Hugging Face's logo

zhiyang1
/
SZ_multinode_vlm_dc-ae-f32c32-1.0-768_patch-1_group-7

Safetensors
Model card Files
xet
Community
zhiyang1 commited on
Commit
9c15394
·
verified ·
1 Parent(s): 06c61ac

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
+ "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": 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:f34245096501a61bc9784bf508281ee890e21da5bc73080e7e58baa8e1c039c1
3
+ size 4559801712
checkpoint-29000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c3495f414059f73f5724c960e5793b35fb6936e6576b2d21bf0e43f3b82e43fb
3
+ size 6628781666
checkpoint-29000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c42b85121173a7c382da32d30367cb20c1aa42c52d6e060c8c98c62104f29925
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 sha256:1003ded4a1c010beeb74d89de21da8c99fda0675530fb9456d233a3f4b420499
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 sha256:219422d4a25b98c32765f39dca94e567665785f7010a863b4c550412bc412c8d
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:f6c27f813aede1f963b36d08b2bef7e43365b77054a7c9156de99b8b1ea89fe4
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:f9312d577c2cf821a693b6006b5013fc5770c37e60b9ff9f01235f56ffd87df1
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 sha256:4de4061dcc22034afdd53919593bdc9d8d179ea75894593cbdc0ed9688c3061d
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 sha256:179c93316df0095b01e5e3052f61cf36d1ac655f145d9682f28713ac5c525fbe
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 sha256:819cc17255de0933ffb573a456584a064973cc65d08eca55908f977d47d0efe5
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:f8e1bd07e5adc47b27fd218389475c1ab3f138c98dfbdbf90b3c749bb2556889
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 sha256:4d37f8be58328af4e0d30f1a17abfa4ff1bb8827a7ac492cd1e7dec13f6b33b8
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:f8bee62ccd73ff424e80fdd2dc4e4ba7136c77560c0f7ed8af42ff21a9536324
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 sha256:4b2eb74b530438a73fc7c26ea559b658816b7d2e31a0c981a28f4b6873646dae
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 sha256:1db2c9cc2a7fb4f76d898e5558ffbfdda212b947f2454a198d16477768dc9e07
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 sha256:3845744135551dc64786321c83dcab36cc12014c65a92408a5a5cef79ad39897
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 sha256:82e632656c1ef08629839fbd1a4475c31328af72ae169c4461845e05befe7ead
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 sha256:9b766dbadb2c3efef102fd2b6e3079d10cf2d6455107e89b2a5c1a02b6dd4aad
3
+ size 15984
checkpoint-29000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:862239c6f43180e1ebaba3d02b5a29bf8f0fa3158ab3de127eef40442f930f26
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 sha256:102463cd53c1e00aa7f1ac848ee73a439fb8002d44362b74cb632b89929e7568
3
+ size 6008
checkpoint-30000/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": 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 sha256:91db938f4004f0274b1b021eef938fd05743e9130d730273d29de0f2297b0f21
3
+ size 4559801712
checkpoint-30000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:66fb37ec3f1131cd7df29a0700226f19670bbdc948687ecc9decaaa1becd2652
3
+ size 6628781666
checkpoint-30000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bc0c52e838781a246967b2066c92af103a650e31e825637d3d464367e1c4b267
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:cd765b064e93e38d1edc355177940b5aeb27605b3ad335020aa3cdbc7028fd44
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 sha256:362723cc85306cd3825a556c83493156bd41ef98efdbb3e5197ac048b303593b
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 sha256:3c15940a464f31bab2cd3b4fbe4875aa189070fd94419efd9ca0c7972a2dac09
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 sha256:95766a961dde00d93600f243c0e8b383fd5091c9e7161e57f1b1ce6eed9f189f
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:b7aacc5822e019b046c7705486fc56ac6fe5f3b2a93455462c59215e69aba595
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 sha256:43b46aab6e7087ebd2e8f019a400f65831f8bcf5defac70892e16aed7640798e
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:c4d12cd9b1fd9471df85472eb5f47e931c70f038f491e153e1a033f6ae475c93
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:d306405866f49a6bdbf0c19a97b4865ba3d7b1577827e2de918c60c61f49de1f
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:e3215e000bd34768d9303018ef1a3752c9b0209f4dc6ac8e2ee2317ea5a32120
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 sha256:99868a17ddf901c7c6473802aba2cc67baada00b0e58a099191f92a042cd736b
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 sha256:800ab1b70cc104b2eaceeb987c578a402591f83cc67506a42db3c67c30887e53
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 sha256:09073290728aa5189c3f2a9931777e2ef2f10ddd42208c24cc0f73a2979f5691
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 sha256:3875f1257d01c04c1c435ac112d6eb85df3199e4ec2d27fc3705a8b8b3dbbe63
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:b90f75f6831493a953e8bba459e3073181574671111756b55e8989d5b24217d7
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:a77bd4db5ac8e91973901ed5e533f813af2c51b7019c99e373a2bdaa793d06bf
3
+ size 15984
checkpoint-30000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0959c4b195ce996e434173b2d27c0800e341bddd8ae3a02bbc42759fa4263cc6
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 sha256:102463cd53c1e00aa7f1ac848ee73a439fb8002d44362b74cb632b89929e7568
3
+ size 6008