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
a421bb4
·
verified ·
1 Parent(s): e36ff06

Upload folder using huggingface_hub

Browse files
Files changed (46) hide show
  1. checkpoint-54000/config.json +79 -0
  2. checkpoint-54000/generation_config.json +6 -0
  3. checkpoint-54000/model.safetensors +3 -0
  4. checkpoint-54000/optimizer.pt +3 -0
  5. checkpoint-54000/rng_state_0.pth +3 -0
  6. checkpoint-54000/rng_state_1.pth +3 -0
  7. checkpoint-54000/rng_state_10.pth +3 -0
  8. checkpoint-54000/rng_state_11.pth +3 -0
  9. checkpoint-54000/rng_state_12.pth +3 -0
  10. checkpoint-54000/rng_state_13.pth +3 -0
  11. checkpoint-54000/rng_state_14.pth +3 -0
  12. checkpoint-54000/rng_state_15.pth +3 -0
  13. checkpoint-54000/rng_state_2.pth +3 -0
  14. checkpoint-54000/rng_state_3.pth +3 -0
  15. checkpoint-54000/rng_state_4.pth +3 -0
  16. checkpoint-54000/rng_state_5.pth +3 -0
  17. checkpoint-54000/rng_state_6.pth +3 -0
  18. checkpoint-54000/rng_state_7.pth +3 -0
  19. checkpoint-54000/rng_state_8.pth +3 -0
  20. checkpoint-54000/rng_state_9.pth +3 -0
  21. checkpoint-54000/scheduler.pt +3 -0
  22. checkpoint-54000/trainer_state.json +0 -0
  23. checkpoint-54000/training_args.bin +3 -0
  24. checkpoint-56000/config.json +79 -0
  25. checkpoint-56000/generation_config.json +6 -0
  26. checkpoint-56000/model.safetensors +3 -0
  27. checkpoint-56000/optimizer.pt +3 -0
  28. checkpoint-56000/rng_state_0.pth +3 -0
  29. checkpoint-56000/rng_state_1.pth +3 -0
  30. checkpoint-56000/rng_state_10.pth +3 -0
  31. checkpoint-56000/rng_state_11.pth +3 -0
  32. checkpoint-56000/rng_state_12.pth +3 -0
  33. checkpoint-56000/rng_state_13.pth +3 -0
  34. checkpoint-56000/rng_state_14.pth +3 -0
  35. checkpoint-56000/rng_state_15.pth +3 -0
  36. checkpoint-56000/rng_state_2.pth +3 -0
  37. checkpoint-56000/rng_state_3.pth +3 -0
  38. checkpoint-56000/rng_state_4.pth +3 -0
  39. checkpoint-56000/rng_state_5.pth +3 -0
  40. checkpoint-56000/rng_state_6.pth +3 -0
  41. checkpoint-56000/rng_state_7.pth +3 -0
  42. checkpoint-56000/rng_state_8.pth +3 -0
  43. checkpoint-56000/rng_state_9.pth +3 -0
  44. checkpoint-56000/scheduler.pt +3 -0
  45. checkpoint-56000/trainer_state.json +0 -0
  46. checkpoint-56000/training_args.bin +3 -0
checkpoint-54000/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-54000/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-54000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:05c579fc48c9f68694168003ae43f47e86e471a71f5f9205df2601ac9cb55bc5
3
+ size 4539487822
checkpoint-54000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c6279c682c10dff5e2eb4b7a988d9d2b1936eff9a2ea3cdf86ddf8a28c2cd249
3
+ size 6632345482
checkpoint-54000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7d3a5bbed8d84e8044b3d01f138d6c3c090e189983f0fc1f40670bee370f1ed2
3
+ size 15984
checkpoint-54000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4e61a34bd66ceac0abac518d6a1c76437f03357f759eac7de5e425acdd7faed8
3
+ size 15984
checkpoint-54000/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:12bfe95bab32f6e6c985d698cec55d06f338aa388956ff35fd156c514a2ac019
3
+ size 15997
checkpoint-54000/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9b68235b80a587f2311169212c9926b74775b60c0c16da9160a8063584cbd769
3
+ size 15997
checkpoint-54000/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fde5b8fe6aacd37ba3032bc5a165c9dc1da3fcb025355793e3681705f8233af5
3
+ size 15997
checkpoint-54000/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9fd554b71070056476dd9a589ecd93ca1806c93d857b1a5cfd58a3c077a3aba6
3
+ size 15997
checkpoint-54000/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:76b28504751c5725be41732d6ba409ba59ffbc8d4f4cfaba258718a7f7af03e7
3
+ size 15997
checkpoint-54000/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ce5d280ef6dd41dd9c888290f8310f09c5783d84f26847999b73d4c85f8c9942
3
+ size 15997
checkpoint-54000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8a45ec62d38b98b991161c08e5f5e05aa815bf5671d8e26579ec12cebeefeaa2
3
+ size 15984
checkpoint-54000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:87c8706c3e112103229813650b47f1c23cd7440f1a633996e070e6dcae0b17db
3
+ size 15984
checkpoint-54000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:28d44c57fba1520d8dcc103322145fbde9c327269e667e10a36f2d4727aac302
3
+ size 15984
