blk.22
blk.22.attn_k.weight
Q8_0
Q8_0
[2048, 256]
blk.22.attn_norm.weight
F32
F32
[2048]
blk.22.attn_output.weight
Q8_0
Q8_0
[1024, 2048]
blk.22.attn_q.weight
Q8_0
Q8_0
[2048, 1024]
blk.22.attn_v.weight
Q8_0
Q8_0
[2048, 256]
blk.22.ffn_down.weight
Q8_0
Q8_0
[4608, 2048]
blk.22.ffn_gate.weight
Q8_0
Q8_0
[2048, 4608]
blk.22.ffn_norm
F32
F32
[2048]
blk.22.ffn_up.weight
Q8_0
Q8_0
[2048, 4608]
blk.22.ssm_a
F32
F32
[1, 48]
blk.22.ssm_conv1d.bias
F32
F32
[3584]
blk.22.ssm_conv1d.weight
F32
F32
[4, 3584]
blk.22.ssm_d
F32
F32
[1, 48]
blk.22.ssm_dt.bias
F32
F32
[48]
blk.22.ssm_in.weight
Q8_0
Q8_0
[2048, 6704]
blk.22.ssm_norm.weight
F32
F32
[3072]
blk.22.ssm_out.weight
Q8_0
Q8_0
[3072, 2048]
blk.23
blk.23.attn_k.weight
Q8_0
Q8_0
[2048, 256]
blk.23.attn_norm.weight
F32
F32
[2048]
blk.23.attn_output.weight
Q8_0
Q8_0
[1024, 2048]
blk.23.attn_q.weight
Q8_0
Q8_0
[2048, 1024]
blk.23.attn_v.weight
Q8_0
Q8_0
[2048, 256]
blk.23.ffn_down.weight
Q8_0
Q8_0
[4608, 2048]
blk.23.ffn_gate.weight
Q8_0
Q8_0
[2048, 4608]
blk.23.ffn_norm
F32
F32
[2048]
blk.23.ffn_up.weight
Q8_0
Q8_0
[2048, 4608]
blk.23.ssm_a
F32
F32
[1, 48]
blk.23.ssm_conv1d.bias
F32
F32
[3584]
blk.23.ssm_conv1d.weight
F32
F32
[4, 3584]
blk.23.ssm_d
F32
F32
[1, 48]
blk.23.ssm_dt.bias
F32
F32
[48]
blk.23.ssm_in.weight
Q8_0
Q8_0
[2048, 6704]
blk.23.ssm_norm.weight
F32
F32
[3072]
blk.23.ssm_out.weight
Q8_0
Q8_0
[3072, 2048]
output.weight
Q8_0
Q8_0
[2048, 65537]
output_norm.weight
F32
F32
[2048]