Hugging Face's logo

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

Safetensors
Model card Files
xet
Community
zhiyang1 commited on
Commit
492c56f
·
verified ·
1 Parent(s): 311e7fa

Upload folder using huggingface_hub

Browse files
Files changed (46) hide show
  1. checkpoint-59000/config.json +79 -0
  2. checkpoint-59000/generation_config.json +6 -0
  3. checkpoint-59000/model.safetensors +3 -0
  4. checkpoint-59000/optimizer.pt +3 -0
  5. checkpoint-59000/rng_state_0.pth +3 -0
  6. checkpoint-59000/rng_state_1.pth +3 -0
  7. checkpoint-59000/rng_state_10.pth +3 -0
  8. checkpoint-59000/rng_state_11.pth +3 -0
  9. checkpoint-59000/rng_state_12.pth +3 -0
  10. checkpoint-59000/rng_state_13.pth +3 -0
  11. checkpoint-59000/rng_state_14.pth +3 -0
  12. checkpoint-59000/rng_state_15.pth +3 -0
  13. checkpoint-59000/rng_state_2.pth +3 -0
  14. checkpoint-59000/rng_state_3.pth +3 -0
  15. checkpoint-59000/rng_state_4.pth +3 -0
  16. checkpoint-59000/rng_state_5.pth +3 -0
  17. checkpoint-59000/rng_state_6.pth +3 -0
  18. checkpoint-59000/rng_state_7.pth +3 -0
  19. checkpoint-59000/rng_state_8.pth +3 -0
  20. checkpoint-59000/rng_state_9.pth +3 -0
  21. checkpoint-59000/scheduler.pt +3 -0
  22. checkpoint-59000/trainer_state.json +0 -0
  23. checkpoint-59000/training_args.bin +3 -0
  24. checkpoint-60000/config.json +79 -0
  25. checkpoint-60000/generation_config.json +6 -0
  26. checkpoint-60000/model.safetensors +3 -0
  27. checkpoint-60000/optimizer.pt +3 -0
  28. checkpoint-60000/rng_state_0.pth +3 -0
  29. checkpoint-60000/rng_state_1.pth +3 -0
  30. checkpoint-60000/rng_state_10.pth +3 -0
  31. checkpoint-60000/rng_state_11.pth +3 -0
  32. checkpoint-60000/rng_state_12.pth +3 -0
  33. checkpoint-60000/rng_state_13.pth +3 -0
  34. checkpoint-60000/rng_state_14.pth +3 -0
  35. checkpoint-60000/rng_state_15.pth +3 -0
  36. checkpoint-60000/rng_state_2.pth +3 -0
  37. checkpoint-60000/rng_state_3.pth +3 -0
  38. checkpoint-60000/rng_state_4.pth +3 -0
  39. checkpoint-60000/rng_state_5.pth +3 -0
  40. checkpoint-60000/rng_state_6.pth +3 -0
  41. checkpoint-60000/rng_state_7.pth +3 -0
  42. checkpoint-60000/rng_state_8.pth +3 -0
  43. checkpoint-60000/rng_state_9.pth +3 -0
  44. checkpoint-60000/scheduler.pt +3 -0
  45. checkpoint-60000/trainer_state.json +0 -0
  46. checkpoint-60000/training_args.bin +3 -0
checkpoint-59000/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-59000/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-59000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5e6d53c4b7ebf7e4f0403beb225d0c82c645454db01e03e060f4a7ddc8aa3994
3
+ size 4410723984
checkpoint-59000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e0b65f6361749f205f1ec614a75cad35f279404afc4f2f19e7afefd47901b1e1
3
+ size 6330255386
checkpoint-59000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:442ccd3889a2bc596983d82a812036ba30bf5a8959a4a64b6d0c5b3725066d84
3
+ size 15984
checkpoint-59000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:db6abdd8a59066c214ed9b06897bf574c3b3d5a4ea61bc3653007a449cc09b6e
3
+ size 15984
checkpoint-59000/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:07c6f4508fed71b6b124cb771d2d0355643fb460e04df32d3327283697917c73
3
+ size 15997
checkpoint-59000/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3f5c3bc8b08e0acea5cc9d47c53ef3be2dfb4883856e29f34ef8a7e703287fbc
3
+ size 15997
checkpoint-59000/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0c8ca347ccb9354636a7c05da2403722b351b46ee2755534ec3d6b7fe5f31ac0
3
+ size 15997
checkpoint-59000/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7a3ed6c72f0d3fcff3df14c1d4e01129d7d7e111fb949b92847da27a6bf8832e
3
+ size 15997
checkpoint-59000/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f2adc420227e6739eaf3299d7247d08872d1d978d941a04b9ea6f28c05d871da
3
+ size 15997
checkpoint-59000/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:03947911e8897d1b418f204991027b6c3f6db59f632b33744535b3c613623d1f
3
+ size 15997
checkpoint-59000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7de1573069f03054d0694ca2ad799678e9d2ffc4bb394a03779b91af79a1eefe
3
+ size 15984
checkpoint-59000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:929b042647c516ff452f275ab78210dd1d51aa897b283492881149d8d63ed5c4
3
+ size 15984
checkpoint-59000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:00946040331af7194af0f7920b8e383afdae276f135ed1b15262b449f6fe946e
