Models
Docs
Pricing
Sign in
Download
Models
Download
Docs
Pricing
Sign in
AKSHIT_05
/
phi15
:latest
65
Downloads
Updated
5 months ago
A high-quality 8-bit (Q8_0) quantization of Microsoft's Phi-1.5 (1.3B). Optimized for local coding assistance and logical reasoning with a tiny 1.8GB memory footprint. Features custom stop tokens to prevent base-model hallucinations.
A high-quality 8-bit (Q8_0) quantization of Microsoft's Phi-1.5 (1.3B). Optimized for local coding assistance and logical reasoning with a tiny 1.8GB memory footprint. Features custom stop tokens to prevent base-model hallucinations.
Cancel
phi15:latest
...
/
model
d44836d19a32 · 1.5GB
Metadata
general.architecture
phi2
phi2
general.file_type
Q8_0
Q8_0
phi2.attention.head_count
32
32
phi2.attention.head_count_kv
32
32
phi2.attention.layer_norm_epsilon
1e-05
1e-05
phi2.block_count
24
24
phi2.context_length
2048
2048
phi2.embedding_length
2048
2048
phi2.feed_forward_length
8192
8192
phi2.rope.dimension_count
32
32
tokenizer.ggml.add_bos_token
false
false
tokenizer.ggml.bos_token_id
50256
50256
tokenizer.ggml.eos_token_id
50256
50256
tokenizer.ggml.merges
[Ġ t, Ġ a, h e, i n, r e, ...]
[Ġ t, Ġ a, h e, i n, r e, ...]
tokenizer.ggml.model
gpt2
gpt2
tokenizer.ggml.pre
phi-2
phi-2
tokenizer.ggml.token_type
[1, 1, 1, 1, 1, ...]
[1, 1, 1, 1, 1, ...]
tokenizer.ggml.tokens
[!, ", #, $, %, ...]
[!, ", #, $, %, ...]
tokenizer.ggml.unknown_token_id
50256
50256
mradermacher.convert_type
hf
hf
mradermacher.quantize_version
2
2
mradermacher.quantized_at
2025-02-22T22:06:41+01:00
2025-02-22T22:06:41+01:00
mradermacher.quantized_by
mradermacher
mradermacher
mradermacher.quantized_on
rain
rain
Tensor
Name
Type
Shape
token_embd.weight
Q8_0
Q8_0
[2048, 51200]
blk.0
blk.0.attn_k.bias
F32
F32
[2048]
blk.0.attn_k.weight
Q8_0
Q8_0
[2048, 2048]
blk.0.attn_norm.bias
F32
F32
[2048]
blk.0.attn_norm.weight
F32
F32
[2048]
blk.0.attn_output.bias
F32
F32
[2048]
blk.0.attn_output.weight
Q8_0
Q8_0
[2048, 2048]
blk.0.attn_q.bias
F32
F32
[2048]
blk.0.attn_q.weight
Q8_0
Q8_0
[2048, 2048]
blk.0.attn_v.bias
F32
F32
[2048]
blk.0.attn_v.weight
Q8_0
Q8_0
[2048, 2048]
blk.0.ffn_down.bias
F32
F32
[2048]
blk.0.ffn_down.weight
