Upload folder using huggingface_hub
Browse filesThis view is limited to 50 files because it contains too many changes.
See raw diff
- checkpoint-19250/config.json +79 -0
- checkpoint-19250/generation_config.json +6 -0
- checkpoint-19250/model-00001-of-00002.safetensors +3 -0
- checkpoint-19250/model-00002-of-00002.safetensors +3 -0
- checkpoint-19250/model.safetensors.index.json +0 -0
- checkpoint-19250/optimizer.pt +3 -0
- checkpoint-19250/rng_state_0.pth +3 -0
- checkpoint-19250/rng_state_1.pth +3 -0
- checkpoint-19250/rng_state_2.pth +3 -0
- checkpoint-19250/rng_state_3.pth +3 -0
- checkpoint-19250/rng_state_4.pth +3 -0
- checkpoint-19250/rng_state_5.pth +3 -0
- checkpoint-19250/rng_state_6.pth +3 -0
- checkpoint-19250/rng_state_7.pth +3 -0
- checkpoint-19250/scheduler.pt +3 -0
- checkpoint-19250/trainer_state.json +0 -0
- checkpoint-19250/training_args.bin +3 -0
- checkpoint-19500/config.json +79 -0
- checkpoint-19500/generation_config.json +6 -0
- checkpoint-19500/model-00001-of-00002.safetensors +3 -0
- checkpoint-19500/model-00002-of-00002.safetensors +3 -0
- checkpoint-19500/model.safetensors.index.json +0 -0
- checkpoint-19500/optimizer.pt +3 -0
- checkpoint-19500/rng_state_0.pth +3 -0
- checkpoint-19500/rng_state_1.pth +3 -0
- checkpoint-19500/rng_state_2.pth +3 -0
- checkpoint-19500/rng_state_3.pth +3 -0
- checkpoint-19500/rng_state_4.pth +3 -0
- checkpoint-19500/rng_state_5.pth +3 -0
- checkpoint-19500/rng_state_6.pth +3 -0
- checkpoint-19500/rng_state_7.pth +3 -0
- checkpoint-19500/scheduler.pt +3 -0
- checkpoint-19500/trainer_state.json +0 -0
- checkpoint-19500/training_args.bin +3 -0
- checkpoint-19750/config.json +79 -0
- checkpoint-19750/generation_config.json +6 -0
- checkpoint-19750/model-00001-of-00002.safetensors +3 -0
- checkpoint-19750/model-00002-of-00002.safetensors +3 -0
- checkpoint-19750/model.safetensors.index.json +0 -0
- checkpoint-19750/optimizer.pt +3 -0
- checkpoint-19750/rng_state_0.pth +3 -0
- checkpoint-19750/rng_state_1.pth +3 -0
- checkpoint-19750/rng_state_2.pth +3 -0
- checkpoint-19750/rng_state_3.pth +3 -0
- checkpoint-19750/rng_state_4.pth +3 -0
- checkpoint-19750/rng_state_5.pth +3 -0
- checkpoint-19750/rng_state_6.pth +3 -0
- checkpoint-19750/rng_state_7.pth +3 -0
- checkpoint-19750/scheduler.pt +3 -0
- checkpoint-19750/trainer_state.json +0 -0
checkpoint-19250/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": 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_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": true,
|
65 |
+
"use_residual_attn": false,
|
66 |
+
"use_sliding_window": false,
|
67 |
+
"vae_path": "mit-han-lab/dc-ae-f128c512-mix-1.0-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-19250/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-19250/model-00001-of-00002.safetensors
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:d963944c6f83f3d13c06a87357244ca3a548819eb779578ee2689c8fd761d913
|
3 |
+
size 4952988672
|
checkpoint-19250/model-00002-of-00002.safetensors
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:b8c02f7e4a464e69c0a3a147527a057743862a4f8da45dc01fcf3ff2cd6f2326
|
3 |
+
size 1735052360
|
checkpoint-19250/model.safetensors.index.json
ADDED
The diff for this file is too large to render.
