Hugging Face's logo

zhiyang1
/
SZ_multinode_vlm_dc-vae-f32c32-768_group-7_no-self-attn-lora_aug_residual_attn

Safetensors
Model card Files
xet
Community
zhiyang1 commited on
Commit
446dc51
·
verified ·
1 Parent(s): 6a4ba5f

Upload folder using huggingface_hub

Browse files
Files changed (24) hide show
  1. .gitattributes +1 -0
  2. checkpoint-94000/config.json +79 -0
  3. checkpoint-94000/generation_config.json +6 -0
  4. checkpoint-94000/model.safetensors +3 -0
  5. checkpoint-94000/optimizer.pt +3 -0
  6. checkpoint-94000/rng_state_0.pth +3 -0
  7. checkpoint-94000/rng_state_1.pth +3 -0
  8. checkpoint-94000/rng_state_10.pth +3 -0
  9. checkpoint-94000/rng_state_11.pth +3 -0
  10. checkpoint-94000/rng_state_12.pth +3 -0
  11. checkpoint-94000/rng_state_13.pth +3 -0
  12. checkpoint-94000/rng_state_14.pth +3 -0
  13. checkpoint-94000/rng_state_15.pth +3 -0
  14. checkpoint-94000/rng_state_2.pth +3 -0
  15. checkpoint-94000/rng_state_3.pth +3 -0
  16. checkpoint-94000/rng_state_4.pth +3 -0
  17. checkpoint-94000/rng_state_5.pth +3 -0
  18. checkpoint-94000/rng_state_6.pth +3 -0
  19. checkpoint-94000/rng_state_7.pth +3 -0
  20. checkpoint-94000/rng_state_8.pth +3 -0
  21. checkpoint-94000/rng_state_9.pth +3 -0
  22. checkpoint-94000/scheduler.pt +3 -0
  23. checkpoint-94000/trainer_state.json +3 -0
  24. checkpoint-94000/training_args.bin +3 -0
.gitattributes CHANGED
@@ -42,3 +42,4 @@ checkpoint-84000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
42
  checkpoint-88000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
43
  checkpoint-90000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
44
  checkpoint-92000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
 
 
42
  checkpoint-88000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
43
  checkpoint-90000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
44
  checkpoint-92000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
45
+ checkpoint-94000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
checkpoint-94000/config.json ADDED
@@ -0,0 +1,79 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "ar_steps": 1,
3
+ "architectures": [
4
+ "DiffVLMDiffusion"
5
+ ],
6
+ "attention_dropout": 0.0,
7
+ "bos_token_id": 151643,
8
+ "condition_layer": -1,
9
+ "eos_token_id": 151645,
10
+ "hidden_act": "silu",
11
+ "hidden_size": 1536,
12
+ "image_token_id": 151655,
13
+ "img_cross_attention_dim": 2048,
14
+ "img_diffuser_depth": 2,
15
+ "img_ffn_dim_multiplier": null,
16
+ "img_hidden_size": 1536,
17
+ "img_multiple_of": 256,
18
+ "img_norm_eps": 1e-05,
19
+ "img_num_attention_heads": 12,
20
+ "img_num_kv_heads": 12,
21
+ "img_qk_norm": true,
22
+ "in_channels": 32,
23
+ "initializer_range": 0.02,
24
+ "inject_img_diffuser": false,
25
+ "input_size": 32,
26
+ "intermediate_size": 8960,
27
+ "layer_group_size": 7,
28
+ "layerwise_start_idx": 0,
29
+ "lora_alpha": 128,
30
+ "lora_bias": "none",
31
+ "lora_dropout": 0.05,
32
+ "lora_enable": false,
33
+ "lora_r": 64,
34
+ "max_position_embeddings": 32768,
35
+ "max_window_layers": 28,
36
+ "model_type": "qwen2_vl",
37
+ "non_linearity": 1,
38
+ "norm_elementwise_affine": true,
39
+ "num_attention_heads": 12,
40
+ "num_hidden_layers": 28,
41
+ "num_key_value_heads": 2,
42
+ "patch_size": 2,
43
+ "repa_coeff": 0.1,
44
+ "repa_layers": "2",
45
+ "repa_shared": false,
46
+ "rms_norm_eps": 1e-06,
47
+ "rope_scaling": {
48
+ "mrope_section": [
49
+ 16,
50
+ 24,
51
+ 24
52
+ ],
53
+ "rope_type": "default",
54
+ "type": "default"
55
+ },
56
+ "rope_theta": 1000000.0,
57
+ "sample_size": 128,
58
+ "sampling_steps": 28,
59
+ "sliding_window": null,
60
+ "tie_word_embeddings": true,
61
+ "torch_dtype": "bfloat16",
62
+ "transformers_version": "4.47.0",
63
+ "use_cache": true,
64
+ "use_repa": false,
65
+ "use_residual_attn": true,
66
+ "use_sliding_window": false,
67
+ "vae_path": "mit-han-lab/dc-ae-f32c32-sana-1.1-diffusers",
68
+ "video_token_id": 151656,
69
+ "vision_config": {
70
+ "hidden_size": 1536,
71
+ "in_chans": 3,
72
+ "model_type": "qwen2_vl",
73
+ "spatial_patch_size": 14
74
+ },
75
+ "vision_end_token_id": 151653,
76
+ "vision_start_token_id": 151652,
77
+ "vision_token_id": 151654,
78
+ "vocab_size": 151936
79
+ }
checkpoint-94000/generation_config.json ADDED
@@ -0,0 +1,6 @@
 
