Hugging Face's logo

ThinkMorph
/
ThinkMorph-SSL-Instruct

Any-to-Any
bagel
custom_code
Model card Files
xet
 Community
kuvvi commited on
Commit
d1fc329
·
verified ·
1 Parent(s): 04c8474

Upload release files

Browse files
Files changed (12) hide show
  1. README.md +127 -0
  2. added_tokens.json +25 -0
  3. ae.safetensors +3 -0
  4. config.json +202 -0
  5. ema.safetensors +3 -0
  6. generation_config.json +14 -0
  7. llm_config.json +86 -0
  8. merges.txt +0 -0
  9. special_tokens_map.json +31 -0
  10. tokenizer_config.json +215 -0
  11. vit_config.json +74 -0
  12. vocab.json +0 -0
README.md ADDED
@@ -0,0 +1,127 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ license: apache-2.0
3
+ base_model: ByteDance-Seed/BAGEL-7B-MoT
4
+ pipeline_tag: any-to-any
5
+ library_name: bagel
6
+ tags:
7
+ - bagel
8
+ - bagel-mot
9
+ - thinkmorph-ssl
10
+ - primer
11
+ - multimodal
12
+ - interleaved-reasoning
13
+ - spatial-reasoning
14
+ datasets:
15
+ - ThinkMorph/Pretext-270K
16
+ - ThinkMorph/Instruct-181K
17
+ ---
18
+
19
+ # ThinkMorph-SSL
20
+
21
+ ThinkMorph-SSL is a unified multimodal model that reasons with interleaved text and
22
+ generated images. It is trained from BAGEL-7B-MoT with the two-stage PRIMER recipe:
23
+
24
+ 1. **Stage 1 (pretext).** Training on [Pretext-270K](https://huggingface.co/datasets/ThinkMorph/Pretext-270K),
25
+ which has 270,896 instruction-free text-image-text traces built from source images.
26
+ This repository starts from the Stage 1 EMA weights at step 5,500, published separately as
27
+ [ThinkMorph-SSL-Stage1](https://huggingface.co/ThinkMorph/ThinkMorph-SSL-Stage1).
28
+ 2. **Stage 2 (instruction).** Training on [Instruct-181K](https://huggingface.co/datasets/ThinkMorph/Instruct-181K),
29
+ which has 181,291 question-driven interleaved traces.
30
+
31
+ This repository holds the **Stage 2 EMA weights at step 5,500**. They are the weights the
32
+ paper reports.
33
+
34
+ ## Files
35
+
36
+ | File | Bytes | SHA256 |
37
+ | --- | ---: | --- |
38
+ | `ema.safetensors` | 29,351,872,574 | `6497ea617dea4193d06135b97f00bc10bbdee1733789e489cae0f3c57141123f` |
39
+ | `ae.safetensors` | 167,664,678 | `e49d48aa3474df744ad718d11736d87f38d52bf7ca57dee6b8417978bfbfc9ad` |
40
+ | `config.json` | 5,175 | `4bc85a4daaef591f0b539216d09a1ee2294a1f3a3bda9f95885c13ed191ae701` |
41
+ | `llm_config.json` | 2,136 | `cb884fc65d117b8e5ec3856090224cd739c511e01c5c1de488cede5a54c8e994` |
42
+ | `vit_config.json` | 1,830 | `78664013bf6778ad4ecaebe78c153ef7fd46f7638fa78c0df1cb17d5bd7df0ab` |
43
+ | `generation_config.json` | 243 | `3a8f9087e486054c8a4a08dae2e5a3ba62e23da212b5b8c08bc42cb983c3459f` |
44
+ | `vocab.json` | 3,383,407 | `87a257b04b17642a0688c98cd1df89c398bda4fee532d6f88b38a659ecb4ac8d` |
45
+ | `merges.txt` | 1,671,853 | `8831e4f1a044471340f7c0a83d7bd71306a5b867e95fd870f74d0c5308a904d5` |
46
+ | `added_tokens.json` | 630 | `a0582f48ed8e795fda5b7b8eb2d98b492842dc229e78b66793bf545ce7ec52c8` |
47
+ | `special_tokens_map.json` | 613 | `76862e765266b85aa9459767e33cbaf13970f327a0e88d1c65846c2ddd3a1ecd` |
48
+ | `tokenizer_config.json` | 4,864 | `b7af0d4c9b917db05150006b466cfa2e8424b3915ff5fe2023a6f512a5e16dd5` |
49
+
50
+ - `ema.safetensors` holds the EMA weights in bfloat16: 1,451 tensors, 14,675,840,307
51
+ parameters. The training model also had a 26th SigLIP layer. BAGEL inference never uses
52
+ that layer (`vit_select_layer = -2`), so it is not stored, as in the original BAGEL release.
