Upload folder using huggingface_hub
Browse filesThis view is limited to 50 files because it contains too many changes.
See raw diff
- checkpoint-10000/config.json +79 -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/rng_state_4.pth +3 -0
- checkpoint-10000/rng_state_5.pth +3 -0
- checkpoint-10000/rng_state_6.pth +3 -0
- checkpoint-10000/rng_state_7.pth +3 -0
- checkpoint-10000/scheduler.pt +3 -0
- checkpoint-10000/trainer_state.json +0 -0
- checkpoint-10000/training_args.bin +3 -0
- checkpoint-9250/config.json +79 -0
- checkpoint-9250/generation_config.json +6 -0
- checkpoint-9250/model-00001-of-00002.safetensors +3 -0
- checkpoint-9250/model-00002-of-00002.safetensors +3 -0
- checkpoint-9250/model.safetensors.index.json +0 -0
- checkpoint-9250/optimizer.pt +3 -0
- checkpoint-9250/rng_state_0.pth +3 -0
- checkpoint-9250/rng_state_1.pth +3 -0
- checkpoint-9250/rng_state_2.pth +3 -0
- checkpoint-9250/rng_state_3.pth +3 -0
- checkpoint-9250/rng_state_4.pth +3 -0
- checkpoint-9250/rng_state_5.pth +3 -0
- checkpoint-9250/rng_state_6.pth +3 -0
- checkpoint-9250/rng_state_7.pth +3 -0
- checkpoint-9250/scheduler.pt +3 -0
- checkpoint-9250/trainer_state.json +0 -0
- checkpoint-9250/training_args.bin +3 -0
- checkpoint-9500/config.json +79 -0
- checkpoint-9500/generation_config.json +6 -0
- checkpoint-9500/model-00001-of-00002.safetensors +3 -0
- checkpoint-9500/model-00002-of-00002.safetensors +3 -0
- checkpoint-9500/model.safetensors.index.json +0 -0
- checkpoint-9500/optimizer.pt +3 -0
- checkpoint-9500/rng_state_0.pth +3 -0
- checkpoint-9500/rng_state_1.pth +3 -0
- checkpoint-9500/rng_state_2.pth +3 -0
- checkpoint-9500/rng_state_3.pth +3 -0
- checkpoint-9500/rng_state_4.pth +3 -0
- checkpoint-9500/rng_state_5.pth +3 -0
- checkpoint-9500/rng_state_6.pth +3 -0
- checkpoint-9500/rng_state_7.pth +3 -0
- checkpoint-9500/scheduler.pt +3 -0
- checkpoint-9500/trainer_state.json +0 -0
checkpoint-10000/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-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:d0006ad253f51ead00ff95b4e2ac642731d286a73edea7c31058e4516cd17537
|
3 |
+
size 4952988672
|
checkpoint-10000/model-00002-of-00002.safetensors
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:b29a64eb589a8652044e3e02821ca8b9b2905356ea4bb9c698836d0732533cc0
|
3 |
+
size 1735052360
|
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 sha256:894780c57260d9303c6fa56be6c4636942dc065aa706dbb58fc72d2688cdc72c
|
3 |
+
size 6996771115
|
checkpoint-10000/rng_state_0.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:5f16841229fb0b965f1f7d32ac7e1435d5fa16e89416f153dff5f2a32fe77d5d
|
3 |
+
size 15984
|
checkpoint-10000/rng_state_1.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:e0565d6bde2cd3a0d6b7c4e2e906bab728ab3745c36475ca3b6215df3b04f089
|
3 |
+
size 15984
|
checkpoint-10000/rng_state_2.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:ae11e374972d83bdcd4a30eaf3f4cfb40d80b1f9ee0de9fdc2f1286e6869d2df
|
3 |
+
size 15984
|
checkpoint-10000/rng_state_3.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:b0b725b685d05fde0b2b5bde178a808fef558969c4f71fa04c847f1260f01126
|
3 |
+
size 15984
|
checkpoint-10000/rng_state_4.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:62c1b167499a66c7e9b533ccb41cd064da0b3804c9cc1fdc0c0b3e2b706335cd
|
3 |
+
size 15984
|
checkpoint-10000/rng_state_5.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:732b389ba6d14e295ce573d4060b8f54d962be4930ce228b5ba0628400517d87
|
3 |
+
size 15984
|
checkpoint-10000/rng_state_6.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:d57e80db5fd577da27b2bac7bf3655df38dafaedd4ed1d1c37c34b3da48db6d8
|
3 |
+
size 15984
|
checkpoint-10000/rng_state_7.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:ef8caa8b923d46cf50994cc194f39a72f60ba98fa80ebc25bc31e937d5c63520
|
3 |
+
size 15984
|
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:d389910406a9592beeec7fa0fa3547dc72f25b192f2b9d695f697348c94850af
|
3 |
+
size 5944
|
checkpoint-9250/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-9250/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-9250/model-00001-of-00002.safetensors
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:06a2f137ca4d34cea1cd0abc98e9ba47fa19da49737e6d6942f50d20479352a2
|
3 |
+
size 4952988672
|
checkpoint-9250/model-00002-of-00002.safetensors
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:24ee3b0ed3d6545528d2c31d48691b988222a1639430a3455efbd50f49409c1c
|
3 |
+
size 1735052360
|
checkpoint-9250/model.safetensors.index.json
ADDED
The diff for this file is too large to render.
