Hugging Face's logo

ibm-granite
/
granite-docling-258M

Image-Text-to-Text
Transformers
Safetensors
English
idefics3
image-to-text
text-generation
documents
code
formula
chart
ocr
layout
table
document-parse
docling
granite
extraction
math
conversational
Model card Files
xet
Community
43
asnassar commited on
Commit
6298576
·
verified ·
1 Parent(s): 341382d

v0.0.2 (#3)

Browse files

- v0.0.2 (ecba62ead2399d4df4d3f69a180419e09d2dbb67)

Files changed (6) hide show
  1. config.json +3 -10
  2. generation_config.json +2 -2
  3. merges.txt +0 -0
  4. model.safetensors +1 -1
  5. tokenizer_config.json +3 -3
  6. trainer_state.json +0 -0
config.json CHANGED
@@ -5,20 +5,14 @@
5
  ],
6
  "attention_bias": false,
7
  "attention_dropout": 0.0,
8
- "bos_token_id": 100257,
9
- "eos_token_id": 100257,
10
- "freeze_lm_head": true,
11
- "freeze_text_layers": true,
12
- "freeze_text_module_exceptions": [],
13
- "freeze_vision_layers": true,
14
- "freeze_vision_module_exceptions": [],
15
  "head_dim": 64,
16
  "hidden_act": "silu",
17
  "hidden_size": 576,
18
  "image_token_id": 100270,
19
  "initializer_range": 0.02,
20
  "intermediate_size": 1536,
21
- "lm_head_path": "/data1/checkpoints/v06-granite165m/step_2000000_ckp/",
22
  "max_position_embeddings": 8192,
23
  "mlp_bias": false,
24
  "model_type": "idefics3",
@@ -47,7 +41,7 @@
47
  "scale_factor": 4,
48
  "text_config": {
49
  "architectures": [
50
- "VLlama3ForCausalLM"
51
  ],
52
  "attention_bias": false,
53
  "attention_dropout": 0.0,
@@ -68,7 +62,6 @@
68
  "rms_norm_eps": 1e-05,
69
  "rope_scaling": null,
70
  "rope_theta": 10000.0,
71
- "tie_word_embeddings": true,
72
  "torch_dtype": "bfloat16",
73
  "use_cache": true,
74
  "vocab_size": 100480
 
5
  ],
6
  "attention_bias": false,
7
  "attention_dropout": 0.0,
8
+ "bos_token_id": 100264,
9
+ "eos_token_id": 100338,
 
 
 
 
 
10
  "head_dim": 64,
11
  "hidden_act": "silu",
12
  "hidden_size": 576,
13
  "image_token_id": 100270,
14
  "initializer_range": 0.02,
15
  "intermediate_size": 1536,
 
16
  "max_position_embeddings": 8192,
17
  "mlp_bias": false,
18
  "model_type": "idefics3",
 
41
  "scale_factor": 4,
42
  "text_config": {
43
  "architectures": [
44
+ "llama"
45
  ],
46
  "attention_bias": false,
47
  "attention_dropout": 0.0,
 
62
  "rms_norm_eps": 1e-05,
63
  "rope_scaling": null,
64
  "rope_theta": 10000.0,
 
65
  "torch_dtype": "bfloat16",
66
  "use_cache": true,
67
  "vocab_size": 100480
generation_config.json CHANGED
@@ -1,7 +1,7 @@
1
  {
2
  "_from_model_config": true,
3
- "bos_token_id": 100257,
4
- "eos_token_id": 100257,
5
  "pad_token_id": 128002,
6
  "transformers_version": "4.53.0.dev0"
7
  }
 
1
  {
2
  "_from_model_config": true,
3
+ "bos_token_id": 100264,
4
+ "eos_token_id": 100338,
5
  "pad_token_id": 128002,
6
  "transformers_version": "4.53.0.dev0"
7
  }
merges.txt ADDED
The diff for this file is too large to render. See raw diff
 
model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:33e612f587edfeaa9f5e8c42b20e52c5a74fdb47a4ec2cc569d005c6b1d400b5
3
  size 630993616
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:92d6fa41f9edcb5ef51f9ac1844edb11dfdaf403473351d1ddbee0e883e42f25
3
  size 630993616
tokenizer_config.json CHANGED
@@ -1064,12 +1064,12 @@
1064
  "<image>",
1065
  "<end_of_utterance>"
1066
  ],
1067
- "bos_token": "<|end_of_text|>",
1068
  "clean_up_tokenization_spaces": false,
1069
- "eos_token": "<|end_of_text|>",
1070
  "errors": "replace",
1071
  "extra_special_tokens": {},
1072
- "model_max_length": 1000000000000000019884624838656,
1073
  "pad_token": "<|pad|>",
1074
  "padding_side": "left",
1075
  "processor_class": "Idefics3Processor",
 
1064
  "<image>",
1065
  "<end_of_utterance>"
1066
  ],
1067
+ "bos_token": "<|start_of_role|>",
1068
  "clean_up_tokenization_spaces": false,
1069
+ "eos_token": "<|end_of_utterance|>",
1070
  "errors": "replace",
1071
  "extra_special_tokens": {},
1072
+ "model_max_length": 8192,
1073
  "pad_token": "<|pad|>",
1074
  "padding_side": "left",
1075
  "processor_class": "Idefics3Processor",
trainer_state.json DELETED
The diff for this file is too large to render. See raw diff