Models
Docs
Pricing
Sign in
Download
Models
Download
Docs
Pricing
Sign in
borekboissy
/
Millesime-2026
:4b-f16
1
Download
Updated
2 days ago
Des modèles compacts et rapides, conçus pour bien écrire et bien comprendre le français, avec une solide culture générale sur l'histoire, la langue et la société françaises.
Des modèles compacts et rapides, conçus pour bien écrire et bien comprendre le français, avec une solide culture générale sur l'histoire, la langue et la société françaises.
Cancel
tools
Millesime-2026:4b-f16
...
/
model
6301c2ef398f · 8.1GB
Metadata
general.architecture
qwen3
qwen3
general.file_type
F16
F16
qwen3.attention.head_count
32
32
qwen3.attention.head_count_kv
8
8
qwen3.attention.key_length
128
128
qwen3.attention.layer_norm_rms_epsilon
1e-06
1e-06
qwen3.attention.value_length
128
128
qwen3.block_count
36
36
qwen3.context_length
262144
262144
qwen3.embedding_length
2560
2560
qwen3.feed_forward_length
9728
9728
qwen3.rope.freq_base
5e+06
5e+06
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
[!, ", #, $, %, ...]
[!, ", #, $, %, ...]
mradermacher.convert_type
hf
hf
mradermacher.quantize_version
2
2
mradermacher.quantized_at
2026-08-24T07:10:37+02:00
2026-08-24T07:10:37+02:00
mradermacher.quantized_by
mradermacher
mradermacher
mradermacher.quantized_on
marco
marco
Tensor
Name
Type
Shape
token_embd.weight
F16
F16
[2560, 151936]
blk.0
blk.0.attn_k.weight
F16
F16
[2560, 1024]
blk.0.attn_k_norm.weight
F32
F32
[128]
blk.0.attn_norm.weight
F32
F32
[2560]
blk.0.attn_output.weight
F16
F16
[4096, 2560]
blk.0.attn_q.weight
F16
F16
[2560, 4096]
blk.0.attn_q_norm.weight
F32
F32
[128]
blk.0.attn_v.weight
F16
F16
[2560, 1024]
blk.0.ffn_down.weight
F16
F16
[9728, 2560]
blk.0.ffn_gate.weight
F16
F16
[2560, 9728]
blk.0.ffn_norm.weight
F32
F32
[2560]
blk.0.ffn_up.weight
F16
F16
[2560, 9728]
blk.1
blk.1.attn_k.weight
F16
F16
[2560, 1024]
blk.1.attn_k_norm.weight
F32
F32
[128]
blk.1.attn_norm.weight
F32
F32
[2560]
blk.1.attn_output.weight
F16
F16
[4096, 2560]
blk.1.attn_q.weight
F16
F16
[2560, 4096]
blk.1.attn_q_norm.weight
F32
F32
[128]
blk.1.attn_v.weight
F16
F16
[2560, 1024]
blk.1.ffn_down.weight
F16
F16
[9728, 2560]
blk.1.ffn_gate.weight
F16
F16
[2560, 9728]
blk.1.ffn_norm.weight
F32
F32
[2560]
blk.1.ffn_up.weight
F16
F16
[2560, 9728]
blk.2
blk.2.attn_k.weight
F16
F16
[2560, 1024]
blk.2.attn_k_norm.weight
F32
F32
[128]
blk.2.attn_norm.weight
F32
F32
[2560]
blk.2.attn_output.weight
F16
F16
[4096, 2560]
blk.2.attn_q.weight
F16
F16
[2560, 4096]
blk.2.attn_q_norm.weight
F32
F32
[128]
blk.2.attn_v.weight
F16
F16
[2560, 1024]
blk.2.ffn_down.weight
