Upload folder using huggingface_hub
Browse files- .gitattributes +1 -0
- checkpoint-112000/config.json +79 -0
- checkpoint-112000/generation_config.json +6 -0
- checkpoint-112000/model.safetensors +3 -0
- checkpoint-112000/optimizer.pt +3 -0
- checkpoint-112000/rng_state_0.pth +3 -0
- checkpoint-112000/rng_state_1.pth +3 -0
- checkpoint-112000/rng_state_10.pth +3 -0
- checkpoint-112000/rng_state_11.pth +3 -0
- checkpoint-112000/rng_state_12.pth +3 -0
- checkpoint-112000/rng_state_13.pth +3 -0
- checkpoint-112000/rng_state_14.pth +3 -0
- checkpoint-112000/rng_state_15.pth +3 -0
- checkpoint-112000/rng_state_2.pth +3 -0
- checkpoint-112000/rng_state_3.pth +3 -0
- checkpoint-112000/rng_state_4.pth +3 -0
- checkpoint-112000/rng_state_5.pth +3 -0
- checkpoint-112000/rng_state_6.pth +3 -0
- checkpoint-112000/rng_state_7.pth +3 -0
- checkpoint-112000/rng_state_8.pth +3 -0
- checkpoint-112000/rng_state_9.pth +3 -0
- checkpoint-112000/scheduler.pt +3 -0
- checkpoint-112000/trainer_state.json +3 -0
- checkpoint-112000/training_args.bin +3 -0
.gitattributes
CHANGED
|
@@ -52,3 +52,4 @@ checkpoint-108000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
|
|
| 52 |
checkpoint-109000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
|
| 53 |
checkpoint-110000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
|
| 54 |
checkpoint-111000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
|
|
|
|
|
|
| 52 |
checkpoint-109000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
|
| 53 |
checkpoint-110000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
|
| 54 |
checkpoint-111000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
|
| 55 |
+
checkpoint-112000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
|
checkpoint-112000/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-112000/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-112000/model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e46b107104175091b3b4cdc6ad4faa8be5064db2b9af74ac2bbc15a38e943380
|
| 3 |
+
size 4538452374
|
checkpoint-112000/optimizer.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:08303b6245b8bef95b23fc48f625f98ec3cdccfabba74afd09ed0c0f3d421d94
|
| 3 |
+
size 6630551906
|
checkpoint-112000/rng_state_0.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:4faf5160bb2ed754706d3ed37f1fe1cfd9bf611360f63017457f5dbdd41e178f
|
| 3 |
+
size 15984
|
checkpoint-112000/rng_state_1.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:7fe4f0473f981dff8f38af1cb7f8c17ba910281d56e5e530cc305126219b615f
|
| 3 |
+
size 15984
|
checkpoint-112000/rng_state_10.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:74e1666949245f6995770847f1ef9cfc09d8c77cddb38f9e9ac034e195b414e4
|
| 3 |
+
size 15997
|
checkpoint-112000/rng_state_11.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9da169a1356230c996b2d0afa72ba7755f72b90ceba1380068d85454aee9c752
|
| 3 |
+
size 15997
|
checkpoint-112000/rng_state_12.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e2d82269fb9ed69be270d7eb46ec8a332d522db261a4731f2b3ef1b4de9889d9
|
| 3 |
+
size 15997
|
checkpoint-112000/rng_state_13.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f7994f6317ed2eb229dcb1f4b21405c5b7635e452b19fdd885c0b993a6b7a199
|
| 3 |
+
size 15997
|
checkpoint-112000/rng_state_14.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:3fc3637c621c4a77ebba35c623221adb1be61b0e769abad7e2f8fd756670b50a
|
| 3 |
+
size 15997
|
checkpoint-112000/rng_state_15.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:8be3e5abbd24818adbda0a93048d6740dacf66ce9cc7d97e4dace001bbfbab79
|
| 3 |
+
size 15997
|
checkpoint-112000/rng_state_2.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:2513c4c39420030773f09cdad6f53070de061922482dfbb166952aa56ea79e55
|
| 3 |
+
size 15984
|
checkpoint-112000/rng_state_3.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:38113cf8547185d7ffe5e1c721906de493d382348093980b29abdf3eda79f959
|
| 3 |
+
size 15984
|
checkpoint-112000/rng_state_4.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c66c765add5ade73331c401ce949010e0354099ca04e57bcf2909f9f817d03ae
|
| 3 |
+
size 15984
|
checkpoint-112000/rng_state_5.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:bef821d5d2956b2ad7a514f8f0125a336687062e7426d718d8aa7f7b15aec05d
|
| 3 |
+
size 15984
|
checkpoint-112000/rng_state_6.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:ff576b6d408cfcd681cafd88187176ae08897747e3de48b0a0bb875280485d5f
|
| 3 |
+
size 15984
|
checkpoint-112000/rng_state_7.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f7f3b08e328cf3eb7fa3ea55cd2b2f0a63a05b144384ba8fd0b0ebac57c6791b
|
| 3 |
+
size 15984
|
checkpoint-112000/rng_state_8.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d960d1a83c647453f28a550281116c348506fd6e136a83b8bd829ae6da6fe0c0
|
| 3 |
+
size 15984
|
checkpoint-112000/rng_state_9.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:eb8c30f4458ced87868e29c3605a8e7637b1fb2c926e162aa8c8d76946367e4e
|
| 3 |
+
size 15984
|
checkpoint-112000/scheduler.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f1640db2d091822bba5ed32d3be53af870a975ecec77eb05684d1223e1ee8080
|
| 3 |
+
size 1064
|
checkpoint-112000/trainer_state.json
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:639a45ea877e89837704db68f7b9525a8fc354017376cbd8ee277e9b40616f6e
|
| 3 |
+
size 17333467
|
checkpoint-112000/training_args.bin
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:70df7d9f76be9a229b53ed3625af3d9b5db3e9b8cb82c60f170490f4749ef1ac
|
| 3 |
+
size 5944
|