Skip to content

Commit 32f5f2e

Browse files
committed
Add OLMo2 json definition
1 parent 83f6b07 commit 32f5f2e

File tree

1 file changed

+57
-0
lines changed

1 file changed

+57
-0
lines changed

Diff for: mergekit/_data/architectures/olmo2.json

+57
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,57 @@
1+
{
2+
"model_type": "olmo2",
3+
"architectures": [
4+
"Olmo2ForCausalLM"
5+
],
6+
"pre_weights": [
7+
{
8+
"name": "model.embed_tokens.weight",
9+
"is_embed": true
10+
}
11+
],
12+
"num_layers_config_key": "num_hidden_layers",
13+
"layer_templates": {
14+
"weights": [
15+
{
16+
"name": "model.layers.${layer_index}.self_attn.q_proj.weight"
17+
},
18+
{
19+
"name": "model.layers.${layer_index}.self_attn.k_proj.weight"
20+
},
21+
{
22+
"name": "model.layers.${layer_index}.self_attn.v_proj.weight"
23+
},
24+
{
25+
"name": "model.layers.${layer_index}.self_attn.o_proj.weight"
26+
},
27+
{
28+
"name": "model.layers.${layer_index}.post_attention_layernorm.weight"
29+
},
30+
{
31+
"name": "model.layers.${layer_index}.mlp.up_proj.weight"
32+
},
33+
{
34+
"name": "model.layers.${layer_index}.mlp.gate_proj.weight"
35+
},
36+
{
37+
"name": "model.layers.${layer_index}.mlp.down_proj.weight"
38+
},
39+
{
40+
"name": "model.layers.${layer_index}.post_feedforward_layernorm.weight"
41+
}
42+
]
43+
},
44+
"post_weights": [
45+
{
46+
"name": "model.norm.weight"
47+
},
48+
{
49+
"name": "lm_head.weight",
50+
"is_embed": true,
51+
"optional": true,
52+
"tied_names": [
53+
"model.embed_tokens.weight"
54+
]
55+
}
56+
]
57+
}

0 commit comments

Comments
 (0)