F16
F16
[9728, 2560]
blk.2.ffn_gate.weight
F16
F16
[2560, 9728]
blk.2.ffn_norm.weight
F32
F32
[2560]
blk.2.ffn_up.weight
F16
F16
[2560, 9728]
blk.3
blk.3.attn_k.weight
F16
F16
[2560, 1024]
blk.3.attn_k_norm.weight
F32
F32
[128]
blk.3.attn_norm.weight
F32
F32
[2560]
blk.3.attn_output.weight
F16
F16
[4096, 2560]
blk.3.attn_q.weight
F16
F16
[2560, 4096]
blk.3.attn_q_norm.weight
F32
F32
[128]
blk.3.attn_v.weight
F16
F16
[2560, 1024]
blk.3.ffn_down.weight
F16
F16
[9728, 2560]
blk.3.ffn_gate.weight
F16
F16
[2560, 9728]
blk.3.ffn_norm.weight
F32
F32
[2560]
blk.3.ffn_up.weight
F16
F16
[2560, 9728]
blk.4
blk.4.attn_k.weight
F16
F16
[2560, 1024]
blk.4.attn_k_norm.weight
F32
F32
[128]
blk.4.attn_norm.weight
F32
F32
[2560]
blk.4.attn_output.weight
F16
F16
[4096, 2560]
blk.4.attn_q.weight
F16
F16
[2560, 4096]
blk.4.attn_q_norm.weight
F32
F32
[128]
blk.4.attn_v.weight
F16
F16
[2560, 1024]
blk.4.ffn_down.weight
F16
F16
[9728, 2560]
blk.4.ffn_gate.weight
F16
F16
[2560, 9728]
blk.4.ffn_norm.weight
F32
F32
[2560]
blk.4.ffn_up.weight
F16
F16
[2560, 9728]
blk.5
blk.5.attn_k.weight
F16
F16
[2560, 1024]
blk.5.attn_k_norm.weight
F32
F32
[128]
blk.5.attn_norm.weight
F32
F32
[2560]
blk.5.attn_output.weight
F16
F16
[4096, 2560]
blk.5.attn_q.weight
F16
F16
[2560, 4096]
blk.5.attn_q_norm.weight
F32
F32
[128]
blk.5.attn_v.weight
F16
F16
[2560, 1024]
blk.5.ffn_down.weight
F16
F16
[9728, 2560]
blk.5.ffn_gate.weight
F16
F16
[2560, 9728]
blk.5.ffn_norm.weight
F32
F32
[2560]
blk.5.ffn_up.weight
F16
F16
[2560, 9728]
blk.6
blk.6.attn_k.weight
F16
F16
[2560, 1024]
blk.6.attn_k_norm.weight
F32
F32
[128]
blk.6.attn_norm.weight
F32
F32
[2560]
blk.6.attn_output.weight
F16
F16
[4096, 2560]
blk.6.attn_q.weight
F16
F16
[2560, 4096]
blk.6.attn_q_norm.weight
F32
F32
[128]
blk.6.attn_v.weight
F16
F16
[2560, 1024]
blk.6.ffn_down.weight
F16
F16
[9728, 2560]
blk.6.ffn_gate.weight
F16
F16
[2560, 9728]
blk.6.ffn_norm.weight
F32
F32
[2560]
blk.6.ffn_up.weight
F16
F16
[2560, 9728]
blk.7
blk.7.attn_k.weight
F16
F16
[2560, 1024]
blk.7.attn_k_norm.weight
F32
F32
[128]
blk.7.attn_norm.weight
F32
F32
[2560]
blk.7.attn_output.weight
F16
F16
[4096, 2560]
blk.7.attn_q.weight
F16
F16
[2560, 4096]
blk.7.attn_q_norm.weight
F32
F32
[128]
blk.7.attn_v.weight
F16
F16
[2560, 1024]
blk.7.ffn_down.weight
F16
F16
[9728, 2560]
blk.7.ffn_gate.weight
F16
F16
[2560, 9728]
blk.7.ffn_norm.weight
F32
F32
[2560]
blk.7.ffn_up.weight
F16
F16
[2560, 9728]
blk.8
blk.8.attn_k.weight
F16
F16
[2560, 1024]
blk.8.attn_k_norm.weight
F32
F32
[128]
blk.8.attn_norm.weight
F32
F32
[2560]
blk.8.attn_output.weight
F16
F16
