Hugging Face's logo

PtRain
/
AbductiveMLLM

PEFT
Safetensors
Model card Files
xet
Community
PtRain commited on
Commit
d35a579
·
verified ·
1 Parent(s): dd3ab3f

Upload folder using huggingface_hub

Browse files
Files changed (9) hide show
  1. README.md +51 -0
  2. qwen2vl_7b_var_lora1/adapter_config.json +37 -0
  3. qwen2vl_7b_var_lora1/adapter_model.safetensors +3 -0
  4. qwen2vl_7b_var_select3_lora2/adapter_config.json +99 -0
  5. qwen2vl_7b_var_select3_lora2/adapter_model.safetensors +3 -0
  6. qwen2vl_7b_youcookii_lora1/adapter_config.json +37 -0
  7. qwen2vl_7b_youcookii_lora1/adapter_model.safetensors +3 -0
  8. qwen2vl_7b_youcookii_select3_lora2/adapter_config.json +99 -0
  9. qwen2vl_7b_youcookii_select3_lora2/adapter_model.safetensors +3 -0
README.md ADDED
@@ -0,0 +1,51 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ ---
2
+ base_model: Qwen2-VL-7B-Instruct
3
+ library_name: peft
4
+ ---
5
+
6
+ # Model Card
7
+
8
+ ## Model Details
9
+ - **Finetuned from model:** Qwen2-VL-7B-Instruct
10
+ - **Finetuned dataset:** VAR, YoucookII
11
+ - **Paper:** ABDUCTIVEMLLM: Boosting Visual Abductive Reasoning Within MLLMs
12
+
13
+
14
+ ## How to Get Started with the Model
15
+
16
+ Use the code below to get started with the model.
17
+
18
+ ```python
19
+ model = Qwen2VLForConditionalGeneration.from_pretrained(
20
+ 'path/to/your/Qwen2-VL-7B-Instruct',
21
+ torch_dtype='auto',
22
+ attn_implementation="flash_attention_2"
23
+ )
24
+
25
+ # merge lora adapter from Training Stage-1
26
+ lora_model_dir = 'qwen2vl_7b_var_lora1'
27
+ model = PeftModel.from_pretrained(model, model_id=lora_model_dir)
28
+ model = model.merge_and_unload()
29
+
30
+ # merge lora adapter from Training Stage-2
31
+ lora2_dir = 'qwen2vl_7b_var_select3_lora2'
32
+ model = PeftModel.from_pretrained(model, model_id=lora2_dir)
33
+ model = model.merge_and_unload()
34
+
35
+ model.generate()
36
+ ```
37
+
38
+
39
+ ## Citation
40
+ ```
41
+ @article{chang2026abductivemllm,
42
+ title={AbductiveMLLM: Boosting Visual Abductive Reasoning Within MLLMs},
43
+ author={Chang, Boyu and Wang, Qi and Guo, Xi and Nan, Zhixiong and Yao, Yazhou and Zhou, Tianfei},
44
+ journal={arXiv preprint arXiv:2601.02771},
45
+ year={2026}
46
+ }
47
+ ```
48
+
49
+
50
+
51
+
qwen2vl_7b_var_lora1/adapter_config.json ADDED
@@ -0,0 +1,37 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "alpha_pattern": {},
3
+ "auto_mapping": null,
4
+ "base_model_name_or_path": "dVAR/LLM_Weights/Qwen2-VL-7B-Instruct",
5
+ "bias": "none",
6
+ "eva_config": null,
7
+ "exclude_modules": null,
8
+ "fan_in_fan_out": false,
9
+ "inference_mode": true,
10
+ "init_lora_weights": true,
11
+ "layer_replication": null,
12
+ "layers_pattern": null,
13
+ "layers_to_transform": null,
14
+ "loftq_config": {},
15
+ "lora_alpha": 32,
16
+ "lora_bias": false,
17
+ "lora_dropout": 0.05,
18
+ "megatron_config": null,
19
+ "megatron_core": "megatron.core",
20
+ "modules_to_save": null,
21
+ "peft_type": "LORA",
22
+ "r": 16,
23
+ "rank_pattern": {},
24
+ "revision": null,
25
+ "target_modules": [
26
+ "k_proj",
27
+ "o_proj",
28
