Hugging Face's logo

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

Safetensors
Model card Files
xet
Community
zhiyang1 commited on
Commit
2a9d5be
·
verified ·
1 Parent(s): 6dd6c0c

Upload folder using huggingface_hub

Browse files
Files changed (46) hide show
  1. checkpoint-23000/config.json +79 -0
  2. checkpoint-23000/generation_config.json +6 -0
  3. checkpoint-23000/model.safetensors +3 -0
  4. checkpoint-23000/optimizer.pt +3 -0
  5. checkpoint-23000/rng_state_0.pth +3 -0
  6. checkpoint-23000/rng_state_1.pth +3 -0
  7. checkpoint-23000/rng_state_10.pth +3 -0
  8. checkpoint-23000/rng_state_11.pth +3 -0
  9. checkpoint-23000/rng_state_12.pth +3 -0
  10. checkpoint-23000/rng_state_13.pth +3 -0
  11. checkpoint-23000/rng_state_14.pth +3 -0
  12. checkpoint-23000/rng_state_15.pth +3 -0
  13. checkpoint-23000/rng_state_2.pth +3 -0
  14. checkpoint-23000/rng_state_3.pth +3 -0
  15. checkpoint-23000/rng_state_4.pth +3 -0
  16. checkpoint-23000/rng_state_5.pth +3 -0
  17. checkpoint-23000/rng_state_6.pth +3 -0
  18. checkpoint-23000/rng_state_7.pth +3 -0
  19. checkpoint-23000/rng_state_8.pth +3 -0
  20. checkpoint-23000/rng_state_9.pth +3 -0
  21. checkpoint-23000/scheduler.pt +3 -0
  22. checkpoint-23000/trainer_state.json +0 -0
  23. checkpoint-23000/training_args.bin +3 -0
  24. checkpoint-24000/config.json +79 -0
  25. checkpoint-24000/generation_config.json +6 -0
  26. checkpoint-24000/model.safetensors +3 -0
  27. checkpoint-24000/optimizer.pt +3 -0
  28. checkpoint-24000/rng_state_0.pth +3 -0
  29. checkpoint-24000/rng_state_1.pth +3 -0
  30. checkpoint-24000/rng_state_10.pth +3 -0
  31. checkpoint-24000/rng_state_11.pth +3 -0
  32. checkpoint-24000/rng_state_12.pth +3 -0
  33. checkpoint-24000/rng_state_13.pth +3 -0
  34. checkpoint-24000/rng_state_14.pth +3 -0
  35. checkpoint-24000/rng_state_15.pth +3 -0
  36. checkpoint-24000/rng_state_2.pth +3 -0
  37. checkpoint-24000/rng_state_3.pth +3 -0
  38. checkpoint-24000/rng_state_4.pth +3 -0
  39. checkpoint-24000/rng_state_5.pth +3 -0
  40. checkpoint-24000/rng_state_6.pth +3 -0
  41. checkpoint-24000/rng_state_7.pth +3 -0
  42. checkpoint-24000/rng_state_8.pth +3 -0
  43. checkpoint-24000/rng_state_9.pth +3 -0
  44. checkpoint-24000/scheduler.pt +3 -0
  45. checkpoint-24000/trainer_state.json +0 -0
  46. checkpoint-24000/training_args.bin +3 -0
checkpoint-23000/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-23000/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-23000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f465db3dd9fa9535307a52ca0957c9e8044b26d17d720192c24786573ec2c4e4
3
+ size 4539487822
checkpoint-23000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4a6af7b17be2e396e1601e40a59399a30d83fbc1122e1c280b89766d521dbcad
3
+ size 6632345482
checkpoint-23000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9d1e061d615d6832037e138c506d57a963c8eab02787ebeabb70e69b70b86fe3
3
+ size 15984
checkpoint-23000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:918eede64ed5f3a6fa871811b2fa2d2eabd4b0017f639b5688fd04202b4cc564
3
+ size 15984
checkpoint-23000/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:050a6ba45b8a455e386ba4a554dbc29137401c48cc502d978051f8e022cd60e3
3
+ size 15997
checkpoint-23000/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1aebfdcc44912b975a74709153b5190a691a18433c6596b0b26d3a3408606804
3
+ size 15997
checkpoint-23000/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ce09142415aa153bf130946c54e71abf09f3acd6f1a2e48b24dfcdbe8e4b3b8f
3
+ size 15997
checkpoint-23000/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:aa4401bf6336aa30fe2552d03182c7c383b0b99b1da8580a6285dcba0b6fee8b
3
+ size 15997
checkpoint-23000/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:81f96ec3335ad7dd4040599901c6f8dfd1dbbf62cc2fa709dfd72ed533828e32
3
+ size 15997
checkpoint-23000/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b784069cd18f84117ecca180c4154e45c12aeaf924cddd324f395efbcf7cad24
3
+ size 15997
checkpoint-23000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7ec8de4a342ad075608a25ca1c382ced851cc62c73a33ab804e14187359317e8
3
+ size 15984
checkpoint-23000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1f1889a3d40209dac537254b9bfb0ae0c48d03c973d8f12c4db71449645f2c15
3
+ size 15984
checkpoint-23000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8aafb622344ef14a96509a6579fbddc843c6cf400a3c02e4fb461dd77cd7e867
