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
b1931a6
·
verified ·
1 Parent(s): 0cd228d

Upload folder using huggingface_hub

Browse files
Files changed (46) hide show
  1. checkpoint-38000/config.json +79 -0
  2. checkpoint-38000/generation_config.json +6 -0
  3. checkpoint-38000/model.safetensors +3 -0
  4. checkpoint-38000/optimizer.pt +3 -0
  5. checkpoint-38000/rng_state_0.pth +3 -0
  6. checkpoint-38000/rng_state_1.pth +3 -0
  7. checkpoint-38000/rng_state_10.pth +3 -0
  8. checkpoint-38000/rng_state_11.pth +3 -0
  9. checkpoint-38000/rng_state_12.pth +3 -0
  10. checkpoint-38000/rng_state_13.pth +3 -0
  11. checkpoint-38000/rng_state_14.pth +3 -0
  12. checkpoint-38000/rng_state_15.pth +3 -0
  13. checkpoint-38000/rng_state_2.pth +3 -0
  14. checkpoint-38000/rng_state_3.pth +3 -0
  15. checkpoint-38000/rng_state_4.pth +3 -0
  16. checkpoint-38000/rng_state_5.pth +3 -0
  17. checkpoint-38000/rng_state_6.pth +3 -0
  18. checkpoint-38000/rng_state_7.pth +3 -0
  19. checkpoint-38000/rng_state_8.pth +3 -0
  20. checkpoint-38000/rng_state_9.pth +3 -0
  21. checkpoint-38000/scheduler.pt +3 -0
  22. checkpoint-38000/trainer_state.json +0 -0
  23. checkpoint-38000/training_args.bin +3 -0
  24. checkpoint-39000/config.json +79 -0
  25. checkpoint-39000/generation_config.json +6 -0
  26. checkpoint-39000/model.safetensors +3 -0
  27. checkpoint-39000/optimizer.pt +3 -0
  28. checkpoint-39000/rng_state_0.pth +3 -0
  29. checkpoint-39000/rng_state_1.pth +3 -0
  30. checkpoint-39000/rng_state_10.pth +3 -0
  31. checkpoint-39000/rng_state_11.pth +3 -0
  32. checkpoint-39000/rng_state_12.pth +3 -0
  33. checkpoint-39000/rng_state_13.pth +3 -0
  34. checkpoint-39000/rng_state_14.pth +3 -0
  35. checkpoint-39000/rng_state_15.pth +3 -0
  36. checkpoint-39000/rng_state_2.pth +3 -0
  37. checkpoint-39000/rng_state_3.pth +3 -0
  38. checkpoint-39000/rng_state_4.pth +3 -0
  39. checkpoint-39000/rng_state_5.pth +3 -0
  40. checkpoint-39000/rng_state_6.pth +3 -0
  41. checkpoint-39000/rng_state_7.pth +3 -0
  42. checkpoint-39000/rng_state_8.pth +3 -0
  43. checkpoint-39000/rng_state_9.pth +3 -0
  44. checkpoint-39000/scheduler.pt +3 -0
  45. checkpoint-39000/trainer_state.json +0 -0
  46. checkpoint-39000/training_args.bin +3 -0
checkpoint-38000/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-38000/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-38000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8d807cf3a791b14ee361e0f37178bb82ae6341ce7d04d44e5cf2b8e85caf8952
3
+ size 4538452374
checkpoint-38000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d0671b1d2689fae022019ad21cabb9f9c5fff04a54381a3fe55a2b4519ce4683
3
+ size 6630551906
checkpoint-38000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7bcac203003b1ed4d7bb695d0ab4a122beb03876cd3e3ac3fc6568c768e6233e
3
+ size 15984
checkpoint-38000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d0c328f4c4558ccd6bf5c406f5a863fa92f4d162aea1320be49fc7a13020765a
3
+ size 15984
checkpoint-38000/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1b9527395627abd75a9e3c1ff2271365ef7226434aca91c05563a18e8e4d44ac
3
+ size 15997
checkpoint-38000/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bd3c9c777973e53ffdf3e64c286f03f684d34374c256244a358f8aef9e7ee157
3
+ size 15997
checkpoint-38000/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0d8a88f7835168e6a9a08fb129c4f78724e701a31edee7501c8685169bff5e82
3
+ size 15997
checkpoint-38000/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4e2e9f06a96cbe98108bd0ed1a712f6b8af14f21936d74314e269c33f60ff848
3
+ size 15997
checkpoint-38000/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:25f6fdcd5a56895a4b2349d62a32efe492d81697157fa7a3d5bb7976b257512a
3
+ size 15997
checkpoint-38000/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fe844ebd5ede52ba7710ebf302e7d8bb64a5688d5d5cd3b57ff2c1283395dc7b
3
+ size 15997
checkpoint-38000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c532e3dac9451dcc79b8d3378bb6312f21ad9935f7c8e0dc05f3d4f32a11d00b
3
+ size 15984
checkpoint-38000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:94adddf43e3b80edd81a5a453dc87495fc50c1e10571aca1b68cecf25b5a99ed
3
+ size 15984
checkpoint-38000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:06a1acf2a2fdb16ca6e8b9b482832855b13fdca11fe0adabf7ccf443d7fdf834
3
+ size 15984
