Upload folder using huggingface_hub
Browse filesThis view is limited to 50 files because it contains too many changes.
See raw diff
- checkpoint-29250/config.json +79 -0
- checkpoint-29250/generation_config.json +6 -0
- checkpoint-29250/model-00001-of-00002.safetensors +3 -0
- checkpoint-29250/model-00002-of-00002.safetensors +3 -0
- checkpoint-29250/model.safetensors.index.json +0 -0
- checkpoint-29250/optimizer.pt +3 -0
- checkpoint-29250/rng_state_0.pth +3 -0
- checkpoint-29250/rng_state_1.pth +3 -0
- checkpoint-29250/rng_state_2.pth +3 -0
- checkpoint-29250/rng_state_3.pth +3 -0
- checkpoint-29250/rng_state_4.pth +3 -0
- checkpoint-29250/rng_state_5.pth +3 -0
- checkpoint-29250/rng_state_6.pth +3 -0
- checkpoint-29250/rng_state_7.pth +3 -0
- checkpoint-29250/scheduler.pt +3 -0
- checkpoint-29250/trainer_state.json +0 -0
- checkpoint-29250/training_args.bin +3 -0
- checkpoint-29500/config.json +79 -0
- checkpoint-29500/generation_config.json +6 -0
- checkpoint-29500/model-00001-of-00002.safetensors +3 -0
- checkpoint-29500/model-00002-of-00002.safetensors +3 -0
- checkpoint-29500/model.safetensors.index.json +0 -0
- checkpoint-29500/optimizer.pt +3 -0
- checkpoint-29500/rng_state_0.pth +3 -0
- checkpoint-29500/rng_state_1.pth +3 -0
- checkpoint-29500/rng_state_2.pth +3 -0
- checkpoint-29500/rng_state_3.pth +3 -0
- checkpoint-29500/rng_state_4.pth +3 -0
- checkpoint-29500/rng_state_5.pth +3 -0
- checkpoint-29500/rng_state_6.pth +3 -0
- checkpoint-29500/rng_state_7.pth +3 -0
- checkpoint-29500/scheduler.pt +3 -0
- checkpoint-29500/trainer_state.json +0 -0
- checkpoint-29500/training_args.bin +3 -0
- checkpoint-29750/config.json +79 -0
- checkpoint-29750/generation_config.json +6 -0
- checkpoint-29750/model-00001-of-00002.safetensors +3 -0
- checkpoint-29750/model-00002-of-00002.safetensors +3 -0
- checkpoint-29750/model.safetensors.index.json +0 -0
- checkpoint-29750/optimizer.pt +3 -0
- checkpoint-29750/rng_state_0.pth +3 -0
- checkpoint-29750/rng_state_1.pth +3 -0
- checkpoint-29750/rng_state_2.pth +3 -0
- checkpoint-29750/rng_state_3.pth +3 -0
- checkpoint-29750/rng_state_4.pth +3 -0
- checkpoint-29750/rng_state_5.pth +3 -0
- checkpoint-29750/rng_state_6.pth +3 -0
- checkpoint-29750/rng_state_7.pth +3 -0
- checkpoint-29750/scheduler.pt +3 -0
- checkpoint-29750/trainer_state.json +0 -0
checkpoint-29250/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-29250/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-29250/model-00001-of-00002.safetensors
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:7308adbcedc6e9487225701709087047273b391640b434d1e264ede9e9bed860
|
3 |
+
size 4952988672
|
checkpoint-29250/model-00002-of-00002.safetensors
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:a75c1d39f484b5fe1617325f952c1a4aa6ca081cb69e233a8823992c56b84319
|
3 |
+
size 1735052360
|
checkpoint-29250/model.safetensors.index.json
ADDED
The diff for this file is too large to render.
See raw diff
|
|
checkpoint-29250/optimizer.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:d3f016ad9638d535a155881698f83ac52b0a1eb980cbffec05f75923f0d20798
|
3 |
+
size 6996771115
|
checkpoint-29250/rng_state_0.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:454b0bcaf4865298e1b84a7e6fdc3071e79c013dd036e203bd081f45413b96a6
|
3 |
+
size 15984
|
checkpoint-29250/rng_state_1.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:e6178ade609f55486d0fd96cebfe140545a1e116d5b13267a20bb28ca312c38c
|
3 |
+
size 15984
|
checkpoint-29250/rng_state_2.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:9f2195b320c0b5b899dd3dfc54615b539e684dce2c828fb26a862ca1620abf77
|
3 |
+
size 15984
|
checkpoint-29250/rng_state_3.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:0c5b27edfb4a213624369ee763651b76a0f08fd2ba9b508553f3b0ea4b422c3b
|
3 |
+
size 15984
|
checkpoint-29250/rng_state_4.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:a6f78bb45b9ec95ac8d33d4438aa4fad5a7820b7f6ee9d60f6f3001d7bcd1e14
|
3 |
+
size 15984
|
checkpoint-29250/rng_state_5.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:502a4fc929b1ecb2b7522c91e88626421dd061c16db438c68306019776e2bfbd
|
3 |
+
size 15984
|
checkpoint-29250/rng_state_6.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:aa942c57f957c98731da054ebab3960b4687857012e2398f4dc74e9500ada8f9
|
3 |
+
size 15984
|
checkpoint-29250/rng_state_7.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:8b8345fc87f1decadbda6cbbd7c61fbe6cc60d640877cfa2e28860f692395fb7
|
3 |
+
size 15984
|
checkpoint-29250/scheduler.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:e7062f7acacdb41fcd6067b1893d9095de282acbe2d6f7bbab47bf662777c4a7
|
3 |
+
size 1064
|
checkpoint-29250/trainer_state.json
ADDED
The diff for this file is too large to render.
