Diffusers
Safetensors
dontKnow23456 commited on
Commit
8a42411
·
verified ·
1 Parent(s): d2c6e8a

Native OPD-v5.0 opdv5-libero-long-offline-only-2000s-fullep: student update 100 ema

Browse files
checkpoints/opd-v4-native-libero-long/opdv5-libero-long-offline-only-2000s-fullep/20261106T122632Z-1145c934/students/update-000100-ema/adapter.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:49ecc581a09c6d7cf0357e12e1e1ff806dc3c9e7899417affdc0881429db797b
3
+ size 1824633583
checkpoints/opd-v4-native-libero-long/opdv5-libero-long-offline-only-2000s-fullep/20261106T122632Z-1145c934/students/update-000100-ema/export.json ADDED
@@ -0,0 +1,2884 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "schema": "native-opd-v4-student-v3",
3
+ "variant": "ema",
4
+ "source_commit": "7c6ffa9bfc4b83582cafc860fab4c82cc7deeeeb",
5
+ "base_repo": "robbyant/lingbot-va-posttrain-libero-long",
6
+ "base_revision": "0e89d1e753019988aba484e8da2dc0810e264d9f",
7
+ "contract_hash": "16274088a0749b47b83dc735f7d1a8c2f6527aa97a4c7bfc349f12d4fb3dba71",
8
+ "config": {
9
+ "native": {
10
+ "suite": "libero_10",
11
+ "task_order_index": 0,
12
+ "control_hz": 20,
13
+ "image_height": 128,
14
+ "image_width": 128,
15
+ "cameras": [
16
+ "observation.images.agentview_rgb",
17
+ "observation.images.eye_in_hand_rgb"
18
+ ],
19
+ "latent_channels": 48,
20
+ "latent_height": 8,
21
+ "latent_width": 16,
22
+ "frame_groups": 4,
23
+ "action_dim": 30,
24
+ "active_channels": [
25
+ 0,
26
+ 1,
27
+ 2,
28
+ 3,
29
+ 4,
30
+ 5,
31
+ 6
32
+ ],
33
+ "actions_per_group": 4,
34
+ "patch_size": [
35
+ 1,
36
+ 2,
37
+ 2
38
+ ],
39
+ "attention_window": 30,
40
+ "video_shift": 5.0,
41
+ "action_shift": 0.05,
42
+ "video_cfg": 5.0,
43
+ "action_cfg": 1.0,
44
+ "horizon_protocol": "native_chunk800_v1",
45
+ "q01": [
46
+ -0.6589285731315613,
47
+ -0.84375,
48
+ -0.9375,
49
+ -0.12107142806053162,
50
+ -0.15964286029338837,
51
+ -0.26571428775787354,
52
+ -1.0
53
+ ],
54
+ "q99": [
55
+ 0.8999999761581421,
56
+ 0.8544642925262451,
57
+ 0.9375,
58
+ 0.17142857611179352,
59
+ 0.1842857152223587,
60
+ 0.34392857551574707,
61
+ 1.0
62
+ ]
63
+ },
64
+ "method": {
65
+ "objective": "regression",
66
+ "state_steps": 1,
67
+ "action_steps": 2,
68
+ "teacher_video_steps": 20,
69
+ "teacher_action_steps": 50,
70
+ "lora_rank": 256,
71
+ "lora_alpha": 256,
72
+ "lora_dropout": 0.0,
73
+ "onpolicy_contexts_per_rank": 3,
74
+ "offline_contexts_per_rank": 6,
75
+ "replay_rounds": 0,
76
+ "replay_contexts_per_rank": 0,
77
+ "collect_once_rounds": 100,
78
+ "data_mode": "offline_only",
79
+ "action_target_plan": "teacher",
80
+ "student_updates": 2000,
81
+ "world_size": 8,
82
+ "student_lr": 3e-05,
83
+ "weight_decay": 0.01,
84
+ "grad_clip": 1.0,
85
+ "huber_c": 0.001,
86
+ "action_loss_weight": 1.0
87
+ },
88
+ "experiment": {
89
+ "seed": 42,
90
+ "adam_betas": [
91
+ 0.9,
92
+ 0.999
93
+ ],
94
+ "adam_epsilon": 1e-08,
95
+ "lr_schedule": "student_linear_warmup",
96
+ "student_init_source": "official_libero_long",
97
+ "stage1_training_executed": false,
98
+ "stage1_optimizer_updates": 0,
99
+ "admission_sr_threshold": null,
100
+ "checkpoint_every": 100,
101
+ "retain_checkpoints": 2,
102
+ "student_snapshots": [
103
+ 100,
104
+ 200,
105
+ 300,
106
+ 400,
107
+ 500,
108
+ 600,
109
+ 700,
110
+ 800,
111
+ 900
112
+ ],
113
+ "training_profile": "libero_long_eval_aligned_chunks_v2",
114
+ "training_chunks": [
115
+ 50,
116
+ 50,
117
+ 50,
118
+ 50,
119
+ 50,
120
+ 50,
121
+ 50,
122
+ 50,
123
+ 50,
124
+ 50
125
+ ],
126
+ "task_sampling": "global_confirmed_episode_starts_round_robin",
127
+ "max_active_per_task": 2,
128
+ "diagnostic_updates": 50,
129
+ "train_cases_per_task": 100,
130
+ "validation_cases_per_task": 20,
131
+ "test_cases_per_task": 50,
132
+ "evaluation_video_cases_per_task": 2,
133
+ "student_lr_warmup_steps": 20,
134
+ "ema_decay": 0.99,
135
+ "ema_start_update": 50,
136
+ "protected_checkpoints": [
137
+ 500,
138
+ 900
139
+ ],
140
+ "heldout_contexts": 64,
141
+ "heldout_every": 50,
142
+ "offline_max_prefix": 49,
143
+ "async_collection": false
144
+ },
145
+ "machine": {
146
+ "model_root": "/shared/yue/opd4/wam-policy/native-opd-v4/resources/lingbot-va-posttrain-libero-long",
147
+ "resource_manifest": "/shared/yue/opd4/b2/configs/opd_v4/libero_long/resources.json",
148
+ "cpu_python": "/shared/yue/opd4/wam-policy/native-opd-v4/envs/libero-cpu/bin/python",
149
+ "worker_config": "/shared/yue/opd4/wam-policy/native-opd-v4/envs/libero-cpu/worker-env.json",
150
+ "cpu_runtime_lock": "/shared/yue/opd4/wam-policy/native-opd-v4/resources/cpu-runtime-lock.json",
151
+ "reset_manifest": "/shared/yue/opd4/wam-policy/native-opd-v4/resources/libero-resets/manifest.json",
152
+ "output_dir": "/shared/yue/opd4/wam-policy/native-opd-v4/runs/opdv5-libero-long-offline-only-2000s-fullep/train",
153
+ "offline_cache": "/shared/yue/opd4/wam-policy/native-opd-v4/runs/opdv5-libero-long-offline-only-2000s-fullep/cache",
154
+ "device": "cuda",
155
+ "network_dtype": "bfloat16",
156
+ "worker_timeout_s": 180,
157
+ "process_group_timeout_s": 1800,
158
+ "worker_threads": 1,
159
+ "cpu_affinity": [],
160
+ "activation_checkpointing": false,
161
+ "visualization_every": 400
162
+ },
163
+ "profile": "formal",
164
+ "schema": "native-libero-streaming-opd-v5-regression-v1",
165
+ "legacy_contract": null
166
+ },
167
+ "resources": {
168
+ "repo": "robbyant/lingbot-va-posttrain-libero-long",
169
+ "revision": "0e89d1e753019988aba484e8da2dc0810e264d9f",
170
+ "manifest_hash": "91514a05cebce085949208d389eed5a4322bd4c026d719f8944099a05fde7aad",
171
+ "source_commit": "7c6ffa9bfc4b83582cafc860fab4c82cc7deeeeb",
172
+ "implementation_hash": "32e8c15d2c37000583df6fc14576e505dbc765406eb12bf0b8d0e5f0bd4fa1ce",
173
+ "cpu_runtime_identity": "3283ff6607481e48e1c93cf8f621eb3aecf3a7a16866aa3e41230a1279927256",
174
+ "reset_manifest_hash": "ca5b712b8a97f6282eae7ec809081167a7fecfa53213ce801fbf12c6221be0c6",
175
+ "prompt_file_hash": "d7e3c8d414466f64de724fc497d3bd992da674cf4e6d9e081a68500d14401c63",
176
+ "offline_cache_sha256": "57d0d6222243a900dcc9f67fb316dfefca16b1e5ec7e3c97ef80098a79376968"
177
+ },
178
+ "student_update": 100,
179
+ "ema_decay": 0.99,
180
+ "ema_start_update": 50,
181
+ "lora": [
182
+ {
183
+ "target": "blocks.0.attn1.to_q",
184
+ "rank": 256,
185
+ "alpha": 256.0,
186
+ "dropout": 0.0,
187
+ "A": [
188
+ 256,
189
+ 3072
190
+ ],
191
+ "B": [
192
+ 3072,
193
+ 256
194
+ ],
195
+ "master_dtype": "torch.float32"
196
+ },
197
+ {
198
+ "target": "blocks.0.attn1.to_k",
199
+ "rank": 256,
200
+ "alpha": 256.0,
201
+ "dropout": 0.0,
