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
721a74f
·
verified ·
1 Parent(s): 92345cb

Upload folder using huggingface_hub

Browse files
Files changed (46) hide show
  1. checkpoint-10000/config.json +79 -0
  2. checkpoint-10000/generation_config.json +6 -0
  3. checkpoint-10000/model.safetensors +3 -0
  4. checkpoint-10000/optimizer.pt +3 -0
  5. checkpoint-10000/rng_state_0.pth +3 -0
  6. checkpoint-10000/rng_state_1.pth +3 -0
  7. checkpoint-10000/rng_state_10.pth +3 -0
  8. checkpoint-10000/rng_state_11.pth +3 -0
  9. checkpoint-10000/rng_state_12.pth +3 -0
  10. checkpoint-10000/rng_state_13.pth +3 -0
  11. checkpoint-10000/rng_state_14.pth +3 -0
  12. checkpoint-10000/rng_state_15.pth +3 -0
  13. checkpoint-10000/rng_state_2.pth +3 -0
  14. checkpoint-10000/rng_state_3.pth +3 -0
  15. checkpoint-10000/rng_state_4.pth +3 -0
  16. checkpoint-10000/rng_state_5.pth +3 -0
  17. checkpoint-10000/rng_state_6.pth +3 -0
  18. checkpoint-10000/rng_state_7.pth +3 -0
  19. checkpoint-10000/rng_state_8.pth +3 -0
  20. checkpoint-10000/rng_state_9.pth +3 -0
  21. checkpoint-10000/scheduler.pt +3 -0
  22. checkpoint-10000/trainer_state.json +0 -0
  23. checkpoint-10000/training_args.bin +3 -0
  24. checkpoint-8000/config.json +79 -0
  25. checkpoint-8000/generation_config.json +6 -0
  26. checkpoint-8000/model.safetensors +3 -0
  27. checkpoint-8000/optimizer.pt +3 -0
  28. checkpoint-8000/rng_state_0.pth +3 -0
  29. checkpoint-8000/rng_state_1.pth +3 -0
  30. checkpoint-8000/rng_state_10.pth +3 -0
  31. checkpoint-8000/rng_state_11.pth +3 -0
  32. checkpoint-8000/rng_state_12.pth +3 -0
  33. checkpoint-8000/rng_state_13.pth +3 -0
  34. checkpoint-8000/rng_state_14.pth +3 -0
  35. checkpoint-8000/rng_state_15.pth +3 -0
  36. checkpoint-8000/rng_state_2.pth +3 -0
  37. checkpoint-8000/rng_state_3.pth +3 -0
  38. checkpoint-8000/rng_state_4.pth +3 -0
  39. checkpoint-8000/rng_state_5.pth +3 -0
  40. checkpoint-8000/rng_state_6.pth +3 -0
  41. checkpoint-8000/rng_state_7.pth +3 -0
  42. checkpoint-8000/rng_state_8.pth +3 -0
  43. checkpoint-8000/rng_state_9.pth +3 -0
  44. checkpoint-8000/scheduler.pt +3 -0
  45. checkpoint-8000/trainer_state.json +0 -0
  46. checkpoint-8000/training_args.bin +3 -0
checkpoint-10000/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-10000/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-10000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c5484e53b12992e94ab467fd08b77a1d73a76dbf1cabd32529d7e60d63f60621
3
+ size 4539487822
checkpoint-10000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0987f2514b16d0a9f0b5d2da8015cea79d56b16d0181a4bb18a697ee99504654
3
+ size 6632345482
checkpoint-10000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bc8870a72b91cb59819327165b35b13319629ccf3522595d63c8f7be0698573a
3
+ size 15984
checkpoint-10000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:720d017e1d779aadbe421faa9bb77acab47ee90bc9b12375a3d1f26c5407353b
3
+ size 15984
checkpoint-10000/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7f1d798f8ab3909263545e20fb58b0b8d8cdae01b2c82db3b30deda69185f4fb
3
+ size 15997
checkpoint-10000/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:02660e386dd509f8f5bc725753e16a272bdc5122c73b8e42ff3fd7e74dff8fd4
3
+ size 15997
checkpoint-10000/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:05b3a342a29be05af31e66809e490cb3c70f24492a1b7b46515b99bc185ca40b
3
+ size 15997
checkpoint-10000/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ab7061606a35198b5cbc72b378e26a19b15f1a54f33108a10b6b34b2cd01f021
3
+ size 15997
checkpoint-10000/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4d5db91371fb1b5d8460abeee680ce988f798d37c073394d632ef4c3a8ba3b30
3
+ size 15997
checkpoint-10000/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0600d3a6516bde80a3978434f6494ebb8e54103aa7b3cb4a176f9d3c8a2f0849
3
+ size 15997
checkpoint-10000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d20e17d151c022814afda24cf16cb8e139e87c3374ea778652e77713868c3705
3
+ size 15984
checkpoint-10000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a203c33d3228dd64c747b51d9d04dbe429f86a28e64b0481e60b48a79220a363
3
+ size 15984
checkpoint-10000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5c2c468db3c4210df285e4a1f2d33b1cd8027479341ec4bc910ed6d304cc009b
3
+ size 15984