[4096, 2560]
blk.8.attn_q.weight
F16
F16
[2560, 4096]
blk.8.attn_q_norm.weight
F32
F32
[128]
blk.8.attn_v.weight
F16
F16
[2560, 1024]
blk.8.ffn_down.weight
F16
F16
[9728, 2560]
blk.8.ffn_gate.weight
F16
F16
[2560, 9728]
blk.8.ffn_norm.weight
F32
F32
[2560]
blk.8.ffn_up.weight
F16
F16
[2560, 9728]
blk.9
blk.9.attn_k.weight
F16
F16
[2560, 1024]
blk.9.attn_k_norm.weight
F32
F32
[128]
blk.9.attn_norm.weight
F32
F32
[2560]
blk.9.attn_output.weight
F16
F16
[4096, 2560]
blk.9.attn_q.weight
F16
F16
[2560, 4096]
blk.9.attn_q_norm.weight
F32
F32
[128]
blk.9.attn_v.weight
F16
F16
[2560, 1024]
blk.9.ffn_down.weight
F16
F16
[9728, 2560]
blk.9.ffn_gate.weight
F16
F16
[2560, 9728]
blk.9.ffn_norm.weight
F32
F32
[2560]
blk.9.ffn_up.weight
F16
F16
[2560, 9728]
blk.10
blk.10.attn_k.weight
F16
F16
[2560, 1024]
blk.10.attn_k_norm.weight
F32
F32
[128]
blk.10.attn_norm.weight
F32
F32
[2560]
blk.10.attn_output.weight
F16
F16
[4096, 2560]
blk.10.attn_q.weight
F16
F16
[2560, 4096]
blk.10.attn_q_norm.weight
F32
F32
[128]
blk.10.attn_v.weight
F16
F16
[2560, 1024]
blk.10.ffn_down.weight
F16
F16
[9728, 2560]
blk.10.ffn_gate.weight
F16
F16
[2560, 9728]
blk.10.ffn_norm.weight
F32
F32
[2560]
blk.10.ffn_up.weight
F16
F16
[2560, 9728]
blk.11
blk.11.attn_k.weight
F16
F16
[2560, 1024]
blk.11.attn_k_norm.weight
F32
F32
[128]
blk.11.attn_norm.weight
F32
F32
[2560]
blk.11.attn_output.weight
F16
F16
[4096, 2560]
blk.11.attn_q.weight
F16
F16
[2560, 4096]
blk.11.attn_q_norm.weight
F32
F32
[128]
blk.11.attn_v.weight
F16
F16
[2560, 1024]
blk.11.ffn_down.weight
F16
F16
[9728, 2560]
blk.11.ffn_gate.weight
F16
F16
[2560, 9728]
blk.11.ffn_norm.weight
F32
F32
[2560]
blk.11.ffn_up.weight
F16
F16
[2560, 9728]
blk.12
blk.12.attn_k.weight
F16
F16
[2560, 1024]
blk.12.attn_k_norm.weight
F32
F32
[128]
blk.12.attn_norm.weight
F32
F32
[2560]
blk.12.attn_output.weight
F16
F16
[4096, 2560]
blk.12.attn_q.weight
F16
F16
[2560, 4096]
blk.12.attn_q_norm.weight
F32
F32
[128]
blk.12.attn_v.weight
F16
F16
[2560, 1024]
blk.12.ffn_down.weight
F16
F16
[9728, 2560]
blk.12.ffn_gate.weight
F16
F16
[2560, 9728]
blk.12.ffn_norm.weight
F32
F32
[2560]
blk.12.ffn_up.weight
F16
F16
[2560, 9728]
blk.13
blk.13.attn_k.weight
F16
F16
[2560, 1024]
blk.13.attn_k_norm.weight
F32
F32
[128]
blk.13.attn_norm.weight
F32
F32
[2560]
blk.13.attn_output.weight
F16
F16
[4096, 2560]
blk.13.attn_q.weight
F16
F16
[2560, 4096]
blk.13.attn_q_norm.weight
F32
F32
[128]
blk.13.attn_v.weight
F16
F16
[2560, 1024]
blk.13.ffn_down.weight
F16
F16
[9728, 2560]
blk.13.ffn_gate.weight
F16
F16
[2560, 9728]
blk.13.ffn_norm.weight
F32
F32
[2560]
