Hugging Face's logo

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

Safetensors
Model card Files
xet
Community
zhiyang1 commited on
Commit
e1e8d10
·
verified ·
1 Parent(s): 3e70f14

Upload folder using huggingface_hub

Browse files
Files changed (46) hide show
  1. checkpoint-63000/config.json +79 -0
  2. checkpoint-63000/generation_config.json +6 -0
  3. checkpoint-63000/model.safetensors +3 -0
  4. checkpoint-63000/optimizer.pt +3 -0
  5. checkpoint-63000/rng_state_0.pth +3 -0
  6. checkpoint-63000/rng_state_1.pth +3 -0
  7. checkpoint-63000/rng_state_10.pth +3 -0
  8. checkpoint-63000/rng_state_11.pth +3 -0
  9. checkpoint-63000/rng_state_12.pth +3 -0
  10. checkpoint-63000/rng_state_13.pth +3 -0
  11. checkpoint-63000/rng_state_14.pth +3 -0
  12. checkpoint-63000/rng_state_15.pth +3 -0
  13. checkpoint-63000/rng_state_2.pth +3 -0
  14. checkpoint-63000/rng_state_3.pth +3 -0
  15. checkpoint-63000/rng_state_4.pth +3 -0
  16. checkpoint-63000/rng_state_5.pth +3 -0
  17. checkpoint-63000/rng_state_6.pth +3 -0
  18. checkpoint-63000/rng_state_7.pth +3 -0
  19. checkpoint-63000/rng_state_8.pth +3 -0
  20. checkpoint-63000/rng_state_9.pth +3 -0
  21. checkpoint-63000/scheduler.pt +3 -0
  22. checkpoint-63000/trainer_state.json +0 -0
  23. checkpoint-63000/training_args.bin +3 -0
  24. checkpoint-64000/config.json +79 -0
  25. checkpoint-64000/generation_config.json +6 -0
  26. checkpoint-64000/model.safetensors +3 -0
  27. checkpoint-64000/optimizer.pt +3 -0
  28. checkpoint-64000/rng_state_0.pth +3 -0
  29. checkpoint-64000/rng_state_1.pth +3 -0
  30. checkpoint-64000/rng_state_10.pth +3 -0
  31. checkpoint-64000/rng_state_11.pth +3 -0
  32. checkpoint-64000/rng_state_12.pth +3 -0
  33. checkpoint-64000/rng_state_13.pth +3 -0
  34. checkpoint-64000/rng_state_14.pth +3 -0
  35. checkpoint-64000/rng_state_15.pth +3 -0
  36. checkpoint-64000/rng_state_2.pth +3 -0
  37. checkpoint-64000/rng_state_3.pth +3 -0
  38. checkpoint-64000/rng_state_4.pth +3 -0
  39. checkpoint-64000/rng_state_5.pth +3 -0
  40. checkpoint-64000/rng_state_6.pth +3 -0
  41. checkpoint-64000/rng_state_7.pth +3 -0
  42. checkpoint-64000/rng_state_8.pth +3 -0
  43. checkpoint-64000/rng_state_9.pth +3 -0
  44. checkpoint-64000/scheduler.pt +3 -0
  45. checkpoint-64000/trainer_state.json +0 -0
  46. checkpoint-64000/training_args.bin +3 -0
checkpoint-63000/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-63000/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-63000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:86563b2f623c43431dadef75b991259585a20ad097e5fea6cab2be046e347dad
3
+ size 4539487822
checkpoint-63000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f9c488d8f82c6a1107cccb3fb7e2ec84b810362fdb13d7a9b228fd79491308c5
3
+ size 6632345482
checkpoint-63000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:efdc3291860500d90830348f7fb90acd8914256555fe7e794b6b981fccc8a2cc
3
+ size 15984
checkpoint-63000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cfcb2e0d885c81917989eea1b14bd7d225177fa6784e2371474564123641ff27
3
+ size 15984
checkpoint-63000/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1383c81e5ad20c6278085e8b328c0ef39bd587c08005fbfd3afdc2563513b35e
3
+ size 15997
checkpoint-63000/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b01c7e33b84953a05902f42d3bc2c4287101af1eed2755902f56c2b62510ddfd
3
+ size 15997
checkpoint-63000/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d237dea7e3a6f76be189f2613cb49511c1a9ea48ccec9e7db7a64c9f4a4862cc
3
+ size 15997
checkpoint-63000/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:612f581b8d53b10752f2223145af5cc9ee7d8262590bd8eecce5f5308d1adcee
3
+ size 15997
checkpoint-63000/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:af3d70928630ac5b507250ef644ec5460d11f8ee2b27610a51701edb61b19361
3
+ size 15997
checkpoint-63000/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b9362b722033a19df2567b66e947829a1c5796394ec331ffd00670b2b7d8cf45
3
+ size 15997
checkpoint-63000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:37959dde7f6ad94b918583df336006498600dcf868a6d800a60683b700257c53
3
+ size 15984
checkpoint-63000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1fb4d570cb29e347b4abb2453d58f4d74334fdc49b20d85bf68294707f3e7293
3
+ size 15984
checkpoint-63000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a901239c8f2a1d26637ab122d651c37a94b1f282d3447bd88402f087c329ffea
