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
f1137d0
·
verified ·
1 Parent(s): 319eb68

Upload folder using huggingface_hub

Browse files
Files changed (46) hide show
  1. checkpoint-59000/config.json +79 -0
  2. checkpoint-59000/generation_config.json +6 -0
  3. checkpoint-59000/model.safetensors +3 -0
  4. checkpoint-59000/optimizer.pt +3 -0
  5. checkpoint-59000/rng_state_0.pth +3 -0
  6. checkpoint-59000/rng_state_1.pth +3 -0
  7. checkpoint-59000/rng_state_10.pth +3 -0
  8. checkpoint-59000/rng_state_11.pth +3 -0
  9. checkpoint-59000/rng_state_12.pth +3 -0
  10. checkpoint-59000/rng_state_13.pth +3 -0
  11. checkpoint-59000/rng_state_14.pth +3 -0
  12. checkpoint-59000/rng_state_15.pth +3 -0
  13. checkpoint-59000/rng_state_2.pth +3 -0
  14. checkpoint-59000/rng_state_3.pth +3 -0
  15. checkpoint-59000/rng_state_4.pth +3 -0
  16. checkpoint-59000/rng_state_5.pth +3 -0
  17. checkpoint-59000/rng_state_6.pth +3 -0
  18. checkpoint-59000/rng_state_7.pth +3 -0
  19. checkpoint-59000/rng_state_8.pth +3 -0
  20. checkpoint-59000/rng_state_9.pth +3 -0
  21. checkpoint-59000/scheduler.pt +3 -0
  22. checkpoint-59000/trainer_state.json +0 -0
  23. checkpoint-59000/training_args.bin +3 -0
  24. checkpoint-60000/config.json +79 -0
  25. checkpoint-60000/generation_config.json +6 -0
  26. checkpoint-60000/model.safetensors +3 -0
  27. checkpoint-60000/optimizer.pt +3 -0
  28. checkpoint-60000/rng_state_0.pth +3 -0
  29. checkpoint-60000/rng_state_1.pth +3 -0
  30. checkpoint-60000/rng_state_10.pth +3 -0
  31. checkpoint-60000/rng_state_11.pth +3 -0
  32. checkpoint-60000/rng_state_12.pth +3 -0
  33. checkpoint-60000/rng_state_13.pth +3 -0
  34. checkpoint-60000/rng_state_14.pth +3 -0
  35. checkpoint-60000/rng_state_15.pth +3 -0
  36. checkpoint-60000/rng_state_2.pth +3 -0
  37. checkpoint-60000/rng_state_3.pth +3 -0
  38. checkpoint-60000/rng_state_4.pth +3 -0
  39. checkpoint-60000/rng_state_5.pth +3 -0
  40. checkpoint-60000/rng_state_6.pth +3 -0
  41. checkpoint-60000/rng_state_7.pth +3 -0
  42. checkpoint-60000/rng_state_8.pth +3 -0
  43. checkpoint-60000/rng_state_9.pth +3 -0
  44. checkpoint-60000/scheduler.pt +3 -0
  45. checkpoint-60000/trainer_state.json +0 -0
  46. checkpoint-60000/training_args.bin +3 -0
checkpoint-59000/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-59000/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-59000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:85b814ef7543b4eb42cc508d9b667795d5a12f2fa5c437cc3f1e9f4a0d1e6330
3
+ size 4538452374
checkpoint-59000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:78936892719a78e8baf52a475d0ab8c6ac7a84cbe384413df6827deb3f8acb25
3
+ size 6630551906
checkpoint-59000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f4c6a178034c49d38c7def4263c6f02478e86c1f8db3c1f62c33035745d84b4e
3
+ size 15984
checkpoint-59000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:70433d98d10c934494c71f1f488ffe43aa7e0c275dc60caae249855337b5d4de
3
+ size 15984
checkpoint-59000/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:df5763068904eac8e7ef1f59af4d04cd2a14988123460f2474ef4fe828b68449
3
+ size 15997
checkpoint-59000/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5071ddae07971b801f3ea4552f4ca15984612c832bb0eeb4957e6058ab8ca8df
3
+ size 15997
checkpoint-59000/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:87ed9c8099126ef81aa3a9bfa7a3c7b6cd785eff81f268b9a03e623d4b3d4bc2
3
+ size 15997
checkpoint-59000/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:672c8d756893583d47b39e29efe72c7148ec22de2b1d421791d8a90fb6048f2d
3
+ size 15997
checkpoint-59000/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0a334e2fde4b5a5353bf63e29d42fe09848d34f9f1e49e9b40985dfde7666b70
3
+ size 15997
checkpoint-59000/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:027c0559cd5e595df0d3fb0a806a1bb69749fe3d920122bcac74746650507d85
3
+ size 15997
checkpoint-59000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:38632d140f5d7b2a19e46181607ec81fecd2fa0e5a255804f0464704a4785836
3
+ size 15984
checkpoint-59000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:841863faf21da478fb388256e7b3e9bd72f6999d3f3862369a4c472b59a3c17f
3
+ size 15984
checkpoint-59000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1028f219aaf3574660b88ac778a51a2bad8b38fe57781383653c6a7cb9e71071
3
+ size 15984
