Hugging Face's logo

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

Safetensors
Model card Files
xet
Community
zhiyang1 commited on
Commit
805c0b6
·
verified ·
1 Parent(s): 054e158

Upload folder using huggingface_hub

Browse files
Files changed (46) hide show
  1. checkpoint-37000/config.json +79 -0
  2. checkpoint-37000/generation_config.json +6 -0
  3. checkpoint-37000/model.safetensors +3 -0
  4. checkpoint-37000/optimizer.pt +3 -0
  5. checkpoint-37000/rng_state_0.pth +3 -0
  6. checkpoint-37000/rng_state_1.pth +3 -0
  7. checkpoint-37000/rng_state_10.pth +3 -0
  8. checkpoint-37000/rng_state_11.pth +3 -0
  9. checkpoint-37000/rng_state_12.pth +3 -0
  10. checkpoint-37000/rng_state_13.pth +3 -0
  11. checkpoint-37000/rng_state_14.pth +3 -0
  12. checkpoint-37000/rng_state_15.pth +3 -0
  13. checkpoint-37000/rng_state_2.pth +3 -0
  14. checkpoint-37000/rng_state_3.pth +3 -0
  15. checkpoint-37000/rng_state_4.pth +3 -0
  16. checkpoint-37000/rng_state_5.pth +3 -0
  17. checkpoint-37000/rng_state_6.pth +3 -0
  18. checkpoint-37000/rng_state_7.pth +3 -0
  19. checkpoint-37000/rng_state_8.pth +3 -0
  20. checkpoint-37000/rng_state_9.pth +3 -0
  21. checkpoint-37000/scheduler.pt +3 -0
  22. checkpoint-37000/trainer_state.json +0 -0
  23. checkpoint-37000/training_args.bin +3 -0
  24. checkpoint-38000/config.json +79 -0
  25. checkpoint-38000/generation_config.json +6 -0
  26. checkpoint-38000/model.safetensors +3 -0
  27. checkpoint-38000/optimizer.pt +3 -0
  28. checkpoint-38000/rng_state_0.pth +3 -0
  29. checkpoint-38000/rng_state_1.pth +3 -0
  30. checkpoint-38000/rng_state_10.pth +3 -0
  31. checkpoint-38000/rng_state_11.pth +3 -0
  32. checkpoint-38000/rng_state_12.pth +3 -0
  33. checkpoint-38000/rng_state_13.pth +3 -0
  34. checkpoint-38000/rng_state_14.pth +3 -0
  35. checkpoint-38000/rng_state_15.pth +3 -0
  36. checkpoint-38000/rng_state_2.pth +3 -0
  37. checkpoint-38000/rng_state_3.pth +3 -0
  38. checkpoint-38000/rng_state_4.pth +3 -0
  39. checkpoint-38000/rng_state_5.pth +3 -0
  40. checkpoint-38000/rng_state_6.pth +3 -0
  41. checkpoint-38000/rng_state_7.pth +3 -0
  42. checkpoint-38000/rng_state_8.pth +3 -0
  43. checkpoint-38000/rng_state_9.pth +3 -0
  44. checkpoint-38000/scheduler.pt +3 -0
  45. checkpoint-38000/trainer_state.json +0 -0
  46. checkpoint-38000/training_args.bin +3 -0
checkpoint-37000/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-37000/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-37000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f1cd367581b5cdab248daa182d91454a0e4b7e2d25f0c9a9100922eb7cc32dae
3
+ size 4410723984
checkpoint-37000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7bf8133c16c6223ac58803a48646919e9473dd8193c45e370afd08b2ee53b97f
3
+ size 6330255386
checkpoint-37000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1d29f099e73e1769a847954544e5de26347781ceef2d304fc68a49f5134b94c4
3
+ size 15984
checkpoint-37000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:17b6c2451354b3f05a7cdbb0a2750910e2369eb1b1ac8fa5bb489d21b22e601e
3
+ size 15984
checkpoint-37000/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ff2c5e1189cc17ea3e55cee8634463554944312e0b66f22d90bf627ec1a37c8c
3
+ size 15997
checkpoint-37000/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:db91b11c17700c59dd9390958d3bfa5e6b9edfb2c330c072f786d81ba61359d3
3
+ size 15997
checkpoint-37000/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3b426fa3fa8a43d03157c82564155738c4a8defb9685f8b78a4c975482b61601
3
+ size 15997
checkpoint-37000/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:dfe43cbf1439cc2b25de6043a8a018294c2fec9c8fad3d46ecbf878a546627d4
3
+ size 15997
checkpoint-37000/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:28e951a335b455b2ab256e2cc0cd41058bd620f83d72dba6956e8819303b120d
3
+ size 15997
checkpoint-37000/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d686128e6936d61381e46053292457918ab872aa30171fe7f33ea9fbf05e40b9
3
+ size 15997