202
+ "A": [
203
+ 256,
204
+ 3072
205
+ ],
206
+ "B": [
207
+ 3072,
208
+ 256
209
+ ],
210
+ "master_dtype": "torch.float32"
211
+ },
212
+ {
213
+ "target": "blocks.0.attn1.to_v",
214
+ "rank": 256,
215
+ "alpha": 256.0,
216
+ "dropout": 0.0,
217
+ "A": [
218
+ 256,
219
+ 3072
220
+ ],
221
+ "B": [
222
+ 3072,
223
+ 256
224
+ ],
225
+ "master_dtype": "torch.float32"
226
+ },
227
+ {
228
+ "target": "blocks.0.attn1.to_out.0",
229
+ "rank": 256,
230
+ "alpha": 256.0,
231
+ "dropout": 0.0,
232
+ "A": [
233
+ 256,
234
+ 3072
235
+ ],
236
+ "B": [
237
+ 3072,
238
+ 256
239
+ ],
240
+ "master_dtype": "torch.float32"
241
+ },
242
+ {
243
+ "target": "blocks.0.ffn.net.0.proj",
244
+ "rank": 256,
245
+ "alpha": 256.0,
246
+ "dropout": 0.0,
247
+ "A": [
248
+ 256,
249
+ 3072
250
+ ],
251
+ "B": [
252
+ 14336,
253
+ 256
254
+ ],
255
+ "master_dtype": "torch.float32"
256
+ },
257
+ {
258
+ "target": "blocks.0.ffn.net.2",
259
+ "rank": 256,
260
+ "alpha": 256.0,
261
+ "dropout": 0.0,
262
+ "A": [
263
+ 256,
264
+ 14336
265
+ ],
266
+ "B": [
267
+ 3072,
268
+ 256
269
+ ],
270
+ "master_dtype": "torch.float32"
271
+ },
272
+ {
273
+ "target": "blocks.1.attn1.to_q",
274
+ "rank": 256,
275
+ "alpha": 256.0,
276
+ "dropout": 0.0,
277
+ "A": [
278
+ 256,
279
+ 3072
280
+ ],
281
+ "B": [
282
+ 3072,
283
+ 256
284
+ ],
285
+ "master_dtype": "torch.float32"
286
+ },
287
+ {
288
+ "target": "blocks.1.attn1.to_k",
289
+ "rank": 256,
290
+ "alpha": 256.0,
291
+ "dropout": 0.0,
292
+ "A": [
293
+ 256,
294
+ 3072
295
+ ],
296
+ "B": [
297
+ 3072,
298
+ 256
299
+ ],
300
+ "master_dtype": "torch.float32"
301
+ },
302
+ {
303
+ "target": "blocks.1.attn1.to_v",
304
+ "rank": 256,
305
+ "alpha": 256.0,
306
+ "dropout": 0.0,
307
+ "A": [
308
+ 256,
309
+ 3072
310
+ ],
311
+ "B": [
312
+ 3072,
313
+ 256
314
+ ],
315
+ "master_dtype": "torch.float32"
316
+ },
317
+ {
318
+ "target": "blocks.1.attn1.to_out.0",
319
+ "rank": 256,
320
+ "alpha": 256.0,
321
+ "dropout": 0.0,
322
+ "A": [
323
+ 256,
324
+ 3072
325
+ ],
326
+ "B": [
327
+ 3072,
328
+ 256
329
+ ],
330
+ "master_dtype": "torch.float32"
331
+ },
332
+ {
333
+ "target": "blocks.1.ffn.net.0.proj",
334
+ "rank": 256,
335
+ "alpha": 256.0,
336
+ "dropout": 0.0,
337
+ "A": [
338
+ 256,
339
+ 3072
340
+ ],
341
+ "B": [
342
+ 14336,
343
+ 256
344
+ ],
345
+ "master_dtype": "torch.float32"
346
+ },
347
+ {
348
+ "target": "blocks.1.ffn.net.2",
349
+ "rank": 256,
350
+ "alpha": 256.0,
351
+ "dropout": 0.0,
352
+ "A": [
353
+ 256,
354
+ 14336
355
+ ],
356
+ "B": [
357
+ 3072,
358
+ 256
359
+ ],
360
+ "master_dtype": "torch.float32"
361
+ },
362
+ {
363
+ "target": "blocks.2.attn1.to_q",
364
+ "rank": 256,
365
+ "alpha": 256.0,
366
+ "dropout": 0.0,
367
+ "A": [
368
+ 256,
369
+ 3072
370
+ ],
371
+ "B": [
372
+ 3072,
373
+ 256
374
+ ],
375
+ "master_dtype": "torch.float32"
376
+ },
377
+ {
378
+ "target": "blocks.2.attn1.to_k",
379
+ "rank": 256,
380
+ "alpha": 256.0,
381
+ "dropout": 0.0,
382
+ "A": [
383
+ 256,
384
+ 3072
385
+ ],
386
+ "B": [
387
+ 3072,
388
+ 256
389
+ ],
390
+ "master_dtype": "torch.float32"
391
+ },
392
+ {
393
+ "target": "blocks.2.attn1.to_v",
394
+ "rank": 256,
395
+ "alpha": 256.0,
396
+ "dropout": 0.0,
397
+ "A": [
398
+ 256,
399
+ 3072
400
+ ],
401
+ "B": [
402
+ 3072,
403
+ 256
404
+ ],
405
+ "master_dtype": "torch.float32"
406
+ },
407
+ {
408
+ "target": "blocks.2.attn1.to_out.0",
409
+ "rank": 256,
410
+ "alpha": 256.0,
411
+ "dropout": 0.0,
412
+ "A": [
413
+ 256,
414
+ 3072
415
+ ],
416
+ "B": [
417
+ 3072,
418
+ 256
419
+ ],
420
+ "master_dtype": "torch.float32"
421
+ },
422
+ {
423
+ "target": "blocks.2.ffn.net.0.proj",
424
+ "rank": 256,
425
+ "alpha": 256.0,
426
+ "dropout": 0.0,
427
+ "A": [
428
+ 256,
429
+ 3072
430
+ ],
431
+ "B": [
432
+ 14336,
433
+ 256
434
+ ],
435
+ "master_dtype": "torch.float32"
436
+ },
437
+ {
438
+ "target": "blocks.2.ffn.net.2",
439
+ "rank": 256,
440
+ "alpha": 256.0,
441
+ "dropout": 0.0,
442
+ "A": [
443
+ 256,
444
+ 14336
445
+ ],
446
+ "B": [
447
+ 3072,
448
+ 256
449
+ ],
450
+ "master_dtype": "torch.float32"
451
+ },
452
+ {
453
+ "target": "blocks.3.attn1.to_q",
454
+ "rank": 256,
455
+ "alpha": 256.0,
456
+ "dropout": 0.0,
457
+ "A": [
458
+ 256,
459
+ 3072
460
+ ],
461
+ "B": [
462
+ 3072,
463
+ 256
464
+ ],
465
+ "master_dtype": "torch.float32"
466
+ },
467
+ {
468
+ "target": "blocks.3.attn1.to_k",
469
+ "rank": 256,
470
+ "alpha": 256.0,
471
+ "dropout": 0.0,
472
+ "A": [
473
+ 256,
474
+ 3072
475
+ ],
476
+ "B": [
477
+ 3072,
478
+ 256
479
+ ],
480
+ "master_dtype": "torch.float32"
481
+ },
482
+ {
483
+ "target": "blocks.3.attn1.to_v",
484
+ "rank": 256,
485
+ "alpha": 256.0,
486
+ "dropout": 0.0,
487
+ "A": [
488
+ 256,
489
+ 3072
490
+ ],
491
+ "B": [
492
+ 3072,
493
+ 256
494
+ ],
495
+ "master_dtype": "torch.float32"
496
+ },
497
+ {
498
+ "target": "blocks.3.attn1.to_out.0",
499
+ "rank": 256,
500
+ "alpha": 256.0,
501
+ "dropout": 0.0,
502
+ "A": [
503
+ 256,
504
+ 3072
505
+ ],
506
+ "B": [
507
+ 3072,
508
+ 256
509
+ ],
510
+ "master_dtype": "torch.float32"
511
+ },
512
+ {
513
+ "target": "blocks.3.ffn.net.0.proj",
514
+ "rank": 256,
515
+ "alpha": 256.0,
516
+ "dropout": 0.0,
517
+ "A": [
518
+ 256,
519
+ 3072
520
+ ],
521
+ "B": [
522
+ 14336,
523
+ 256
524
+ ],
525
+ "master_dtype": "torch.float32"
526
+ },
527
+ {
528
+ "target": "blocks.3.ffn.net.2",
529
+ "rank": 256,
530
+ "alpha": 256.0,
531
+ "dropout": 0.0,
532
+ "A": [
533
+ 256,
534
+ 14336
535
+ ],
536
+ "B": [
537
+ 3072,
538
+ 256
539
+ ],
540
+ "master_dtype": "torch.float32"
541
+ },
542
+ {
543
+ "target": "blocks.4.attn1.to_q",
544
+ "rank": 256,
545
+ "alpha": 256.0,
546
+ "dropout": 0.0,
547
+ "A": [
548
+ 256,
549
+ 3072
550
+ ],
551
+ "B": [
552
+ 3072,
553
+ 256
554
+ ],
555
+ "master_dtype": "torch.float32"
556
+ },
557
+ {
558
+ "target": "blocks.4.attn1.to_k",
559
+ "rank": 256,
560
+ "alpha": 256.0,
561
+ "dropout": 0.0,
562
+ "A": [
563
+ 256,
564
+ 3072
565
+ ],
566
+ "B": [
567
+ 3072,
568
+ 256
569
+ ],
570
+ "master_dtype": "torch.float32"
571
+ },
572
+ {
573
+ "target": "blocks.4.attn1.to_v",
574
+ "rank": 256,
575
+ "alpha": 256.0,
576
+ "dropout": 0.0,
577
+ "A": [
578
+ 256,
579
+ 3072
580
+ ],
581
+ "B": [
582
+ 3072,
583
+ 256
584
+ ],
585
+ "master_dtype": "torch.float32"
586
+ },
587
+ {
588
+ "target": "blocks.4.attn1.to_out.0",
589
+ "rank": 256,
590
+ "alpha": 256.0,
591
+ "dropout": 0.0,
592
+ "A": [
593
+ 256,
594
+ 3072
595
+ ],
596
+ "B": [
597
+ 3072,
598
+ 256
599
+ ],
600
+ "master_dtype": "torch.float32"
601
+ },
602
+ {
603
+ "target": "blocks.4.ffn.net.0.proj",