checkpoint-54000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:715bb5d2e2676ee4607684484399274997cb3339302f1d635d4b524766fc06d1
3
+ size 15984
checkpoint-54000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a1b18948948785760833b011a3f6c4f12bb2eeb1960996864cee9496c68baadd
3
+ size 15984
checkpoint-54000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c67a653987eeeb02525745ed0db8088312c819a564c7bd65147c431ae4fe63aa
3
+ size 15984
checkpoint-54000/rng_state_8.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8597e21e28e4fdfd5e432c59942563d7b52612fabef2459b21a6ed6eba9aa20a
3
+ size 15984
checkpoint-54000/rng_state_9.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:05e5ec1d9fda3a934df1b264d8e60ba41038117f5c6a88d47101732a23b60e2d
3
+ size 15984
checkpoint-54000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e6aebc3bb5c108540a2853191acf62c549320d9066d440094c7133f164a83ffd
3
+ size 1064
checkpoint-54000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-54000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cddf6aa1e019b6cb76c4feec899aca5b127f86adc727f0978decf2c3331de1c5
3
+ size 6008
checkpoint-56000/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-56000/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-56000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:98fbe143f9d593dd6e7f5aabeaeef57390d3144742676e575bf0615298c56e3c
3
+ size 4539487822
checkpoint-56000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:29d8aa03e7a8b879af3b1e4b29c6bbf5f13d1d40c31d9c75444ed9cfaf23bdbd
3
+ size 6632345482
checkpoint-56000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ffec09bb78829c4acffaac0a8fc4cb418dd523d047adbbeb14fc885016f8564f
3
+ size 15984
checkpoint-56000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c7163cddf7e2947c68f49eac8455d3507f9afa2bbb40a932320eb3d28b1f3503
3
+ size 15984
checkpoint-56000/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:aaadadddcace53bde44675dd456540c9ba98e0f0157386b3c9a3186c911c046a
3
+ size 15997
checkpoint-56000/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:269ba3971bfa0ca2f0f183b63e5c128e494860fe7b47d43a88f39c2d5f1c31a2
3
+ size 15997
checkpoint-56000/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e99242af3f58dc365682e4424c33ff45c5bdeb1c80374b555e2ac7c000de7d4e
3
+ size 15997
checkpoint-56000/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9442475df6e8cbe421af94033faab5cdccc271989df47a8d339ce07ee70c1c29
3
+ size 15997
checkpoint-56000/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0abf8885297761e39909c7e2640c782b7b262c61e6537893e84bc871f5cd8040
3
+ size 15997
checkpoint-56000/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9f7d54357f746f12a5f7221c5a0864153b5cb4562ac94296c03cde21195769b0
3
+ size 15997
checkpoint-56000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:325cb9b4a67b20f6e530749f7dbcc8e76460c98b0af05af8b388af67dcbda07a
3
+ size 15984
checkpoint-56000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9cdcb543d0802f45f920754d457efa42d2cb6174e2cd32305ae019f55bbc2a28
3
+ size 15984
checkpoint-56000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5fc1b5206490726babbd34c187b31bd82d256e57ef9e01d107f4eff9d9c05213
3
+ size 15984
checkpoint-56000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:33786dfac8f2fb74ee7597f3aebc48f91434c8b38a79a360ed26672caf759b70
3
+ size 15984
checkpoint-56000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cf8d360b877926bddc1be11f484a4f4dda8cf7f0be06618c454be601f420f57a
3
+ size 15984
checkpoint-56000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:28905654e302bd83cd7c3db99804ce53cb43837ca6e9501a5643aa815f7b1754
3
+ size 15984
checkpoint-56000/rng_state_8.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fea4dc0775367760a2672fe71283563aa5e8fc3438873abdec416fbbcc3e3510
3
+ size 15984
checkpoint-56000/rng_state_9.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7dadd59c2e06e3369203344fed79e5ded772221d3f5d9097287b7bbbe13c5949
3
+ size 15984
checkpoint-56000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ba4f33232eebcc1e79f822b2dde4537e011cb3134ce17d90d2885a430e71f68c
3
+ size 1064
checkpoint-56000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-56000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bde04bf3cdff2d41f8698e3699f67bcd79fd8473a265eefb05e79bc74507b63d
3
+ size 6008