enmingzhangzz commited on
Commit
6474444
·
verified ·
1 Parent(s): 1897405

Publish scopd-offpolicy

Browse files
models/scopd-offpolicy/README.md ADDED
@@ -0,0 +1,9 @@
 
 
 
 
 
 
 
 
 
 
1
+ # SCOPD Off-Policy
2
+
3
+ LoRA adapter for Qwen2.5-VL-7B-Instruct.
4
+
5
+ Objective: forward KL on teacher-generated responses.
6
+
7
+ Selection: all response tokens.
8
+
9
+ See the [collection model card](../../README.md) for download and evaluation.
models/scopd-offpolicy/adapter_config.json ADDED
@@ -0,0 +1,75 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "alora_invocation_tokens": null,
3
+ "alpha_pattern": {},
4
+ "arrow_config": null,
5
+ "auto_mapping": null,
6
+ "base_model_name_or_path": "Qwen/Qwen2.5-VL-7B-Instruct",
7
+ "bias": "none",
8
+ "corda_config": null,
9
+ "ensure_weight_tying": false,
10
+ "eva_config": null,
11
+ "exclude_modules": null,
12
+ "fan_in_fan_out": false,
13
+ "inference_mode": true,
14
+ "init_lora_weights": true,
15
+ "layer_replication": null,
16
+ "layers_pattern": "layers",
17
+ "layers_to_transform": [
18
+ 0,
19
+ 1,
20
+ 2,
21
+ 3,
22
+ 4,
23
+ 5,
24
+ 6,
25
+ 7,
26
+ 8,
27
+ 9,
28
+ 10,
29
+ 11,
30
+ 12,
31
+ 13,
32
+ 14,
33
+ 15,
34
+ 16,
35
+ 17,
36
+ 18,
37
+ 19,
38
+ 20,
39
+ 21,
40
+ 22,
41
+ 23,
42
+ 24,
43
+ 25,
44
+ 26,
45
+ 27
46
+ ],
47
+ "loftq_config": {},
48
+ "lora_alpha": 32,
49
+ "lora_bias": false,
50
+ "lora_dropout": 0.0,
51
+ "megatron_config": null,
52
+ "megatron_core": "megatron.core",
53
+ "modules_to_save": null,
54
+ "peft_type": "LORA",
55
+ "peft_version": "0.18.1",
56
+ "qalora_group_size": 16,
57
+ "r": 16,
58
+ "rank_pattern": {},
59
+ "revision": "cc594898137f460bfe9f0759e9844b3ce807cfb5",
60
+ "target_modules": [
61
+ "q_proj",
62
+ "k_proj",
63
+ "down_proj",
64
+ "gate_proj",
65
+ "up_proj",
66
+ "v_proj",
67
+ "o_proj"
68
+ ],
69
+ "target_parameters": null,
70
+ "task_type": "CAUSAL_LM",
71
+ "trainable_token_indices": null,
72
+ "use_dora": false,
73
+ "use_qalora": false,
74
+ "use_rslora": false
75
+ }
models/scopd-offpolicy/adapter_model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:dbf02470f09a41bfe8bad808a17f96ba892c582e7510d12daf2df0e1047b9281
3
+ size 161539072
models/scopd-offpolicy/provenance.json ADDED
@@ -0,0 +1,21 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "id": "scopd-offpolicy",
3
+ "label": "SCOPD Off-Policy",
4
+ "objective": "forward KL on teacher-generated responses",
5
+ "selection": "all response tokens",
6
+ "adapter_sha256": "dbf02470f09a41bfe8bad808a17f96ba892c582e7510d12daf2df0e1047b9281",
7
+ "base_model": "Qwen/Qwen2.5-VL-7B-Instruct",
8
+ "base_revision": "cc594898137f460bfe9f0759e9844b3ce807cfb5",
9
+ "merged": false,
10
+ "format": "peft_lora",
11
+ "dtype": "float32",
12
+ "inference_dtype": "bfloat16",
13
+ "questions_processed": 10240,
14
+ "path": "models/scopd-offpolicy",
15
+ "tensor_count": 392,
16
+ "parameter_count": 40370176,
17
+ "weight_bytes": 161480704,
18
+ "lora_rank": 16,
19
+ "lora_alpha": 32,
20
+ "lora_dropout": 0.0
21
+ }