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
b32a02d
·
verified ·
1 Parent(s): e1cf58a

Upload folder using huggingface_hub

Browse files
Files changed (24) hide show
  1. .gitattributes +1 -0
  2. checkpoint-252000/config.json +79 -0
  3. checkpoint-252000/generation_config.json +6 -0
  4. checkpoint-252000/model.safetensors +3 -0
  5. checkpoint-252000/optimizer.pt +3 -0
  6. checkpoint-252000/rng_state_0.pth +3 -0
  7. checkpoint-252000/rng_state_1.pth +3 -0
  8. checkpoint-252000/rng_state_10.pth +3 -0
  9. checkpoint-252000/rng_state_11.pth +3 -0
  10. checkpoint-252000/rng_state_12.pth +3 -0
  11. checkpoint-252000/rng_state_13.pth +3 -0
  12. checkpoint-252000/rng_state_14.pth +3 -0
  13. checkpoint-252000/rng_state_15.pth +3 -0
  14. checkpoint-252000/rng_state_2.pth +3 -0
  15. checkpoint-252000/rng_state_3.pth +3 -0
  16. checkpoint-252000/rng_state_4.pth +3 -0
  17. checkpoint-252000/rng_state_5.pth +3 -0
  18. checkpoint-252000/rng_state_6.pth +3 -0
  19. checkpoint-252000/rng_state_7.pth +3 -0
  20. checkpoint-252000/rng_state_8.pth +3 -0
  21. checkpoint-252000/rng_state_9.pth +3 -0
  22. checkpoint-252000/scheduler.pt +3 -0
  23. checkpoint-252000/trainer_state.json +3 -0
  24. checkpoint-252000/training_args.bin +3 -0
.gitattributes CHANGED
@@ -97,3 +97,4 @@ checkpoint-240000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
97
  checkpoint-242000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
98
  checkpoint-248000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
99
  checkpoint-250000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
 
 
97
  checkpoint-242000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
98
  checkpoint-248000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
99
  checkpoint-250000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
100
+ checkpoint-252000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
checkpoint-252000/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-252000/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-252000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:493a9116afda8436069ec98c02ba87daec5d01754a204cce2301d4d1c3da9765
3
+ size 4539487822
checkpoint-252000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5b3ba92a334b2efa92154104acc0111ebdf9ef3956e98e7e5f1882729a98b226
3
+ size 6632345482
checkpoint-252000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:543dab9ae881a1bb6f5fc2aa1a345d83bb2e97e1b431d2a0c1aa6c768bb360b0
3
+ size 15984
checkpoint-252000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:648ac04276f26af59d65d354374f324774b39bbc70dc36048c56daa091c96ad9
3
+ size 15984
checkpoint-252000/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9b717c7cf11868dfbe866fb26a195306043270741002671d04cad438aa8f353a
3
+ size 15997
checkpoint-252000/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1c8a09285d21756999a247d2fc5a85838f1a16ec3b4362e3cc5e539c0823e554
3
+ size 15997
checkpoint-252000/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:681084a9ed69a1a5ff797168f501b8ffaef183e8c44afe7fdc51329e26b59c9b
3
+ size 15997
checkpoint-252000/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6edca28e3996758ac94b48d7724654d3c9f8436cdcc58458439e94c700f9e0f9
3
+ size 15997
checkpoint-252000/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a2745f590141813c0b5c99cf12ab195eddd38e67fac4e924aa6ccbcf7bf48800
3
+ size 15997
checkpoint-252000/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:207db54fb460c35f666fc0e30dad407144ba2bdc088d19461ee9075b9b8f0bf5
3
+ size 15997
checkpoint-252000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:96f4d42eb8bd4a6564b60b258502d0a87be774f4db1d8a0e09fb7ebb3e4e1cc9
3
+ size 15984
checkpoint-252000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:25e58c173d973448a70325343d090f56f74a6d1cb7bcf0b6290e96e194dcf11b
3
+ size 15984
checkpoint-252000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d3eafdd1e258cf37daf7c4717adcad7b342805977058e84944aad811617f6b50
3
+ size 15984
checkpoint-252000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a24d21b9d0c340afdb16b68bf76345654cbdb8089baa7b5cd9fa0193a3563d65
3
+ size 15984
checkpoint-252000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:760f4b64d866624fd526fcb1f0a5d9e41ed7074b311ced9b1e19174546d31451
3
+ size 15984
checkpoint-252000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:39dc0db4bae4a6a70d7017bde190e3a9887811e99632ab7658036d77b8faac17
3
+ size 15984
checkpoint-252000/rng_state_8.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f0267d700c8d4bb864cf7f353c7c5bb034d2062e7618389d8add164f7d0f885f
3
+ size 15984
checkpoint-252000/rng_state_9.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b05442037df90331a12fffc51352ea198ef48d70c4ccb60332bcadc11d6d8542
3
+ size 15984
checkpoint-252000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2849d81f4cea367c73f08914d7472ce9b72bc77cb3ca9143b9cdfa860b66edf2
3
+ size 1064
checkpoint-252000/trainer_state.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:763bc49c7680d8a2fa357ed76a1e239f77884e78444cf749438ec895cb700958
3
+ size 38965832
checkpoint-252000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ffebe12ce6335b28f56567d1cd253b171d211189a5e173a06044f40faec9a83a
3
+ size 6008