checkpoint-10000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f65eec327f7081bee85be2133ee1a3e4f38cd85f9e3d8f7054659a3b54bef4c2
3
+ size 15984
checkpoint-10000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fd1e0809080c2cc48a56daa9b66e5e1fcdddd49476e8caf7d7545ef29247b9e1
3
+ size 15984
checkpoint-10000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:38c1c128d27a990108defff7b99f321676072e44cff5408dafc5fa37c3de6928
3
+ size 15984
checkpoint-10000/rng_state_8.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7b41f9c80c50d18ce91dba088284ce93a40ab444e0b3453a568e7b69cde178c9
3
+ size 15984
checkpoint-10000/rng_state_9.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9f7f5958bf5f5ea685a3b081ac3ab6d27ae589dc2e7907b4eae91802b027a666
3
+ size 15984
checkpoint-10000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:45e1e9caa8d5e0a6b6164550b191ba0b0c93227138124bd184e22ebbea7e7e37
3
+ size 1064
checkpoint-10000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-10000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0f4774c8b9c62660f29613e59b7f68f345c3b171eb7663537271946ac3968e3e
3
+ size 6008
checkpoint-8000/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-8000/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-8000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f2f72f108471cee3ce6e74c04b343ac484f6fcfe1c7a03aaf9af801f892c4962
3
+ size 4539487822
checkpoint-8000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:21d749d71e2c9d690439c6cd75a35e4ccaa288bef7e8803bf9f8e452194f2e7b
3
+ size 6632330442
checkpoint-8000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:425a2323dc4f9dd0c96b6949b01207e075250218aa126dd05b609ef6f69941dd
3
+ size 15984
checkpoint-8000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f71ccc72effc9e9d3304f7717dfcd66b6bd8e0bbcc6670c128bce64f3874b75d
3
+ size 15984
checkpoint-8000/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f25e83f35cfe02e490da1330e193b9b6582717ee46fe09c3963c6e6a6f41bc95
3
+ size 15997
checkpoint-8000/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ecf58c124539eefaa591b4a25d0efa5cc145bb8bb4d50c62a89a4ce4f7366d00
3
+ size 15997
checkpoint-8000/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:133369f3739a1344813b293b428f9f5bc937d466be252548728d6e070947b10d
3
+ size 15997
checkpoint-8000/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:41b18751dd22751bee53632a75729c511998ae18014399f6445e406f0fa66191
3
+ size 15997
checkpoint-8000/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f5bc16740d23aaa22a9b2a73dfa16e86a769d14540791353749f56115812bf48
3
+ size 15997
checkpoint-8000/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d10090973a658f25f99e285bf64cedef137aa576f7978ce682eed5d78611c90a
3
+ size 15997
checkpoint-8000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2aa4dcf33235fef427cf2653e032e956bf90e6830c110cbdf32b714265bd1435
3
+ size 15984
checkpoint-8000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bd9efbee0e2cb5f44ed338414513e22809913eb6bd818e8f6ef84f905190740a
3
+ size 15984
checkpoint-8000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:416e81e6a0525b6ff4943db451c26291366e3c811b0e560e3ec8852c90b1b9bc
3
+ size 15984
checkpoint-8000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2fc2665525cb243f83362757d87741dd569208a2bae36410106deea9c7330985
3
+ size 15984
checkpoint-8000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:efd244d60037b975465ba5814c8d6a4d7faa0dea8c0684fa36a3cf9b8bd30429
3
+ size 15984
checkpoint-8000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:45221dd068f44a76ee5cb569ae6b7e8a3195367b4d2f09283f36c04a651704fd
3
+ size 15984
checkpoint-8000/rng_state_8.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5630dc77dacaa8524f6ead98a723632fec5080e5a5d20f54eade394d9cb6236e
3
+ size 15984
checkpoint-8000/rng_state_9.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:27dfd99ad8ff0881c5c180ee7d3072e9885c8d075afe89ce6dadb1314e8e74a5
3
+ size 15984
checkpoint-8000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2fb1162c23ac67eadc0ff5bceaf1e92fc579d3e568d0261afd39c9a66eea467d
3
+ size 1064
checkpoint-8000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-8000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:920e5ed93f6ab1e0eb452ee346be913d1b38a73645bd466fea0aa2c7fd89f34a
3
+ size 6008