See raw diff
|
|
checkpoint-19250/optimizer.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:ce0d654c3e06367a0e9837d18216a391749a5ce6209c6e8c61fd609b211ca96e
|
3 |
+
size 6996771115
|
checkpoint-19250/rng_state_0.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:6716930d2d754e882f2a34aff853f19a72c87fb83e28124f4139360be28be0d5
|
3 |
+
size 15984
|
checkpoint-19250/rng_state_1.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:8bd99c8b5d5e6562ee04fe8a687ad983b9a42778a7f62889e9abc230b94dedaa
|
3 |
+
size 15984
|
checkpoint-19250/rng_state_2.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:5bf198da129d299db22fbc0f22b9fb607b85a0d5452040a5785de595aef23539
|
3 |
+
size 15984
|
checkpoint-19250/rng_state_3.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:94992c34fadde612206763a6b77cd59502b2d367bd1f93cda0c282e4ff1056e6
|
3 |
+
size 15984
|
checkpoint-19250/rng_state_4.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:1faa64b325908cfe1fea579947791077d947bfb1be118dd094b36455ea231ea6
|
3 |
+
size 15984
|
checkpoint-19250/rng_state_5.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:15e60fe8306e9643967e37b0605d7ee50d2fdc44b1ca3d0d63125d3db7c56f79
|
3 |
+
size 15984
|
checkpoint-19250/rng_state_6.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:63e52b9cdff73bf009802ea29094664e2776f8081fa81339b5c2a9599326b181
|
3 |
+
size 15984
|
checkpoint-19250/rng_state_7.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:243a3edbfe780670dd7dfa04632a387f118ecc574f7a6368c48945cc6355fcd0
|
3 |
+
size 15984
|
checkpoint-19250/scheduler.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:4f05e855310a7bfa40316df7452e3307b5608794551da30f7f731560bd86b8ce
|
3 |
+
size 1064
|
checkpoint-19250/trainer_state.json
ADDED
The diff for this file is too large to render.
See raw diff
|
|
checkpoint-19250/training_args.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:d389910406a9592beeec7fa0fa3547dc72f25b192f2b9d695f697348c94850af
|
3 |
+
size 5944
|
checkpoint-19500/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": 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_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": true,
|
65 |
+
"use_residual_attn": false,
|
66 |
+
"use_sliding_window": false,
|
67 |
+
"vae_path": "mit-han-lab/dc-ae-f128c512-mix-1.0-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-19500/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-19500/model-00001-of-00002.safetensors
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:437cb1be48ef241486af301b4a9ee51b089fcf3ba95995f10f06456a26d75909
|
3 |
+
size 4952988672
|
checkpoint-19500/model-00002-of-00002.safetensors
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:13d55e2c678246678fab1c8df8a269ccb467f629f8ce5de1dfb01370a53157ce
|
3 |
+
size 1735052360
|
checkpoint-19500/model.safetensors.index.json
ADDED
The diff for this file is too large to render.
See raw diff
|
|
checkpoint-19500/optimizer.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:02056d13df2049616503dd859ac5d1e493db2b8c26b2b47c8a354e7d223e3b04
|
3 |
+
size 6996771115
|
checkpoint-19500/rng_state_0.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:aae6df6a92c10c3cdffc0cc6b476ea82738d6e1c5e96537319cd9e5c6fd095aa
|
3 |
+
size 15984
|
checkpoint-19500/rng_state_1.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:37b0f7b93eba94ea5efc7d0c963125618b86423f9a1d4c1e2abd059506ef2a04
|
3 |
+
size 15984
|
checkpoint-19500/rng_state_2.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:c5a8ae8029a47e37635d5bf48db542cb015c1d7fcbf975320f566fffa14e844c
|
3 |
+
size 15984
|
checkpoint-19500/rng_state_3.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:be65ea1d1bc64b821f0091c77b3fea7b1c886ef6f7546320d18a7bf8a7a90327
|
3 |
+
size 15984
|
checkpoint-19500/rng_state_4.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:eaee6d8af270d6cd03cf24432baac1ae2543aef44b53cb0fa500ea831d0e7577
|
3 |
+
size 15984
|
checkpoint-19500/rng_state_5.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:df2bbfbd4a1e694e5259b74bf404c928e70b222b270adce2f78ae62b8a3acceb
|
3 |
+
size 15984
|
checkpoint-19500/rng_state_6.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:7ae15ccd1eba92a2805fb643377020a779a49d4e063f24584f993c0546ddc90e
|
3 |
+
size 15984
|
checkpoint-19500/rng_state_7.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:beddb2af756b4f442af8f764dacdd76bd0b519303f5b83a097d455ecf14878cb
|
3 |
+
size 15984
|
checkpoint-19500/scheduler.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:4afa35f4b32b417eec751a4bb43398efb94f15390672a3160295126c10272c6b
|
3 |
+
size 1064
|
checkpoint-19500/trainer_state.json
ADDED
The diff for this file is too large to render.