+ "up_proj",
29
+ "gate_proj",
30
+ "down_proj",
31
+ "q_proj",
32
+ "v_proj"
33
+ ],
34
+ "task_type": "CAUSAL_LM",
35
+ "use_dora": false,
36
+ "use_rslora": false
37
+ }
qwen2vl_7b_var_lora1/adapter_model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5d0a12fca5b399467581381ef926a5a33ed9a0c8b17def9957e1c5f0aa6bb8ed
3
+ size 80792880
qwen2vl_7b_var_select3_lora2/adapter_config.json ADDED
@@ -0,0 +1,99 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "alpha_pattern": {},
3
+ "auto_mapping": null,
4
+ "base_model_name_or_path": "dVAR/LLM_Weights/Qwen2-VL-7B-Instruct",
5
+ "bias": "none",
6
+ "eva_config": null,
7
+ "exclude_modules": [
8
+ "text_projection",
9
+ "clip_text_encoder.text_model.encoder.layers.7.self_attn.out_proj",
10
+ "clip_text_encoder.text_model.encoder.layers.8.self_attn",
11
+ "clip_text_encoder.text_model.encoder.layers.2.self_attn.k_proj",
12
+ "clip_text_encoder.text_model.encoder.layers.2.self_attn.v_proj",
13
+ "clip_text_encoder.text_model.encoder.layers.9.self_attn",
14
+ "clip_text_encoder.text_model.encoder.layers.11.self_attn.out_proj",
15
+ "clip_text_encoder.text_model.encoder.layers.3.self_attn.out_proj",
16
+ "clip_text_encoder.text_model.encoder.layers.0.self_attn.out_proj",
17
+ "clip_text_encoder.text_model.encoder.layers.9.self_attn.v_proj",
18
+ "clip_text_encoder.text_model.encoder.layers.11.self_attn.k_proj",
19
+ "clip_text_encoder.text_model.encoder.layers.3.self_attn.k_proj",
20
+ "clip_text_encoder.text_model.encoder.layers.8.self_attn.v_proj",
21
+ "clip_text_encoder.text_model.encoder.layers.5.self_attn",
22
+ "clip_text_encoder.text_model.encoder.layers.3.self_attn.q_proj",
23
+ "clip_text_encoder.text_model.encoder.layers.3.self_attn.v_proj",
24
+ "clip_text_encoder.text_model.encoder.layers.2.self_attn.out_proj",
25
+ "clip_text_encoder.text_model.encoder.layers.7.self_attn",
26
+ "clip_text_encoder.text_model.encoder.layers.2.self_attn.q_proj",
27
+ "clip_text_encoder.text_model.encoder.layers.4.self_attn.k_proj",
28
+ "clip_text_encoder.text_model.encoder.layers.1.self_attn",
29
+ "clip_text_encoder.text_model.encoder.layers.5.self_attn.k_proj",
30
+ "clip_text_encoder.text_model.encoder.layers.10.self_attn.q_proj",
31
+ "clip_text_encoder.text_model.encoder.layers.10.self_attn.out_proj",
32
+ "clip_text_encoder.text_model.encoder.layers.4.self_attn.out_proj",
33
+ "clip_text_encoder.text_model.encoder.layers.4.self_attn",
34
+ "clip_text_encoder.text_model.encoder.layers.10.self_attn.k_proj",
35
+ "clip_text_encoder.text_model.encoder.layers.6.self_attn.v_proj",
36
+ "clip_text_encoder.text_model.encoder.layers.7.self_attn.q_proj",
37
+ "clip_text_encoder.text_model.encoder.layers.9.self_attn.k_proj",
38
+ "clip_text_encoder.text_model.encoder.layers.0.self_attn.k_proj",
39
+ "clip_text_encoder.text_model.encoder.layers.6.self_attn",
40
+ "clip_text_encoder.text_model.encoder.layers.10.self_attn",
41
+ "clip_text_encoder.text_model.encoder.layers.11.self_attn",
42
+ "clip_text_encoder.text_model.encoder.layers.1.self_attn.q_proj",
43
+ "clip_text_encoder.text_model.encoder.layers.1.self_attn.out_proj",