Q8_0
Q8_0
[8192, 2048]
blk.0.ffn_up.bias
F32
F32
[8192]
blk.0.ffn_up.weight
Q8_0
Q8_0
[2048, 8192]
blk.1
blk.1.attn_k.bias
F32
F32
[2048]
blk.1.attn_k.weight
Q8_0
Q8_0
[2048, 2048]
blk.1.attn_norm.bias
F32
F32
[2048]
blk.1.attn_norm.weight
F32
F32
[2048]
blk.1.attn_output.bias
F32
F32
[2048]
blk.1.attn_output.weight
Q8_0
Q8_0
[2048, 2048]
blk.1.attn_q.bias
F32
F32
[2048]
blk.1.attn_q.weight
Q8_0
Q8_0
[2048, 2048]
blk.1.attn_v.bias
F32
F32
[2048]
blk.1.attn_v.weight
Q8_0
Q8_0
[2048, 2048]
blk.1.ffn_down.bias
F32
F32
[2048]
blk.1.ffn_down.weight
Q8_0
Q8_0
[8192, 2048]
blk.1.ffn_up.bias
F32
F32
[8192]
blk.1.ffn_up.weight
Q8_0
Q8_0
[2048, 8192]
blk.2
blk.2.attn_k.bias
F32
F32
[2048]
blk.2.attn_k.weight
Q8_0
Q8_0
[2048, 2048]
blk.2.attn_norm.bias
F32
F32
[2048]
blk.2.attn_norm.weight
F32
F32
[2048]
blk.2.attn_output.bias
F32
F32
[2048]
blk.2.attn_output.weight
Q8_0
Q8_0
[2048, 2048]
blk.2.attn_q.bias
F32
F32
[2048]
blk.2.attn_q.weight
Q8_0
Q8_0
[2048, 2048]
blk.2.attn_v.bias
F32
F32
[2048]
blk.2.attn_v.weight
Q8_0
Q8_0
[2048, 2048]
blk.2.ffn_down.bias
F32
F32
[2048]
blk.2.ffn_down.weight
Q8_0
Q8_0
[8192, 2048]
blk.2.ffn_up.bias
F32
F32
[8192]
blk.2.ffn_up.weight
Q8_0
Q8_0
[2048, 8192]
blk.3
blk.3.attn_k.bias
F32
F32
[2048]
blk.3.attn_k.weight
Q8_0
Q8_0
[2048, 2048]
blk.3.attn_norm.bias
F32
F32
[2048]
blk.3.attn_norm.weight
F32
F32
[2048]
blk.3.attn_output.bias
F32
F32
[2048]
blk.3.attn_output.weight
Q8_0
Q8_0
[2048, 2048]
blk.3.attn_q.bias
F32
F32
[2048]
blk.3.attn_q.weight
Q8_0
Q8_0
[2048, 2048]
blk.3.attn_v.bias
F32
F32
[2048]
blk.3.attn_v.weight
Q8_0
Q8_0
[2048, 2048]
blk.3.ffn_down.bias
F32
F32
[2048]
blk.3.ffn_down.weight
Q8_0
Q8_0
[8192, 2048]
blk.3.ffn_up.bias
F32
F32
[8192]
blk.3.ffn_up.weight
Q8_0
Q8_0
[2048, 8192]
blk.4
blk.4.attn_k.bias
F32
F32
[2048]
blk.4.attn_k.weight
Q8_0
Q8_0
[2048, 2048]
blk.4.attn_norm.bias
F32
F32
[2048]
blk.4.attn_norm.weight
F32
F32
[2048]
blk.4.attn_output.bias
F32
F32
[2048]
blk.4.attn_output.weight
Q8_0
Q8_0
[2048, 2048]
blk.4.attn_q.bias
F32
F32
[2048]
blk.4.attn_q.weight
Q8_0
Q8_0
[2048, 2048]
blk.4.attn_v.bias
F32
F32
[2048]
blk.4.attn_v.weight
Q8_0
Q8_0
[2048, 2048]
blk.4.ffn_down.bias
F32
F32
[2048]