blk.13.ffn_up.weight
F16
F16
[2560, 9728]
blk.14
blk.14.attn_k.weight
F16
F16
[2560, 1024]
blk.14.attn_k_norm.weight
F32
F32
[128]
blk.14.attn_norm.weight
F32
F32
[2560]
blk.14.attn_output.weight
F16
F16
[4096, 2560]
blk.14.attn_q.weight
F16
F16
[2560, 4096]
blk.14.attn_q_norm.weight
F32
F32
[128]
blk.14.attn_v.weight
F16
F16
[2560, 1024]
blk.14.ffn_down.weight
F16
F16
[9728, 2560]
blk.14.ffn_gate.weight
F16
F16
[2560, 9728]
blk.14.ffn_norm.weight
F32
F32
[2560]
blk.14.ffn_up.weight
F16
F16
[2560, 9728]
blk.15
blk.15.attn_k.weight
F16
F16
[2560, 1024]
blk.15.attn_k_norm.weight
F32
F32
[128]
blk.15.attn_norm.weight
F32
F32
[2560]
blk.15.attn_output.weight
F16
F16
[4096, 2560]
blk.15.attn_q.weight
F16
F16
[2560, 4096]
blk.15.attn_q_norm.weight
F32
F32
[128]
blk.15.attn_v.weight
F16
F16
[2560, 1024]
blk.15.ffn_down.weight
F16
F16
[9728, 2560]
blk.15.ffn_gate.weight
F16
F16
[2560, 9728]
blk.15.ffn_norm.weight
F32
F32
[2560]
blk.15.ffn_up.weight
F16
F16
[2560, 9728]
blk.16
blk.16.attn_k.weight
F16
F16
[2560, 1024]
blk.16.attn_k_norm.weight
F32
F32
[128]
blk.16.attn_norm.weight
F32
F32
[2560]
blk.16.attn_output.weight
F16
F16
[4096, 2560]
blk.16.attn_q.weight
F16
F16
[2560, 4096]
blk.16.attn_q_norm.weight
F32
F32
[128]
blk.16.attn_v.weight
F16
F16
[2560, 1024]
blk.16.ffn_down.weight
F16
F16
[9728, 2560]
blk.16.ffn_gate.weight
F16
F16
[2560, 9728]
blk.16.ffn_norm.weight
F32
F32
[2560]
blk.16.ffn_up.weight
F16
F16
[2560, 9728]
blk.17
blk.17.attn_k.weight
F16
F16
[2560, 1024]
blk.17.attn_k_norm.weight
F32
F32
[128]
blk.17.attn_norm.weight
F32
F32
[2560]
blk.17.attn_output.weight
F16
F16
[4096, 2560]
blk.17.attn_q.weight
F16
F16
[2560, 4096]
blk.17.attn_q_norm.weight
F32
F32
[128]
blk.17.attn_v.weight
F16
F16
[2560, 1024]
blk.17.ffn_down.weight
F16
F16
[9728, 2560]
blk.17.ffn_gate.weight
F16
F16
[2560, 9728]
blk.17.ffn_norm.weight
F32
F32
[2560]
blk.17.ffn_up.weight
F16
F16
[2560, 9728]
blk.18
blk.18.attn_k.weight
F16
F16
[2560, 1024]
blk.18.attn_k_norm.weight
F32
F32
[128]
blk.18.attn_norm.weight
F32
F32
[2560]
blk.18.attn_output.weight
F16
F16
[4096, 2560]
blk.18.attn_q.weight
F16
F16
[2560, 4096]
blk.18.attn_q_norm.weight
F32
F32
[128]
blk.18.attn_v.weight
F16
F16
[2560, 1024]
blk.18.ffn_down.weight
F16
F16
[9728, 2560]
blk.18.ffn_gate.weight
F16
F16
[2560, 9728]
blk.18.ffn_norm.weight
F32
F32
[2560]
blk.18.ffn_up.weight
F16
F16
[2560, 9728]
blk.19
blk.19.attn_k.weight
F16
F16
[2560, 1024]
blk.19.attn_k_norm.weight
F32
F32
[128]
blk.19.attn_norm.weight
F32
F32
[2560]
blk.19.attn_output.weight
F16
F16
[4096, 2560]
blk.19.attn_q.weight
F16
F16
[2560, 4096]
