Hugging Face's logo

HuggingFaceTB
/
SmolVLM2-2.2B-Instruct

Video-Text-to-Text
Transformers
Safetensors
English
smolvlm
image-text-to-text
Inference Endpoints
Model card Files Community
8
mfarre HF staff commited on
Commit
70402a4
·
verified ·
1 Parent(s): 67b63ec

Upload 3 files

Browse files
Files changed (3) hide show
  1. config.json +2 -2
  2. preprocessor_config.json +6 -1
  3. processor_config.json +2 -2
config.json CHANGED
@@ -4,7 +4,7 @@
4
  "SmolVLMForConditionalGeneration"
5
  ],
6
  "image_token_id": 49190,
7
- "model_type": "idefics3",
8
  "pad_token_id": 128002,
9
  "scale_factor": 3,
10
  "text_config": {
@@ -119,7 +119,7 @@
119
  "max_image_size": {
120
  "longest_edge": 384
121
  },
122
- "model_type": "idefics3_vision",
123
  "num_hidden_layers": 27,
124
  "patch_size": 14,
125
  "size": {
 
4
  "SmolVLMForConditionalGeneration"
5
  ],
6
  "image_token_id": 49190,
7
+ "model_type": "smolvlm",
8
  "pad_token_id": 128002,
9
  "scale_factor": 3,
10
  "text_config": {
 
119
  "max_image_size": {
120
  "longest_edge": 384
121
  },
122
+ "model_type": "smolvlm_vision",
123
  "num_hidden_layers": 27,
124
  "patch_size": 14,
125
  "size": {
preprocessor_config.json CHANGED
@@ -19,7 +19,12 @@
19
  "max_image_size": {
20
  "longest_edge": 384
21
  },
22
- "processor_class": "Idefics3Processor",
 
 
 
 
 
23
  "resample": 1,
24
  "rescale_factor": 0.00392156862745098,
25
  "size": {
 
19
  "max_image_size": {
20
  "longest_edge": 384
21
  },
22
+ "video_sampling": {
23
+ "fps": 1,
24
+ "max_frames": 64,
25
+ "video_size": {"longest_edge": 384}
26
+ },
27
+ "processor_class": "SmolVLMProcessor",
28
  "resample": 1,
29
  "rescale_factor": 0.00392156862745098,
30
  "size": {
processor_config.json CHANGED
@@ -1,4 +1,4 @@
1
  {
2
  "image_seq_len": 81,
3
- "processor_class": "Idefics3Processor"
4
- }
 
1
  {
2
  "image_seq_len": 81,
3
+ "processor_class": "SmolVLMProcessor"
4
+ }