See raw diff
|
|
checkpoint-19500/training_args.bin
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:d389910406a9592beeec7fa0fa3547dc72f25b192f2b9d695f697348c94850af
|
3 |
+
size 5944
|
checkpoint-19750/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": 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_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": true,
|
65 |
+
"use_residual_attn": false,
|
66 |
+
"use_sliding_window": false,
|
67 |
+
"vae_path": "mit-han-lab/dc-ae-f128c512-mix-1.0-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-19750/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-19750/model-00001-of-00002.safetensors
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:83f92c4f85349adfc5ccee1deaf5b30cc5019d544becd1a1c0e3a179f517ab92
|
3 |
+
size 4952988672
|
checkpoint-19750/model-00002-of-00002.safetensors
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:2ac9e754e43292a6de944534b9fe64c313622d1fed3ae2fb0128411ce3600466
|
3 |
+
size 1735052360
|
checkpoint-19750/model.safetensors.index.json
ADDED
The diff for this file is too large to render.
See raw diff
|
|
checkpoint-19750/optimizer.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:925f7b37216beab5598ff5abd2f2f3aae354c23aff0a87cde00994bfbf7505f1
|
3 |
+
size 6996771115
|
checkpoint-19750/rng_state_0.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:23a01162a010fd3e6bb5d9d47d9c28dab80ea63f01b517c8cced3a82b159dfac
|
3 |
+
size 15984
|
checkpoint-19750/rng_state_1.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:d61361b2948d3f4d9f36725bb939925d3fd6cda0bd08e572e29b09cfe8c46a84
|
3 |
+
size 15984
|
checkpoint-19750/rng_state_2.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:572b82c53a77dfb3429eca7c9dd6d0ac94158b17179efc7d500f56e6333660fb
|
3 |
+
size 15984
|
checkpoint-19750/rng_state_3.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:0806c796d6419ae1dbb54beda00a1cdfdac862c43a25fe2dee0f8abe4dd4bd99
|
3 |
+
size 15984
|
checkpoint-19750/rng_state_4.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:34fa4c4d04ca26eecdc1c1c3824638e3238c35a6f3ddb04fd241914c10a8843f
|
3 |
+
size 15984
|
checkpoint-19750/rng_state_5.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:b28f6aaa4b8e43800445c9ed8a112fbd285380b7ed6fb5c85c9ecf0313891880
|
3 |
+
size 15984
|
checkpoint-19750/rng_state_6.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:800197cd2c6b192307919fcb2beb52307f777c369e562105e7b166b9fce6eb1b
|
3 |
+
size 15984
|
checkpoint-19750/rng_state_7.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:e0e10711237744a6e756da67a9b0b72a1df96d327d1d7a6ffc526ad4d801489b
|
3 |
+
size 15984
|
checkpoint-19750/scheduler.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:0cf42ee96064c9ca1d57e07f00fc20d9087286fe5d6a6918f1c5a4f1503347be
|
3 |
+
size 1064
|
checkpoint-19750/trainer_state.json
ADDED
The diff for this file is too large to render.
See raw diff
|
|