Upload folder using huggingface_hub
Browse files- checkpoint-40000/config.json +79 -0
- checkpoint-40000/generation_config.json +6 -0
- checkpoint-40000/model.safetensors +3 -0
- checkpoint-40000/optimizer.pt +3 -0
- checkpoint-40000/rng_state_0.pth +3 -0
- checkpoint-40000/rng_state_1.pth +3 -0
- checkpoint-40000/rng_state_10.pth +3 -0
- checkpoint-40000/rng_state_11.pth +3 -0
- checkpoint-40000/rng_state_12.pth +3 -0
- checkpoint-40000/rng_state_13.pth +3 -0
- checkpoint-40000/rng_state_14.pth +3 -0
- checkpoint-40000/rng_state_15.pth +3 -0
- checkpoint-40000/rng_state_2.pth +3 -0
- checkpoint-40000/rng_state_3.pth +3 -0
- checkpoint-40000/rng_state_4.pth +3 -0
- checkpoint-40000/rng_state_5.pth +3 -0
- checkpoint-40000/rng_state_6.pth +3 -0
- checkpoint-40000/rng_state_7.pth +3 -0
- checkpoint-40000/rng_state_8.pth +3 -0
- checkpoint-40000/rng_state_9.pth +3 -0
- checkpoint-40000/scheduler.pt +3 -0
- checkpoint-40000/trainer_state.json +0 -0
- checkpoint-40000/training_args.bin +3 -0
checkpoint-40000/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-40000/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-40000/model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9db2fceaa5e3273bc260d38dbeaddd24d1f05860c5ba45c9865485f23c34d87b
|
| 3 |
+
size 4538452374
|
checkpoint-40000/optimizer.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:cffd9040999c1f66b6fba0dc84c0c21de2979b91f3314c531104f663987ab008
|
| 3 |
+
size 6630551906
|
checkpoint-40000/rng_state_0.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d023e1bcb994670e858c1ead5bbac26e275f696a1d93b3c294c0e99342d0a20d
|
| 3 |
+
size 15984
|
checkpoint-40000/rng_state_1.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:094fd881b264612a9a8f5f75651d332a77322654d666c58c51f2d8705fb81aa8
|
| 3 |
+
size 15984
|
checkpoint-40000/rng_state_10.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:aba5b5664d1a0a5a69e22aa2ce5b58d0daf80b0a6bf67122cb52ecbb931eeace
|
| 3 |
+
size 15997
|
checkpoint-40000/rng_state_11.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6c22e6ddc376d21f5ef349d9121e311d93769db8112ae0b6ee3a494315dc6834
|
| 3 |
+
size 15997
|
checkpoint-40000/rng_state_12.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:857fd03768c20300dd92312984954a11839d1ec6f8cae0c07606e4d36a469fbf
|
| 3 |
+
size 15997
|
checkpoint-40000/rng_state_13.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:69c475ef6c703d4c21db3192f3675601e4a4db9fa4c9a8714d03a168261964de
|
| 3 |
+
size 15997
|
checkpoint-40000/rng_state_14.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:dfa8dad177c79c421ad8ca52506140b307b53ed64dd7f49a20fc0b892d98d65b
|
| 3 |
+
size 15997
|
checkpoint-40000/rng_state_15.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:07a3c6f13ccfdc874221e3249f76368f30e49415e9455f33cbd1294d1465bca2
|
| 3 |
+
size 15997
|
checkpoint-40000/rng_state_2.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:4db15fbe56be6ab2f5e261b81bb51d80941aab230ec5535ea36ddb6c7a0950a4
|
| 3 |
+
size 15984
|
checkpoint-40000/rng_state_3.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b1b7002f17220a7665b30a3461ddd733f8ecaacd821b10d47f1dde55cc29757f
|
| 3 |
+
size 15984
|
checkpoint-40000/rng_state_4.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:cc2f92c23d4b9b259e127c6ab7b2b978bec8aa88369a6f40d3efca17c5fe3d88
|
| 3 |
+
size 15984
|
checkpoint-40000/rng_state_5.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6a8fd96e0d17a2944d4ffda8af0ee7fd842f0b7a2d1a7753a47c004561389c73
|
| 3 |
+
size 15984
|
checkpoint-40000/rng_state_6.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:58f4dc2a8747161c26ad12fd1b30f88d146a617a9b6ef2d4c6703a6daedfef57
|
| 3 |
+
size 15984
|
checkpoint-40000/rng_state_7.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:fe9d86501586688c81063794ab09cc9f1943be8a1dadc59ff395631bc309fe51
|
| 3 |
+
size 15984
|
checkpoint-40000/rng_state_8.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1380c6de65344a9b0ddb2b55efa80b81f98d7b9efe96d5e2c111eb4b41ace463
|
| 3 |
+
size 15984
|
checkpoint-40000/rng_state_9.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:c92f286b082b1a8bcaaf8c15aefb527e220be38b5fb888f0bce818971b3b1131
|
| 3 |
+
size 15984
|
checkpoint-40000/scheduler.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:2a6ddc8cb2af1ec91280ca0f9f919ebd0c98af2adfa8a2cbfd6da4e14464b7e1
|
| 3 |
+
size 1064
|
checkpoint-40000/trainer_state.json
ADDED
|
The diff for this file is too large to render.
See raw diff
|
|
|
checkpoint-40000/training_args.bin
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:bce5e60ec03f6f98d07099ab6a6b73499c1e204935550e90d9577897d3d3bf40
|
| 3 |
+
size 5944
|