blk.4.ffn_down.weight
Q8_0
Q8_0
[8192, 2048]
blk.4.ffn_up.bias
F32
F32
[8192]
blk.4.ffn_up.weight
Q8_0
Q8_0
[2048, 8192]
blk.5
blk.5.attn_k.bias
F32
F32
[2048]
blk.5.attn_k.weight
Q8_0
Q8_0
[2048, 2048]
blk.5.attn_norm.bias
F32
F32
[2048]
blk.5.attn_norm.weight
F32
F32
[2048]
blk.5.attn_output.bias
F32
F32
[2048]
blk.5.attn_output.weight
Q8_0
Q8_0
[2048, 2048]
blk.5.attn_q.bias
F32
F32
[2048]
blk.5.attn_q.weight
Q8_0
Q8_0
[2048, 2048]
blk.5.attn_v.bias
F32
F32
[2048]
blk.5.attn_v.weight
Q8_0
Q8_0
[2048, 2048]
blk.5.ffn_down.bias
F32
F32
[2048]
blk.5.ffn_down.weight
Q8_0
Q8_0
[8192, 2048]
blk.5.ffn_up.bias
F32
F32
[8192]
blk.5.ffn_up.weight
Q8_0
Q8_0
[2048, 8192]
blk.6
blk.6.attn_k.bias
F32
F32
[2048]
blk.6.attn_k.weight
Q8_0
Q8_0
[2048, 2048]
blk.6.attn_norm.bias
F32
F32
[2048]
blk.6.attn_norm.weight
F32
F32
[2048]
blk.6.attn_output.bias
F32
F32
[2048]
blk.6.attn_output.weight
Q8_0
Q8_0
[2048, 2048]
blk.6.attn_q.bias
F32
F32
[2048]
blk.6.attn_q.weight
Q8_0
Q8_0
[2048, 2048]
blk.6.attn_v.bias
F32
F32
[2048]
blk.6.attn_v.weight
Q8_0
Q8_0
[2048, 2048]
blk.6.ffn_down.bias
F32
F32
[2048]
blk.6.ffn_down.weight
Q8_0
Q8_0
[8192, 2048]
blk.6.ffn_up.bias
F32
F32
[8192]
blk.6.ffn_up.weight
Q8_0
Q8_0
[2048, 8192]
blk.7
blk.7.attn_k.bias
F32
F32
[2048]
blk.7.attn_k.weight
Q8_0
Q8_0
[2048, 2048]
blk.7.attn_norm.bias
F32
F32
[2048]
blk.7.attn_norm.weight
F32
F32
[2048]
blk.7.attn_output.bias
F32
F32
[2048]
blk.7.attn_output.weight
Q8_0
Q8_0
[2048, 2048]
blk.7.attn_q.bias
F32
F32
[2048]
blk.7.attn_q.weight
Q8_0
Q8_0
[2048, 2048]
blk.7.attn_v.bias
F32
F32
[2048]
blk.7.attn_v.weight
Q8_0
Q8_0
[2048, 2048]
blk.7.ffn_down.bias
F32
F32
[2048]
blk.7.ffn_down.weight
Q8_0
Q8_0
[8192, 2048]
blk.7.ffn_up.bias
F32
F32
[8192]
blk.7.ffn_up.weight
Q8_0
Q8_0
[2048, 8192]
blk.8
blk.8.attn_k.bias
F32
F32
[2048]
blk.8.attn_k.weight
Q8_0
Q8_0
[2048, 2048]
blk.8.attn_norm.bias
F32
F32
[2048]
blk.8.attn_norm.weight
F32
F32
[2048]
blk.8.attn_output.bias
F32
F32
[2048]
blk.8.attn_output.weight
Q8_0
Q8_0
[2048, 2048]
blk.8.attn_q.bias
F32
F32
[2048]
blk.8.attn_q.weight
Q8_0
Q8_0
[2048, 2048]
blk.8.attn_v.bias
F32
F32
[2048]
blk.8.attn_v.weight
Q8_0
Q8_0
[2048, 2048]
blk.8.ffn_down.bias
