Upload ColPali (#1)
Browse files- Upload ColPali (9f6110e72ebb37949b08b8f86cfadb2c6819d83b)
- config.json +15 -15
- model.safetensors +3 -0
config.json
CHANGED
|
@@ -1,6 +1,6 @@
|
|
| 1 |
{
|
| 2 |
-
"_name_or_path": "
|
| 3 |
-
"_vocab_size":
|
| 4 |
"architectures": [
|
| 5 |
"ColPali"
|
| 6 |
],
|
|
@@ -10,30 +10,30 @@
|
|
| 10 |
"image_token_index": 257152,
|
| 11 |
"model_type": "paligemma",
|
| 12 |
"pad_token_id": 0,
|
| 13 |
-
"projection_dim":
|
| 14 |
"text_config": {
|
| 15 |
-
"hidden_size":
|
| 16 |
-
"intermediate_size":
|
| 17 |
"model_type": "gemma",
|
| 18 |
"num_attention_heads": 8,
|
| 19 |
-
"
|
| 20 |
-
"
|
|
|
|
| 21 |
"num_key_value_heads": 1,
|
| 22 |
"torch_dtype": "float32",
|
| 23 |
"vocab_size": 257216
|
| 24 |
},
|
| 25 |
-
"torch_dtype": "
|
| 26 |
"transformers_version": "4.45.2",
|
| 27 |
"vision_config": {
|
| 28 |
-
"hidden_size":
|
| 29 |
-
"
|
| 30 |
-
"intermediate_size": 4304,
|
| 31 |
"model_type": "siglip_vision_model",
|
| 32 |
-
"num_attention_heads":
|
| 33 |
-
"num_hidden_layers":
|
| 34 |
-
"num_image_tokens":
|
| 35 |
"patch_size": 14,
|
| 36 |
-
"projection_dim":
|
| 37 |
"projector_hidden_act": "gelu_fast",
|
| 38 |
"vision_use_head": false
|
| 39 |
}
|
|
|
|
| 1 |
{
|
| 2 |
+
"_name_or_path": "trl-internal-testing/tiny-random-paligemma",
|
| 3 |
+
"_vocab_size": 257216,
|
| 4 |
"architectures": [
|
| 5 |
"ColPali"
|
| 6 |
],
|
|
|
|
| 10 |
"image_token_index": 257152,
|
| 11 |
"model_type": "paligemma",
|
| 12 |
"pad_token_id": 0,
|
| 13 |
+
"projection_dim": 32,
|
| 14 |
"text_config": {
|
| 15 |
+
"hidden_size": 32,
|
| 16 |
+
"intermediate_size": 32,
|
| 17 |
"model_type": "gemma",
|
| 18 |
"num_attention_heads": 8,
|
| 19 |
+
"num_attn_heads": 1,
|
| 20 |
+
"num_hidden_layers": 1,
|
| 21 |
+
"num_image_tokens": 256,
|
| 22 |
"num_key_value_heads": 1,
|
| 23 |
"torch_dtype": "float32",
|
| 24 |
"vocab_size": 257216
|
| 25 |
},
|
| 26 |
+
"torch_dtype": "float16",
|
| 27 |
"transformers_version": "4.45.2",
|
| 28 |
"vision_config": {
|
| 29 |
+
"hidden_size": 32,
|
| 30 |
+
"intermediate_size": 32,
|
|
|
|
| 31 |
"model_type": "siglip_vision_model",
|
| 32 |
+
"num_attention_heads": 1,
|
| 33 |
+
"num_hidden_layers": 1,
|
| 34 |
+
"num_image_tokens": 256,
|
| 35 |
"patch_size": 14,
|
| 36 |
+
"projection_dim": 32,
|
| 37 |
"projector_hidden_act": "gelu_fast",
|
| 38 |
"vision_use_head": false
|
| 39 |
}
|
model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e3824a4177e663718a679e763219e4ae6c56107c9276044f8ff086695ea631ea
|
| 3 |
+
size 16845392
|