44
+ "clip_text_encoder.text_model.encoder.layers.0.self_attn.q_proj",
45
+ "clip_text_encoder.text_model.encoder.layers.8.self_attn.out_proj",
46
+ "clip_text_encoder.text_model.encoder.layers.0.self_attn.v_proj",
47
+ "clip_text_encoder.text_model.encoder.layers.5.self_attn.q_proj",
48
+ "clip_text_encoder.text_model.encoder.layers.1.self_attn.k_proj",
49
+ "clip_text_encoder.text_model.encoder.layers.5.self_attn.v_proj",
50
+ "clip_text_encoder.text_model.encoder.layers.9.self_attn.out_proj",
51
+ "clip_text_encoder.text_model.encoder.layers.8.self_attn.q_proj",
52
+ "clip_text_encoder.text_model.encoder.layers.10.self_attn.v_proj",
53
+ "clip_text_encoder.text_model.encoder.layers.6.self_attn.q_proj",
54
+ "clip_text_encoder.text_model.encoder.layers.2.self_attn",
55
+ "clip_text_encoder.text_model.encoder.layers.7.self_attn.v_proj",
56
+ "clip_text_encoder.text_model.encoder.layers.11.self_attn.q_proj",
57
+ "clip_text_encoder.text_model.encoder.layers.0.self_attn",
58
+ "clip_text_encoder.text_model.encoder.layers.6.self_attn.k_proj",
59
+ "clip_text_encoder.text_model.encoder.layers.9.self_attn.q_proj",
60
+ "clip_text_encoder.text_model.encoder.layers.6.self_attn.out_proj",
61
+ "clip_text_encoder.text_model.encoder.layers.1.self_attn.v_proj",
62
+ "clip_text_encoder.text_model.encoder.layers.11.self_attn.v_proj",
63
+ "clip_text_encoder.text_model.encoder.layers.8.self_attn.k_proj",
64
+ "clip_text_encoder.text_model.encoder.layers.3.self_attn",
65
+ "clip_text_encoder.text_model.encoder.layers.5.self_attn.out_proj",
66
+ "clip_text_encoder.text_model.encoder.layers.7.self_attn.k_proj",
67
+ "clip_text_encoder.text_model.encoder.layers.4.self_attn.q_proj",
68
+ "clip_text_encoder.text_model.encoder.layers.4.self_attn.v_proj"
69
+ ],
70
+ "fan_in_fan_out": false,
71
+ "inference_mode": true,
72
+ "init_lora_weights": true,
73
+ "layer_replication": null,
74
+ "layers_pattern": null,
75
+ "layers_to_transform": null,
76
+ "loftq_config": {},
77
+ "lora_alpha": 32,
78
+ "lora_bias": false,
79
+ "lora_dropout": 0.05,
80
+ "megatron_config": null,
81
+ "megatron_core": "megatron.core",
82
+ "modules_to_save": null,
83
+ "peft_type": "LORA",
84
+ "r": 16,
85
+ "rank_pattern": {},
86
+ "revision": null,
87
+ "target_modules": [
88
+ "down_proj",
89
+ "v_proj",
90
+ "o_proj",
91
+ "up_proj",
92
+ "k_proj",
93
+ "gate_proj",
94
+ "q_proj"
95
+ ],
96
+ "task_type": "CAUSAL_LM",
97
+ "use_dora": false,
98
+ "use_rslora": false
99
+ }
qwen2vl_7b_var_select3_lora2/adapter_model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:759802f9cc8c1ede48c21f0aaed0a133667d90724e4a8ca770a9d687a736a103
3
+ size 161533192
qwen2vl_7b_youcookii_lora1/adapter_config.json ADDED
@@ -0,0 +1,37 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "alpha_pattern": {},
3
+ "auto_mapping": null,
4
+ "base_model_name_or_path": "dVAR/LLM_Weights/Qwen2-VL-7B-Instruct",
5
+ "bias": "none",
6
+ "eva_config": null,
7
+ "exclude_modules": null,
8
+ "fan_in_fan_out": false,
9
+ "inference_mode": true,
10
+ "init_lora_weights": true,
11
+ "layer_replication": null,