604
+ "rank": 256,
605
+ "alpha": 256.0,
606
+ "dropout": 0.0,
607
+ "A": [
608
+ 256,
609
+ 3072
610
+ ],
611
+ "B": [
612
+ 14336,
613
+ 256
614
+ ],
615
+ "master_dtype": "torch.float32"
616
+ },
617
+ {
618
+ "target": "blocks.4.ffn.net.2",
619
+ "rank": 256,
620
+ "alpha": 256.0,
621
+ "dropout": 0.0,
622
+ "A": [
623
+ 256,
624
+ 14336
625
+ ],
626
+ "B": [
627
+ 3072,
628
+ 256
629
+ ],
630
+ "master_dtype": "torch.float32"
631
+ },
632
+ {
633
+ "target": "blocks.5.attn1.to_q",
634
+ "rank": 256,
635
+ "alpha": 256.0,
636
+ "dropout": 0.0,
637
+ "A": [
638
+ 256,
639
+ 3072
640
+ ],
641
+ "B": [
642
+ 3072,
643
+ 256
644
+ ],
645
+ "master_dtype": "torch.float32"
646
+ },
647
+ {
648
+ "target": "blocks.5.attn1.to_k",
649
+ "rank": 256,
650
+ "alpha": 256.0,
651
+ "dropout": 0.0,
652
+ "A": [
653
+ 256,
654
+ 3072
655
+ ],
656
+ "B": [
657
+ 3072,
658
+ 256
659
+ ],
660
+ "master_dtype": "torch.float32"
661
+ },
662
+ {
663
+ "target": "blocks.5.attn1.to_v",
664
+ "rank": 256,
665
+ "alpha": 256.0,
666
+ "dropout": 0.0,
667
+ "A": [
668
+ 256,
669
+ 3072
670
+ ],
671
+ "B": [
672
+ 3072,
673
+ 256
674
+ ],
675
+ "master_dtype": "torch.float32"
676
+ },
677
+ {
678
+ "target": "blocks.5.attn1.to_out.0",
679
+ "rank": 256,
680
+ "alpha": 256.0,
681
+ "dropout": 0.0,
682
+ "A": [
683
+ 256,
684
+ 3072
685
+ ],
686
+ "B": [
687
+ 3072,
688
+ 256
689
+ ],
690
+ "master_dtype": "torch.float32"
691
+ },
692
+ {
693
+ "target": "blocks.5.ffn.net.0.proj",
694
+ "rank": 256,
695
+ "alpha": 256.0,
696
+ "dropout": 0.0,
697
+ "A": [
698
+ 256,
699
+ 3072
700
+ ],
701
+ "B": [
702
+ 14336,
703
+ 256
704
+ ],
705
+ "master_dtype": "torch.float32"
706
+ },
707
+ {
708
+ "target": "blocks.5.ffn.net.2",
709
+ "rank": 256,
710
+ "alpha": 256.0,
711
+ "dropout": 0.0,
712
+ "A": [
713
+ 256,
714
+ 14336
715
+ ],
716
+ "B": [
717
+ 3072,
718
+ 256
719
+ ],
720
+ "master_dtype": "torch.float32"
721
+ },
722
+ {
723
+ "target": "blocks.6.attn1.to_q",
724
+ "rank": 256,
725
+ "alpha": 256.0,
726
+ "dropout": 0.0,
727
+ "A": [
728
+ 256,
729
+ 3072
730
+ ],
731
+ "B": [
732
+ 3072,
733
+ 256
734
+ ],
735
+ "master_dtype": "torch.float32"
736
+ },
737
+ {
738
+ "target": "blocks.6.attn1.to_k",
739
+ "rank": 256,
740
+ "alpha": 256.0,
741
+ "dropout": 0.0,
742
+ "A": [
743
+ 256,
744
+ 3072
745
+ ],
746
+ "B": [
747
+ 3072,
748
+ 256
749
+ ],
750
+ "master_dtype": "torch.float32"
751
+ },
752
+ {
753
+ "target": "blocks.6.attn1.to_v",
754
+ "rank": 256,
755
+ "alpha": 256.0,
756
+ "dropout": 0.0,
757
+ "A": [
758
+ 256,
759
+ 3072
760
+ ],
761
+ "B": [
762
+ 3072,
763
+ 256
764
+ ],
765
+ "master_dtype": "torch.float32"
766
+ },
767
+ {
768
+ "target": "blocks.6.attn1.to_out.0",
769
+ "rank": 256,
770
+ "alpha": 256.0,
771
+ "dropout": 0.0,
772
+ "A": [
773
+ 256,
774
+ 3072
775
+ ],
776
+ "B": [
777
+ 3072,
778
+ 256
779
+ ],
780
+ "master_dtype": "torch.float32"
781
+ },
782
+ {
783
+ "target": "blocks.6.ffn.net.0.proj",
784
+ "rank": 256,
785
+ "alpha": 256.0,
786
+ "dropout": 0.0,
787
+ "A": [
788
+ 256,
789
+ 3072
790
+ ],
791
+ "B": [
792
+ 14336,
793
+ 256
794
+ ],
795
+ "master_dtype": "torch.float32"
796
+ },
797
+ {
798
+ "target": "blocks.6.ffn.net.2",
799
+ "rank": 256,
800
+ "alpha": 256.0,
801
+ "dropout": 0.0,
802
+ "A": [
803
+ 256,
804
+ 14336
805
+ ],
806
+ "B": [
807
+ 3072,
808
+ 256
809
+ ],
810
+ "master_dtype": "torch.float32"
811
+ },
812
+ {
813
+ "target": "blocks.7.attn1.to_q",
814
+ "rank": 256,
815
+ "alpha": 256.0,
816
+ "dropout": 0.0,
817
+ "A": [
818
+ 256,
819
+ 3072
820
+ ],
821
+ "B": [
822
+ 3072,
823
+ 256
824
+ ],
825
+ "master_dtype": "torch.float32"
826
+ },
827
+ {
828
+ "target": "blocks.7.attn1.to_k",
829
+ "rank": 256,
830
+ "alpha": 256.0,
831
+ "dropout": 0.0,
832
+ "A": [
833
+ 256,
834
+ 3072
835
+ ],
836
+ "B": [
837
+ 3072,
838
+ 256
839
+ ],
840
+ "master_dtype": "torch.float32"
841
+ },
842
+ {
843
+ "target": "blocks.7.attn1.to_v",
844
+ "rank": 256,
845
+ "alpha": 256.0,
846
+ "dropout": 0.0,
847
+ "A": [
848
+ 256,
849
+ 3072
850
+ ],
851
+ "B": [
852
+ 3072,
853
+ 256
854
+ ],
855
+ "master_dtype": "torch.float32"
856
+ },
857
+ {
858
+ "target": "blocks.7.attn1.to_out.0",
859
+ "rank": 256,
860
+ "alpha": 256.0,
861
+ "dropout": 0.0,
862
+ "A": [
863
+ 256,
864
+ 3072
865
+ ],
866
+ "B": [
867
+ 3072,
868
+ 256
869
+ ],
870
+ "master_dtype": "torch.float32"
871
+ },
872
+ {
873
+ "target": "blocks.7.ffn.net.0.proj",
874
+ "rank": 256,
875
+ "alpha": 256.0,
876
+ "dropout": 0.0,
877
+ "A": [
878
+ 256,
879
+ 3072
880
+ ],
881
+ "B": [
882
+ 14336,
883
+ 256
884
+ ],
885
+ "master_dtype": "torch.float32"
886
+ },
887
+ {
888
+ "target": "blocks.7.ffn.net.2",
889
+ "rank": 256,
890
+ "alpha": 256.0,
891
+ "dropout": 0.0,
892
+ "A": [
893
+ 256,
894
+ 14336
895
+ ],
896
+ "B": [
897
+ 3072,
898
+ 256
899
+ ],
900
+ "master_dtype": "torch.float32"
901
+ },
902
+ {
903
+ "target": "blocks.8.attn1.to_q",
904
+ "rank": 256,
905
+ "alpha": 256.0,
906
+ "dropout": 0.0,
907
+ "A": [
908
+ 256,
909
+ 3072
910
+ ],
911
+ "B": [
912
+ 3072,
913
+ 256
914
+ ],
915
+ "master_dtype": "torch.float32"
916
+ },
917
+ {
918
+ "target": "blocks.8.attn1.to_k",
919
+ "rank": 256,
920
+ "alpha": 256.0,
921
+ "dropout": 0.0,
922
+ "A": [
923
+ 256,
924
+ 3072
925
+ ],
926
+ "B": [
927
+ 3072,
928
+ 256
929
+ ],
930
+ "master_dtype": "torch.float32"
931
+ },
932
+ {
933
+ "target": "blocks.8.attn1.to_v",
934
+ "rank": 256,
935
+ "alpha": 256.0,
936
+ "dropout": 0.0,
937
+ "A": [
938
+ 256,
939
+ 3072
940
+ ],
941
+ "B": [
942
+ 3072,
943
+ 256
944
+ ],
945
+ "master_dtype": "torch.float32"
946
+ },
947
+ {
948
+ "target": "blocks.8.attn1.to_out.0",
949
+ "rank": 256,
950
+ "alpha": 256.0,
951
+ "dropout": 0.0,
952
+ "A": [
953
+ 256,
954
+ 3072
955
+ ],
956
+ "B": [
957
+ 3072,
958
+ 256
959
+ ],
960
+ "master_dtype": "torch.float32"
961
+ },
962
+ {
963
+ "target": "blocks.8.ffn.net.0.proj",
964
+ "rank": 256,
965
+ "alpha": 256.0,
966
+ "dropout": 0.0,
967
+ "A": [
968
+ 256,
969
+ 3072
970
+ ],
971
+ "B": [
972
+ 14336,
973
+ 256
974
+ ],
975
+ "master_dtype": "torch.float32"
976
+ },
977
+ {
978
+ "target": "blocks.8.ffn.net.2",
979
+ "rank": 256,
980
+ "alpha": 256.0,
981
+ "dropout": 0.0,
982
+ "A": [
983
+ 256,
984
+ 14336
985
+ ],
986
+ "B": [
987
+ 3072,
988
+ 256
989
+ ],
990
+ "master_dtype": "torch.float32"
991
+ },
992
+ {
993
+ "target": "blocks.9.attn1.to_q",
994
+ "rank": 256,
995
+ "alpha": 256.0,
996
+ "dropout": 0.0,
997
+ "A": [
998
+ 256,
999
+ 3072
1000
+ ],
1001
+ "B": [
1002
+ 3072,
1003
+ 256
1004
+ ],
1005