3
+ size 15984
checkpoint-23000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0566f7d4aed5953afe40e60306575d8bdc4d8f22284be49a5e6c96a903557cce
3
+ size 15984
checkpoint-23000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f20e248836718ea41635e19f44515c44f4e338f9240546ece53e88f1a63e1359
3
+ size 15984
checkpoint-23000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:148533d6ab0a0085b5187f0d6ea1aaf4cc29753db923614f95ee8c438df2aa63
3
+ size 15984
checkpoint-23000/rng_state_8.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3f2f7394be2be2ba512dd11a0271c16136fc94335485d262e25ddd27e6fe10e0
3
+ size 15984
checkpoint-23000/rng_state_9.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:770fee3e3f97cba7ecf240864910daf5bb215ad621f3cd5b330f0d14ffe53da8
3
+ size 15984
checkpoint-23000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:71bb3071a7eb693cbc1a7384b9e95b3b5edd1471482347f2ed059c4ee6cb6c20
3
+ size 1064
checkpoint-23000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-23000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0f6c94967502fdc8e21bceaa12d058ab60a58bfce3b107b4e1a746a756541d34
3
+ size 6008
checkpoint-24000/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-24000/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-24000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c11d21d9d400b5f4e19bbd9c58a0f6b55b6e3ca96f6513f2fc5f535d92da1def
3
+ size 4539487822
checkpoint-24000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:05833efbe69e844555b9926945fbed0f733f8c82bfde123da2fc2778eefdb2d3
3
+ size 6632345482
checkpoint-24000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bc08f3fd7bfe69739e37a81a20d63bf11d80b4199fe54713a0e9b895c9bddf5f
3
+ size 15984
checkpoint-24000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:394ea25912c50c71086ec5206b4be332bfb4e150e0d6b8b00bd0442067bb4b59
3
+ size 15984
checkpoint-24000/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:739f60b73187d47ea3fd8733565892a3ff4a5c7ebd24226a083ba107cb26b40b
3
+ size 15997
checkpoint-24000/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:27f7bb40f35f2a723e62e270143b1838820461f3b34035ab3e4e516dd34df1f4
3
+ size 15997
checkpoint-24000/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c2d10b0a2253ec489ca0e7577170a7166c3695320e961ad8a37163f7cfc0920d
3
+ size 15997
checkpoint-24000/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f48a329ce380bcff1bc5a40ee54359b6bae9bbab7d38df5bec6edfc024e4fb72
3
+ size 15997
checkpoint-24000/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:200effcaa26dcf674863d8ffbfe598a9d95c8e6e399b18d90901bac25c66ee58
3
+ size 15997
checkpoint-24000/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:750520099105729bb8bdec286021ccc05230970ad00bf2268bacff9e21b35dab
3
+ size 15997
checkpoint-24000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0a599267ac77ecd81694a736d7819211d5dfef157f550fc6e015a6a4eee9239b
3
+ size 15984
checkpoint-24000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b857dc78bd71e979ea83df6d4d33e4fcc4023b8763fa56871a5cb4604790ff36
3
+ size 15984
checkpoint-24000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e84f5599de5a8551dd7a79b962818b0d6005daa21f5dbdbb50b2cecca0891b0a
3
+ size 15984
checkpoint-24000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:254be8495a49cf549a3ed63eccbc5c88168bd7bdbe2bf5cfceea8e6e9cad1fc2
3
+ size 15984
checkpoint-24000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:130975a3b5bb8bf5cac04b6e1269bf3170a796031a7fd2cb3592f2841e9877af
3
+ size 15984
checkpoint-24000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3b2191d106af93abfe2c7f9a03ff4919d47df9edd4faa24c5888ede8540857d5
3
+ size 15984
checkpoint-24000/rng_state_8.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c39b5c5ba159758da16182da07ee46bef2479ca19d555d3bf79f45b636757804
3
+ size 15984
checkpoint-24000/rng_state_9.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:968472e2701bce789bdc7171cee2f394bb13cc829c0be2f4e46c699f9a9d392a
3
+ size 15984
checkpoint-24000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9b2d044fecc747319e5e895ba410137a50bf1f796597ff0a282dcdc29549b036
3
+ size 1064
checkpoint-24000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-24000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7851ed44acd1c3a21a49b619cbce79aa203ef0caaa4fea19146d2eda46b7d550
3
+ size 6008