Hugging Face's logo

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

Safetensors
Model card Files
xet
Community
zhiyang1 commited on
Commit
b0bc7c3
·
verified ·
1 Parent(s): 2c4cf44

Upload folder using huggingface_hub

Browse files
Files changed (46) hide show
  1. checkpoint-48000/config.json +79 -0
  2. checkpoint-48000/generation_config.json +6 -0
  3. checkpoint-48000/model.safetensors +3 -0
  4. checkpoint-48000/optimizer.pt +3 -0
  5. checkpoint-48000/rng_state_0.pth +3 -0
  6. checkpoint-48000/rng_state_1.pth +3 -0
  7. checkpoint-48000/rng_state_10.pth +3 -0
  8. checkpoint-48000/rng_state_11.pth +3 -0
  9. checkpoint-48000/rng_state_12.pth +3 -0
  10. checkpoint-48000/rng_state_13.pth +3 -0
  11. checkpoint-48000/rng_state_14.pth +3 -0
  12. checkpoint-48000/rng_state_15.pth +3 -0
  13. checkpoint-48000/rng_state_2.pth +3 -0
  14. checkpoint-48000/rng_state_3.pth +3 -0
  15. checkpoint-48000/rng_state_4.pth +3 -0
  16. checkpoint-48000/rng_state_5.pth +3 -0
  17. checkpoint-48000/rng_state_6.pth +3 -0
  18. checkpoint-48000/rng_state_7.pth +3 -0
  19. checkpoint-48000/rng_state_8.pth +3 -0
  20. checkpoint-48000/rng_state_9.pth +3 -0
  21. checkpoint-48000/scheduler.pt +3 -0
  22. checkpoint-48000/trainer_state.json +0 -0
  23. checkpoint-48000/training_args.bin +3 -0
  24. checkpoint-49000/config.json +79 -0
  25. checkpoint-49000/generation_config.json +6 -0
  26. checkpoint-49000/model.safetensors +3 -0
  27. checkpoint-49000/optimizer.pt +3 -0
  28. checkpoint-49000/rng_state_0.pth +3 -0
  29. checkpoint-49000/rng_state_1.pth +3 -0
  30. checkpoint-49000/rng_state_10.pth +3 -0
  31. checkpoint-49000/rng_state_11.pth +3 -0
  32. checkpoint-49000/rng_state_12.pth +3 -0
  33. checkpoint-49000/rng_state_13.pth +3 -0
  34. checkpoint-49000/rng_state_14.pth +3 -0
  35. checkpoint-49000/rng_state_15.pth +3 -0
  36. checkpoint-49000/rng_state_2.pth +3 -0
  37. checkpoint-49000/rng_state_3.pth +3 -0
  38. checkpoint-49000/rng_state_4.pth +3 -0
  39. checkpoint-49000/rng_state_5.pth +3 -0
  40. checkpoint-49000/rng_state_6.pth +3 -0
  41. checkpoint-49000/rng_state_7.pth +3 -0
  42. checkpoint-49000/rng_state_8.pth +3 -0
  43. checkpoint-49000/rng_state_9.pth +3 -0
  44. checkpoint-49000/scheduler.pt +3 -0
  45. checkpoint-49000/trainer_state.json +0 -0
  46. checkpoint-49000/training_args.bin +3 -0
checkpoint-48000/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": false,
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-48000/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-48000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5de5c159e5ea496ae23ad993f0551c45c93f5c48bd562cac3b48e0421ab67be9
3
+ size 4538452374
checkpoint-48000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b248ec1fd65a2442e1db00834e591777130dda37be0d3889b185be8d2054a51f
3
+ size 6630551906
checkpoint-48000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7c08b5ae272d174c16e96c578fb68b690831247654b5e4252884ab7a35471bb0
3
+ size 15984
checkpoint-48000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b8108a9cde2ed8fd9f046f44aed7581598660f57101ca3d93f168a046a45e2bd
3
+ size 15984
checkpoint-48000/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b8502feaa5acb7fcaa038fee911541ddc4903b3247ce10cfba7d8e8ddbc023af
3
+ size 15997
checkpoint-48000/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2e8023dd836a77416e4de8e42127bf2c3ddb3a9aec2230a39a064bdd97f5b9ef
3
+ size 15997
checkpoint-48000/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1429e3a94714ed05d56846c8d0dbc319b53651557ad57cae485a1cb43df82fb8
3
+ size 15997
checkpoint-48000/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b0c45d4c61ed00459417f83716ea3152d547a95bf90adf1602a18506a53f8b51
3
+ size 15997
checkpoint-48000/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9d255384fa48aeab3b91750d43fe337d661f4ae412becb02f32584c6d0411b9b
3
+ size 15997
checkpoint-48000/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0640416cec5b98f900677c2bc6b9cf5ef6a817e66e934a2c2fcb72df99bf6ed2
3
+ size 15997
checkpoint-48000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2c82ea3f49889e4d739c4178c9acff24e43945980c0d916e590f5345cc8bbb05
3
+ size 15984
checkpoint-48000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0067029d02c8ca97c63cfa9a62a099c45b8a88767d3a422e712bdaa97dee2dbe
3
+ size 15984
checkpoint-48000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8b514a20709e013ad1f330c964970a5082540aa86a48787874aa5417fe730139