53
+ - `ae.safetensors` is the VAE used in training (about 160 MB). It is the one from
54
+ `lmms-lab/BAGEL-7B-MoT-ver.LE`. Do not replace it with the 320 MB VAE of the public BAGEL
55
+ release.
56
+ - Use the tokenizer files shipped here. They include the added special tokens used in
57
+ training.
58
+
59
+ ## Loading
60
+
61
+ The directory has the same layout as the original BAGEL-7B-MoT release, and the weights file
62
+ has the same name, `ema.safetensors`. To load:
63
+
64
+ - **Official BAGEL inference code** (<https://github.com/ByteDance-Seed/Bagel>): point
65
+ `model_path` at a local copy of this repository. The loader reads `llm_config.json`,
66
+ `vit_config.json`, `ae.safetensors` and `ema.safetensors`.
67
+ - **PRIMER code release:** `Bagel.from_pretrained(model_path, config, training_config=...)`
68
+ detects `ema.safetensors` and loads it directly. The evaluation wrapper `wrapper: Bagel`
69
+ also reads `ema.safetensors` when there is no `model.safetensors`.
70
+
71
+ ```bash
72
+ hf download ThinkMorph/ThinkMorph-SSL --local-dir ThinkMorph-SSL
73
+ ```
74
+
75
+ ## Training
76
+
77
+ | Item | Stage 1 (pretext) | Stage 2 (instruction) |
78
+ | --- | --- | --- |
79
+ | Run | `0503_stage0_270k_32gpu` | `0505_s0s2_173k_48gpus` |
80
+ | Initialization | BAGEL-7B-MoT (`lmms-lab/BAGEL-7B-MoT-ver.LE` layout) | Stage 1 EMA, step 5,500 |
81
+ | Data | Pretext-270K: 270,896 train rows; the val files are not used | Instruct-181K: all 21 subsets, train and val files |
82
+ | GPUs | 32 (4 nodes × 8) | 48 (6 nodes × 8) |
83
+ | Learning rate | 2.83e-5 (warmup 1,000) | 2e-5 (warmup 1,000) |
84
+ | Configured horizon | 10,000 steps | 12,000 steps |
85
+ | Released save | EMA, step 5,500 | EMA, step 5,500 |
86
+ | EMA | decay 0.9998 from step 1,000, fp32 shadow | decay 0.9998 from step 1,000, fp32 shadow |
87
+ | Packing | dynamic, up to 16,384 tokens | dynamic, up to 20,480 tokens |
88
+
89
+ The losses are text cross-entropy and image flow matching, both with weight 1. The VAE is
90
+ frozen, and the SigLIP encoder is trained. Stage 2 keeps only traces that have a first
91
+ reasoning image and an answer (`require_reasoning_image_0_and_answer`), and it trains on
92
+ the val files too (`include_val_in_training`).
93
+
94
+ ## Evaluation
95
+
96
+ The scores will be filled in after the paper's main table is final.
97
+
98
+ | Benchmark group | ThinkMorph-SSL |
99
+ | --- | --- |
100
+ | Visible evidence | TBD |
101
+ | Spatial structure | TBD |
102
+ | Scene reconstruction | TBD |
103
+
104
+ ## Intended use and limitations
105
+
106
+ This is a research model. It answers visual questions by writing text and generating
107
+ intermediate images.
108
+
109
+ - The intermediate images can leave out or change evidence that the question depends on.