+ "master_dtype": "torch.float32"
1006
+ },
1007
+ {
1008
+ "target": "blocks.9.attn1.to_k",
1009
+ "rank": 256,
1010
+ "alpha": 256.0,
1011
+ "dropout": 0.0,
1012
+ "A": [
1013
+ 256,
1014
+ 3072
1015
+ ],
1016
+ "B": [
1017
+ 3072,
1018
+ 256
1019
+ ],
1020
+ "master_dtype": "torch.float32"
1021
+ },
1022
+ {
1023
+ "target": "blocks.9.attn1.to_v",
1024
+ "rank": 256,
1025
+ "alpha": 256.0,
1026
+ "dropout": 0.0,
1027
+ "A": [
1028
+ 256,
1029
+ 3072
1030
+ ],
1031
+ "B": [
1032
+ 3072,
1033
+ 256
1034
+ ],
1035
+ "master_dtype": "torch.float32"
1036
+ },
1037
+ {
1038
+ "target": "blocks.9.attn1.to_out.0",
1039
+ "rank": 256,
1040
+ "alpha": 256.0,
1041
+ "dropout": 0.0,
1042
+ "A": [
1043
+ 256,
1044
+ 3072
1045
+ ],
1046
+ "B": [
1047
+ 3072,
1048
+ 256
1049
+ ],
1050
+ "master_dtype": "torch.float32"
1051
+ },
1052
+ {
1053
+ "target": "blocks.9.ffn.net.0.proj",
1054
+ "rank": 256,
1055
+ "alpha": 256.0,
1056
+ "dropout": 0.0,
1057
+ "A": [
1058
+ 256,
1059
+ 3072
1060
+ ],
1061
+ "B": [
1062
+ 14336,
1063
+ 256
1064
+ ],
1065
+ "master_dtype": "torch.float32"
1066
+ },
1067
+ {
1068
+ "target": "blocks.9.ffn.net.2",
1069
+ "rank": 256,
1070
+ "alpha": 256.0,
1071
+ "dropout": 0.0,
1072
+ "A": [
1073
+ 256,
1074
+ 14336
1075
+ ],
1076
+ "B": [
1077
+ 3072,
1078
+ 256
1079
+ ],
1080
+ "master_dtype": "torch.float32"
1081
+ },
1082
+ {
1083
+ "target": "blocks.10.attn1.to_q",
1084
+ "rank": 256,
1085
+ "alpha": 256.0,
1086
+ "dropout": 0.0,
1087
+ "A": [
1088
+ 256,
1089
+ 3072
1090
+ ],
1091
+ "B": [
1092
+ 3072,
1093
+ 256
1094
+ ],
1095
+ "master_dtype": "torch.float32"
1096
+ },
1097
+ {
1098
+ "target": "blocks.10.attn1.to_k",
1099
+ "rank": 256,
1100
+ "alpha": 256.0,
1101
+ "dropout": 0.0,
1102
+ "A": [
1103
+ 256,
1104
+ 3072
1105
+ ],
1106
+ "B": [
1107
+ 3072,
1108
+ 256
1109
+ ],
1110
+ "master_dtype": "torch.float32"
1111
+ },
1112
+ {
1113
+ "target": "blocks.10.attn1.to_v",
1114
+ "rank": 256,
1115
+ "alpha": 256.0,
1116
+ "dropout": 0.0,
1117
+ "A": [
1118
+ 256,
1119
+ 3072
1120
+ ],
1121
+ "B": [
1122
+ 3072,
1123
+ 256
1124
+ ],
1125
+ "master_dtype": "torch.float32"
1126
+ },
1127
+ {
1128
+ "target": "blocks.10.attn1.to_out.0",
1129
+ "rank": 256,
1130
+ "alpha": 256.0,
1131
+ "dropout": 0.0,
1132
+ "A": [
1133
+ 256,
1134
+ 3072
1135
+ ],
1136
+ "B": [
1137
+ 3072,
1138
+ 256
1139
+ ],
1140
+ "master_dtype": "torch.float32"
1141
+ },
1142
+ {
1143
+ "target": "blocks.10.ffn.net.0.proj",
1144
+ "rank": 256,
1145
+ "alpha": 256.0,
1146
+ "dropout": 0.0,
1147
+ "A": [
1148
+ 256,
1149
+ 3072
1150
+ ],
1151
+ "B": [
1152
+ 14336,
1153
+ 256
1154
+ ],
1155
+ "master_dtype": "torch.float32"
1156
+ },
1157
+ {
1158
+ "target": "blocks.10.ffn.net.2",
1159
+ "rank": 256,
1160
+ "alpha": 256.0,
1161
+ "dropout": 0.0,
1162
+ "A": [
1163
+ 256,
1164
+ 14336
1165
+ ],
1166
+ "B": [
1167
+ 3072,
1168
+ 256
1169
+ ],
1170
+ "master_dtype": "torch.float32"
1171
+ },
1172
+ {
1173
+ "target": "blocks.11.attn1.to_q",
1174
+ "rank": 256,
1175
+ "alpha": 256.0,
1176
+ "dropout": 0.0,
1177
+ "A": [
1178
+ 256,
1179
+ 3072
1180
+ ],
1181
+ "B": [
1182
+ 3072,
1183
+ 256
1184
+ ],
1185
+ "master_dtype": "torch.float32"
1186
+ },
1187
+ {
1188
+ "target": "blocks.11.attn1.to_k",
1189
+ "rank": 256,
1190
+ "alpha": 256.0,
1191
+ "dropout": 0.0,
1192
+ "A": [
1193
+ 256,
1194
+ 3072
1195
+ ],
1196
+ "B": [
1197
+ 3072,
1198
+ 256
1199
+ ],
1200
+ "master_dtype": "torch.float32"
1201
+ },
1202
+ {
1203
+ "target": "blocks.11.attn1.to_v",
1204
+ "rank": 256,
1205
+ "alpha": 256.0,
1206
+ "dropout": 0.0,
1207
+ "A": [
1208
+ 256,
1209
+ 3072
1210
+ ],
1211
+ "B": [
1212
+ 3072,
1213
+ 256
1214
+ ],
1215
+ "master_dtype": "torch.float32"
1216
+ },
1217
+ {
1218
+ "target": "blocks.11.attn1.to_out.0",
1219
+ "rank": 256,
1220
+ "alpha": 256.0,
1221
+ "dropout": 0.0,
1222
+ "A": [
1223
+ 256,
1224
+ 3072
1225
+ ],
1226
+ "B": [
1227
+ 3072,
1228
+ 256
1229
+ ],
1230
+ "master_dtype": "torch.float32"
1231
+ },
1232
+ {
1233
+ "target": "blocks.11.ffn.net.0.proj",
1234
+ "rank": 256,
1235
+ "alpha": 256.0,
1236
+ "dropout": 0.0,
1237
+ "A": [
1238
+ 256,
1239
+ 3072
1240
+ ],
1241
+ "B": [
1242
+ 14336,
1243
+ 256
1244
+ ],
1245
+ "master_dtype": "torch.float32"
1246
+ },
1247
+ {
1248
+ "target": "blocks.11.ffn.net.2",
1249
+ "rank": 256,
1250
+ "alpha": 256.0,
1251
+ "dropout": 0.0,
1252
+ "A": [
1253
+ 256,
1254
+ 14336
1255
+ ],
1256
+ "B": [
1257
+ 3072,
1258
+ 256
1259
+ ],
1260
+ "master_dtype": "torch.float32"
1261
+ },
1262
+ {
1263
+ "target": "blocks.12.attn1.to_q",
1264
+ "rank": 256,
1265
+ "alpha": 256.0,
1266
+ "dropout": 0.0,
1267
+ "A": [
1268
+ 256,
1269
+ 3072
1270
+ ],
1271
+ "B": [
1272
+ 3072,
1273
+ 256
1274
+ ],
1275
+ "master_dtype": "torch.float32"
1276
+ },
1277
+ {
1278
+ "target": "blocks.12.attn1.to_k",
1279
+ "rank": 256,
1280
+ "alpha": 256.0,
1281
+ "dropout": 0.0,
1282
+ "A": [
1283
+ 256,
1284
+ 3072
1285
+ ],
1286
+ "B": [
1287
+ 3072,
1288
+ 256
1289
+ ],
1290
+ "master_dtype": "torch.float32"
1291
+ },
1292
+ {
1293
+ "target": "blocks.12.attn1.to_v",
1294
+ "rank": 256,
1295
+ "alpha": 256.0,
1296
+ "dropout": 0.0,
1297
+ "A": [
1298
+ 256,
1299
+ 3072
1300
+ ],
1301
+ "B": [
1302
+ 3072,
1303
+ 256
1304
+ ],
1305
+ "master_dtype": "torch.float32"
1306
+ },
1307
+ {
1308
+ "target": "blocks.12.attn1.to_out.0",
1309
+ "rank": 256,
1310
+ "alpha": 256.0,
1311
+ "dropout": 0.0,
1312
+ "A": [
1313
+ 256,
1314
+ 3072
1315
+ ],
1316
+ "B": [
1317
+ 3072,
1318
+ 256
1319
+ ],
1320
+ "master_dtype": "torch.float32"
1321
+ },
1322
+ {
1323
+ "target": "blocks.12.ffn.net.0.proj",
1324
+ "rank": 256,
1325
+ "alpha": 256.0,
1326
+ "dropout": 0.0,
1327
+ "A": [
1328
+ 256,
1329
+ 3072
1330
+ ],
1331
+ "B": [
1332
+ 14336,
1333
+ 256
1334
+ ],
1335
+ "master_dtype": "torch.float32"
1336
+ },
1337
+ {
1338
+ "target": "blocks.12.ffn.net.2",
1339
+ "rank": 256,
1340
+ "alpha": 256.0,
1341
+ "dropout": 0.0,
1342
+ "A": [
1343
+ 256,
1344
+ 14336
1345
+ ],
1346
+ "B": [
1347
+ 3072,
1348
+ 256
1349
+ ],
1350
+ "master_dtype": "torch.float32"
1351
+ },
1352
+ {
1353
+ "target": "blocks.13.attn1.to_q",
1354
+ "rank": 256,
1355
+ "alpha": 256.0,
1356
+ "dropout": 0.0,
1357
+ "A": [
1358
+ 256,
1359
+ 3072
1360
+ ],
1361
+ "B": [
1362
+ 3072,
1363
+ 256
1364
+ ],
1365
+ "master_dtype": "torch.float32"
1366
+ },
1367
+ {
1368
+ "target": "blocks.13.attn1.to_k",
1369
+ "rank": 256,
1370
+ "alpha": 256.0,
1371
+ "dropout": 0.0,
1372
+ "A": [
1373
+ 256,
1374
+ 3072
1375
+ ],
1376
+ "B": [
1377
+ 3072,
1378
+ 256
1379
+ ],
1380
+ "master_dtype": "torch.float32"
1381
+ },
