Hugging Face's logo

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

Safetensors
Model card Files
xet
Community
zhiyang1 commited on
Commit
054e158
·
verified ·
1 Parent(s): 1119d2a

Upload folder using huggingface_hub

Browse files
Files changed (46) hide show
  1. checkpoint-29000/config.json +79 -0
  2. checkpoint-29000/generation_config.json +6 -0
  3. checkpoint-29000/model.safetensors +3 -0
  4. checkpoint-29000/optimizer.pt +3 -0
  5. checkpoint-29000/rng_state_0.pth +3 -0
  6. checkpoint-29000/rng_state_1.pth +3 -0
  7. checkpoint-29000/rng_state_10.pth +3 -0
  8. checkpoint-29000/rng_state_11.pth +3 -0
  9. checkpoint-29000/rng_state_12.pth +3 -0
  10. checkpoint-29000/rng_state_13.pth +3 -0
  11. checkpoint-29000/rng_state_14.pth +3 -0
  12. checkpoint-29000/rng_state_15.pth +3 -0
  13. checkpoint-29000/rng_state_2.pth +3 -0
  14. checkpoint-29000/rng_state_3.pth +3 -0
  15. checkpoint-29000/rng_state_4.pth +3 -0
  16. checkpoint-29000/rng_state_5.pth +3 -0
  17. checkpoint-29000/rng_state_6.pth +3 -0
  18. checkpoint-29000/rng_state_7.pth +3 -0
  19. checkpoint-29000/rng_state_8.pth +3 -0
  20. checkpoint-29000/rng_state_9.pth +3 -0
  21. checkpoint-29000/scheduler.pt +3 -0
  22. checkpoint-29000/trainer_state.json +0 -0
  23. checkpoint-29000/training_args.bin +3 -0
  24. checkpoint-30000/config.json +79 -0
  25. checkpoint-30000/generation_config.json +6 -0
  26. checkpoint-30000/model.safetensors +3 -0
  27. checkpoint-30000/optimizer.pt +3 -0
  28. checkpoint-30000/rng_state_0.pth +3 -0
  29. checkpoint-30000/rng_state_1.pth +3 -0
  30. checkpoint-30000/rng_state_10.pth +3 -0
  31. checkpoint-30000/rng_state_11.pth +3 -0
  32. checkpoint-30000/rng_state_12.pth +3 -0
  33. checkpoint-30000/rng_state_13.pth +3 -0
  34. checkpoint-30000/rng_state_14.pth +3 -0
  35. checkpoint-30000/rng_state_15.pth +3 -0
  36. checkpoint-30000/rng_state_2.pth +3 -0
  37. checkpoint-30000/rng_state_3.pth +3 -0
  38. checkpoint-30000/rng_state_4.pth +3 -0
  39. checkpoint-30000/rng_state_5.pth +3 -0
  40. checkpoint-30000/rng_state_6.pth +3 -0
  41. checkpoint-30000/rng_state_7.pth +3 -0
  42. checkpoint-30000/rng_state_8.pth +3 -0
  43. checkpoint-30000/rng_state_9.pth +3 -0
  44. checkpoint-30000/scheduler.pt +3 -0
  45. checkpoint-30000/trainer_state.json +0 -0
  46. checkpoint-30000/training_args.bin +3 -0
checkpoint-29000/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-29000/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-29000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d88fc2b939e0bb71c34bf577438abeea0e5985324bc676e8955e4e2eeb7727df
3
+ size 4410723984
checkpoint-29000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5edac9bed071a84183628b346dc77442e4800b3668d0d3d664517d4e652822f0
3
+ size 6330255386
checkpoint-29000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:221de0e4f7367e99e3901ea2b2d6eb8c366f8f1fe7b8d72acd8bbecba20f3ade
3
+ size 15984
checkpoint-29000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c34426cb89fbc9e06c8fa62ffbf24afdfe65f1ffe6be058d4726b036b47e6545
3
+ size 15984
checkpoint-29000/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:b82388fb71a33d149b56c3c8242e1f89e27c53d34e3c4518f9b17f5582431c31
3
+ size 15997
checkpoint-29000/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:24e40d999abb6fa027f5906431ac7794251569b58b05a0aeec57ba810dba7ae9
3
+ size 15997
checkpoint-29000/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1bee4ec2f760c505114570c58c20307330769aa6079b8d8e433982760d3b04d0
3
+ size 15997
checkpoint-29000/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:053ffc151f39a4ccecd9c5f7bf24358b398d88727c69c31b09bafcf1c175044d
3
+ size 15997
checkpoint-29000/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8a19d5fd186341785e6474a73533c415bc17c79a2e81e4d7205a559569f57788
3
+ size 15997
checkpoint-29000/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:cdb8cfd5f3d2473d160b0b36c7bb41d1129c50dc9adb8a5f190331ef3a588782
3
+ size 15997
checkpoint-29000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:974183016f21e2cc5bd7bb2fd90da6ce82cef5d1dbc452a397bde09b8a9594d9
3
+ size 15984
checkpoint-29000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c9bb87754ac3171957485b64a3cb298ef9d542856f1b6cbe259d13a471666e0a
3
+ size 15984
checkpoint-29000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:95f8993afcc59e674cb8fcde343781fe1c276d28e1b2f144dbbc738b7ec56d5b