110
+ - Default inference generates images, which costs a lot of compute.
111
+ - The earlier `ThinkMorph/ThinkMorph-7B` is a different model and cannot stand in for this
112
+ one.
113
+
114
+ ## License
115
+
116
+ The weights are released under Apache-2.0, following BAGEL-7B-MoT. The datasets carry their
117
+ own source terms; see their cards.
118
+
119
+ ## Citation
120
+
121
+ ```bibtex
122
+ @misc{thinkmorphssl2026,
123
+ title = {TBD},
124
+ author = {TBD},
125
+ year = {2026}
126
+ }
127
+ ```
added_tokens.json ADDED
@@ -0,0 +1,25 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "</tool_call>": 151658,
3
+ "<gen_image>": 151665,
4
+ "<tool_call>": 151657,
5
+ "<|box_end|>": 151649,
6
+ "<|box_start|>": 151648,
7
+ "<|endoftext|>": 151643,
8
+ "<|file_sep|>": 151664,
9
+ "<|fim_middle|>": 151660,
10
+ "<|fim_pad|>": 151662,
11
+ "<|fim_prefix|>": 151659,
12
+ "<|fim_suffix|>": 151661,
13
+ "<|im_end|>": 151645,
14
+ "<|im_start|>": 151644,
15
+ "<|image_pad|>": 151655,
16
+ "<|object_ref_end|>": 151647,
17
+ "<|object_ref_start|>": 151646,
18
+ "<|quad_end|>": 151651,
19
+ "<|quad_start|>": 151650,
20
+ "<|repo_name|>": 151663,
21
+ "<|video_pad|>": 151656,
22
+ "<|vision_end|>": 151653,
23
+ "<|vision_pad|>": 151654,
24
+ "<|vision_start|>": 151652
25
+ }
ae.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e49d48aa3474df744ad718d11736d87f38d52bf7ca57dee6b8417978bfbfc9ad
3
+ size 167664678
config.json ADDED
@@ -0,0 +1,202 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "architectures": [
3
+ "Bagel"
4
+ ],
5
+ "auto_map": {
6
+ "AutoConfig": "configuration_bagel.BagelConfig"
7
+ },
8
+ "ce_loss_reweighting": true,
9
+ "ce_weight": 1.0,
10
+ "connector_act": "gelu_pytorch_tanh",
11
+ "dtype": "float32",
12
+ "interpolate_pos": false,
13
+ "latent_patch_size": 2,
14
+ "llm_config": {
15
+ "_name_or_path": "",
16
+ "add_cross_attention": false,
17
+ "architectures": [
18
+ "Qwen2ForCausalLM"
19
+ ],
20
+ "attention_dropout": 0.0,
21
+ "bad_words_ids": null,
22
+ "begin_suppress_tokens": null,
23
+ "bos_token_id": 151643,
24
+ "chunk_size_feed_forward": 0,
25
+ "cross_attention_hidden_size": null,
26
+ "decoder_start_token_id": null,
27
+ "diversity_penalty": 0.0,
28
+ "do_sample": false,
29
+ "dtype": "bfloat16",
30
+ "early_stopping": false,
31
+ "encoder_no_repeat_ngram_size": 0,
32
+ "eos_token_id": 151645,
33
+ "exponential_decay_length_penalty": null,
34
+ "finetuning_task": null,
35
+ "forced_bos_token_id": null,
36
+ "forced_eos_token_id": null,
37
+ "freeze_und": false,
38
+ "hidden_act": "silu",
39
+ "hidden_size": 3584,
40
+ "id2label": {
41
+ "0": "LABEL_0",
42
+ "1": "LABEL_1"
43
+ },
44
+ "initializer_range": 0.02,
45
+ "intermediate_size": 18944,
46
+ "is_causal": true,
47
+ "is_decoder": false,
48
+ "is_encoder_decoder": false,
49
+ "label2id": {
50
+ "LABEL_0": 0,
51
+ "LABEL_1": 1
52
+ },
53
+ "layer_module": "Qwen2MoTDecoderLayer",
54
+ "length_penalty": 1.0,
55
+ "max_length": 20,
56
+ "max_position_embeddings": 32768,