checkpoint-59000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b896e3a40d07c28dd7c70f05e6414c748f9a48942c8c1eb2c55c96e88a0f33ec
3
+ size 15984
checkpoint-59000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1d823118287a310abe43836b13e707611fc13606c870e037fe0f29074a9a6b3d
3
+ size 15984
checkpoint-59000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e66af39a1d9f90799d5ee66bf7211a201439d72860163bce0c44b7d15e28b6f9
3
+ size 15984
checkpoint-59000/rng_state_8.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6c406df489eb34e544e37b7e8359b03a37fd755db9bbb328b650688562fa4b82
3
+ size 15984
checkpoint-59000/rng_state_9.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:357368f5db7e6cc805b08e410af77843e4300546c11498f41f4f8a364e0a1a0c
3
+ size 15984
checkpoint-59000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:22637f13eec9427f7d55ff3ae6c2f1637419da7810dada29432fc646ff7c12a0
3
+ size 1064
checkpoint-59000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-59000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d14473743eee6e55194ab3222005765bbd12554ccc51d93ccbe97fe4986de8b7
3
+ size 6008
checkpoint-60000/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-60000/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-60000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7095b2cc83fdec7076cae1ed3ec45150c01325c06c0663d5f57eb9ea41276f77
3
+ size 4538452374
checkpoint-60000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:381ca7e18afdef0db9fa42c1e0021cc6c2c4e7f95fc871a3be5f743f7a295fa5
3
+ size 6630551906
checkpoint-60000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7a2d62e706bcf331443f8cdc21fe360895cba5cbab6bfdb132bb9abf2387326a
3
+ size 15984
checkpoint-60000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:86c7f84949d7f72538d287e63f6ddc175462ab2b41a874f1c74c061c537c1c05
3
+ size 15984
checkpoint-60000/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:39eab69a46de29f76073c83b6a6b580ee77d38086437ec1a825dcc93c4c32c0d
3
+ size 15997
checkpoint-60000/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:14e8a874141d73d1fc498350094b11191635b0525729ee8ae5131cc6a339dbbb
3
+ size 15997
checkpoint-60000/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3dff1976f1533d494662d5e26eab4218124730fc7b2be22e1d6899ba3e68220c
3
+ size 15997
checkpoint-60000/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ffbdb60f2eac2d4286cca7ca0961013ba7d3460d8a124d3a8ecd3f1713808bdf
3
+ size 15997
checkpoint-60000/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:864ae24d74b0863b33149939f95a122a2758ec57f5c28fa08fe9cf3c42ec32b2
3
+ size 15997
checkpoint-60000/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:616271d81d23d8d5dcf330326373f4af667b4b5adff4945e68915a6f97eb16a7
3
+ size 15997
checkpoint-60000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bde2d41a16e7fa0cd8e76ec0600585ea99f92926da9f2092d8956371b812c2ae
3
+ size 15984
checkpoint-60000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b5f2914082bc1f7ef254cff695890300d0644df432903cfc18abff6ccb54629a
3
+ size 15984
checkpoint-60000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:107752498ab953bf5360783ee6221110b1130b2754eb3145acc8f8bcbfcd3e52
3
+ size 15984
checkpoint-60000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7d8b8c1996905d5a5e20e49a93f3c9851621f78d7e8a4f1c137bc3f9fdb21ff6
3
+ size 15984
checkpoint-60000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2bd84bfd9fe1b42e34b1ff8eabaa4a9c405b9a0e8f8a4fa3918d19bd479730e3
3
+ size 15984
checkpoint-60000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f9e0b2093b21a275e3ee795cb5982f63727f605c062e7b2b0d41d0c8bb19a39f
3
+ size 15984
checkpoint-60000/rng_state_8.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4cb22837f0b429397a89351375ef51e8423b1b187f086f60b77d819872189747
3
+ size 15984
checkpoint-60000/rng_state_9.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b4ad71b4bcc85b655dc0764d0c943fc11e172d0465a5ff8bd2e709202a88f4fd
3
+ size 15984
checkpoint-60000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:372c00dc306c2e58c8faeebfcb1afeddb696e813703161b02285eb9c7b472469
3
+ size 1064
checkpoint-60000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-60000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d14473743eee6e55194ab3222005765bbd12554ccc51d93ccbe97fe4986de8b7
3
+ size 6008