F32
F32
[2048]
blk.8.ffn_down.weight
Q8_0
Q8_0
[8192, 2048]
blk.8.ffn_up.bias
F32
F32
[8192]
blk.8.ffn_up.weight
Q8_0
Q8_0
[2048, 8192]
blk.9
blk.9.attn_k.bias
F32
F32
[2048]
blk.9.attn_k.weight
Q8_0
Q8_0
[2048, 2048]
blk.9.attn_norm.bias
F32
F32
[2048]
blk.9.attn_norm.weight
F32
F32
[2048]
blk.9.attn_output.bias
F32
F32
[2048]
blk.9.attn_output.weight
Q8_0
Q8_0
[2048, 2048]
blk.9.attn_q.bias
F32
F32
[2048]
blk.9.attn_q.weight
Q8_0
Q8_0
[2048, 2048]
blk.9.attn_v.bias
F32
F32
[2048]
blk.9.attn_v.weight
Q8_0
Q8_0
[2048, 2048]
blk.9.ffn_down.bias
F32
F32
[2048]
blk.9.ffn_down.weight
Q8_0
Q8_0
[8192, 2048]
blk.9.ffn_up.bias
F32
F32
[8192]
blk.9.ffn_up.weight
Q8_0
Q8_0
[2048, 8192]
blk.10
blk.10.attn_k.bias
F32
F32
[2048]
blk.10.attn_k.weight
Q8_0
Q8_0
[2048, 2048]
blk.10.attn_norm.bias
F32
F32
[2048]
blk.10.attn_norm.weight
F32
F32
[2048]
blk.10.attn_output.bias
F32
F32
[2048]
blk.10.attn_output.weight
Q8_0
Q8_0
[2048, 2048]
blk.10.attn_q.bias
F32
F32
[2048]
blk.10.attn_q.weight
Q8_0
Q8_0
[2048, 2048]
blk.10.attn_v.bias
F32
F32
[2048]
blk.10.attn_v.weight
Q8_0
Q8_0
[2048, 2048]
blk.10.ffn_down.bias
F32
F32
[2048]
blk.10.ffn_down.weight
Q8_0
Q8_0
[8192, 2048]
blk.10.ffn_up.bias
F32
F32
[8192]
blk.10.ffn_up.weight
Q8_0
Q8_0
[2048, 8192]
blk.11
blk.11.attn_k.bias
F32
F32
[2048]
blk.11.attn_k.weight
Q8_0
Q8_0
[2048, 2048]
blk.11.attn_norm.bias
F32
F32
[2048]
blk.11.attn_norm.weight
F32
F32
[2048]
blk.11.attn_output.bias
F32
F32
[2048]
blk.11.attn_output.weight
Q8_0
Q8_0
[2048, 2048]
blk.11.attn_q.bias
F32
F32
[2048]
blk.11.attn_q.weight
Q8_0
Q8_0
[2048, 2048]
blk.11.attn_v.bias
F32
F32
[2048]
blk.11.attn_v.weight
Q8_0
Q8_0
[2048, 2048]
blk.11.ffn_down.bias
F32
F32
[2048]
blk.11.ffn_down.weight
Q8_0
Q8_0
[8192, 2048]
blk.11.ffn_up.bias
F32
F32
[8192]
blk.11.ffn_up.weight
Q8_0
Q8_0
[2048, 8192]
blk.12
blk.12.attn_k.bias
F32
F32
[2048]
blk.12.attn_k.weight
Q8_0
Q8_0
[2048, 2048]
blk.12.attn_norm.bias
F32
F32
[2048]
blk.12.attn_norm.weight
F32
F32
[2048]
blk.12.attn_output.bias
F32
F32
[2048]
blk.12.attn_output.weight
Q8_0
Q8_0
[2048, 2048]
blk.12.attn_q.bias
F32
F32
[2048]
blk.12.attn_q.weight
Q8_0
Q8_0
[2048, 2048]
blk.12.attn_v.bias
F32
F32
[2048]
