Upload folder using huggingface_hub
Browse files- .gitattributes +1 -0
- checkpoint-110000/config.json +79 -0
- checkpoint-110000/generation_config.json +6 -0
- checkpoint-110000/model.safetensors +3 -0
- checkpoint-110000/optimizer.pt +3 -0
- checkpoint-110000/rng_state_0.pth +3 -0
- checkpoint-110000/rng_state_1.pth +3 -0
- checkpoint-110000/rng_state_10.pth +3 -0
- checkpoint-110000/rng_state_11.pth +3 -0
- checkpoint-110000/rng_state_12.pth +3 -0
- checkpoint-110000/rng_state_13.pth +3 -0
- checkpoint-110000/rng_state_14.pth +3 -0
- checkpoint-110000/rng_state_15.pth +3 -0
- checkpoint-110000/rng_state_2.pth +3 -0
- checkpoint-110000/rng_state_3.pth +3 -0
- checkpoint-110000/rng_state_4.pth +3 -0
- checkpoint-110000/rng_state_5.pth +3 -0
- checkpoint-110000/rng_state_6.pth +3 -0
- checkpoint-110000/rng_state_7.pth +3 -0
- checkpoint-110000/rng_state_8.pth +3 -0
- checkpoint-110000/rng_state_9.pth +3 -0
- checkpoint-110000/scheduler.pt +3 -0
- checkpoint-110000/trainer_state.json +3 -0
- checkpoint-110000/training_args.bin +3 -0
.gitattributes
CHANGED
|
@@ -50,3 +50,4 @@ checkpoint-99000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
|
|
| 50 |
checkpoint-101000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
|
| 51 |
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
|
|
|
|
|
|
| 50 |
checkpoint-101000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
|
| 51 |
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
|
checkpoint-110000/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": true,
|
| 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-110000/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-110000/model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:eb065cd951a50b8765f59613afb2ea8c49c664dfbdecf4dcf60e1bdb0679dab0
|
| 3 |
+
size 4539487822
|
checkpoint-110000/optimizer.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:2481999781bd00cbed2712271a6b4c3478b9c7f4b89991a11571bafbbcb658a0
|
| 3 |
+
size 6632345482
|
checkpoint-110000/rng_state_0.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:2d932cf2804d1b6534a39d121086e8cc05445456479e0afc3a57b76971f1be95
|
| 3 |
+
size 15984
|
checkpoint-110000/rng_state_1.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:659c330f72a282864697a7c3d3fe9d6453d53f1c69bfd0d85761c514363a31b5
|
| 3 |
+
size 15984
|
checkpoint-110000/rng_state_10.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:0fdb12cf1330d2428bedd9286a62744fef0756547643644a9d5f208f9c1c38cf
|
| 3 |
+
size 15997
|
checkpoint-110000/rng_state_11.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f26432be5084feb14253fd08d7d4ce7b1cbbb2deb03bcb52bdf8d2512d3f0fb0
|
| 3 |
+
size 15997
|
checkpoint-110000/rng_state_12.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:da924c093604aae800bd4d24afac36b1b3bce0b03a06da54b879a24b66a22ade
|
| 3 |
+
size 15997
|
checkpoint-110000/rng_state_13.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:827a96a445246b90f1bd5b70fba0b059a72a6a6c2f9d88b8b46744948cdcc286
|
| 3 |
+
size 15997
|
checkpoint-110000/rng_state_14.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:81959d3fe812f0614599e5d9f69b483822388e818d12de94a081febe6b81e209
|
| 3 |
+
size 15997
|
checkpoint-110000/rng_state_15.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:2727b8d8ab37eb693a90ba1f6f82e49e49960d70bf5f79ead9b22becc2f64cad
|
| 3 |
+
size 15997
|
checkpoint-110000/rng_state_2.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9c014c2ef1abaa44de7987baeaed1b1037b6d4b7284671d89c971e0d2a70d909
|
| 3 |
+
size 15984
|
checkpoint-110000/rng_state_3.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:4243d6b642bf27a2b53f1b37e46cb9a4f6fd7ccc688f2de06c8456593af4d67b
|
| 3 |
+
size 15984
|
checkpoint-110000/rng_state_4.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:308ff8dbfafc3681a03689b3fe0bd409d64172d1283381f8b810abdef8208c4a
|
| 3 |
+
size 15984
|
checkpoint-110000/rng_state_5.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:6fdfde12a1c1fe79af3a5ca5a84ce3057405cd3d89dd118c3b56f0a847153f91
|
| 3 |
+
size 15984
|
checkpoint-110000/rng_state_6.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e592453318b76180724070a9efd0927e19d8cd7b10eed1f8f40f7a7f8591fb4d
|
| 3 |
+
size 15984
|
checkpoint-110000/rng_state_7.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:fdad7b6807c4da2262547907f2b741c8d09e77e4ee2bb78baad4438f1fe0a853
|
| 3 |
+
size 15984
|
checkpoint-110000/rng_state_8.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b5027ceaf22fc80b5d400cee635b9b6381ba5ad1b195a512b174326e669e340a
|
| 3 |
+
size 15984
|
checkpoint-110000/rng_state_9.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d51ea7594f072e307faa56cbd1f396ee61d0e48e9a551283a9452e7322b5ec5c
|
| 3 |
+
size 15984
|
checkpoint-110000/scheduler.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:9c4f0b2439dd7fa4ae2aceb98ade9f541d95d2f057f7ce4f443e7cd4efa5cd45
|
| 3 |
+
size 1064
|
checkpoint-110000/trainer_state.json
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:22b5832e07ae0a61103eb1b1a8e84e4e95c2c21e929040c5210e34e2df4e9016
|
| 3 |
+
size 16980403
|
checkpoint-110000/training_args.bin
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:dc7bfc04a705bee1f1b062e2451f2b3f42aab3395d1c2f42da0e09219053db01
|
| 3 |
+
size 6008
|