Hugging Face's logo

zhiyang1
/
SZ_multinode_dc-ae-f32c32-1.0-768_baseline

Safetensors
Model card Files
xet
Community
zhiyang1 commited on
Commit
053b309
·
verified ·
1 Parent(s): f3441e9

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-9000/config.json +79 -0
  25. checkpoint-9000/generation_config.json +6 -0
  26. checkpoint-9000/model.safetensors +3 -0
  27. checkpoint-9000/optimizer.pt +3 -0
  28. checkpoint-9000/rng_state_0.pth +3 -0
  29. checkpoint-9000/rng_state_1.pth +3 -0
  30. checkpoint-9000/rng_state_10.pth +3 -0
  31. checkpoint-9000/rng_state_11.pth +3 -0
  32. checkpoint-9000/rng_state_12.pth +3 -0
  33. checkpoint-9000/rng_state_13.pth +3 -0
  34. checkpoint-9000/rng_state_14.pth +3 -0
  35. checkpoint-9000/rng_state_15.pth +3 -0
  36. checkpoint-9000/rng_state_2.pth +3 -0
  37. checkpoint-9000/rng_state_3.pth +3 -0
  38. checkpoint-9000/rng_state_4.pth +3 -0
  39. checkpoint-9000/rng_state_5.pth +3 -0
  40. checkpoint-9000/rng_state_6.pth +3 -0
  41. checkpoint-9000/rng_state_7.pth +3 -0
  42. checkpoint-9000/rng_state_8.pth +3 -0
  43. checkpoint-9000/rng_state_9.pth +3 -0
  44. checkpoint-9000/scheduler.pt +3 -0
  45. checkpoint-9000/trainer_state.json +0 -0
  46. checkpoint-9000/training_args.bin +3 -0
checkpoint-10000/config.json ADDED
@@ -0,0 +1,79 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "ar_steps": 1,
3
+ "architectures": [
4
+ "DiffVLMBaseline"
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": 1,
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-in-1.0-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:e5ce1c9d3fd5fd42ecaf4b41cdf2abb360d8e95713098e52394c0cd1137d18c6
3
+ size 4410723984
checkpoint-10000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3dab96c74c923eeb58f920e10408cd5a5041b1f93546294c2a460790361ea510
3
+ size 6330248922
checkpoint-10000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:998791b8eb46c207c5220d4bda3e8826a467d8e7514984872ba70855f39e65b7
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 sha256:83a27aa5b1d186061c0bc38c0d7a52e0edee969e17e7d6b989d279f765805d82
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 sha256:9bf688eab3063e93d439b5ff94eae92cb5a7d34f7954f227dcc5d26230655a90
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 sha256:5b9cee80bfa96ba505fff9000f301bfa9940352154f2bbc36512a4ccb95b7827
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:ff32ddfde5316ecfdbc893499cef2afb821b960352dbb0878c0d04f9861fc374
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 sha256:38b7ec595ecf052f92879b03e4b416c7bba8f1ccefd45553e4837fe8f6575a48
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 sha256:4783e40b4e6a971e40e2e509edcb8b8006c64cd53ac3289c2879f455adcc2618
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 sha256:1137e90486f07469572ab4ae6e664dc19d76321d4ef066e92ee6f1e9ece458d6
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 sha256:1e51bdfd3a2005ba5b858a86f18daf909000854f0ea68a0908035c196f2d2d6e
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 sha256:0fcfd3544f7d7784ffe3eb79102c902dd06eac7ea25c40be6f287582e6b5321b
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 sha256:0be27f972520269103e28d3a6cf2a253f8f3387205265edb3094207373628dd9
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:ef5536a805ef8fedc8ffc2473241f9c1b31c7de9dbfb8df143b2523e08032be9
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:fc45e8028f06bff14b810abc1bae0e3eba7754369a681a235e7d296068ff712e
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 sha256:5681416f5bd3243cd76d1f48fa8ba26ad848d1ea8eb0569232c9b51f5fb4311e
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 sha256:27f8da779f7cd1c2cfcb26833885c83ac645ab487418d1a30b8230db1b7e30be
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 sha256:40863cb164ae5c5df6ff81d7e440c0f121740a35af3d4c7106adefd0f1209d3f