blk.19.attn_q_norm.weight
F32
F32
[128]
blk.19.attn_v.weight
F16
F16
[2560, 1024]
blk.19.ffn_down.weight
F16
F16
[9728, 2560]
blk.19.ffn_gate.weight
F16
F16
[2560, 9728]
blk.19.ffn_norm.weight
F32
F32
[2560]
blk.19.ffn_up.weight
F16
F16
[2560, 9728]
blk.20
blk.20.attn_k.weight
F16
F16
[2560, 1024]
blk.20.attn_k_norm.weight
F32
F32
[128]
blk.20.attn_norm.weight
F32
F32
[2560]
blk.20.attn_output.weight
F16
F16
[4096, 2560]
blk.20.attn_q.weight
F16
F16
[2560, 4096]
blk.20.attn_q_norm.weight
F32
F32
[128]
blk.20.attn_v.weight
F16
F16
[2560, 1024]
blk.20.ffn_down.weight
F16
F16
[9728, 2560]
blk.20.ffn_gate.weight
F16
F16
[2560, 9728]
blk.20.ffn_norm.weight
F32
F32
[2560]
blk.20.ffn_up.weight
F16
F16
[2560, 9728]
blk.21
blk.21.attn_k.weight
F16
F16
[2560, 1024]
blk.21.attn_k_norm.weight
F32
F32
[128]
blk.21.attn_norm.weight
F32
F32
[2560]
blk.21.attn_output.weight
F16
F16
[4096, 2560]
blk.21.attn_q.weight
F16
F16
[2560, 4096]
blk.21.attn_q_norm.weight
F32
F32
[128]
blk.21.attn_v.weight
F16
F16
[2560, 1024]
blk.21.ffn_down.weight
F16
F16
[9728, 2560]
blk.21.ffn_gate.weight
F16
F16
[2560, 9728]
blk.21.ffn_norm.weight
F32
F32
[2560]
blk.21.ffn_up.weight
F16
F16
[2560, 9728]
blk.22
blk.22.attn_k.weight
F16
F16
[2560, 1024]
blk.22.attn_k_norm.weight
F32
F32
[128]
blk.22.attn_norm.weight
F32
F32
[2560]
blk.22.attn_output.weight
F16
F16
[4096, 2560]
blk.22.attn_q.weight
F16
F16
[2560, 4096]
blk.22.attn_q_norm.weight
F32
F32
[128]
blk.22.attn_v.weight
F16
F16
[2560, 1024]
blk.22.ffn_down.weight
F16
F16
[9728, 2560]
blk.22.ffn_gate.weight
F16
F16
[2560, 9728]
blk.22.ffn_norm.weight
F32
F32
[2560]
blk.22.ffn_up.weight
F16
F16
[2560, 9728]
blk.23
blk.23.attn_k.weight
F16
F16
[2560, 1024]
blk.23.attn_k_norm.weight
F32
F32
[128]
blk.23.attn_norm.weight
F32
F32
[2560]
blk.23.attn_output.weight
F16
F16
[4096, 2560]
blk.23.attn_q.weight
F16
F16
[2560, 4096]
blk.23.attn_q_norm.weight
F32
F32
[128]
blk.23.attn_v.weight
F16
F16
[2560, 1024]
blk.23.ffn_down.weight
F16
F16
[9728, 2560]
blk.23.ffn_gate.weight
F16
F16
[2560, 9728]
blk.23.ffn_norm.weight
F32
F32
[2560]
blk.23.ffn_up.weight
F16
F16
[2560, 9728]
blk.24
blk.24.attn_k.weight
F16
F16
[2560, 1024]
blk.24.attn_k_norm.weight
F32
F32
[128]
blk.24.attn_norm.weight
F32
F32
[2560]
blk.24.attn_output.weight
F16
F16
[4096, 2560]
blk.24.attn_q.weight
F16
F16
[2560, 4096]
blk.24.attn_q_norm.weight
F32
F32
[128]
blk.24.attn_v.weight
F16
F16
[2560, 1024]
blk.24.ffn_down.weight
F16
F16
[9728, 2560]
blk.24.ffn_gate.weight
F16
F16
[2560, 9728]
blk.24.ffn_norm.weight
F32
F32
[2560]
blk.24.ffn_up.weight
F16
F16
[2560, 9728]