blk.12.attn_v.weight
Q8_0
Q8_0
[2048, 2048]
blk.12.ffn_down.bias
F32
F32
[2048]
blk.12.ffn_down.weight
Q8_0
Q8_0
[8192, 2048]
blk.12.ffn_up.bias
F32
F32
[8192]
blk.12.ffn_up.weight
Q8_0
Q8_0
[2048, 8192]
blk.13
blk.13.attn_k.bias
F32
F32
[2048]
blk.13.attn_k.weight
Q8_0
Q8_0
[2048, 2048]
blk.13.attn_norm.bias
F32
F32
[2048]
blk.13.attn_norm.weight
F32
F32
[2048]
blk.13.attn_output.bias
F32
F32
[2048]
blk.13.attn_output.weight
Q8_0
Q8_0
[2048, 2048]
blk.13.attn_q.bias
F32
F32
[2048]
blk.13.attn_q.weight
Q8_0
Q8_0
[2048, 2048]
blk.13.attn_v.bias
F32
F32
[2048]
blk.13.attn_v.weight
Q8_0
Q8_0
[2048, 2048]
blk.13.ffn_down.bias
F32
F32
[2048]
blk.13.ffn_down.weight
Q8_0
Q8_0
[8192, 2048]
blk.13.ffn_up.bias
F32
F32
[8192]
blk.13.ffn_up.weight
Q8_0
Q8_0
[2048, 8192]
blk.14
blk.14.attn_k.bias
F32
F32
[2048]
blk.14.attn_k.weight
Q8_0
Q8_0
[2048, 2048]
blk.14.attn_norm.bias
F32
F32
[2048]
blk.14.attn_norm.weight
F32
F32
[2048]
blk.14.attn_output.bias
F32
F32
[2048]
blk.14.attn_output.weight
Q8_0
Q8_0
[2048, 2048]
blk.14.attn_q.bias
F32
F32
[2048]
blk.14.attn_q.weight
Q8_0
Q8_0
[2048, 2048]
blk.14.attn_v.bias
F32
F32
[2048]
blk.14.attn_v.weight
Q8_0
Q8_0
[2048, 2048]
blk.14.ffn_down.bias
F32
F32
[2048]
blk.14.ffn_down.weight
Q8_0
Q8_0
[8192, 2048]
blk.14.ffn_up.bias
F32
F32
[8192]
blk.14.ffn_up.weight
Q8_0
Q8_0
[2048, 8192]
blk.15
blk.15.attn_k.bias
F32
F32
[2048]
blk.15.attn_k.weight
Q8_0
Q8_0
[2048, 2048]
blk.15.attn_norm.bias
F32
F32
[2048]
blk.15.attn_norm.weight
F32
F32
[2048]
blk.15.attn_output.bias
F32
F32
[2048]
blk.15.attn_output.weight
Q8_0
Q8_0
[2048, 2048]
blk.15.attn_q.bias
F32
F32
[2048]
blk.15.attn_q.weight
Q8_0
Q8_0
[2048, 2048]
blk.15.attn_v.bias
F32
F32
[2048]
blk.15.attn_v.weight
Q8_0
Q8_0
[2048, 2048]
blk.15.ffn_down.bias
F32
F32
[2048]
blk.15.ffn_down.weight
Q8_0
Q8_0
[8192, 2048]
blk.15.ffn_up.bias
F32
F32
[8192]
blk.15.ffn_up.weight
Q8_0
Q8_0
[2048, 8192]
blk.16
blk.16.attn_k.bias
F32
F32
[2048]
blk.16.attn_k.weight
Q8_0
Q8_0
[2048, 2048]
blk.16.attn_norm.bias
F32
F32
[2048]
blk.16.attn_norm.weight
F32
F32
[2048]
blk.16.attn_output.bias
F32
F32
[2048]
blk.16.attn_output.weight
Q8_0
Q8_0
[2048, 2048]
blk.16.attn_q.bias
F32
F32
[2048]