3
+ size 15984
checkpoint-48000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:dba98d398c9e01ae1bd3f92d0f4ea775f4f5174bebfb8f4a9aa005496b4d9f46
3
+ size 15984
checkpoint-48000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0e1afc3cf8c1f5d2d43ca91a794d6126ca8ed51d0ef0fc9f18e7ac2c6b1fc03f
3
+ size 15984
checkpoint-48000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3bfeafa1d05ade12a0e102a837816371919413a4a47c648a5a5a921003712772
3
+ size 15984
checkpoint-48000/rng_state_8.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c88983f1e11891e43444e42e4f0855a6210408d2ef721e6e9006e9a8d0761a13
3
+ size 15984
checkpoint-48000/rng_state_9.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c0120fc0f63786c005a486c7b8bf4afd2f36b515b8f47ec46c2c41044a994dc2
3
+ size 15984
checkpoint-48000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:52e261ae24ab4937b650b8012575f7670cc08b7154604736c56b393ecce9ce9a
3
+ size 1064
checkpoint-48000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-48000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:73b9002f0d0669c7d76a11456b4a3d055f0b63567dc395f60c74e2df70deccf9
3
+ size 6008
checkpoint-49000/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": false,
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-49000/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-49000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1eaede1b379435f21ce1d83b08f86a18024ccb8725cb754307884027ce560e88
3
+ size 4538452374
checkpoint-49000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:86ab7ead535ef665e664779571772837cc8329ee0d25b459d76300ee5293a590
3
+ size 6630551906
checkpoint-49000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c462a44e4801be2e8796ca3b7d543ab2d0bf2de04f6ed9e2a5b3ba4c57996c89
3
+ size 15984
checkpoint-49000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1612b5dc247b54bafd228039126f12faf2b72af59afa278f49a58fa0126fd4cc
3
+ size 15984
checkpoint-49000/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:78f13101267effa1375269da070994d5585b6de58e4641299931cfadca7de228
3
+ size 15997
checkpoint-49000/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:78411a42c54df731c653a6d0d5c79d6f94feb9f039095b247c3f5e3e752eb74a
3
+ size 15997
checkpoint-49000/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:698276b74a18df2c451f425936f49f7be05fc4948424f26653c6e0d27576e2d9
3
+ size 15997
checkpoint-49000/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d0d039d25aaa2ca1866971f345b7a6ad0ff5e56f6e0dd182d3cc149e0b0b0066
3
+ size 15997
checkpoint-49000/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7c178fad60c75bcfb91fdc5d7d096eeb6ef984e19cc610297344466e60d1a130
3
+ size 15997
checkpoint-49000/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:04cb3e06ac4907a22481f2f63cda886d1bc8d4867097ac1d3587b689596e3d75
3
+ size 15997
checkpoint-49000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:dc599d4c09085a30f1441e089ae8798daa292ea480960b939fb1b4f1d3c23eb6
3
+ size 15984
checkpoint-49000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ecf917a7c192b0174df63556de4d01228520ba8e8c741a0af1c6afb8a8a3913f
3
+ size 15984
checkpoint-49000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ad3aa0eb751cf6a3600afa7343d6c1477ea976117f5ecb6995f99e80e46477df
3
+ size 15984
checkpoint-49000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6ef34713d465999e44156d5fc69bab2e04d6382388feed85bfd26d79aff6b4ae
3
+ size 15984
checkpoint-49000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3cee757cf845097b6756cc6016465bd7efc283875d6761b4d3799b061e130c50
3
+ size 15984
checkpoint-49000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7b1d1ae22e6a0d6693b51aad14ff5d6d2f19056b3d9ac6afc8be3ae3ee795e71
3
+ size 15984
checkpoint-49000/rng_state_8.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3100ee40f773ab6f4ad7e5e19138eb54f87148966fde7894bc239314c963bf32
3
+ size 15984
checkpoint-49000/rng_state_9.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2eabacfc577ad5bf56c9ff3d7120ef70381b7a5a86798ff68e2b002ed92554af
3
+ size 15984
checkpoint-49000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9279b9d8d91ce8102afc6ff2cbbcb1534217c1c028e8bc8acc02e7c5263bc88f
3
+ size 1064
checkpoint-49000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-49000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b541906e8650abce1db8ac188680f7fb82ed963726ffcb009158d733373e8741
3
+ size 5944