Upload folder using huggingface_hub
Browse files- checkpoint-10000/config.json +77 -0
- checkpoint-10000/generation_config.json +6 -0
- checkpoint-10000/model-00001-of-00002.safetensors +3 -0
- checkpoint-10000/model-00002-of-00002.safetensors +3 -0
- checkpoint-10000/model.safetensors.index.json +0 -0
- checkpoint-10000/optimizer.pt +3 -0
- checkpoint-10000/rng_state_0.pth +3 -0
- checkpoint-10000/rng_state_1.pth +3 -0
- checkpoint-10000/rng_state_2.pth +3 -0
- checkpoint-10000/rng_state_3.pth +3 -0
- checkpoint-10000/scheduler.pt +3 -0
- checkpoint-10000/trainer_state.json +0 -0
- checkpoint-10000/training_args.bin +3 -0
checkpoint-10000/config.json
ADDED
@@ -0,0 +1,77 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
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": 6,
|
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": 512,
|
23 |
+
"initializer_range": 0.02,
|
24 |
+
"inject_img_diffuser": false,
|
25 |
+
"input_size": 64,
|
26 |
+
"intermediate_size": 8960,
|
27 |
+
"layer_group_size": 7,
|
28 |
+
"layerwise_start_idx": 0,
|
29 |
+
"lora_alpha": 256,
|
30 |
+
"lora_bias": "none",
|
31 |
+
"lora_dropout": 0.05,
|
32 |
+
"lora_enable": false,
|
33 |
+
"lora_r": 128,
|
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_layer": 4,
|
44 |
+
"rms_norm_eps": 1e-06,
|
45 |
+
"rope_scaling": {
|
46 |
+
"mrope_section": [
|
47 |
+
16,
|
48 |
+
24,
|
49 |
+
24
|
50 |
+
],
|
51 |
+
"rope_type": "default",
|
52 |
+
"type": "default"
|
53 |
+
},
|
54 |
+
"rope_theta": 1000000.0,
|
55 |
+
"sample_size": 128,
|
56 |
+
"sampling_steps": 28,
|
57 |
+
"sliding_window": null,
|
58 |
+
"tie_word_embeddings": true,
|
59 |
+
"torch_dtype": "bfloat16",
|
60 |
+
"transformers_version": "4.47.0",
|
61 |
+
"use_cache": true,
|
62 |
+
"use_repa": false,
|
63 |
+
"use_residual_attn": true,
|
64 |
+
"use_sliding_window": false,
|
65 |
+
"vae_path": "mit-han-lab/dc-ae-f128c512-mix-1.0-diffusers",
|
66 |
+
"video_token_id": 151656,
|
67 |
+
"vision_config": {
|
68 |
+
"hidden_size": 1536,
|
69 |
+
"in_chans": 3,
|
70 |
+
"model_type": "qwen2_vl",
|
71 |
+
"spatial_patch_size": 14
|
72 |
+
},
|
73 |
+
"vision_end_token_id": 151653,
|
74 |
+
"vision_start_token_id": 151652,
|
75 |
+
"vision_token_id": 151654,
|
76 |
+
"vocab_size": 151936
|
77 |
+
}
|
checkpoint-10000/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-10000/model-00001-of-00002.safetensors
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:39f84ba39f0f85910e074dc0ccb4081131a4915430e70b1cb97392c40b9b2eb3
|
3 |
+
size 4952991712
|
checkpoint-10000/model-00002-of-00002.safetensors
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:3b1b0e0a51eef0f669df60602ef26789b1be93d8d37e125ee2ef8eb0e259773a
|
3 |
+
size 1530007672
|
checkpoint-10000/model.safetensors.index.json
ADDED
The diff for this file is too large to render.
See raw diff
|
|
checkpoint-10000/optimizer.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:93db33bd66a5df034c198c3f3093518293b521f4e2bd494ca26fa48fd736b660
|
3 |
+
size 6933060634
|
checkpoint-10000/rng_state_0.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:049bb34a78196ec7c0efe030d497e46314a50da5768c9996b6df39d41824587d
|
3 |
+
size 14960
|
checkpoint-10000/rng_state_1.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:fc7e88a5ae03f9fdaad2209ed2e696fa4b550d993dd3e5fa60cfc1d38e19d732
|
3 |
+
size 14960
|
checkpoint-10000/rng_state_2.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:b4c9cce86c5910d4c38aab1ed89ee6dcd5a62e38f49c8cecc2b8994eb2c18d05
|
3 |
+
size 14960
|
checkpoint-10000/rng_state_3.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:8f2c6d9905d9f87dd4904a6c7b639872c686d583e367401cb8ceb58be8cc7854
|
3 |
+
size 14960
|
checkpoint-10000/scheduler.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:c1a2ebdf221fb8b90a9f3123f8ab542370743f5c880ae73b57834885b2112820
|
3 |
+
size 1064
|
checkpoint-10000/trainer_state.json
ADDED
The diff for this file is too large to render.
See raw diff
|
|
checkpoint-10000/training_args.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:fea4d2237550ac02fab09f592114431bcac00eac1150b45379b7aa134ef4566d
|
3 |
+
size 5880
|