See raw diff
|
|
checkpoint-9250/optimizer.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:2bee428c9f91fb58ecfef630f8db03eb92f92f6603f0940641c85be8ccdea17f
|
3 |
+
size 6996771115
|
checkpoint-9250/rng_state_0.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:ca4a0b1d5d6ba4caf67ded36d7258956dca5c8143726e1a2a0ee1266d973cbdf
|
3 |
+
size 15984
|
checkpoint-9250/rng_state_1.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:b4b25898cbbda7133a0eb66dbe7e69917501d5cc880039a982eef60fe7a9d230
|
3 |
+
size 15984
|
checkpoint-9250/rng_state_2.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:06612122fc25f3085620611f8d49352b5bbf378e30233c4526bac09626c2ef30
|
3 |
+
size 15984
|
checkpoint-9250/rng_state_3.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:6fa7f04fa990a1826e2252e3528515f519ed183e67109249de6e6dc74143ce83
|
3 |
+
size 15984
|
checkpoint-9250/rng_state_4.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:d1c6eb98aae57ef08a9c88a3d37292e8e6bb5ee6d9dd18f31e268c01c8c7be78
|
3 |
+
size 15984
|
checkpoint-9250/rng_state_5.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:b2352161281b58d104aa11993e1f0d09ea29372421fcfc0ea090e64ee90e6a13
|
3 |
+
size 15984
|
checkpoint-9250/rng_state_6.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:dae9a5d53b64bd7adef052f783fea408c941571bcf5b0d6373b46b3332ac514a
|
3 |
+
size 15984
|
checkpoint-9250/rng_state_7.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:6c615f22eaf16eece2284c7586dc858908fd2c760294a7a7d94a6f6bebc52c52
|
3 |
+
size 15984
|
checkpoint-9250/scheduler.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:e72f122a583de946d6809f712966515ade23d20f19693b28731603d1b9299538
|
3 |
+
size 1064
|
checkpoint-9250/trainer_state.json
ADDED
The diff for this file is too large to render.
See raw diff
|
|
checkpoint-9250/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-9500/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-9500/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-9500/model-00001-of-00002.safetensors
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:bf6526254c896fea5f02b737a804ab8d6f8debb5de8c2be1f92fcc4e582f5c2b
|
3 |
+
size 4952988672
|
checkpoint-9500/model-00002-of-00002.safetensors
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:ef05efb8b7d420fca2b5050a5c568bc8b263afede2d75fe35a40318f97524c91
|
3 |
+
size 1735052360
|
checkpoint-9500/model.safetensors.index.json
ADDED
The diff for this file is too large to render.
See raw diff
|
|
checkpoint-9500/optimizer.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:b0177ef317e72a40c2cedfa0c94906787590d4f151131864f347f4c8a90c16fc
|
3 |
+
size 6996771115
|
checkpoint-9500/rng_state_0.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:c600100eef04f44914a457c95f2b0f2ea4267c89dede0aa931a9f84523a7885d
|
3 |
+
size 15984
|
checkpoint-9500/rng_state_1.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:bd413427e07876f80152e66f17c47c1d1fc1c2a1f38067f18add62ef6b075105
|
3 |
+
size 15984
|
checkpoint-9500/rng_state_2.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:c9399b5d5c2f62c12fc749d94e34e7c00d61b96ba124971caf1c2b0822fd8649
|
3 |
+
size 15984
|
checkpoint-9500/rng_state_3.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:749bc7cd236b6bcb097db0d6633cd8a147ac230f1b3389656fa80e7b6389809d
|
3 |
+
size 15984
|
checkpoint-9500/rng_state_4.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:4d38bb8eda0c5ca1121a00631c3d4e4c97c6d69b4cfa81e9ce20b9c4d3ffefa7
|
3 |
+
size 15984
|
checkpoint-9500/rng_state_5.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:ef1d7c0a9618e5e8b975f82e410480ac42c8507ffb232161521bf8236f4cac3a
|
3 |
+
size 15984
|
checkpoint-9500/rng_state_6.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:717b3cb4794482757921a628a67c144a7911b0312ab759d66a5c794bfa22b465
|
3 |
+
size 15984
|
checkpoint-9500/rng_state_7.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:e8c8138584fd46ea0d815b045872ba272f63468d5859696fd06ee65adbe32933
|
3 |
+
size 15984
|
checkpoint-9500/scheduler.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:364b13a9335179fa7a4f4ad975663545687496bf00fbcdcd859b4047e3552052
|
3 |
+
size 1064
|
checkpoint-9500/trainer_state.json
ADDED
The diff for this file is too large to render.
See raw diff
|
|