1382
+ {
1383
+ "target": "blocks.13.attn1.to_v",
1384
+ "rank": 256,
1385
+ "alpha": 256.0,
1386
+ "dropout": 0.0,
1387
+ "A": [
1388
+ 256,
1389
+ 3072
1390
+ ],
1391
+ "B": [
1392
+ 3072,
1393
+ 256
1394
+ ],
1395
+ "master_dtype": "torch.float32"
1396
+ },
1397
+ {
1398
+ "target": "blocks.13.attn1.to_out.0",
1399
+ "rank": 256,
1400
+ "alpha": 256.0,
1401
+ "dropout": 0.0,
1402
+ "A": [
1403
+ 256,
1404
+ 3072
1405
+ ],
1406
+ "B": [
1407
+ 3072,
1408
+ 256
1409
+ ],
1410
+ "master_dtype": "torch.float32"
1411
+ },
1412
+ {
1413
+ "target": "blocks.13.ffn.net.0.proj",
1414
+ "rank": 256,
1415
+ "alpha": 256.0,
1416
+ "dropout": 0.0,
1417
+ "A": [
1418
+ 256,
1419
+ 3072
1420
+ ],
1421
+ "B": [
1422
+ 14336,
1423
+ 256
1424
+ ],
1425
+ "master_dtype": "torch.float32"
1426
+ },
1427
+ {
1428
+ "target": "blocks.13.ffn.net.2",
1429
+ "rank": 256,
1430
+ "alpha": 256.0,
1431
+ "dropout": 0.0,
1432
+ "A": [
1433
+ 256,
1434
+ 14336
1435
+ ],
1436
+ "B": [
1437
+ 3072,
1438
+ 256
1439
+ ],
1440
+ "master_dtype": "torch.float32"
1441
+ },
1442
+ {
1443
+ "target": "blocks.14.attn1.to_q",
1444
+ "rank": 256,
1445
+ "alpha": 256.0,
1446
+ "dropout": 0.0,
1447
+ "A": [
1448
+ 256,
1449
+ 3072
1450
+ ],
1451
+ "B": [
1452
+ 3072,
1453
+ 256
1454
+ ],
1455
+ "master_dtype": "torch.float32"
1456
+ },
1457
+ {
1458
+ "target": "blocks.14.attn1.to_k",
1459
+ "rank": 256,
1460
+ "alpha": 256.0,
1461
+ "dropout": 0.0,
1462
+ "A": [
1463
+ 256,
1464
+ 3072
1465
+ ],
1466
+ "B": [
1467
+ 3072,
1468
+ 256
1469
+ ],
1470
+ "master_dtype": "torch.float32"
1471
+ },
1472
+ {
1473
+ "target": "blocks.14.attn1.to_v",
1474
+ "rank": 256,
1475
+ "alpha": 256.0,
1476
+ "dropout": 0.0,
1477
+ "A": [
1478
+ 256,
1479
+ 3072
1480
+ ],
1481
+ "B": [
1482
+ 3072,
1483
+ 256
1484
+ ],
1485
+ "master_dtype": "torch.float32"
1486
+ },
1487
+ {
1488
+ "target": "blocks.14.attn1.to_out.0",
1489
+ "rank": 256,
1490
+ "alpha": 256.0,
1491
+ "dropout": 0.0,
1492
+ "A": [
1493
+ 256,
1494
+ 3072
1495
+ ],
1496
+ "B": [
1497
+ 3072,
1498
+ 256
1499
+ ],
1500
+ "master_dtype": "torch.float32"
1501
+ },
1502
+ {
1503
+ "target": "blocks.14.ffn.net.0.proj",
1504
+ "rank": 256,
1505
+ "alpha": 256.0,
1506
+ "dropout": 0.0,
1507
+ "A": [
1508
+ 256,
1509
+ 3072
1510
+ ],
1511
+ "B": [
1512
+ 14336,
1513
+ 256
1514
+ ],
1515
+ "master_dtype": "torch.float32"
1516
+ },
1517
+ {
1518
+ "target": "blocks.14.ffn.net.2",
1519
+ "rank": 256,
1520
+ "alpha": 256.0,
1521
+ "dropout": 0.0,
1522
+ "A": [
1523
+ 256,
1524
+ 14336
1525
+ ],
1526
+ "B": [
1527
+ 3072,
1528
+ 256
1529
+ ],
1530
+ "master_dtype": "torch.float32"
1531
+ },
1532
+ {
1533
+ "target": "blocks.15.attn1.to_q",
1534
+ "rank": 256,
1535
+ "alpha": 256.0,
1536
+ "dropout": 0.0,
1537
+ "A": [
1538
+ 256,
1539
+ 3072
1540
+ ],
1541
+ "B": [
1542
+ 3072,
1543
+ 256
1544
+ ],
1545
+ "master_dtype": "torch.float32"
1546
+ },
1547
+ {
1548
+ "target": "blocks.15.attn1.to_k",
1549
+ "rank": 256,
1550
+ "alpha": 256.0,
1551
+ "dropout": 0.0,
1552
+ "A": [
1553
+ 256,
1554
+ 3072
1555
+ ],
1556
+ "B": [
1557
+ 3072,
1558
+ 256
1559
+ ],
1560
+ "master_dtype": "torch.float32"
1561
+ },
1562
+ {
1563
+ "target": "blocks.15.attn1.to_v",
1564
+ "rank": 256,
1565
+ "alpha": 256.0,
1566
+ "dropout": 0.0,
1567
+ "A": [
1568
+ 256,
1569
+ 3072
1570
+ ],
1571
+ "B": [
1572
+ 3072,
1573
+ 256
1574
+ ],
1575
+ "master_dtype": "torch.float32"
1576
+ },
1577
+ {
1578
+ "target": "blocks.15.attn1.to_out.0",
1579
+ "rank": 256,
1580
+ "alpha": 256.0,
1581
+ "dropout": 0.0,
1582
+ "A": [
1583
+ 256,
1584
+ 3072
1585
+ ],
1586
+ "B": [
1587
+ 3072,
1588
+ 256
1589
+ ],
1590
+ "master_dtype": "torch.float32"
1591
+ },
1592
+ {
1593
+ "target": "blocks.15.ffn.net.0.proj",
1594
+ "rank": 256,
1595
+ "alpha": 256.0,
1596
+ "dropout": 0.0,
1597
+ "A": [
1598
+ 256,
1599
+ 3072
1600
+ ],
1601
+ "B": [
1602
+ 14336,
1603
+ 256
1604
+ ],
1605
+ "master_dtype": "torch.float32"
1606
+ },
1607
+ {
1608
+ "target": "blocks.15.ffn.net.2",
1609
+ "rank": 256,
1610
+ "alpha": 256.0,
1611
+ "dropout": 0.0,
1612
+ "A": [
1613
+ 256,
1614
+ 14336
1615
+ ],
1616
+ "B": [
1617
+ 3072,
1618
+ 256
1619
+ ],
1620
+ "master_dtype": "torch.float32"
1621
+ },
1622
+ {
1623
+ "target": "blocks.16.attn1.to_q",
1624
+ "rank": 256,
1625
+ "alpha": 256.0,
1626
+ "dropout": 0.0,
1627
+ "A": [
1628
+ 256,
1629
+ 3072
1630
+ ],
1631
+ "B": [
1632
+ 3072,
1633
+ 256
1634
+ ],
1635
+ "master_dtype": "torch.float32"
1636
+ },
1637
+ {
1638
+ "target": "blocks.16.attn1.to_k",
1639
+ "rank": 256,
1640
+ "alpha": 256.0,
1641
+ "dropout": 0.0,
1642
+ "A": [
1643
+ 256,
1644
+ 3072
1645
+ ],
1646
+ "B": [
1647
+ 3072,
1648
+ 256
1649
+ ],
1650
+ "master_dtype": "torch.float32"
1651
+ },
1652
+ {
1653
+ "target": "blocks.16.attn1.to_v",
1654
+ "rank": 256,
1655
+ "alpha": 256.0,
1656
+ "dropout": 0.0,
1657
+ "A": [
1658
+ 256,
1659
+ 3072
1660
+ ],
1661
+ "B": [
1662
+ 3072,
1663
+ 256
1664
+ ],
1665
+ "master_dtype": "torch.float32"
1666
+ },
1667
+ {
1668
+ "target": "blocks.16.attn1.to_out.0",
1669
+ "rank": 256,
1670
+ "alpha": 256.0,
1671
+ "dropout": 0.0,
1672
+ "A": [
1673
+ 256,
1674
+ 3072
1675
+ ],
1676
+ "B": [
1677
+ 3072,
1678
+ 256
1679
+ ],
1680
+ "master_dtype": "torch.float32"
1681
+ },
1682
+ {
1683
+ "target": "blocks.16.ffn.net.0.proj",
1684
+ "rank": 256,
1685
+ "alpha": 256.0,
1686
+ "dropout": 0.0,
1687
+ "A": [
1688
+ 256,
1689
+ 3072
1690
+ ],
1691
+ "B": [
1692
+ 14336,
1693
+ 256
1694
+ ],
1695
+ "master_dtype": "torch.float32"
1696
+ },
1697
+ {
1698
+ "target": "blocks.16.ffn.net.2",
1699
+ "rank": 256,
1700
+ "alpha": 256.0,
1701
+ "dropout": 0.0,
1702
+ "A": [
1703
+ 256,
1704
+ 14336
1705
+ ],
1706
+ "B": [
1707
+ 3072,
1708
+ 256
1709
+ ],
1710
+ "master_dtype": "torch.float32"
1711
+ },
1712
+ {
1713
+ "target": "blocks.17.attn1.to_q",
1714
+ "rank": 256,
1715
+ "alpha": 256.0,
1716
+ "dropout": 0.0,
1717
+ "A": [
1718
+ 256,
1719
+ 3072
1720
+ ],
1721
+ "B": [
1722
+ 3072,
1723
+ 256
1724
+ ],
1725
+ "master_dtype": "torch.float32"
1726
+ },
1727
+ {
1728
+ "target": "blocks.17.attn1.to_k",
1729
+ "rank": 256,
1730
+ "alpha": 256.0,
1731
+ "dropout": 0.0,
1732
+ "A": [
1733
+ 256,
1734
+ 3072
1735
+ ],
1736
+ "B": [
1737
+ 3072,
1738
+ 256
1739
+ ],
1740
+ "master_dtype": "torch.float32"
1741
+ },
1742
+ {
1743
+ "target": "blocks.17.attn1.to_v",
1744
+ "rank": 256,
1745
+ "alpha": 256.0,
1746
+ "dropout": 0.0,
1747
+ "A": [
1748
+ 256,
1749
+ 3072
1750
+ ],
1751
+ "B": [
1752
+ 3072,
1753
+ 256
1754
+ ],
1755
+ "master_dtype": "torch.float32"
1756
+ },
1757
+ {
1758