12
+ "layers_pattern": null,
13
+ "layers_to_transform": null,
14
+ "loftq_config": {},
15
+ "lora_alpha": 32,
16
+ "lora_bias": false,
17
+ "lora_dropout": 0.05,
18
+ "megatron_config": null,
19
+ "megatron_core": "megatron.core",
20
+ "modules_to_save": null,
21
+ "peft_type": "LORA",
22
+ "r": 16,
23
+ "rank_pattern": {},
24
+ "revision": null,
25
+ "target_modules": [
26
+ "k_proj",
27
+ "q_proj",
28
+ "up_proj",
29
+ "o_proj",
30
+ "down_proj",
31
+ "v_proj",
32
+ "gate_proj"
33
+ ],
34
+ "task_type": "CAUSAL_LM",
35
+ "use_dora": false,
36
+ "use_rslora": false
37
+ }
qwen2vl_7b_youcookii_lora1/adapter_model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:47153c635282573c65397d3f253a5a29e4c2d1d25707baaee62e2076f383f05a
3
+ size 80792880
qwen2vl_7b_youcookii_select3_lora2/adapter_config.json ADDED
@@ -0,0 +1,99 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "alpha_pattern": {},
3
+ "auto_mapping": null,
4
+ "base_model_name_or_path": "dVAR/LLM_Weights/Qwen2-VL-7B-Instruct",
5
+ "bias": "none",
6
+ "eva_config": null,
7
+ "exclude_modules": [
8
+ "clip_text_encoder.text_model.encoder.layers.1.self_attn.out_proj",
9
+ "clip_text_encoder.text_model.encoder.layers.7.self_attn.v_proj",
10
+ "clip_text_encoder.text_model.encoder.layers.10.self_attn.v_proj",
11
+ "clip_text_encoder.text_model.encoder.layers.4.self_attn.q_proj",
12
+ "clip_text_encoder.text_model.encoder.layers.0.self_attn.v_proj",
13
+ "clip_text_encoder.text_model.encoder.layers.6.self_attn.k_proj",
14
+ "clip_text_encoder.text_model.encoder.layers.4.self_attn.out_proj",
15
+ "clip_text_encoder.text_model.encoder.layers.0.self_attn.k_proj",
16
+ "clip_text_encoder.text_model.encoder.layers.0.self_attn.out_proj",
17
+ "clip_text_encoder.text_model.encoder.layers.3.self_attn.k_proj",
18
+ "clip_text_encoder.text_model.encoder.layers.4.self_attn",
19
+ "clip_text_encoder.text_model.encoder.layers.6.self_attn.v_proj",
20
+ "clip_text_encoder.text_model.encoder.layers.5.self_attn",
21
+ "clip_text_encoder.text_model.encoder.layers.7.self_attn.q_proj",
22
+ "clip_text_encoder.text_model.encoder.layers.8.self_attn",
23
+ "clip_text_encoder.text_model.encoder.layers.2.self_attn",
24
+ "clip_text_encoder.text_model.encoder.layers.11.self_attn.k_proj",
25
+ "clip_text_encoder.text_model.encoder.layers.5.self_attn.q_proj",
26
+ "clip_text_encoder.text_model.encoder.layers.10.self_attn.k_proj",
27
+ "clip_text_encoder.text_model.encoder.layers.0.self_attn",
28
+ "clip_text_encoder.text_model.encoder.layers.8.self_attn.v_proj",
29
+ "clip_text_encoder.text_model.encoder.layers.4.self_attn.k_proj",
30
+ "clip_text_encoder.text_model.encoder.layers.8.self_attn.out_proj",
31
+ "clip_text_encoder.text_model.encoder.layers.9.self_attn",
32
+ "clip_text_encoder.text_model.encoder.layers.11.self_attn.v_proj",
33
+ "clip_text_encoder.text_model.encoder.layers.2.self_attn.k_proj",
34
+ "clip_text_encoder.text_model.encoder.layers.7.self_attn",
35
+ "clip_text_encoder.text_model.encoder.layers.3.self_attn.out_proj",
36
+ "clip_text_encoder.text_model.encoder.layers.11.self_attn.out_proj",