57
+ "max_window_layers": 28,
58
+ "min_length": 0,
59
+ "model_type": "qwen2",
60
+ "no_repeat_ngram_size": 0,
61
+ "num_attention_heads": 28,
62
+ "num_beam_groups": 1,
63
+ "num_beams": 1,
64
+ "num_hidden_layers": 28,
65
+ "num_key_value_heads": 4,
66
+ "num_return_sequences": 1,
67
+ "output_attentions": false,
68
+ "output_hidden_states": false,
69
+ "output_scores": false,
70
+ "pad_token_id": null,
71
+ "prefix": null,
72
+ "problem_type": null,
73
+ "pruned_heads": {},
74
+ "qk_norm": true,
75
+ "remove_invalid_values": false,
76
+ "repetition_penalty": 1.0,
77
+ "return_dict": true,
78
+ "return_dict_in_generate": false,
79
+ "rms_norm_eps": 1e-06,
80
+ "rope_scaling": null,
81
+ "rope_theta": 1000000.0,
82
+ "sep_token_id": null,
83
+ "sliding_window": null,
84
+ "suppress_tokens": null,
85
+ "task_specific_params": null,
86
+ "temperature": 1.0,
87
+ "tf_legacy_loss": false,
88
+ "tie_encoder_decoder": false,
89
+ "tie_word_embeddings": false,
90
+ "tokenizer_class": null,
91
+ "top_k": 50,
92
+ "top_p": 1.0,
93
+ "torchscript": false,
94
+ "typical_p": 1.0,
95
+ "use_bfloat16": false,
96
+ "use_cache": true,
97
+ "use_sliding_window": false,
98
+ "vocab_size": 152064
99
+ },
100
+ "max_latent_size": 64,
101
+ "model_type": "bagel",
102
+ "mse_timestep_weighting": false,
103
+ "mse_weight": 1.0,
104
+ "timestep_shift": 1.0,
105
+ "transformers_version": "4.57.1",
106
+ "vae_config": {
107
+ "ch": 128,
108
+ "ch_mult": [
109
+ 1,
110
+ 2,
111
+ 4,
112
+ 4
113
+ ],
114
+ "downsample": 8,
115
+ "in_channels": 3,
116
+ "num_res_blocks": 2,
117
+ "out_ch": 3,
118
+ "resolution": 256,
119
+ "scale_factor": 0.3611,
120
+ "shift_factor": 0.1159,
121
+ "z_channels": 16
122
+ },
123
+ "visual_gen": true,
124
+ "visual_und": true,
125
+ "vit_config": {
126
+ "_name_or_path": "",
127
+ "add_cross_attention": false,
128
+ "architectures": null,
129
+ "attention_dropout": 0.0,
130
+ "bad_words_ids": null,
131
+ "begin_suppress_tokens": null,
132
+ "bos_token_id": null,
133
+ "chunk_size_feed_forward": 0,
134
+ "cross_attention_hidden_size": null,
135
+ "decoder_start_token_id": null,
136
+ "diversity_penalty": 0.0,
137
+ "do_sample": false,
138
+ "dtype": null,
139
+ "early_stopping": false,
140
+ "encoder_no_repeat_ngram_size": 0,
141
+ "eos_token_id": null,
142
+ "exponential_decay_length_penalty": null,
143
+ "finetuning_task": null,
144
+ "forced_bos_token_id": null,
145
+ "forced_eos_token_id": null,
146
+ "hidden_act": "gelu_pytorch_tanh",
147
+ "hidden_size": 1152,
148
+ "id2label": {
149
+ "0": "LABEL_0",
150
+ "1": "LABEL_1"
151
+ },
152
+ "image_size": 980,
153
+ "intermediate_size": 4304,
154
+ "is_decoder": false,
155
+ "is_encoder_decoder": false,
156
+ "label2id": {
157
+ "LABEL_0": 0,
158
+ "LABEL_1": 1
159
+ },
160
+ "layer_norm_eps": 1e-06,
161
+ "length_penalty": 1.0,
162
+ "max_length": 20,
163
+ "min_length": 0,
164
+ "model_type": "siglip_vision_model",
165
+ "no_repeat_ngram_size": 0,
166
+ "num_attention_heads": 16,
167