+ "target": "blocks.17.attn1.to_out.0",
1759
+ "rank": 256,
1760
+ "alpha": 256.0,
1761
+ "dropout": 0.0,
1762
+ "A": [
1763
+ 256,
1764
+ 3072
1765
+ ],
1766
+ "B": [
1767
+ 3072,
1768
+ 256
1769
+ ],
1770
+ "master_dtype": "torch.float32"
1771
+ },
1772
+ {
1773
+ "target": "blocks.17.ffn.net.0.proj",
1774
+ "rank": 256,
1775
+ "alpha": 256.0,
1776
+ "dropout": 0.0,
1777
+ "A": [
1778
+ 256,
1779
+ 3072
1780
+ ],
1781
+ "B": [
1782
+ 14336,
1783
+ 256
1784
+ ],
1785
+ "master_dtype": "torch.float32"
1786
+ },
1787
+ {
1788
+ "target": "blocks.17.ffn.net.2",
1789
+ "rank": 256,
1790
+ "alpha": 256.0,
1791
+ "dropout": 0.0,
1792
+ "A": [
1793
+ 256,
1794
+ 14336
1795
+ ],
1796
+ "B": [
1797
+ 3072,
1798
+ 256
1799
+ ],
1800
+ "master_dtype": "torch.float32"
1801
+ },
1802
+ {
1803
+ "target": "blocks.18.attn1.to_q",
1804
+ "rank": 256,
1805
+ "alpha": 256.0,
1806
+ "dropout": 0.0,
1807
+ "A": [
1808
+ 256,
1809
+ 3072
1810
+ ],
1811
+ "B": [
1812
+ 3072,
1813
+ 256
1814
+ ],
1815
+ "master_dtype": "torch.float32"
1816
+ },
1817
+ {
1818
+ "target": "blocks.18.attn1.to_k",
1819
+ "rank": 256,
1820
+ "alpha": 256.0,
1821
+ "dropout": 0.0,
1822
+ "A": [
1823
+ 256,
1824
+ 3072
1825
+ ],
1826
+ "B": [
1827
+ 3072,
1828
+ 256
1829
+ ],
1830
+ "master_dtype": "torch.float32"
1831
+ },
1832
+ {
1833
+ "target": "blocks.18.attn1.to_v",
1834
+ "rank": 256,
1835
+ "alpha": 256.0,
1836
+ "dropout": 0.0,
1837
+ "A": [
1838
+ 256,
1839
+ 3072
1840
+ ],
1841
+ "B": [
1842
+ 3072,
1843
+ 256
1844
+ ],
1845
+ "master_dtype": "torch.float32"
1846
+ },
1847
+ {
1848
+ "target": "blocks.18.attn1.to_out.0",
1849
+ "rank": 256,
1850
+ "alpha": 256.0,
1851
+ "dropout": 0.0,
1852
+ "A": [
1853
+ 256,
1854
+ 3072
1855
+ ],
1856
+ "B": [
1857
+ 3072,
1858
+ 256
1859
+ ],
1860
+ "master_dtype": "torch.float32"
1861
+ },
1862
+ {
1863
+ "target": "blocks.18.ffn.net.0.proj",
1864
+ "rank": 256,
1865
+ "alpha": 256.0,
1866
+ "dropout": 0.0,
1867
+ "A": [
1868
+ 256,
1869
+ 3072
1870
+ ],
1871
+ "B": [
1872
+ 14336,
1873
+ 256
1874
+ ],
1875
+ "master_dtype": "torch.float32"
1876
+ },
1877
+ {
1878
+ "target": "blocks.18.ffn.net.2",
1879
+ "rank": 256,
1880
+ "alpha": 256.0,
1881
+ "dropout": 0.0,
1882
+ "A": [
1883
+ 256,
1884
+ 14336
1885
+ ],
1886
+ "B": [
1887
+ 3072,
1888
+ 256
1889
+ ],
1890
+ "master_dtype": "torch.float32"
1891
+ },
1892
+ {
1893
+ "target": "blocks.19.attn1.to_q",
1894
+ "rank": 256,
1895
+ "alpha": 256.0,
1896
+ "dropout": 0.0,
1897
+ "A": [
1898
+ 256,
1899
+ 3072
1900
+ ],
1901
+ "B": [
1902
+ 3072,
1903
+ 256
1904
+ ],
1905
+ "master_dtype": "torch.float32"
1906
+ },
1907
+ {
1908
+ "target": "blocks.19.attn1.to_k",
1909
+ "rank": 256,
1910
+ "alpha": 256.0,
1911
+ "dropout": 0.0,
1912
+ "A": [
1913
+ 256,
1914
+ 3072
1915
+ ],
1916
+ "B": [
1917
+ 3072,
1918
+ 256
1919
+ ],
1920
+ "master_dtype": "torch.float32"
1921
+ },
1922
+ {
1923
+ "target": "blocks.19.attn1.to_v",
1924
+ "rank": 256,
1925
+ "alpha": 256.0,
1926
+ "dropout": 0.0,
1927
+ "A": [
1928
+ 256,
1929
+ 3072
1930
+ ],
1931
+ "B": [
1932
+ 3072,
1933
+ 256
1934
+ ],
1935
+ "master_dtype": "torch.float32"
1936
+ },
1937
+ {
1938
+ "target": "blocks.19.attn1.to_out.0",
1939
+ "rank": 256,
1940
+ "alpha": 256.0,
1941
+ "dropout": 0.0,
1942
+ "A": [
1943
+ 256,
1944
+ 3072
1945
+ ],
1946
+ "B": [
1947
+ 3072,
1948
+ 256
1949
+ ],
1950
+ "master_dtype": "torch.float32"
1951
+ },
1952
+ {
1953
+ "target": "blocks.19.ffn.net.0.proj",
1954
+ "rank": 256,
1955
+ "alpha": 256.0,
1956
+ "dropout": 0.0,
1957
+ "A": [
1958
+ 256,
1959
+ 3072
1960
+ ],
1961
+ "B": [
1962
+ 14336,
1963
+ 256
1964
+ ],
1965
+ "master_dtype": "torch.float32"
1966
+ },
1967
+ {
1968
+ "target": "blocks.19.ffn.net.2",
1969
+ "rank": 256,
1970
+ "alpha": 256.0,
1971
+ "dropout": 0.0,
1972
+ "A": [
1973
+ 256,
1974
+ 14336
1975
+ ],
1976
+ "B": [
1977
+ 3072,
1978
+ 256
1979
+ ],
1980
+ "master_dtype": "torch.float32"
1981
+ },
1982
+ {
1983
+ "target": "blocks.20.attn1.to_q",
1984
+ "rank": 256,
1985
+ "alpha": 256.0,
1986
+ "dropout": 0.0,
1987
+ "A": [
1988
+ 256,
1989
+ 3072
1990
+ ],
1991
+ "B": [
1992
+ 3072,
1993
+ 256
1994
+ ],
1995
+ "master_dtype": "torch.float32"
1996
+ },
1997
+ {
1998
+ "target": "blocks.20.attn1.to_k",
1999
+ "rank": 256,
2000
+ "alpha": 256.0,
2001
+ "dropout": 0.0,
2002
+ "A": [
2003
+ 256,
2004
+ 3072
2005
+ ],
2006
+ "B": [
2007
+ 3072,
2008
+ 256
2009
+ ],
2010
+ "master_dtype": "torch.float32"
2011
+ },
2012
+ {
2013
+ "target": "blocks.20.attn1.to_v",
2014
+ "rank": 256,
2015
+ "alpha": 256.0,
2016
+ "dropout": 0.0,
2017
+ "A": [
2018
+ 256,
2019
+ 3072
2020
+ ],
2021
+ "B": [
2022
+ 3072,
2023
+ 256
2024
+ ],
2025
+ "master_dtype": "torch.float32"
2026
+ },
2027
+ {
2028
+ "target": "blocks.20.attn1.to_out.0",
2029
+ "rank": 256,
2030
+ "alpha": 256.0,
2031
+ "dropout": 0.0,
2032
+ "A": [
2033
+ 256,
2034
+ 3072
2035
+ ],
2036
+ "B": [
2037
+ 3072,
2038
+ 256
2039
+ ],
2040
+ "master_dtype": "torch.float32"
2041
+ },
2042
+ {
2043
+ "target": "blocks.20.ffn.net.0.proj",
2044
+ "rank": 256,
2045
+ "alpha": 256.0,
2046
+ "dropout": 0.0,
2047
+ "A": [
2048
+ 256,
2049
+ 3072
2050
+ ],
2051
+ "B": [
2052
+ 14336,
2053
+ 256
2054
+ ],
2055
+ "master_dtype": "torch.float32"
2056
+ },
2057
+ {
2058
+ "target": "blocks.20.ffn.net.2",
2059
+ "rank": 256,
2060
+ "alpha": 256.0,
2061
+ "dropout": 0.0,
2062
+ "A": [
2063
+ 256,
2064
+ 14336
2065
+ ],
2066
+ "B": [
2067
+ 3072,
2068
+ 256
2069
+ ],
2070
+ "master_dtype": "torch.float32"
2071
+ },
2072
+ {
2073
+ "target": "blocks.21.attn1.to_q",
2074
+ "rank": 256,
2075
+ "alpha": 256.0,
2076
+ "dropout": 0.0,
2077
+ "A": [
2078
+ 256,
2079
+ 3072
2080
+ ],
2081
+ "B": [
2082
+ 3072,
2083
+ 256
2084
+ ],
2085
+ "master_dtype": "torch.float32"
2086
+ },
2087
+ {
2088
+ "target": "blocks.21.attn1.to_k",
2089
+ "rank": 256,
2090
+ "alpha": 256.0,
2091
+ "dropout": 0.0,
2092
+ "A": [
2093
+ 256,
2094
+ 3072
2095
+ ],
2096
+ "B": [
2097
+ 3072,
2098
+ 256
2099
+ ],
2100
+ "master_dtype": "torch.float32"
2101
+ },
2102
+ {
2103
+ "target": "blocks.21.attn1.to_v",
2104
+ "rank": 256,
2105
+ "alpha": 256.0,
2106
+ "dropout": 0.0,
2107
+ "A": [
2108
+ 256,
2109
+ 3072
2110
+ ],
2111
+ "B": [
2112
+ 3072,
2113
+ 256
2114
+ ],
2115
+ "master_dtype": "torch.float32"
2116
+ },
2117
+ {
2118
+ "target": "blocks.21.attn1.to_out.0",
2119
+ "rank": 256,
2120
+ "alpha": 256.0,
2121
+ "dropout": 0.0,
2122
+ "A": [
2123
+ 256,
2124
+ 3072
2125
+ ],
2126
+ "B": [
2127
+ 3072,
2128
+ 256
2129
+ ],
2130
+ "master_dtype": "torch.float32"
2131
+ },
2132
+ {
2133
+ "target": "blocks.21.ffn.net.0.proj",