See raw diff
|
|
checkpoint-29250/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-29500/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-29500/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-29500/model-00001-of-00002.safetensors
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:9cb0338bc1785bc89c577ed723c517b637a39d977baac2a07c7aa6cb8bd14df2
|
3 |
+
size 4952988672
|
checkpoint-29500/model-00002-of-00002.safetensors
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:bcbfbb5da3028be6706b6051385879ab39d85e66d7df384ae68bd739b9d33682
|
3 |
+
size 1735052360
|
checkpoint-29500/model.safetensors.index.json
ADDED
The diff for this file is too large to render.
See raw diff
|
|
checkpoint-29500/optimizer.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:c23a7bf020fe20ab75db7dde12e8c3487b43265e2a159c9e3dd145c0e70e7e03
|
3 |
+
size 6996771115
|
checkpoint-29500/rng_state_0.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:bbd2679e16e93886595f962f018e544f6a88ed0c8c13c90c6313fdce963a6e49
|
3 |
+
size 15984
|
checkpoint-29500/rng_state_1.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:19d984671b2ed77d25e8ad259a5e65847129a7c7531137fbfd46d6876be15503
|
3 |
+
size 15984
|
checkpoint-29500/rng_state_2.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:562470f0f9998f028b1df6ad161388c0fc00fd3798f1533395f80fd59d3fdbe8
|
3 |
+
size 15984
|
checkpoint-29500/rng_state_3.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:75898249479bb92ed9b7d9994e596732aaedc131ba8f80ee2a67eb193e69ac7e
|
3 |
+
size 15984
|
checkpoint-29500/rng_state_4.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:febf925ca4842e95fa0e80952e6f24cd12b787a65f4e24c55c63a68293b63c83
|
3 |
+
size 15984
|
checkpoint-29500/rng_state_5.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:34a1c8c7a8eab2da85ef9f5852374cd03ef0b2638dc6594dfb2645413ab3cb1d
|
3 |
+
size 15984
|
checkpoint-29500/rng_state_6.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:c220030bbdc07e79ad43001028ad91f20606d647c116a62125fba001436ec123
|
3 |
+
size 15984
|
checkpoint-29500/rng_state_7.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:e7abf8b704069ca8b411ab3fe19419e6d22ccc973101edb756613ba435bfe8e1
|
3 |
+
size 15984
|
checkpoint-29500/scheduler.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:e81d68ad83258d11c96543bbff473ebe5420d12cf92d967a77e63fc5c6a2b6ba
|
3 |
+
size 1064
|
checkpoint-29500/trainer_state.json
ADDED
The diff for this file is too large to render.
See raw diff
|
|
checkpoint-29500/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-29750/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-29750/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-29750/model-00001-of-00002.safetensors
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:69c8948e0ab99fddee64d6579b7c7cc007e01132ecec4cc5c0c62d604cc52b83
|
3 |
+
size 4952988672
|
checkpoint-29750/model-00002-of-00002.safetensors
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:ab068ff14b1b6ad3dfa28ff55678aaf04579e78b650237e84548ff6273f785a5
|
3 |
+
size 1735052360
|
checkpoint-29750/model.safetensors.index.json
ADDED
The diff for this file is too large to render.
See raw diff
|
|
checkpoint-29750/optimizer.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:02f917c66e02859c58f9f1fe6f3341fe51e597cc73ce45d1f124e25ab202ba44
|
3 |
+
size 6996771115
|
checkpoint-29750/rng_state_0.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:026440f6b54e853cdc636801ebf5ef238df659737d617a4a947fbecec43cc865
|
3 |
+
size 15984
|
checkpoint-29750/rng_state_1.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:2ae3809a67f1005713a15747d02304628127142bb7e2bf1b9818ebf8c9688595
|
3 |
+
size 15984
|
checkpoint-29750/rng_state_2.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:e7ed6b9a902cd4c58cb21068e5e4443d12074a23f9985f8ad6d2777a7f3316fc
|
3 |
+
size 15984
|
checkpoint-29750/rng_state_3.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:44867ff7ea79f7b0d6bed542cf0475791553ad9cb7a3c1364e05eea696043b3c
|
3 |
+
size 15984
|
checkpoint-29750/rng_state_4.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:1628bc4508eee75cf7607df79234d0860416fdc1dd042619e2f9aa8d1b2a0b48
|
3 |
+
size 15984
|
checkpoint-29750/rng_state_5.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:b26616a98fbdf2cab2fedd21078608ffdfacd4d89ebc4e04d830d7ab48cbd7fc
|
3 |
+
size 15984
|
checkpoint-29750/rng_state_6.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:fcd3a1db033522a3097145a7522b9d2e12ac6a70df941380cafd49644fe437f8
|
3 |
+
size 15984
|
checkpoint-29750/rng_state_7.pth
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:23a5b5e35c7f61784809b7c9a4942ca5428a22e4d80aa6fd7c7b37505a4a7d83
|
3 |
+
size 15984
|
checkpoint-29750/scheduler.pt
ADDED
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
1 |
+
version https://git-lfs.github.com/spec/v1
|
2 |
+
oid sha256:82a4793a82a3365621e3b416fcef19e19c9b5883add5f8bfddc51879ec173e26
|
3 |
+
size 1064
|
checkpoint-29750/trainer_state.json
ADDED
The diff for this file is too large to render.
See raw diff
|
|