+ "num_beam_groups": 1,
168
+ "num_beams": 1,
169
+ "num_channels": 3,
170
+ "num_hidden_layers": 26,
171
+ "num_return_sequences": 1,
172
+ "output_attentions": false,
173
+ "output_hidden_states": false,
174
+ "output_scores": false,
175
+ "pad_token_id": null,
176
+ "patch_size": 14,
177
+ "prefix": null,
178
+ "problem_type": null,
179
+ "pruned_heads": {},
180
+ "remove_invalid_values": false,
181
+ "repetition_penalty": 1.0,
182
+ "return_dict": true,
183
+ "return_dict_in_generate": false,
184
+ "rope": false,
185
+ "sep_token_id": null,
186
+ "suppress_tokens": null,
187
+ "task_specific_params": null,
188
+ "temperature": 1.0,
189
+ "tf_legacy_loss": false,
190
+ "tie_encoder_decoder": false,
191
+ "tie_word_embeddings": true,
192
+ "tokenizer_class": null,
193
+ "top_k": 50,
194
+ "top_p": 1.0,
195
+ "torchscript": false,
196
+ "typical_p": 1.0,
197
+ "use_bfloat16": false
198
+ },
199
+ "vit_max_num_patch_per_side": 70,
200
+ "vit_rope": false,
201
+ "vit_select_layer": -2
202
+ }
ema.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6497ea617dea4193d06135b97f00bc10bbdee1733789e489cae0f3c57141123f
3
+ size 29351872574
generation_config.json ADDED
@@ -0,0 +1,14 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "bos_token_id": 151643,
3
+ "pad_token_id": 151643,
4
+ "do_sample": true,
5
+ "eos_token_id": [
6
+ 151645,
7
+ 151643
8
+ ],
9
+ "repetition_penalty": 1.05,
10
+ "temperature": 0.7,
11
+ "top_p": 0.8,
12
+ "top_k": 20,
13
+ "transformers_version": "4.37.0"
14
+ }
llm_config.json ADDED
@@ -0,0 +1,86 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "",
3
+ "add_cross_attention": false,
4
+ "architectures": [
5
+ "Qwen2ForCausalLM"
6
+ ],
7
+ "attention_dropout": 0.0,
8
+ "bad_words_ids": null,
9
+ "begin_suppress_tokens": null,
10
+ "bos_token_id": 151643,
11
+ "chunk_size_feed_forward": 0,
12
+ "cross_attention_hidden_size": null,
13
+ "decoder_start_token_id": null,
14
+ "diversity_penalty": 0.0,
15
+ "do_sample": false,
16
+ "dtype": "bfloat16",
17
+ "early_stopping": false,
18
+ "encoder_no_repeat_ngram_size": 0,
19
+ "eos_token_id": 151645,
20
+ "exponential_decay_length_penalty": null,
21
+ "finetuning_task": null,
22
+ "forced_bos_token_id": null,
23
+ "forced_eos_token_id": null,
24
+ "freeze_und": false,
25
+ "hidden_act": "silu",
26
+ "hidden_size": 3584,
27
+ "id2label": {
28
+ "0": "LABEL_0",
29
+ "1": "LABEL_1"
30
+ },
31
+ "initializer_range": 0.02,
32
+ "intermediate_size": 18944,
33
+ "is_causal": true,
34
+ "is_decoder": false,
35
+ "is_encoder_decoder": false,
36
+ "label2id": {
37
+ "LABEL_0": 0,
38
+ "LABEL_1": 1
39
+ },
40
+ "layer_module": "Qwen2MoTDecoderLayer",
41
+ "length_penalty": 1.0,
42
+ "max_length": 20,
43
+ "max_position_embeddings": 32768,
44
+ "max_window_layers": 28,
45
+ "min_length": 0,
46
+ "model_type": "qwen2",
47
+ "no_repeat_ngram_size": 0,
48
+ "num_attention_heads": 28,
49
+ "num_beam_groups": 1,
50
+ "num_beams": 1,
51
+ "num_hidden_layers": 28,
52
+ "num_key_value_heads": 4,