2134
+ "rank": 256,
2135
+ "alpha": 256.0,
2136
+ "dropout": 0.0,
2137
+ "A": [
2138
+ 256,
2139
+ 3072
2140
+ ],
2141
+ "B": [
2142
+ 14336,
2143
+ 256
2144
+ ],
2145
+ "master_dtype": "torch.float32"
2146
+ },
2147
+ {
2148
+ "target": "blocks.21.ffn.net.2",
2149
+ "rank": 256,
2150
+ "alpha": 256.0,
2151
+ "dropout": 0.0,
2152
+ "A": [
2153
+ 256,
2154
+ 14336
2155
+ ],
2156
+ "B": [
2157
+ 3072,
2158
+ 256
2159
+ ],
2160
+ "master_dtype": "torch.float32"
2161
+ },
2162
+ {
2163
+ "target": "blocks.22.attn1.to_q",
2164
+ "rank": 256,
2165
+ "alpha": 256.0,
2166
+ "dropout": 0.0,
2167
+ "A": [
2168
+ 256,
2169
+ 3072
2170
+ ],
2171
+ "B": [
2172
+ 3072,
2173
+ 256
2174
+ ],
2175
+ "master_dtype": "torch.float32"
2176
+ },
2177
+ {
2178
+ "target": "blocks.22.attn1.to_k",
2179
+ "rank": 256,
2180
+ "alpha": 256.0,
2181
+ "dropout": 0.0,
2182
+ "A": [
2183
+ 256,
2184
+ 3072
2185
+ ],
2186
+ "B": [
2187
+ 3072,
2188
+ 256
2189
+ ],
2190
+ "master_dtype": "torch.float32"
2191
+ },
2192
+ {
2193
+ "target": "blocks.22.attn1.to_v",
2194
+ "rank": 256,
2195
+ "alpha": 256.0,
2196
+ "dropout": 0.0,
2197
+ "A": [
2198
+ 256,
2199
+ 3072
2200
+ ],
2201
+ "B": [
2202
+ 3072,
2203
+ 256
2204
+ ],
2205
+ "master_dtype": "torch.float32"
2206
+ },
2207
+ {
2208
+ "target": "blocks.22.attn1.to_out.0",
2209
+ "rank": 256,
2210
+ "alpha": 256.0,
2211
+ "dropout": 0.0,
2212
+ "A": [
2213
+ 256,
2214
+ 3072
2215
+ ],
2216
+ "B": [
2217
+ 3072,
2218
+ 256
2219
+ ],
2220
+ "master_dtype": "torch.float32"
2221
+ },
2222
+ {
2223
+ "target": "blocks.22.ffn.net.0.proj",
2224
+ "rank": 256,
2225
+ "alpha": 256.0,
2226
+ "dropout": 0.0,
2227
+ "A": [
2228
+ 256,
2229
+ 3072
2230
+ ],
2231
+ "B": [
2232
+ 14336,
2233
+ 256
2234
+ ],
2235
+ "master_dtype": "torch.float32"
2236
+ },
2237
+ {
2238
+ "target": "blocks.22.ffn.net.2",
2239
+ "rank": 256,
2240
+ "alpha": 256.0,
2241
+ "dropout": 0.0,
2242
+ "A": [
2243
+ 256,
2244
+ 14336
2245
+ ],
2246
+ "B": [
2247
+ 3072,
2248
+ 256
2249
+ ],
2250
+ "master_dtype": "torch.float32"
2251
+ },
2252
+ {
2253
+ "target": "blocks.23.attn1.to_q",
2254
+ "rank": 256,
2255
+ "alpha": 256.0,
2256
+ "dropout": 0.0,
2257
+ "A": [
2258
+ 256,
2259
+ 3072
2260
+ ],
2261
+ "B": [
2262
+ 3072,
2263
+ 256
2264
+ ],
2265
+ "master_dtype": "torch.float32"
2266
+ },
2267
+ {
2268
+ "target": "blocks.23.attn1.to_k",
2269
+ "rank": 256,
2270
+ "alpha": 256.0,
2271
+ "dropout": 0.0,
2272
+ "A": [
2273
+ 256,
2274
+ 3072
2275
+ ],
2276
+ "B": [
2277
+ 3072,
2278
+ 256
2279
+ ],
2280
+ "master_dtype": "torch.float32"
2281
+ },
2282
+ {
2283
+ "target": "blocks.23.attn1.to_v",
2284
+ "rank": 256,
2285
+ "alpha": 256.0,
2286
+ "dropout": 0.0,
2287
+ "A": [
2288
+ 256,
2289
+ 3072
2290
+ ],
2291
+ "B": [
2292
+ 3072,
2293
+ 256
2294
+ ],
2295
+ "master_dtype": "torch.float32"
2296
+ },
2297
+ {
2298
+ "target": "blocks.23.attn1.to_out.0",
2299
+ "rank": 256,
2300
+ "alpha": 256.0,
2301
+ "dropout": 0.0,
2302
+ "A": [
2303
+ 256,
2304
+ 3072
2305
+ ],
2306
+ "B": [
2307
+ 3072,
2308
+ 256
2309
+ ],
2310
+ "master_dtype": "torch.float32"
2311
+ },
2312
+ {
2313
+ "target": "blocks.23.ffn.net.0.proj",
2314
+ "rank": 256,
2315
+ "alpha": 256.0,
2316
+ "dropout": 0.0,
2317
+ "A": [
2318
+ 256,
2319
+ 3072
2320
+ ],
2321
+ "B": [
2322
+ 14336,
2323
+ 256
2324
+ ],
2325
+ "master_dtype": "torch.float32"
2326
+ },
2327
+ {
2328
+ "target": "blocks.23.ffn.net.2",
2329
+ "rank": 256,
2330
+ "alpha": 256.0,
2331
+ "dropout": 0.0,
2332
+ "A": [
2333
+ 256,
2334
+ 14336
2335
+ ],
2336
+ "B": [
2337
+ 3072,
2338
+ 256
2339
+ ],
2340
+ "master_dtype": "torch.float32"
2341
+ },
2342
+ {
2343
+ "target": "blocks.24.attn1.to_q",
2344
+ "rank": 256,
2345
+ "alpha": 256.0,
2346
+ "dropout": 0.0,
2347
+ "A": [
2348
+ 256,
2349
+ 3072
2350
+ ],
2351
+ "B": [
2352
+ 3072,
2353
+ 256
2354
+ ],
2355
+ "master_dtype": "torch.float32"
2356
+ },
2357
+ {
2358
+ "target": "blocks.24.attn1.to_k",
2359
+ "rank": 256,
2360
+ "alpha": 256.0,
2361
+ "dropout": 0.0,
2362
+ "A": [
2363
+ 256,
2364
+ 3072
2365
+ ],
2366
+ "B": [
2367
+ 3072,
2368
+ 256
2369
+ ],
2370
+ "master_dtype": "torch.float32"
2371
+ },
2372
+ {
2373
+ "target": "blocks.24.attn1.to_v",
2374
+ "rank": 256,
2375
+ "alpha": 256.0,
2376
+ "dropout": 0.0,
2377
+ "A": [
2378
+ 256,
2379
+ 3072
2380
+ ],
2381
+ "B": [
2382
+ 3072,
2383
+ 256
2384
+ ],
2385
+ "master_dtype": "torch.float32"
2386
+ },
2387
+ {
2388
+ "target": "blocks.24.attn1.to_out.0",
2389
+ "rank": 256,
2390
+ "alpha": 256.0,
2391
+ "dropout": 0.0,
2392
+ "A": [
2393
+ 256,
2394
+ 3072
2395
+ ],
2396
+ "B": [
2397
+ 3072,
2398
+ 256
2399
+ ],
2400
+ "master_dtype": "torch.float32"
2401
+ },
2402
+ {
2403
+ "target": "blocks.24.ffn.net.0.proj",
2404
+ "rank": 256,
2405
+ "alpha": 256.0,
2406
+ "dropout": 0.0,
2407
+ "A": [
2408
+ 256,
2409
+ 3072
2410
+ ],
2411
+ "B": [
2412
+ 14336,
2413
+ 256
2414
+ ],
2415
+ "master_dtype": "torch.float32"
2416
+ },
2417
+ {
2418
+ "target": "blocks.24.ffn.net.2",
2419
+ "rank": 256,
2420
+ "alpha": 256.0,
2421
+ "dropout": 0.0,
2422
+ "A": [
2423
+ 256,
2424
+ 14336
2425
+ ],
2426
+ "B": [
2427
+ 3072,
2428
+ 256
2429
+ ],
2430
+ "master_dtype": "torch.float32"
2431
+ },
2432
+ {
2433
+ "target": "blocks.25.attn1.to_q",
2434
+ "rank": 256,
2435
+ "alpha": 256.0,
2436
+ "dropout": 0.0,
2437
+ "A": [
2438
+ 256,
2439
+ 3072
2440
+ ],
2441
+ "B": [
2442
+ 3072,
2443
+ 256
2444
+ ],
2445
+ "master_dtype": "torch.float32"
2446
+ },
2447
+ {
2448
+ "target": "blocks.25.attn1.to_k",
2449
+ "rank": 256,
2450
+ "alpha": 256.0,
2451
+ "dropout": 0.0,
2452
+ "A": [
2453
+ 256,
2454
+ 3072
2455
+ ],
2456
+ "B": [
2457
+ 3072,
2458
+ 256
2459
+ ],
2460
+ "master_dtype": "torch.float32"
2461
+ },
2462
+ {
2463
+ "target": "blocks.25.attn1.to_v",
2464
+ "rank": 256,
2465
+ "alpha": 256.0,
2466
+ "dropout": 0.0,
2467
+ "A": [
2468
+ 256,
2469
+ 3072
2470
+ ],
2471
+ "B": [
2472
+ 3072,
2473
+ 256
2474
+ ],
2475
+ "master_dtype": "torch.float32"
2476
+ },
2477
+ {
2478
+ "target": "blocks.25.attn1.to_out.0",
2479
+ "rank": 256,
2480
+ "alpha": 256.0,
2481
+ "dropout": 0.0,
2482
+ "A": [
2483
+ 256,
2484
+ 3072
2485
+ ],
2486
+ "B": [
2487
+ 3072,
2488
+ 256
2489
+ ],
2490
+ "master_dtype": "torch.float32"
2491
+ },
2492
+ {
2493
+ "target": "blocks.25.ffn.net.0.proj",
2494
+ "rank": 256,
2495
+ "alpha": 256.0,
2496
+ "dropout": 0.0,
2497
+ "A": [
2498
+ 256,
2499
+ 3072
2500
+ ],
2501
+ "B": [
2502
+ 14336,
2503
+ 256
2504
+ ],
2505
+ "master_dtype": "torch.float32"
2506
+ },
2507
+ {
2508
+ "target": "blocks.25.ffn.net.2",
2509
+ "rank": 256,
2510