3
+ size 15984
checkpoint-10000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8cecba43d7459ec5ce9f816c2e7f8c9a6c0974f0e469d7bde6bc33484cab04b5
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 sha256:522879493545089489589ec6ed17886b1183c5e595b82676e5ea483690ac40c9
3
+ size 5944
checkpoint-9000/config.json ADDED
@@ -0,0 +1,79 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ {
2
+ "ar_steps": 1,
3
+ "architectures": [
4
+ "DiffVLMBaseline"
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": 1,
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-in-1.0-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-9000/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-9000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:26b7c850d76b87e3a8a4bcc0f9efe037cb0b6d644854173588031cbe177b99cc
3
+ size 4410723984
checkpoint-9000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8d2a033ff30661c7d295e3b281177be12139d3850759114048d9d8b575d9f070
3
+ size 6330248922
checkpoint-9000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:200121c47a1b42ef7e7df3f6c17c65352ca592d2c129b31f001d2838c8b86021
3
+ size 15984
checkpoint-9000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4be634d2588f117757395b8067f4ad348a831d8cd4caf5f44259a2dd8afe93eb
3
+ size 15984
checkpoint-9000/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:838c6d0ba4a5bf3e75d10d4202dbb25986197921539264e993fce4bb4c233326
3
+ size 15997
checkpoint-9000/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f10d9fea8071ea56ecd7587c30a3e756c8dc4f83cf989daf51789d3cca9a84c7
3
+ size 15997
checkpoint-9000/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0e4b875e9653372c0cb54cb9c3d5b274f4ba98589286b819e2ca61153cc82832
3
+ size 15997
checkpoint-9000/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:93b1f283203568c11e29757b84cdf432ec212ab0f5ffab7a7c04cf2eaab38863
3
+ size 15997
checkpoint-9000/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:19e99be3370107711987258bc3f04dc8e9d0ca068c36169d790694d72fd1c101
3
+ size 15997
checkpoint-9000/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f5a0b88385325398e3f5f327732ee0aae036ff4178f61a2a658cc67ce566ce59
3
+ size 15997
checkpoint-9000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:529b83fd721de9f071b74891e55d4f81b6e8361ab8d1b0e6544bc17318f5ec54
3
+ size 15984
checkpoint-9000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4c4778f532286c90c9e448c321987217bc80def32119ab031d48fd412b66d7a7
3
+ size 15984
checkpoint-9000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:61d34e50691892944dca6f5fc68e8e5cd0c17d0cb98bd16ba05b5412e6259ee5
3
+ size 15984
checkpoint-9000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8567072be9b8c310c6547fd0f7a7153ead53fd05d0af97864fd9aed71fdadce1
3
+ size 15984
checkpoint-9000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3d2a29529a7674f08dda04f52ffa6d94cf4b3af4715c4159e1024a736c51b1ce
3
+ size 15984
checkpoint-9000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:dbbf9f355daae8fb31e7d4b7b9ac0990ad8fc0317e51ec6bcfb53ba33e1933ee
3
+ size 15984
checkpoint-9000/rng_state_8.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d7827875ff21a681e8da73d25dc992afaacfea390351095fa2b2b20eb387610e
3
+ size 15984
checkpoint-9000/rng_state_9.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c83db7e9cc64e02b6816cf29ccc85df203ae5466c79eacab8ff7bae0ccd77b68
3
+ size 15984
checkpoint-9000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e71621b7afdbc391d9e5784b68966c636091af8a60dceeb1c8486a8fbac2cd89
3
+ size 1064
checkpoint-9000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-9000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:522879493545089489589ec6ed17886b1183c5e595b82676e5ea483690ac40c9
3
+ size 5944