checkpoint-38000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:80eb7a1e3ef01d3b63da628745f4f40070caa44295e62f48c63f09b0c55d6157
3
+ size 15984
checkpoint-38000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e08dfbccb2d9195c82d829060942c413a97825f78657dabb6f31e83211598ea4
3
+ size 15984
checkpoint-38000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:72babff89c4d8060194c379bba3b97be1f5926e8d3f755ddeb94f8ad3c6dd61c
3
+ size 15984
checkpoint-38000/rng_state_8.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:748172bb3454c72768ce5ced1ccfb3ee638e23b32fb7d6c10d55eca7b52ecd60
3
+ size 15984
checkpoint-38000/rng_state_9.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cb615af9b8784646c155f5768fc0a8a7522abfac5d7427002033a5a8bb69ce97
3
+ size 15984
checkpoint-38000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f8365b7b873509a6c7cf9b5cc77ed98a004c4d46913094aa05bc9f136ea89c62
3
+ size 1064
checkpoint-38000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-38000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b23422be8986d084cba979977490ea1dba17ccb5378da21bc7967d0f52fcfee6
3
+ size 6008
checkpoint-39000/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-39000/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-39000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:29dc23d3fefc05d88ab34301b10c69c229f34d716823225c8131e356e2ecc694
3
+ size 4538452374
checkpoint-39000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:759fc5aac98702ca09b38bdaaff7f8a51ff9f0963896c850c425defad629dbf9
3
+ size 6630551906
checkpoint-39000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f823dbde5a0ff512e8efc53bf5bde90893548dfe373da88718e48e71225ebf01
3
+ size 15984
checkpoint-39000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6f099c37d0e336163be30797cd017cd9b99c8414c1fd15de3e0b9b857f3ce0e1
3
+ size 15984
checkpoint-39000/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ee0cb9387677a99969d0e1517a6727395186c3c9bdde674955cc2db565e68ae8
3
+ size 15997
checkpoint-39000/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0e518e6c7a64b2c0add4d9eec5d687345f5350c3ec54fc68cff8468868890b02
3
+ size 15997
checkpoint-39000/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:329f7d020b11d0793048ed8d634993def0663e8c226f2235bbe825a7a1e2d6bf
3
+ size 15997
checkpoint-39000/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3cf9146efcb5190895464e01a9c68a5254c907648f8daedfa9d6258b73abcff3
3
+ size 15997
checkpoint-39000/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6400ce59139c68f9278ce3cc00798db8abe471cdb7e86e82c45db2e25e1b9990
3
+ size 15997
checkpoint-39000/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:af705ceb6a92b3859a252787d547238e8ea526c173f4bdb1a78a60edaadc6f58
3
+ size 15997
checkpoint-39000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f4b2449b88303be59e91617903f5c7152f6ace02e2593fd4f8df5c68995c753f
3
+ size 15984
checkpoint-39000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5c4b319a6c721545083a6f0ad1f44eb597cc0ed4cc1e7294a16f55b50a535222
3
+ size 15984
checkpoint-39000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0d0c225bcdfa33957dc1df7ce4b49ece47286d4eeec2bf60fff365d4b37091ce
3
+ size 15984
checkpoint-39000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:32ac7f75e5b85061d4304f9f8c1feb9b9887a3064dfd8d8df7271c4dd83499a1
3
+ size 15984
checkpoint-39000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:35f7c8fd60e230da1260164c85ddd7dbc9f75064c6d31bc25d873b8d32f4a6d9
3
+ size 15984
checkpoint-39000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:59bc46776ef943af7478a85fb4fbc15a5260f4a07f7024f20e3d2b5fc9e7458f
3
+ size 15984
checkpoint-39000/rng_state_8.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:038d2b56fe0bf1f8e48bbbef1b5ec40d177a2799e147703e278e9ee2cbd033b6
3
+ size 15984
checkpoint-39000/rng_state_9.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e5731ec7dfb29437efdecc86dcee2d75f83c801813fd2c96218c871526cb517d
3
+ size 15984
checkpoint-39000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:51db50218a1516f400252b4633addd27930e9db8194e29d2f3267385fb573eee
3
+ size 1064
checkpoint-39000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-39000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9b97687156d332aaefd1ecce1b964b5b1ca2e6ad5330969fad97468aada44ade
3
+ size 5944