53
+ "num_return_sequences": 1,
54
+ "output_attentions": false,
55
+ "output_hidden_states": false,
56
+ "output_scores": false,
57
+ "pad_token_id": null,
58
+ "prefix": null,
59
+ "problem_type": null,
60
+ "pruned_heads": {},
61
+ "qk_norm": true,
62
+ "remove_invalid_values": false,
63
+ "repetition_penalty": 1.0,
64
+ "return_dict": true,
65
+ "return_dict_in_generate": false,
66
+ "rms_norm_eps": 1e-06,
67
+ "rope_scaling": null,
68
+ "rope_theta": 1000000.0,
69
+ "sep_token_id": null,
70
+ "sliding_window": null,
71
+ "suppress_tokens": null,
72
+ "task_specific_params": null,
73
+ "temperature": 1.0,
74
+ "tf_legacy_loss": false,
75
+ "tie_encoder_decoder": false,
76
+ "tie_word_embeddings": false,
77
+ "tokenizer_class": null,
78
+ "top_k": 50,
79
+ "top_p": 1.0,
80
+ "torchscript": false,
81
+ "typical_p": 1.0,
82
+ "use_bfloat16": false,
83
+ "use_cache": true,
84
+ "use_sliding_window": false,
85
+ "vocab_size": 152064
86
+ }
merges.txt ADDED
The diff for this file is too large to render. See raw diff
 
special_tokens_map.json ADDED
@@ -0,0 +1,31 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "additional_special_tokens": [
3
+ "<|im_start|>",
4
+ "<|im_end|>",
5
+ "<|object_ref_start|>",
6
+ "<|object_ref_end|>",
7
+ "<|box_start|>",
8
+ "<|box_end|>",
9
+ "<|quad_start|>",
10
+ "<|quad_end|>",
11
+ "<|vision_start|>",
12
+ "<|vision_end|>",
13
+ "<|vision_pad|>",
14
+ "<|image_pad|>",
15
+ "<|video_pad|>"
16
+ ],
17
+ "eos_token": {
18
+ "content": "<|im_end|>",
19
+ "lstrip": false,
20
+ "normalized": false,
21
+ "rstrip": false,
22
+ "single_word": false
23
+ },
24
+ "pad_token": {
25
+ "content": "<|endoftext|>",
26
+ "lstrip": false,
27
+ "normalized": false,
28
+ "rstrip": false,
29
+ "single_word": false
30
+ }
31
+ }
tokenizer_config.json ADDED
@@ -0,0 +1,215 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "add_bos_token": false,
3
+ "add_prefix_space": false,
4
+ "added_tokens_decoder": {
5
+ "151643": {
6
+ "content": "<|endoftext|>",
7
+ "lstrip": false,
8
+ "normalized": false,
9
+ "rstrip": false,
10
+ "single_word": false,
11
+ "special": true
12
+ },
13
+ "151644": {
14
+ "content": "<|im_start|>",
15
+ "lstrip": false,
16
+ "normalized": false,
17
+ "rstrip": false,
18
+ "single_word": false,
19
+ "special": true
20
+ },
21
+ "151645": {
22
+ "content": "<|im_end|>",
23
+ "lstrip": false,
24
+ "normalized": false,
25
+ "rstrip": false,
26
+ "single_word": false,
27
+ "special": true
28
+ },
29
+ "151646": {
30
+ "content": "<|object_ref_start|>",
31
+ "lstrip": false,
32
+ "normalized": false,
33
+ "rstrip": false,
34
+ "single_word": false,
35
+ "special": true
36
+ },
37
+ "151647": {
38
+ "content": "<|object_ref_end|>",
39
+ "lstrip": false,
40
+ "normalized": false,
41
+ "rstrip": false,
42
+ "single_word": false,
43
+ "special": true
44
+ },
45
+ "151648": {
46
+ "content": "<|box_start|>",
47
+ "lstrip": false,
48
+ "normalized": false,
49
+ "rstrip": false,
50
+ "single_word": false,
51
+ "special": true