blk.16.attn_q.weight
Q8_0
Q8_0
[2048, 2048]
blk.16.attn_v.bias
F32
F32
[2048]
blk.16.attn_v.weight
Q8_0
Q8_0
[2048, 2048]
blk.16.ffn_down.bias
F32
F32
[2048]
blk.16.ffn_down.weight
Q8_0
Q8_0
[8192, 2048]
blk.16.ffn_up.bias
F32
F32
[8192]
blk.16.ffn_up.weight
Q8_0
Q8_0
[2048, 8192]
blk.17
blk.17.attn_k.bias
F32
F32
[2048]
blk.17.attn_k.weight
Q8_0
Q8_0
[2048, 2048]
blk.17.attn_norm.bias
F32
F32
[2048]
blk.17.attn_norm.weight
F32
F32
[2048]
blk.17.attn_output.bias
F32
F32
[2048]
blk.17.attn_output.weight
Q8_0
Q8_0
[2048, 2048]
blk.17.attn_q.bias
F32
F32
[2048]
blk.17.attn_q.weight
Q8_0
Q8_0
[2048, 2048]
blk.17.attn_v.bias
F32
F32
[2048]
blk.17.attn_v.weight
Q8_0
Q8_0
[2048, 2048]
blk.17.ffn_down.bias
F32
F32
[2048]
blk.17.ffn_down.weight
Q8_0
Q8_0
[8192, 2048]
blk.17.ffn_up.bias
F32
F32
[8192]
blk.17.ffn_up.weight
Q8_0
Q8_0
[2048, 8192]
blk.18
blk.18.attn_k.bias
F32
F32
[2048]
blk.18.attn_k.weight
Q8_0
Q8_0
[2048, 2048]
blk.18.attn_norm.bias
F32
F32
[2048]
blk.18.attn_norm.weight
F32
F32
[2048]
blk.18.attn_output.bias
F32
F32
[2048]
blk.18.attn_output.weight
Q8_0
Q8_0
[2048, 2048]
blk.18.attn_q.bias
F32
F32
[2048]
blk.18.attn_q.weight
Q8_0
Q8_0
[2048, 2048]
blk.18.attn_v.bias
F32
F32
[2048]
blk.18.attn_v.weight
Q8_0
Q8_0
[2048, 2048]
blk.18.ffn_down.bias
F32
F32
[2048]
blk.18.ffn_down.weight
Q8_0
Q8_0
[8192, 2048]
blk.18.ffn_up.bias
F32
F32
[8192]
blk.18.ffn_up.weight
Q8_0
Q8_0
[2048, 8192]
blk.19
blk.19.attn_k.bias
F32
F32
[2048]
blk.19.attn_k.weight
Q8_0
Q8_0
[2048, 2048]
blk.19.attn_norm.bias
F32
F32
[2048]
blk.19.attn_norm.weight
F32
F32
[2048]
blk.19.attn_output.bias
F32
F32
[2048]
blk.19.attn_output.weight
Q8_0
Q8_0
[2048, 2048]
blk.19.attn_q.bias
F32
F32
[2048]
blk.19.attn_q.weight
Q8_0
Q8_0
[2048, 2048]
blk.19.attn_v.bias
F32
F32
[2048]
blk.19.attn_v.weight
Q8_0
Q8_0
[2048, 2048]
blk.19.ffn_down.bias
F32
F32
[2048]
blk.19.ffn_down.weight
Q8_0
Q8_0
[8192, 2048]
blk.19.ffn_up.bias
F32
F32
[8192]
blk.19.ffn_up.weight
Q8_0
Q8_0
[2048, 8192]
blk.20
blk.20.attn_k.bias
F32
F32
[2048]
blk.20.attn_k.weight
Q8_0
Q8_0
[2048, 2048]
blk.20.attn_norm.bias
F32
F32
[2048]
blk.20.attn_norm.weight
F32
F32
[2048]
blk.20.attn_output.bias
F32
F32
[2048]