 
 
 
 
 
 
1
+ {
2
+ "_from_model_config": true,
3
+ "bos_token_id": 151643,
4
+ "eos_token_id": 151645,
5
+ "transformers_version": "4.47.0"
6
+ }
checkpoint-94000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b4d68442df73cfe35f031b9cbde35029c11c91e1f3dfa18fb8620c37ee5f86e1
3
+ size 4539487822
checkpoint-94000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:13b093c4f59108b58417cea16ad0f8ada5f7045d1906ba3ff5b62582b530ddb2
3
+ size 6632345482
checkpoint-94000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:620793806a759320f7145834c504356a45a8bd4a27b4902f65b135bfa17d994f
3
+ size 15984
checkpoint-94000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3f1af17911371216e20677614259afb4a584e3425db0509c055a4dd723fef480
3
+ size 15984
checkpoint-94000/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2a17a7092913154521d04c2bcc039b178565e5fb2e53a475806e200a2e80e1bb
3
+ size 15997
checkpoint-94000/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cbc159b814df76a0944754bee17f23f3a4fe4578f3b6aadf8ddfb57284899120
3
+ size 15997
checkpoint-94000/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:66267f59b2eb1907fdb0d1486af9f949da21b203fa476da9a63a932033a64854
3
+ size 15997
checkpoint-94000/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2bf20d2aca99263f5ee9a1c780f86bb14e3d122d4c345c0c161085535d4bf2b9
3
+ size 15997
checkpoint-94000/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6092a627b881c8241cc104bc0e3241d99e490b03ce613dd1f12b8e9a2b9059c1
3
+ size 15997
checkpoint-94000/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d8eae2b889e6af456e4c6bf07170733781501cf54c0ef43d85f4ee3cc7c5f465
3
+ size 15997
checkpoint-94000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c5afde0e21fbc44c580c4cdafaef1e0bdfe8ec2cdb8eca07889211ed384c3940
3
+ size 15984
checkpoint-94000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:341c3ea0e9c64aef07d5e9ad30d066164f09af7bca036f2af4d683c35e64d1ab
3
+ size 15984
checkpoint-94000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d31444f1b3fad9f129d3e6b4661562397661c3754b4fd7336dd9741d3783b133
3
+ size 15984
checkpoint-94000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c1df048339b6b9bef09e9e12d630f6dbcb177e3e69f4788371796935422b923b
3
+ size 15984
checkpoint-94000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0c4aa6d16b4c3fd3ac89c00550ebac1930804aa4a9dffeca4b719bc680cd8255
3
+ size 15984
checkpoint-94000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9acb4dd0bc8467645547321a123d3197e326b382e0747bb4a54aa56f88af19e7
3
+ size 15984
checkpoint-94000/rng_state_8.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cb0f5dea5afcb078b4e8a5d523f8ebef08e7591379cee3a3c3693ca2e9f0e19e
3
+ size 15984
checkpoint-94000/rng_state_9.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2d4d1d14170d4002b7624dc7db4f52ec9290455789cf5b7d7c013d2502d522a7
3
+ size 15984
checkpoint-94000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6271d0f5cdba92bc3b160a7eba4e0b57e4f06e52e86e2115d3a347152299fd39
3
+ size 1064
checkpoint-94000/trainer_state.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:563aa0b0d9f74bff30f31f7d2052af903ccda4ebb5bfa88b173cdb150b043a4c
3
+ size 14497633
checkpoint-94000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c9233bcb38350336f1d8bfa55a4cac9c0295046db04d2cd649007e4d5db119eb
3
+ size 6008