52
+ },
53
+ "151649": {
54
+ "content": "<|box_end|>",
55
+ "lstrip": false,
56
+ "normalized": false,
57
+ "rstrip": false,
58
+ "single_word": false,
59
+ "special": true
60
+ },
61
+ "151650": {
62
+ "content": "<|quad_start|>",
63
+ "lstrip": false,
64
+ "normalized": false,
65
+ "rstrip": false,
66
+ "single_word": false,
67
+ "special": true
68
+ },
69
+ "151651": {
70
+ "content": "<|quad_end|>",
71
+ "lstrip": false,
72
+ "normalized": false,
73
+ "rstrip": false,
74
+ "single_word": false,
75
+ "special": true
76
+ },
77
+ "151652": {
78
+ "content": "<|vision_start|>",
79
+ "lstrip": false,
80
+ "normalized": false,
81
+ "rstrip": false,
82
+ "single_word": false,
83
+ "special": true
84
+ },
85
+ "151653": {
86
+ "content": "<|vision_end|>",
87
+ "lstrip": false,
88
+ "normalized": false,
89
+ "rstrip": false,
90
+ "single_word": false,
91
+ "special": true
92
+ },
93
+ "151654": {
94
+ "content": "<|vision_pad|>",
95
+ "lstrip": false,
96
+ "normalized": false,
97
+ "rstrip": false,
98
+ "single_word": false,
99
+ "special": true
100
+ },
101
+ "151655": {
102
+ "content": "<|image_pad|>",
103
+ "lstrip": false,
104
+ "normalized": false,
105
+ "rstrip": false,
106
+ "single_word": false,
107
+ "special": true
108
+ },
109
+ "151656": {
110
+ "content": "<|video_pad|>",
111
+ "lstrip": false,
112
+ "normalized": false,
113
+ "rstrip": false,
114
+ "single_word": false,
115
+ "special": true
116
+ },
117
+ "151657": {
118
+ "content": "<tool_call>",
119
+ "lstrip": false,
120
+ "normalized": false,
121
+ "rstrip": false,
122
+ "single_word": false,
123
+ "special": false
124
+ },
125
+ "151658": {
126
+ "content": "</tool_call>",
127
+ "lstrip": false,
128
+ "normalized": false,
129
+ "rstrip": false,
130
+ "single_word": false,
131
+ "special": false
132
+ },
133
+ "151659": {
134
+ "content": "<|fim_prefix|>",
135
+ "lstrip": false,
136
+ "normalized": false,
137
+ "rstrip": false,
138
+ "single_word": false,
139
+ "special": false
140
+ },
141
+ "151660": {
142
+ "content": "<|fim_middle|>",
143
+ "lstrip": false,
144
+ "normalized": false,
145
+ "rstrip": false,
146
+ "single_word": false,
147
+ "special": false
148
+ },
149
+ "151661": {
150
+ "content": "<|fim_suffix|>",
151
+ "lstrip": false,
152
+ "normalized": false,
153
+ "rstrip": false,
154
+ "single_word": false,
155
+ "special": false
156
+ },
157
+ "151662": {
158
+ "content": "<|fim_pad|>",
159
+ "lstrip": false,
160
+ "normalized": false,
161
+ "rstrip": false,
162
+ "single_word": false,
163
+ "special": false
164
+ },
165
+ "151663": {
166
+ "content": "<|repo_name|>",
167
+ "lstrip": false,
168
+ "normalized": false,
169
+ "rstrip": false,
170
+ "single_word": false,
171
+ "special": false
172
+ },
173
+ "151664": {
174
+ "content": "<|file_sep|>",
175
+ "lstrip": false,
176
+ "normalized": false,
177
+ "rstrip": false,
178
+ "single_word": false,
179
+ "special": false
180
+ },
181
+ "151665": {
182
+ "content": "<gen_image>",
183
+ "lstrip": false,
184
