Hugging Face's logo

zhiyang1
/
SZ_multinode_vlm_dc-vae-f32c32-768_group-7_no-self-attn-lora

Safetensors
Model card Files
xet
Community
zhiyang1 commited on
Commit
0d5d29a
·
verified ·
1 Parent(s): 0f49ab6

Upload folder using huggingface_hub

Browse files
Files changed (24) hide show
  1. .gitattributes +1 -0
  2. checkpoint-125000/config.json +79 -0
  3. checkpoint-125000/generation_config.json +6 -0
  4. checkpoint-125000/model.safetensors +3 -0
  5. checkpoint-125000/optimizer.pt +3 -0
  6. checkpoint-125000/rng_state_0.pth +3 -0
  7. checkpoint-125000/rng_state_1.pth +3 -0
  8. checkpoint-125000/rng_state_10.pth +3 -0
  9. checkpoint-125000/rng_state_11.pth +3 -0
  10. checkpoint-125000/rng_state_12.pth +3 -0
  11. checkpoint-125000/rng_state_13.pth +3 -0
  12. checkpoint-125000/rng_state_14.pth +3 -0
  13. checkpoint-125000/rng_state_15.pth +3 -0
  14. checkpoint-125000/rng_state_2.pth +3 -0
  15. checkpoint-125000/rng_state_3.pth +3 -0
  16. checkpoint-125000/rng_state_4.pth +3 -0
  17. checkpoint-125000/rng_state_5.pth +3 -0
  18. checkpoint-125000/rng_state_6.pth +3 -0
  19. checkpoint-125000/rng_state_7.pth +3 -0
  20. checkpoint-125000/rng_state_8.pth +3 -0
  21. checkpoint-125000/rng_state_9.pth +3 -0
  22. checkpoint-125000/scheduler.pt +3 -0
  23. checkpoint-125000/trainer_state.json +3 -0
  24. checkpoint-125000/training_args.bin +3 -0
.gitattributes CHANGED
@@ -61,3 +61,4 @@ checkpoint-120000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
61
  checkpoint-122000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
62
  checkpoint-123000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
63
  checkpoint-124000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
 
 
61
  checkpoint-122000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
62
  checkpoint-123000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
63
  checkpoint-124000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
64
+ checkpoint-125000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
checkpoint-125000/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": 2,
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-sana-1.1-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-125000/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-125000/model.safetensors ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:160e17499baa5d966e5ec45b54427153fcf01bb0a65041ac5d3ec9dc4acca9b0
3
+ size 4538452374
checkpoint-125000/optimizer.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a3843114fb01716a3e5088bb6272698448ad410c96e54960ec6c4a7bf12bf15c
3
+ size 6630551906
checkpoint-125000/rng_state_0.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a0354f1a0198422411873df7e9cd888fe1db586b174fdac2583bb953e6f7fec4
3
+ size 15984
checkpoint-125000/rng_state_1.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c08dc39c226ded371412749fb43c899ce19b7e21fa7ce209a4b2dfe54253e45c
3
+ size 15984
checkpoint-125000/rng_state_10.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:104151bd30b0a0e2d3533c5b773807cc66e6072a9888f93c548a88dd508436af
3
+ size 15997
checkpoint-125000/rng_state_11.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:8cb1ff792236a5b5c99e81f8b01063f6af732d2bb07b6a7645e35cf4e50f0f62
3
+ size 15997
checkpoint-125000/rng_state_12.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:10a58ba4c9f967d66cbfba2d3b19d0611d4510c0aeab17856896ecc782110847
3
+ size 15997
checkpoint-125000/rng_state_13.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:343893bf79429b5fd6133f35777c973dcd807527b17e57797df303f3c093ed0a
3
+ size 15997
checkpoint-125000/rng_state_14.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:d74590a0653ec22a2d46399b8dacaa25a9a548b06a8495b423fbc517b5880323
3
+ size 15997
checkpoint-125000/rng_state_15.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:bdd7e892506a01889da0dd2a7d19eb519aa5ea03806ce8c53e6eb2c33e1080a0
3
+ size 15997
checkpoint-125000/rng_state_2.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:13c7e0696d20fd38b89e1971ff0da0cea8021afd4d6ae254b87234fd8a6d67f1
3
+ size 15984
checkpoint-125000/rng_state_3.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:85199885becc9f2a534e4e09931f2822ddcd39c2fd0beb4cf485a5db26aa4ddf
3
+ size 15984
checkpoint-125000/rng_state_4.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:76f610d6024a40f2f74b735cd49e2c0a77e2802d5653810c3e7587c8364e876d
3
+ size 15984
checkpoint-125000/rng_state_5.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:a90a1c5f4d427c7837ac69d8dd11d91b3899e77270cfbd61424c8aba3d406727
3
+ size 15984
checkpoint-125000/rng_state_6.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:0acf3bc10d84f33df4cf156461e2c4cdfedf2bc049ea4f4d565dff04699beda7
3
+ size 15984
checkpoint-125000/rng_state_7.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:93813e61b1489a1a0d2274377a4a65844d00561dceb960455f0bbad05023df32
3
+ size 15984
checkpoint-125000/rng_state_8.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:1a83ba145e1585937c4a2f285dc14c395a81b4f9c6c23eea57a42e1cd4175daa
3
+ size 15984
checkpoint-125000/rng_state_9.pth ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c0bdf65932213b8e22529473892d2bb3fde5d41f41a71981b8aa7c032f505e95
3
+ size 15984
checkpoint-125000/scheduler.pt ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:c756bab95a9572f3a8ffbca2f4930211953eb96adb1b905777cae9b3682000f0
3
+ size 1064
checkpoint-125000/trainer_state.json ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:4978e54c671548bde01e5a29d33e120e41f2bca5afef6d5d2ed1322c442cfbf5
3
+ size 19357013
checkpoint-125000/training_args.bin ADDED
@@ -0,0 +1,3 @@
 
 
 
 
1
+ version https://git-lfs.github.com/spec/v1
2
+ oid sha256:125325f88ee09b414e0114ec48fe561bac0855b70f00588aafbd49db652df2a8
3
+ size 5944