37
+ "text_projection",
38
+ "clip_text_encoder.text_model.encoder.layers.1.self_attn.q_proj",
39
+ "clip_text_encoder.text_model.encoder.layers.3.self_attn",
40
+ "clip_text_encoder.text_model.encoder.layers.4.self_attn.v_proj",
41
+ "clip_text_encoder.text_model.encoder.layers.1.self_attn.k_proj",
42
+ "clip_text_encoder.text_model.encoder.layers.7.self_attn.out_proj",
43
+ "clip_text_encoder.text_model.encoder.layers.1.self_attn",
44
+ "clip_text_encoder.text_model.encoder.layers.9.self_attn.q_proj",
45
+ "clip_text_encoder.text_model.encoder.layers.10.self_attn",
46
+ "clip_text_encoder.text_model.encoder.layers.5.self_attn.k_proj",
47
+ "clip_text_encoder.text_model.encoder.layers.6.self_attn",
48
+ "clip_text_encoder.text_model.encoder.layers.3.self_attn.v_proj",
49
+ "clip_text_encoder.text_model.encoder.layers.0.self_attn.q_proj",
50
+ "clip_text_encoder.text_model.encoder.layers.9.self_attn.k_proj",
51
+ "clip_text_encoder.text_model.encoder.layers.2.self_attn.v_proj",
52
+ "clip_text_encoder.text_model.encoder.layers.10.self_attn.out_proj",
53
+ "clip_text_encoder.text_model.encoder.layers.7.self_attn.k_proj",
54
+ "clip_text_encoder.text_model.encoder.layers.10.self_attn.q_proj",
55
+ "clip_text_encoder.text_model.encoder.layers.5.self_attn.v_proj",
56
+ "clip_text_encoder.text_model.encoder.layers.6.self_attn.q_proj",
57
+ "clip_text_encoder.text_model.encoder.layers.8.self_attn.q_proj",
58
+ "clip_text_encoder.text_model.encoder.layers.9.self_attn.v_proj",
59
+ "clip_text_encoder.text_model.encoder.layers.3.self_attn.q_proj",
60
+ "clip_text_encoder.text_model.encoder.layers.9.self_attn.out_proj",
61
+ "clip_text_encoder.text_model.encoder.layers.11.self_attn.q_proj",
62
+ "clip_text_encoder.text_model.encoder.layers.5.self_attn.out_proj",
63
+ "clip_text_encoder.text_model.encoder.layers.2.self_attn.q_proj",
64
+ "clip_text_encoder.text_model.encoder.layers.2.self_attn.out_proj",
65
+ "clip_text_encoder.text_model.encoder.layers.8.self_attn.k_proj",
66
+ "clip_text_encoder.text_model.encoder.layers.1.self_attn.v_proj",
67
+ "clip_text_encoder.text_model.encoder.layers.11.self_attn",
68
+ "clip_text_encoder.text_model.encoder.layers.6.self_attn.out_proj"
69
+ ],
70
+ "fan_in_fan_out": false,
71
+ "inference_mode": true,
72
+ "init_lora_weights": true,
73
+ "layer_replication": null,
74
+ "layers_pattern": null,
75
+ "layers_to_transform": null,
76
+ "loftq_config": {},
77
+ "lora_alpha": 32,
78
+ "lora_bias": false,
79
+ "lora_dropout": 0.05,
80
+ "megatron_config": null,
81
+ "megatron_core": "megatron.core",
82
+ "modules_to_save": null,
83
+ "peft_type": "LORA",
84
+ "r": 16,
85
+ "rank_pattern": {},
86
+ "revision": null,
87
+ "target_modules": [
88
+ "down_proj",
89
+ "gate_proj",
90
+ "k_proj",
91
+ "v_proj",
92
+ "o_proj",
93
+ "q_proj",
94
+ "up_proj"
95
+ ],
96
+ "task_type": "CAUSAL_LM",
97
+ "use_dora": false,
98
+ "use_rslora": false
99
+ }
qwen2vl_7b_youcookii_select3_lora2/adapter_model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5205f2327872201d2827fde4a8ad1c170b9efd4b6fb986b41c06b477eede8fe5
3
+ size 161533192