checkpoint-37000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:035056048ff3f47173b59eaf840cb8d1b9f2c1b51d1fca84a1571aae78fdabbb
3
+ size 15984
checkpoint-37000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0c1efaf840b594113412a4e3e9da18f0ec4e20fffb8d6b88e83d396bde7b6845
3
+ size 15984
checkpoint-37000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1f24a91165486c236112f1f065d9cdc79920840ad3884b07df3c121a9de7c9bd
3
+ size 15984
checkpoint-37000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b6e5dbd79d29cd8901628004430536ff9499d9e29b1af50252c19180ea237fbc
3
+ size 15984
checkpoint-37000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1851310b514edf2629909a40a70adb2ba53a3292efe51bca935b56789a21ef91
3
+ size 15984
checkpoint-37000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:305366a3ba7788d1db49a3179dc8b961a05bbdc5c1ebe69bde2a7d2d27ae316a
3
+ size 15984
checkpoint-37000/rng_state_8.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:254c57627cd6d9e98d611d79ec1f3e27e812e177804a26f4d2eb240418c9e46b
3
+ size 15984
checkpoint-37000/rng_state_9.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5f2ab4db8f4652f24247597a141183ce390830d89ede85ab20f204a17c8a5b0a
3
+ size 15984
checkpoint-37000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:58fa66c1d0f8aa7323e50560180684f3df163c50f332912f6bd5273db08411b0
3
+ size 1064
checkpoint-37000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-37000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8cb66ea4e0d0158a311c80cec7f365aa6c3617cf37338886f3fb08ba5c2e3f12
3
+ size 5944
checkpoint-38000/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-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 sha256:78b34fb92a1f60b3ca3b2a9e62539f5a0e29bd13421c352354f2f5471ec16c46
3
+ size 4410723984
checkpoint-38000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:369b6dceca61f3edf28daafb6c7e5c220cb37ac1aa695d36f27090452bd9ac3b
3
+ size 6330255386
checkpoint-38000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:09f36494d80da18582fb5ade0aa4d1b6e8b910e4305faef0dc52d93db8d93675
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:eb3817f5e6847b923cced0d67d85999b08d34d411f13a4c9ffa9c2b09ac2146d
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 sha256:811863b353f8c31d86bd3aab484a37363f49c17c550becb8d25d1e911c0f8176
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 sha256:4fde4a09b1861087fc278b749199e5d7e8625b497e435f2a3be8f1a85a62af18
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:de4a99bda7eeb6a57f7d6a32f4188ee8fb828715bdacabc1b5d132c52a265638
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 sha256:3957a7cab2881b2610094be38aa09a0ca79ca6d9a3ca9aa5ce6e9c3205b3e1d9
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 sha256:984b9005976331a422e25181b36a5e9184325fe8d61347740d8c384bfe071d1e
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 sha256:12630f5770a86282fcb6bdbadfb6e6a810e6107c7e17d6731274b3b4653f42d6
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 sha256:0f7b0b96caf6958e9f9492486afcad8850ccbb4e6f5cf0ea90c81ce2b76418b2
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 sha256:46773524a635f9a98b8259faf8e35ec58489604bfe11af8360334723e0c9c314
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 sha256:18a4b30630d51afd4fba2abffd1e9372adf77fa4bf675af84a78d5490879b513
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:ef01d40140f7083d1eeaabce0f58c39aadfcd1427c3d9b3bcf2d304336510da0
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:c48b9347ace626fe862d84dd73e85b40ed06c75a5ce903439820b8fd329fae9c
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 sha256:28c605668faef1bba152aa63c8595a63881d42a37a2ddc4b8dfd8a8414ee6b7f
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 sha256:1c013d5bf2eca8fb6b1bee6c65db2654e1d2c22adcc5482c1f05e24d1391b183
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 sha256:106873fbb5bba8a32cf7fee1b49b1efb4ff505de2c9ba8d0fd1ff2b019c5a6e4
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 sha256:277131efe6248ed8fda169c609c5b4a791b616beb21d222500cd6105692b07b9
3
+ size 5944