blk.20.attn_output.weight
Q8_0
Q8_0
[2048, 2048]
blk.20.attn_q.bias
F32
F32
[2048]
blk.20.attn_q.weight
Q8_0
Q8_0
[2048, 2048]
blk.20.attn_v.bias
F32
F32
[2048]
blk.20.attn_v.weight
Q8_0
Q8_0
[2048, 2048]
blk.20.ffn_down.bias
F32
F32
[2048]
blk.20.ffn_down.weight
Q8_0
Q8_0
[8192, 2048]
blk.20.ffn_up.bias
F32
F32
[8192]
blk.20.ffn_up.weight
Q8_0
Q8_0
[2048, 8192]
blk.21
blk.21.attn_k.bias
F32
F32
[2048]
blk.21.attn_k.weight
Q8_0
Q8_0
[2048, 2048]
blk.21.attn_norm.bias
F32
F32
[2048]
blk.21.attn_norm.weight
F32
F32
[2048]
blk.21.attn_output.bias
F32
F32
[2048]
blk.21.attn_output.weight
Q8_0
Q8_0
[2048, 2048]
blk.21.attn_q.bias
F32
F32
[2048]
blk.21.attn_q.weight
Q8_0
Q8_0
[2048, 2048]
blk.21.attn_v.bias
F32
F32
[2048]
blk.21.attn_v.weight
Q8_0
Q8_0
[2048, 2048]
blk.21.ffn_down.bias
F32
F32
[2048]
blk.21.ffn_down.weight
Q8_0
Q8_0
[8192, 2048]
blk.21.ffn_up.bias
F32
F32
[8192]
blk.21.ffn_up.weight
Q8_0
Q8_0
[2048, 8192]
blk.22
blk.22.attn_k.bias
F32
F32
[2048]
blk.22.attn_k.weight
Q8_0
Q8_0
[2048, 2048]
blk.22.attn_norm.bias
F32
F32
[2048]
blk.22.attn_norm.weight
F32
F32
[2048]
blk.22.attn_output.bias
F32
F32
[2048]
blk.22.attn_output.weight
Q8_0
Q8_0
[2048, 2048]
blk.22.attn_q.bias
F32
F32
[2048]
blk.22.attn_q.weight
Q8_0
Q8_0
[2048, 2048]
blk.22.attn_v.bias
F32
F32
[2048]
blk.22.attn_v.weight
Q8_0
Q8_0
[2048, 2048]
blk.22.ffn_down.bias
F32
F32
[2048]
blk.22.ffn_down.weight
Q8_0
Q8_0
[8192, 2048]
blk.22.ffn_up.bias
F32
F32
[8192]
blk.22.ffn_up.weight
Q8_0
Q8_0
[2048, 8192]
blk.23
blk.23.attn_k.bias
F32
F32
[2048]
blk.23.attn_k.weight
Q8_0
Q8_0
[2048, 2048]
blk.23.attn_norm.bias
F32
F32
[2048]
blk.23.attn_norm.weight
F32
F32
[2048]
blk.23.attn_output.bias
F32
F32
[2048]
blk.23.attn_output.weight
Q8_0
Q8_0
[2048, 2048]
blk.23.attn_q.bias
F32
F32
[2048]
blk.23.attn_q.weight
Q8_0
Q8_0
[2048, 2048]
blk.23.attn_v.bias
F32
F32
[2048]
blk.23.attn_v.weight
Q8_0
Q8_0
[2048, 2048]
blk.23.ffn_down.bias
F32
F32
[2048]
blk.23.ffn_down.weight
Q8_0
Q8_0
[8192, 2048]
blk.23.ffn_up.bias
F32
F32
[8192]
blk.23.ffn_up.weight
Q8_0
Q8_0
[2048, 8192]
output.bias
F32
F32
[51200]
output.weight
Q8_0
Q8_0
[2048, 51200]
output_norm.bias
F32
F32
[2048]
output_norm.weight
F32
F32
[2048]