+ "alpha": 256.0,
2511
+ "dropout": 0.0,
2512
+ "A": [
2513
+ 256,
2514
+ 14336
2515
+ ],
2516
+ "B": [
2517
+ 3072,
2518
+ 256
2519
+ ],
2520
+ "master_dtype": "torch.float32"
2521
+ },
2522
+ {
2523
+ "target": "blocks.26.attn1.to_q",
2524
+ "rank": 256,
2525
+ "alpha": 256.0,
2526
+ "dropout": 0.0,
2527
+ "A": [
2528
+ 256,
2529
+ 3072
2530
+ ],
2531
+ "B": [
2532
+ 3072,
2533
+ 256
2534
+ ],
2535
+ "master_dtype": "torch.float32"
2536
+ },
2537
+ {
2538
+ "target": "blocks.26.attn1.to_k",
2539
+ "rank": 256,
2540
+ "alpha": 256.0,
2541
+ "dropout": 0.0,
2542
+ "A": [
2543
+ 256,
2544
+ 3072
2545
+ ],
2546
+ "B": [
2547
+ 3072,
2548
+ 256
2549
+ ],
2550
+ "master_dtype": "torch.float32"
2551
+ },
2552
+ {
2553
+ "target": "blocks.26.attn1.to_v",
2554
+ "rank": 256,
2555
+ "alpha": 256.0,
2556
+ "dropout": 0.0,
2557
+ "A": [
2558
+ 256,
2559
+ 3072
2560
+ ],
2561
+ "B": [
2562
+ 3072,
2563
+ 256
2564
+ ],
2565
+ "master_dtype": "torch.float32"
2566
+ },
2567
+ {
2568
+ "target": "blocks.26.attn1.to_out.0",
2569
+ "rank": 256,
2570
+ "alpha": 256.0,
2571
+ "dropout": 0.0,
2572
+ "A": [
2573
+ 256,
2574
+ 3072
2575
+ ],
2576
+ "B": [
2577
+ 3072,
2578
+ 256
2579
+ ],
2580
+ "master_dtype": "torch.float32"
2581
+ },
2582
+ {
2583
+ "target": "blocks.26.ffn.net.0.proj",
2584
+ "rank": 256,
2585
+ "alpha": 256.0,
2586
+ "dropout": 0.0,
2587
+ "A": [
2588
+ 256,
2589
+ 3072
2590
+ ],
2591
+ "B": [
2592
+ 14336,
2593
+ 256
2594
+ ],
2595
+ "master_dtype": "torch.float32"
2596
+ },
2597
+ {
2598
+ "target": "blocks.26.ffn.net.2",
2599
+ "rank": 256,
2600
+ "alpha": 256.0,
2601
+ "dropout": 0.0,
2602
+ "A": [
2603
+ 256,
2604
+ 14336
2605
+ ],
2606
+ "B": [
2607
+ 3072,
2608
+ 256
2609
+ ],
2610
+ "master_dtype": "torch.float32"
2611
+ },
2612
+ {
2613
+ "target": "blocks.27.attn1.to_q",
2614
+ "rank": 256,
2615
+ "alpha": 256.0,
2616
+ "dropout": 0.0,
2617
+ "A": [
2618
+ 256,
2619
+ 3072
2620
+ ],
2621
+ "B": [
2622
+ 3072,
2623
+ 256
2624
+ ],
2625
+ "master_dtype": "torch.float32"
2626
+ },
2627
+ {
2628
+ "target": "blocks.27.attn1.to_k",
2629
+ "rank": 256,
2630
+ "alpha": 256.0,
2631
+ "dropout": 0.0,
2632
+ "A": [
2633
+ 256,
2634
+ 3072
2635
+ ],
2636
+ "B": [
2637
+ 3072,
2638
+ 256
2639
+ ],
2640
+ "master_dtype": "torch.float32"
2641
+ },
2642
+ {
2643
+ "target": "blocks.27.attn1.to_v",
2644
+ "rank": 256,
2645
+ "alpha": 256.0,
2646
+ "dropout": 0.0,
2647
+ "A": [
2648
+ 256,
2649
+ 3072
2650
+ ],
2651
+ "B": [
2652
+ 3072,
2653
+ 256
2654
+ ],
2655
+ "master_dtype": "torch.float32"
2656
+ },
2657
+ {
2658
+ "target": "blocks.27.attn1.to_out.0",
2659
+ "rank": 256,
2660
+ "alpha": 256.0,
2661
+ "dropout": 0.0,
2662
+ "A": [
2663
+ 256,
2664
+ 3072
2665
+ ],
2666
+ "B": [
2667
+ 3072,
2668
+ 256
2669
+ ],
2670
+ "master_dtype": "torch.float32"
2671
+ },
2672
+ {
2673
+ "target": "blocks.27.ffn.net.0.proj",
2674
+ "rank": 256,
2675
+ "alpha": 256.0,
2676
+ "dropout": 0.0,
2677
+ "A": [
2678
+ 256,
2679
+ 3072
2680
+ ],
2681
+ "B": [
2682
+ 14336,
2683
+ 256
2684
+ ],
2685
+ "master_dtype": "torch.float32"
2686
+ },
2687
+ {
2688
+ "target": "blocks.27.ffn.net.2",
2689
+ "rank": 256,
2690
+ "alpha": 256.0,
2691
+ "dropout": 0.0,
2692
+ "A": [
2693
+ 256,
2694
+ 14336
2695
+ ],
2696
+ "B": [
2697
+ 3072,
2698
+ 256
2699
+ ],
2700
+ "master_dtype": "torch.float32"
2701
+ },
2702
+ {
2703
+ "target": "blocks.28.attn1.to_q",
2704
+ "rank": 256,
2705
+ "alpha": 256.0,
2706
+ "dropout": 0.0,
2707
+ "A": [
2708
+ 256,
2709
+ 3072
2710
+ ],
2711
+ "B": [
2712
+ 3072,
2713
+ 256
2714
+ ],
2715
+ "master_dtype": "torch.float32"
2716
+ },
2717
+ {
2718
+ "target": "blocks.28.attn1.to_k",
2719
+ "rank": 256,
2720
+ "alpha": 256.0,
2721
+ "dropout": 0.0,
2722
+ "A": [
2723
+ 256,
2724
+ 3072
2725
+ ],
2726
+ "B": [
2727
+ 3072,
2728
+ 256
2729
+ ],
2730
+ "master_dtype": "torch.float32"
2731
+ },
2732
+ {
2733
+ "target": "blocks.28.attn1.to_v",
2734
+ "rank": 256,
2735
+ "alpha": 256.0,
2736
+ "dropout": 0.0,
2737
+ "A": [
2738
+ 256,
2739
+ 3072
2740
+ ],
2741
+ "B": [
2742
+ 3072,
2743
+ 256
2744
+ ],
2745
+ "master_dtype": "torch.float32"
2746
+ },
2747
+ {
2748
+ "target": "blocks.28.attn1.to_out.0",
2749
+ "rank": 256,
2750
+ "alpha": 256.0,
2751
+ "dropout": 0.0,
2752
+ "A": [
2753
+ 256,
2754
+ 3072
2755
+ ],
2756
+ "B": [
2757
+ 3072,
2758
+ 256
2759
+ ],
2760
+ "master_dtype": "torch.float32"
2761
+ },
2762
+ {
2763
+ "target": "blocks.28.ffn.net.0.proj",
2764
+ "rank": 256,
2765
+ "alpha": 256.0,
2766
+ "dropout": 0.0,
2767
+ "A": [
2768
+ 256,
2769
+ 3072
2770
+ ],
2771
+ "B": [
2772
+ 14336,
2773
+ 256
2774
+ ],
2775
+ "master_dtype": "torch.float32"
2776
+ },
2777
+ {
2778
+ "target": "blocks.28.ffn.net.2",
2779
+ "rank": 256,
2780
+ "alpha": 256.0,
2781
+ "dropout": 0.0,
2782
+ "A": [
2783
+ 256,
2784
+ 14336
2785
+ ],
2786
+ "B": [
2787
+ 3072,
2788
+ 256
2789
+ ],
2790
+ "master_dtype": "torch.float32"
2791
+ },
2792
+ {
2793
+ "target": "blocks.29.attn1.to_q",
2794
+ "rank": 256,
2795
+ "alpha": 256.0,
2796
+ "dropout": 0.0,
2797
+ "A": [
2798
+ 256,
2799
+ 3072
2800
+ ],
2801
+ "B": [
2802
+ 3072,
2803
+ 256
2804
+ ],
2805
+ "master_dtype": "torch.float32"
2806
+ },
2807
+ {
2808
+ "target": "blocks.29.attn1.to_k",
2809
+ "rank": 256,
2810
+ "alpha": 256.0,
2811
+ "dropout": 0.0,
2812
+ "A": [
2813
+ 256,
2814
+ 3072
2815
+ ],
2816
+ "B": [
2817
+ 3072,
2818
+ 256
2819
+ ],
2820
+ "master_dtype": "torch.float32"
2821
+ },
2822
+ {
2823
+ "target": "blocks.29.attn1.to_v",
2824
+ "rank": 256,
2825
+ "alpha": 256.0,
2826
+ "dropout": 0.0,
2827
+ "A": [
2828
+ 256,
2829
+ 3072
2830
+ ],
2831
+ "B": [
2832
+ 3072,
2833
+ 256
2834
+ ],
2835
+ "master_dtype": "torch.float32"
2836
+ },
2837
+ {
2838
+ "target": "blocks.29.attn1.to_out.0",
2839
+ "rank": 256,
2840
+ "alpha": 256.0,
2841
+ "dropout": 0.0,
2842
+ "A": [
2843
+ 256,
2844
+ 3072
2845
+ ],
2846
+ "B": [
2847
+ 3072,
2848
+ 256
2849
+ ],
2850
+ "master_dtype": "torch.float32"
2851
+ },
2852
+ {
2853
+ "target": "blocks.29.ffn.net.0.proj",
2854
+ "rank": 256,
2855
+ "alpha": 256.0,
2856
+ "dropout": 0.0,
2857
+ "A": [
2858
+ 256,
2859
+ 3072
2860
+ ],
2861
+ "B": [
2862
+ 14336,
2863
+ 256
2864
+ ],
2865
+ "master_dtype": "torch.float32"
2866
+ },
2867
+ {
2868
+ "target": "blocks.29.ffn.net.2",
2869
+ "rank": 256,
2870
+ "alpha": 256.0,
2871
+ "dropout": 0.0,
2872
+ "A": [
2873
+ 256,
2874
+ 14336
2875
+ ],
2876
+ "B": [
2877
+ 3072,
2878
+ 256
2879
+ ],
2880
+ "master_dtype": "torch.float32"
2881
+ }
2882
+ ],
2883
+ "sha256": "49ecc581a09c6d7cf0357e12e1e1ff806dc3c9e7899417affdc0881429db797b"
2884
+ }