3
+ size 15984
checkpoint-29000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bdbddae73a2f1335619bf877b7833dbcee741195682f40b885f6feeeea4b0a45
3
+ size 15984
checkpoint-29000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5003201f8324896ae0cca53b56679822dbc4f00e9776a649378d3de75201cea3
3
+ size 15984
checkpoint-29000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d04b144a623b24eda4342a48b8a63769c2209f8270e9669ffd6522729329ed29
3
+ size 15984
checkpoint-29000/rng_state_8.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1addf29929e891b9166f6517b152d189a64133cb873d177cee0c8906853aba9e
3
+ size 15984
checkpoint-29000/rng_state_9.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:e83b406a21eb549019ef3ef47726e8b7464b63399c5b35a6389e65d489ca60b1
3
+ size 15984
checkpoint-29000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:17963727d317dfb508460ed1089817ef3f12b052c232b268d68775fd37f294c6
3
+ size 1064
checkpoint-29000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-29000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3dc59fcd4493a1c683d9af10a54a1e2187dbec994357536ac5eacce2c086abfe
3
+ size 5944
checkpoint-30000/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-30000/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-30000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4baad1179b9e45fb17d7572c6bbf98d5ce4ea12ea1c7a6959e03d50a27ff9ec6
3
+ size 4410723984
checkpoint-30000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d904ca73a1571190071ca94737b0a41b15175fa430c394c420c7dd0fbea6a0c4
3
+ size 6330255386
checkpoint-30000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3d9e19de943150c34cd0e7c7c2cb42c191373f3dc7298d2e2ce654bc2d488176
3
+ size 15984
checkpoint-30000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:699c3fa90c4caa6f47b9adc38d881406b9eb7e71185a431c7f39e53221809e53
3
+ size 15984
checkpoint-30000/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d7cb14f44aab11a7244d2e80edd807ea3a3b70ffa3f54561c14bfb7d53da8cb9
3
+ size 15997
checkpoint-30000/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4d9246669b176c2e5a5ae16e9d9212630f0b61305ab79eafa5699352ed45b215
3
+ size 15997
checkpoint-30000/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8585db53af0bf40dd956aa1610dee6ee710309f20f8320e78631b88f0fe34cab
3
+ size 15997
checkpoint-30000/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:997f2356548ab6bf7817663cd1132e7b390d531e3056b9a0663103dd597c1e88
3
+ size 15997
checkpoint-30000/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:2884b62cbfeffaa30a60dd4f3ec2796b8c9e6c71dc0d498fec719ef024773195
3
+ size 15997
checkpoint-30000/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:32132e56bd81caf6a97c3888e50683525b050829e6d228556e9a8cb2e714fc5a
3
+ size 15997
checkpoint-30000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:697e9eaf1fe680b8bccff053e4598f6a90562611b21522ec1db55ce92b12e37a
3
+ size 15984
checkpoint-30000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6ec14328bd926e7439c71cb89446dc432beeb90f8e72a05c93e8f87478a98295
3
+ size 15984
checkpoint-30000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:5d0b59e8b72da2ac1ba4d4948029872d48f65b0f8a8d0e5ca98706b8f32ead40
3
+ size 15984
checkpoint-30000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4f38d4f3ac69caf58b85b6f4234cff399960a72716bb941d6f10601124feefe1
3
+ size 15984
checkpoint-30000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:7c97e18a5b9b88ae396bc6fbfac2afdb045e31b9ad77c8c5930925b31c2b85fd
3
+ size 15984
checkpoint-30000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ecb96f0e6ee345edf498e58c596a5aaa35e03306872de827d6bb036b59772bbb
3
+ size 15984
checkpoint-30000/rng_state_8.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d4e4544e73ac5243c00abe9b0e5cf399f8496cafb297eb9bb9f01c4ba828ed87
3
+ size 15984
checkpoint-30000/rng_state_9.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:aefced54e32451d58ae4a5719106c1c8c6a836db8e9c745d5f00f721e1d566a2
3
+ size 15984
checkpoint-30000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:10251bcfbd6cb1085d5e9df908000ba794cb50dfd74fc6b90c9196b38832fa8f
3
+ size 1064
checkpoint-30000/trainer_state.json ADDED
The diff for this file is too large to render. See raw diff
 
checkpoint-30000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:3dc59fcd4493a1c683d9af10a54a1e2187dbec994357536ac5eacce2c086abfe
3
+ size 5944