blk.25
blk.25.attn_k.weight
F16
F16
[2560, 1024]
blk.25.attn_k_norm.weight
F32
F32
[128]
blk.25.attn_norm.weight
F32
F32
[2560]
blk.25.attn_output.weight
F16
F16
[4096, 2560]
blk.25.attn_q.weight
F16
F16
[2560, 4096]
blk.25.attn_q_norm.weight
F32
F32
[128]
blk.25.attn_v.weight
F16
F16
[2560, 1024]
blk.25.ffn_down.weight
F16
F16
[9728, 2560]
blk.25.ffn_gate.weight
F16
F16
[2560, 9728]
blk.25.ffn_norm.weight
F32
F32
[2560]
blk.25.ffn_up.weight
F16
F16
[2560, 9728]
blk.26
blk.26.attn_k.weight
F16
F16
[2560, 1024]
blk.26.attn_k_norm.weight
F32
F32
[128]
blk.26.attn_norm.weight
F32
F32
[2560]
blk.26.attn_output.weight
F16
F16
[4096, 2560]
blk.26.attn_q.weight
F16
F16
[2560, 4096]
blk.26.attn_q_norm.weight
F32
F32
[128]
blk.26.attn_v.weight
F16
F16
[2560, 1024]
blk.26.ffn_down.weight
F16
F16
[9728, 2560]
blk.26.ffn_gate.weight
F16
F16
[2560, 9728]
blk.26.ffn_norm.weight
F32
F32
[2560]
blk.26.ffn_up.weight
F16
F16
[2560, 9728]
blk.27
blk.27.attn_k.weight
F16
F16
[2560, 1024]
blk.27.attn_k_norm.weight
F32
F32
[128]
blk.27.attn_norm.weight
F32
F32
[2560]
blk.27.attn_output.weight
F16
F16
[4096, 2560]
blk.27.attn_q.weight
F16
F16
[2560, 4096]
blk.27.attn_q_norm.weight
F32
F32
[128]
blk.27.attn_v.weight
F16
F16
[2560, 1024]
blk.27.ffn_down.weight
F16
F16
[9728, 2560]
blk.27.ffn_gate.weight
F16
F16
[2560, 9728]
blk.27.ffn_norm.weight
F32
F32
[2560]
blk.27.ffn_up.weight
F16
F16
[2560, 9728]
blk.28
blk.28.attn_k.weight
F16
F16
[2560, 1024]
blk.28.attn_k_norm.weight
F32
F32
[128]
blk.28.attn_norm.weight
F32
F32
[2560]
blk.28.attn_output.weight
F16
F16
[4096, 2560]
blk.28.attn_q.weight
F16
F16
[2560, 4096]
blk.28.attn_q_norm.weight
F32
F32
[128]
blk.28.attn_v.weight
F16
F16
[2560, 1024]
blk.28.ffn_down.weight
F16
F16
[9728, 2560]
blk.28.ffn_gate.weight
F16
F16
[2560, 9728]
blk.28.ffn_norm.weight
F32
F32
[2560]
blk.28.ffn_up.weight
F16
F16
[2560, 9728]
blk.29
blk.29.attn_k.weight
F16
F16
[2560, 1024]
blk.29.attn_k_norm.weight
F32
F32
[128]
blk.29.attn_norm.weight
F32
F32
[2560]
blk.29.attn_output.weight
F16
F16
[4096, 2560]
blk.29.attn_q.weight
F16
F16
[2560, 4096]
blk.29.attn_q_norm.weight
F32
F32
[128]
blk.29.attn_v.weight
F16
F16
[2560, 1024]
blk.29.ffn_down.weight
F16
F16
[9728, 2560]
blk.29.ffn_gate.weight
F16
F16
[2560, 9728]
blk.29.ffn_norm.weight
F32
F32
[2560]
blk.29.ffn_up.weight
F16
F16
[2560, 9728]
blk.30
blk.30.attn_k.weight
F16
F16
[2560, 1024]
blk.30.attn_k_norm.weight
F32
F32
[128]
blk.30.attn_norm.weight
F32
F32
[2560]
blk.30.attn_output.weight
F16
F16
[4096, 2560]
blk.30.attn_q.weight
F16
F16
[2560, 4096]
blk.30.attn_q_norm.weight
F32
F32