+ "normalized": true,
185
+ "rstrip": false,
186
+ "single_word": false,
187
+ "special": false
188
+ }
189
+ },
190
+ "additional_special_tokens": [
191
+ "<|im_start|>",
192
+ "<|im_end|>",
193
+ "<|object_ref_start|>",
194
+ "<|object_ref_end|>",
195
+ "<|box_start|>",
196
+ "<|box_end|>",
197
+ "<|quad_start|>",
198
+ "<|quad_end|>",
199
+ "<|vision_start|>",
200
+ "<|vision_end|>",
201
+ "<|vision_pad|>",
202
+ "<|image_pad|>",
203
+ "<|video_pad|>"
204
+ ],
205
+ "bos_token": null,
206
+ "clean_up_tokenization_spaces": false,
207
+ "eos_token": "<|im_end|>",
208
+ "errors": "replace",
209
+ "extra_special_tokens": {},
210
+ "model_max_length": 131072,
211
+ "pad_token": "<|endoftext|>",
212
+ "split_special_tokens": false,
213
+ "tokenizer_class": "Qwen2Tokenizer",
214
+ "unk_token": null
215
+ }
vit_config.json ADDED
@@ -0,0 +1,74 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "_name_or_path": "",
3
+ "add_cross_attention": false,
4
+ "architectures": null,
5
+ "attention_dropout": 0.0,
6
+ "bad_words_ids": null,
7
+ "begin_suppress_tokens": null,
8
+ "bos_token_id": null,
9
+ "chunk_size_feed_forward": 0,
10
+ "cross_attention_hidden_size": null,
11
+ "decoder_start_token_id": null,
12
+ "diversity_penalty": 0.0,
13
+ "do_sample": false,
14
+ "dtype": null,
15
+ "early_stopping": false,
16
+ "encoder_no_repeat_ngram_size": 0,
17
+ "eos_token_id": null,
18
+ "exponential_decay_length_penalty": null,
19
+ "finetuning_task": null,
20
+ "forced_bos_token_id": null,
21
+ "forced_eos_token_id": null,
22
+ "hidden_act": "gelu_pytorch_tanh",
23
+ "hidden_size": 1152,
24
+ "id2label": {
25
+ "0": "LABEL_0",
26
+ "1": "LABEL_1"
27
+ },
28
+ "image_size": 980,
29
+ "intermediate_size": 4304,
30
+ "is_decoder": false,
31
+ "is_encoder_decoder": false,
32
+ "label2id": {
33
+ "LABEL_0": 0,
34
+ "LABEL_1": 1
35
+ },
36
+ "layer_norm_eps": 1e-06,
37
+ "length_penalty": 1.0,
38
+ "max_length": 20,
39
+ "min_length": 0,
40
+ "model_type": "siglip_vision_model",
41
+ "no_repeat_ngram_size": 0,
42
+ "num_attention_heads": 16,
43
+ "num_beam_groups": 1,
44
+ "num_beams": 1,
45
+ "num_channels": 3,
46
+ "num_hidden_layers": 26,
47
+ "num_return_sequences": 1,
48
+ "output_attentions": false,
49
+ "output_hidden_states": false,
50
+ "output_scores": false,
51
+ "pad_token_id": null,
52
+ "patch_size": 14,
53
+ "prefix": null,
54
+ "problem_type": null,
55
+ "pruned_heads": {},
56
+ "remove_invalid_values": false,
57
+ "repetition_penalty": 1.0,
58
+ "return_dict": true,
59
+ "return_dict_in_generate": false,
60
+ "rope": false,
61
+ "sep_token_id": null,
62
+ "suppress_tokens": null,
63
+ "task_specific_params": null,
64
+ "temperature": 1.0,
65
+ "tf_legacy_loss": false,
66
+ "tie_encoder_decoder": false,
67
+ "tie_word_embeddings": true,
68
+ "tokenizer_class": null,
69
+ "top_k": 50,
70
+ "top_p": 1.0,
71
+ "torchscript": false,
72
+ "typical_p": 1.0,
73
+ "use_bfloat16": false
74
+ }
vocab.json ADDED
The diff for this file is too large to render. See raw diff