3
+ size 15984
checkpoint-59000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:02d342a1ce9b467c760230a41b5e5096d23c107467d06fe6658109f553191ffc
3
+ size 15984
checkpoint-59000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:865ca723175bd38e36d7d5a9687a3a1869b64dae58f7f24dc64aacea4a2f79c2
3
+ size 15984
checkpoint-59000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:931a9831763b393b0c003a34ec5dea804101687b50490399e23927ae0cb86ed8
3
+ size 15984
checkpoint-59000/rng_state_8.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7835c6b10d642710cbcb573c3feed3e6b1623acb4c5a4b5e26bf6dd2bb2951e3
3
+ size 15984
checkpoint-59000/rng_state_9.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:31ae4b9fbda4d375a3eaf1b3d366bbab93b6706cf06af3313433a35d8505e4c7
3
+ size 15984
checkpoint-59000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:22637f13eec9427f7d55ff3ae6c2f1637419da7810dada29432fc646ff7c12a0
3
+ size 1064
checkpoint-59000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-59000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fb1da560e638b234d61bbff9b9e4a72adcc427150ca807e3f89ec06d6624b6cf
3
+ size 5944
checkpoint-60000/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-60000/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-60000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:f1b2266d3d5fa40c44ea945214c11490733d7dfa2fc29b80ab7c26c02fe77c1e
3
+ size 4410723984
checkpoint-60000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:12ec5f59daa569e01da29a696ae9838cd7c356fa1bdbdab28709693309514b90
3
+ size 6330255386
checkpoint-60000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9598d3d064b76989b3b096f4fa268cdcf2c6accc41e8de9fcfd13033c7f82a1d
3
+ size 15984
checkpoint-60000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0a4cb569ea4024f99309d403fdb90756880ce0b671338609f04a79385ce455a6
3
+ size 15984
checkpoint-60000/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:06a1d0bb30a60ab7394d27b1c5d276ebb8760cca3d738bb90ea149b01fbf2354
3
+ size 15997
checkpoint-60000/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:000f09ef36cbb5672af67322fb31d24a5e0280a24688bc0e5cb69ebf5b9b646d
3
+ size 15997
checkpoint-60000/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d24a56d19c723401d26f86aff5d38dddc69949042f024b4296297bbcab94c998
3
+ size 15997
checkpoint-60000/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fef7d8d692ba2bb76815672d64fa05502f70c4d2232b6d96576327e1658258c1
3
+ size 15997
checkpoint-60000/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:36ecd1206e214fcd40ef3c7a595fb8c338430518181f272f8ad09d8f41dc295c
3
+ size 15997
checkpoint-60000/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9b955633e747dd1cb52d680bb2f03611f1fc747329e647309b43edaa38ab2128
3
+ size 15997
checkpoint-60000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:def8e2e91fb46f921c093cf3e3a401101e1b1ee9919c07a4ceb13b492cc1f6a2
3
+ size 15984
checkpoint-60000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cf6787320a4aa2c63883b4a557e4e9a08bc68280746d539f432977f901ba37c1
3
+ size 15984
checkpoint-60000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:14c610af99e9788dabceb117e2ac8286502212459a664d84846bff48d2baf084
3
+ size 15984
checkpoint-60000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c6b0e7c060607bfe366f5de97cef87cdcd135b77a6000f297d3ee5eb3299110d
3
+ size 15984
checkpoint-60000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:86bd88bd6500ddde691b65fc833bad3a07f29e94232146648273950cc8453c24
3
+ size 15984
checkpoint-60000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ceb2e8ab97d59344d7521d663e821f70193c5b0294b352b4a28371cdd4e1735e
3
+ size 15984
checkpoint-60000/rng_state_8.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fa921de68db3d27a46f80d68d6cd9ff721d8f708bf6c70a48bddc8ee046da988
3
+ size 15984
checkpoint-60000/rng_state_9.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:9a806b3501ce63c6824c5118207f31e5d7db19f041cc6bbe1d97928754c0a18b
3
+ size 15984
checkpoint-60000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:372c00dc306c2e58c8faeebfcb1afeddb696e813703161b02285eb9c7b472469
3
+ size 1064
checkpoint-60000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-60000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:fb1da560e638b234d61bbff9b9e4a72adcc427150ca807e3f89ec06d6624b6cf
3
+ size 5944