[128]
blk.30.attn_v.weight
F16
F16
[2560, 1024]
blk.30.ffn_down.weight
F16
F16
[9728, 2560]
blk.30.ffn_gate.weight
F16
F16
[2560, 9728]
blk.30.ffn_norm.weight
F32
F32
[2560]
blk.30.ffn_up.weight
F16
F16
[2560, 9728]
blk.31
blk.31.attn_k.weight
F16
F16
[2560, 1024]
blk.31.attn_k_norm.weight
F32
F32
[128]
blk.31.attn_norm.weight
F32
F32
[2560]
blk.31.attn_output.weight
F16
F16
[4096, 2560]
blk.31.attn_q.weight
F16
F16
[2560, 4096]
blk.31.attn_q_norm.weight
F32
F32
[128]
blk.31.attn_v.weight
F16
F16
[2560, 1024]
blk.31.ffn_down.weight
F16
F16
[9728, 2560]
blk.31.ffn_gate.weight
F16
F16
[2560, 9728]
blk.31.ffn_norm.weight
F32
F32
[2560]
blk.31.ffn_up.weight
F16
F16
[2560, 9728]
blk.32
blk.32.attn_k.weight
F16
F16
[2560, 1024]
blk.32.attn_k_norm.weight
F32
F32
[128]
blk.32.attn_norm.weight
F32
F32
[2560]
blk.32.attn_output.weight
F16
F16
[4096, 2560]
blk.32.attn_q.weight
F16
F16
[2560, 4096]
blk.32.attn_q_norm.weight
F32
F32
[128]
blk.32.attn_v.weight
F16
F16
[2560, 1024]
blk.32.ffn_down.weight
F16
F16
[9728, 2560]
blk.32.ffn_gate.weight
F16
F16
[2560, 9728]
blk.32.ffn_norm.weight
F32
F32
[2560]
blk.32.ffn_up.weight
F16
F16
[2560, 9728]
blk.33
blk.33.attn_k.weight
F16
F16
[2560, 1024]
blk.33.attn_k_norm.weight
F32
F32
[128]
blk.33.attn_norm.weight
F32
F32
[2560]
blk.33.attn_output.weight
F16
F16
[4096, 2560]
blk.33.attn_q.weight
F16
F16
[2560, 4096]
blk.33.attn_q_norm.weight
F32
F32
[128]
blk.33.attn_v.weight
F16
F16
[2560, 1024]
blk.33.ffn_down.weight
F16
F16
[9728, 2560]
blk.33.ffn_gate.weight
F16
F16
[2560, 9728]
blk.33.ffn_norm.weight
F32
F32
[2560]
blk.33.ffn_up.weight
F16
F16
[2560, 9728]
blk.34
blk.34.attn_k.weight
F16
F16
[2560, 1024]
blk.34.attn_k_norm.weight
F32
F32
[128]
blk.34.attn_norm.weight
F32
F32
[2560]
blk.34.attn_output.weight
F16
F16
[4096, 2560]
blk.34.attn_q.weight
F16
F16
[2560, 4096]
blk.34.attn_q_norm.weight
F32
F32
[128]
blk.34.attn_v.weight
F16
F16
[2560, 1024]
blk.34.ffn_down.weight
F16
F16
[9728, 2560]
blk.34.ffn_gate.weight
F16
F16
[2560, 9728]
blk.34.ffn_norm.weight
F32
F32
[2560]
blk.34.ffn_up.weight
F16
F16
[2560, 9728]
blk.35
blk.35.attn_k.weight
F16
F16
[2560, 1024]
blk.35.attn_k_norm.weight
F32
F32
[128]
blk.35.attn_norm.weight
F32
F32
[2560]
blk.35.attn_output.weight
F16
F16
[4096, 2560]
blk.35.attn_q.weight
F16
F16
[2560, 4096]
blk.35.attn_q_norm.weight
F32
F32
[128]
blk.35.attn_v.weight
F16
F16
[2560, 1024]
blk.35.ffn_down.weight
F16
F16
[9728, 2560]
blk.35.ffn_gate.weight
F16
F16
[2560, 9728]
blk.35.ffn_norm.weight
F32
F32
[2560]
blk.35.ffn_up.weight
F16
F16
[2560, 9728]
output_norm.weight
F32
F32
[2560]