3
+ size 15984
checkpoint-63000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:540fcafa92333ea26e15781beff0b974d6815ba28ab6d5d8cf28013a1ad428eb
3
+ size 15984
checkpoint-63000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8d3bcde18ee358b3b9b4635f0adc58e1b8c2c38e04491233519091b44e095cdd
3
+ size 15984
checkpoint-63000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:29deb40ed82ee0561434b560ca0d04f9921552b74da39665f333ed218caebc42
3
+ size 15984
checkpoint-63000/rng_state_8.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a5e93177ce95250e392af2c5a27a9b69de9df987d9a6227ee332bec67f5e2081
3
+ size 15984
checkpoint-63000/rng_state_9.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6f5dc08dae74eef916af29f213ca8a4db64dbf97530dbc190849d8e80d62081d
3
+ size 15984
checkpoint-63000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:90d9865d74713befcefdc70b34b6033a6cbd61479f058e5a5b15aa91fe3721c9
3
+ size 1064
checkpoint-63000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-63000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b08b9bc1bd1148ed8042b74897cb568a0ac75b375b6a38e1b6a7ea94c2e264c9
3
+ size 6008
checkpoint-64000/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-64000/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-64000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:07da30427fbe336fbceaafeda8bde6ccc1909d149d27b80563881612837a29d2
3
+ size 4539487822
checkpoint-64000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:aba3b17faaebd5a32cbdf1eaa2a50fdbc791e71230f824b6c9cdc6abb48156b2
3
+ size 6632345482
checkpoint-64000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ea9323bf9f3cb3fc303623cc12a229831e689c6a689c2f77f93b0fcb5f40e971
3
+ size 15984
checkpoint-64000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:39deee21f799b2a129c6dd297350cf9193cc431c0b1e15ab0b980e1bf50a6601
3
+ size 15984
checkpoint-64000/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1001fbea05913eb5ad6738503d85be8c6defe0d19a6839eb3f1f091432eb8261
3
+ size 15997
checkpoint-64000/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1c91a95df7bbb39866b630ed510f057d67000e21133f7d6da45642c9481d8471
3
+ size 15997
checkpoint-64000/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d6ad39eb362ecaf3b4a0f3584392c3b424f3598910b61cb76f9fbbb6bf7ffa42
3
+ size 15997
checkpoint-64000/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:77ad52854540223e35e92dafba713f66a6efb55ea54a80153f19adfd3c0fc7a4
3
+ size 15997
checkpoint-64000/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5192d0ec2c82bfbc2c474747fc7a8b40db1080758a3f8bf29ec8cb8f2a38a5b9
3
+ size 15997
checkpoint-64000/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5ced93b26cba561879ce18caeaf5a86ac60576a52e9e96eb1d79535d6cdd14fd
3
+ size 15997
checkpoint-64000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:31e7248d07db79cf59d07a91e34a9dd534f39ca48ab0f1979f4c50043974ba0e
3
+ size 15984
checkpoint-64000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5e4b0757e8c2c591ec0295fcc92ce2c46c5513de4983baad787c947efec4dba2
3
+ size 15984
checkpoint-64000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1752d15f523bf2173cdca3bf663311d4298a408bd085e1db5834b3efbbfc9786
3
+ size 15984
checkpoint-64000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:90854534e75ed3a0e06809b86a51de246caa8816d728ac025ca0d31974abfe7e
3
+ size 15984
checkpoint-64000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:288219d041d728fe722aac0f1dccdf423d4ecd3670b710504c4027505486118f
3
+ size 15984
checkpoint-64000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9f37048f992eba3475f203379d7f5ab0a335d0438d8b643304ea55516d9adc11
3
+ size 15984
checkpoint-64000/rng_state_8.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:05e92c5fc993c9d9fd079a9319cadf5ced72e37987ffb1a31bbc043a994f7894
3
+ size 15984
checkpoint-64000/rng_state_9.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:be3030161b191abfa951d13cfe1e6c95081261b8329e0dce048ec3757b698700
3
+ size 15984
checkpoint-64000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8b11185a1922e0e408f17dd3dbef9808f844ee990abfbde80f97cb3d3e7c14e4
3
+ size 1064
checkpoint-64000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-64000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b9a15bc4b21953997921e1a9d0204be15634938948b0111ac62a29ca5f3d7d67
3
+ size 6008