Upload MolmoForConditionalGeneration
Browse files- config.json +55 -7
- generation_config.json +1 -1
- model-00001-of-00007.safetensors +3 -0
- model-00002-of-00007.safetensors +3 -0
- model-00003-of-00007.safetensors +3 -0
- model-00004-of-00007.safetensors +3 -0
- model-00005-of-00007.safetensors +3 -0
- model-00006-of-00007.safetensors +3 -0
- model-00007-of-00007.safetensors +3 -0
- model.safetensors.index.json +0 -0
config.json
CHANGED
@@ -1,28 +1,76 @@
|
|
1 |
{
|
2 |
-
"_name_or_path": "/raid/pablo/Molmo-7B-D-hf-modular",
|
3 |
"architectures": [
|
4 |
"MolmoForConditionalGeneration"
|
5 |
],
|
6 |
"image_token_index": 152069,
|
7 |
"initializer_range": 0.02,
|
8 |
"model_type": "molmo",
|
|
|
9 |
"pooling_config": {
|
10 |
-
"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
11 |
},
|
12 |
"text_config": {
|
|
|
13 |
"attention_bias": true,
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
14 |
"model_type": "molmo_text",
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
15 |
"use_attention_layer_norm": false,
|
16 |
-
"
|
|
|
|
|
|
|
17 |
},
|
18 |
-
"torch_dtype": "
|
19 |
-
"transformers_version": "4.
|
20 |
"vision_config": {
|
21 |
-
"
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
22 |
},
|
23 |
"vision_feature_layers": [
|
24 |
-2,
|
25 |
-9
|
26 |
],
|
27 |
-
"vision_feature_select_strategy": "default"
|
|
|
28 |
}
|
|
|
1 |
{
|
|
|
2 |
"architectures": [
|
3 |
"MolmoForConditionalGeneration"
|
4 |
],
|
5 |
"image_token_index": 152069,
|
6 |
"initializer_range": 0.02,
|
7 |
"model_type": "molmo",
|
8 |
+
"pad_token_id": 152070,
|
9 |
"pooling_config": {
|
10 |
+
"attention_dropout": 0.0,
|
11 |
+
"head_dim": 64,
|
12 |
+
"hidden_size": 2048,
|
13 |
+
"image_feature_dropout": 0.0,
|
14 |
+
"image_num_patches": 24,
|
15 |
+
"image_padding_embed": "pad_and_partial_pad",
|
16 |
+
"image_pooling_type": "attention_meanq",
|
17 |
+
"initializer_range": 0.02,
|
18 |
+
"model_type": "",
|
19 |
+
"num_attention_heads": 16,
|
20 |
+
"num_key_value_heads": 16,
|
21 |
+
"pad_embed_dim": 2048,
|
22 |
+
"pooling_height": 2,
|
23 |
+
"pooling_width": 2,
|
24 |
+
"projector_hidden_act": "silu",
|
25 |
+
"text_hidden_size": 3584,
|
26 |
+
"text_intermediate_size": 37888
|
27 |
},
|
28 |
"text_config": {
|
29 |
+
"additional_embedding_size": 0,
|
30 |
"attention_bias": true,
|
31 |
+
"attention_dropout": 0.0,
|
32 |
+
"head_dim": 128,
|
33 |
+
"hidden_act": "swiglu",
|
34 |
+
"hidden_size": 3584,
|
35 |
+
"initializer_range": 0.02,
|
36 |
+
"intermediate_size": 37888,
|
37 |
+
"layer_norm_eps": 1e-06,
|
38 |
+
"max_position_embeddings": 4096,
|
39 |
"model_type": "molmo_text",
|
40 |
+
"num_attention_heads": 28,
|
41 |
+
"num_hidden_layers": 28,
|
42 |
+
"num_key_value_heads": 4,
|
43 |
+
"pad_token_id": 152070,
|
44 |
+
"rope_scaling": null,
|
45 |
+
"rope_theta": 1000000.0,
|
46 |
+
"sliding_window": 4096,
|
47 |
"use_attention_layer_norm": false,
|
48 |
+
"use_cache": true,
|
49 |
+
"use_postnorm": false,
|
50 |
+
"use_qk_norm": false,
|
51 |
+
"vocab_size": 152071
|
52 |
},
|
53 |
+
"torch_dtype": "float32",
|
54 |
+
"transformers_version": "4.52.0.dev0",
|
55 |
"vision_config": {
|
56 |
+
"attention_dropout": 0.0,
|
57 |
+
"hidden_act": "quick_gelu",
|
58 |
+
"hidden_size": 1024,
|
59 |
+
"image_size": 576,
|
60 |
+
"initializer_range": 0.02,
|
61 |
+
"intermediate_size": 4096,
|
62 |
+
"layer_norm_eps": 1e-05,
|
63 |
+
"model_type": "molmo_vision_model",
|
64 |
+
"num_attention_heads": 16,
|
65 |
+
"num_hidden_layers": 23,
|
66 |
+
"num_image_positions": 577,
|
67 |
+
"num_key_value_heads": 16,
|
68 |
+
"patch_size": 14
|
69 |
},
|
70 |
"vision_feature_layers": [
|
71 |
-2,
|
72 |
-9
|
73 |
],
|
74 |
+
"vision_feature_select_strategy": "default",
|
75 |
+
"vocab_size": 152071
|
76 |
}
|
generation_config.json
CHANGED
@@ -1,4 +1,4 @@
|
|
1 |
{
|
2 |
"_from_model_config": true,
|
3 |
-
"transformers_version": "4.
|
4 |
}
|
|
|
1 |
{
|
2 |
"_from_model_config": true,
|
3 |
+
"transformers_version": "4.52.0.dev0"
|
4 |
}
|
model-00001-of-00007.safetensors
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:33faa9d5db45f207158c24765c462c07d9f523b5f4aabb56a4abab5f7b415790
|
3 |
+
size 4936632760
|
model-00002-of-00007.safetensors
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:377183545865d53632f4c5b6d045085cb9a705fea5150279f48de09b904d48d7
|
3 |
+
size 4998852560
|
model-00003-of-00007.safetensors
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:c3025afae405d5aafc22d08568f5800610ed89b092fbedb1afba59a4d0705888
|
3 |
+
size 4712543208
|
model-00004-of-00007.safetensors
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:a30711525d2912e5fcc45d8b6c6d6b36a5fb893392e9aa46ed94107ed3af6cc3
|
3 |
+
size 4661162896
|
model-00005-of-00007.safetensors
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:18ee25ec0d9b67899e3e0fb30975a72607bb4a0345382c778a7d412cdd31df6a
|
3 |
+
size 4661162896
|
model-00006-of-00007.safetensors
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:dea8f823824d5bf72ef143d34ae864784cacf767daa0bd7471c456e19975c8e4
|
3 |
+
size 4661162896
|
model-00007-of-00007.safetensors
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:f0c5aeae04bed3bdec13feab3e9880bb2b12d274f83958927678244b7b55743f
|
3 |
+
size 3451041816
|
model.safetensors.index.json
CHANGED
The diff for this file is too large to render.
See raw diff
|
|