Upload folder using huggingface_hub
Browse files- .gitattributes +1 -0
- checkpoint-221000/config.json +79 -0
- checkpoint-221000/generation_config.json +6 -0
- checkpoint-221000/model.safetensors +3 -0
- checkpoint-221000/optimizer.pt +3 -0
- checkpoint-221000/rng_state_0.pth +3 -0
- checkpoint-221000/rng_state_1.pth +3 -0
- checkpoint-221000/rng_state_10.pth +3 -0
- checkpoint-221000/rng_state_11.pth +3 -0
- checkpoint-221000/rng_state_12.pth +3 -0
- checkpoint-221000/rng_state_13.pth +3 -0
- checkpoint-221000/rng_state_14.pth +3 -0
- checkpoint-221000/rng_state_15.pth +3 -0
- checkpoint-221000/rng_state_2.pth +3 -0
- checkpoint-221000/rng_state_3.pth +3 -0
- checkpoint-221000/rng_state_4.pth +3 -0
- checkpoint-221000/rng_state_5.pth +3 -0
- checkpoint-221000/rng_state_6.pth +3 -0
- checkpoint-221000/rng_state_7.pth +3 -0
- checkpoint-221000/rng_state_8.pth +3 -0
- checkpoint-221000/rng_state_9.pth +3 -0
- checkpoint-221000/scheduler.pt +3 -0
- checkpoint-221000/trainer_state.json +3 -0
- checkpoint-221000/training_args.bin +3 -0
.gitattributes
CHANGED
|
@@ -87,3 +87,4 @@ checkpoint-212000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
|
|
| 87 |
checkpoint-213000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
|
| 88 |
checkpoint-219000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
|
| 89 |
checkpoint-220000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
|
|
|
|
|
|
| 87 |
checkpoint-213000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
|
| 88 |
checkpoint-219000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
|
| 89 |
checkpoint-220000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
|
| 90 |
+
checkpoint-221000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
|
checkpoint-221000/config.json
ADDED
|
@@ -0,0 +1,79 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"ar_steps": 1,
|
| 3 |
+
"architectures": [
|
| 4 |
+
"DiffVLMBaseline"
|
| 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": 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": false,
|
| 65 |
+
"use_residual_attn": false,
|
| 66 |
+
"use_sliding_window": false,
|
| 67 |
+
"vae_path": "mit-han-lab/dc-ae-f32c32-in-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-221000/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-221000/model.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:da0284405b867cc51d04c3b590435c7b685c1b9a8a3e26a765b06f1c3f05e9ec
|
| 3 |
+
size 4410723984
|
checkpoint-221000/optimizer.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:f1ed7770eeef46c60dacabd58725cf5837c23df4ab98d44a45b86cdb09c30f1f
|
| 3 |
+
size 6330255386
|
checkpoint-221000/rng_state_0.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:e3afede1f996892713fbb56eba56641345e312a96605a1feb75995901aa8308c
|
| 3 |
+
size 15984
|
checkpoint-221000/rng_state_1.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:3b2c1b6b9d5442e90dec0abf2a5ba703372f5d007e598e86eff66f4c2c4ebf59
|
| 3 |
+
size 15984
|
checkpoint-221000/rng_state_10.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b0a51392f99fc817edfda3145d1dd601ff8ffad23def534d13a22a138443860b
|
| 3 |
+
size 15997
|
checkpoint-221000/rng_state_11.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:dc9eabb6d27ad9ed8e8df97f904ded07b946533d16b479557c363cbcba3b18f5
|
| 3 |
+
size 15997
|
checkpoint-221000/rng_state_12.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:4f9633b0dba239037c48ebe015583f8c87343437f5fc1703584839ae2f80a669
|
| 3 |
+
size 15997
|
checkpoint-221000/rng_state_13.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:38f6baf9c9a51b1e85b8d01c9be736f94b30aeb7f43ca94cdcc2d94682a700df
|
| 3 |
+
size 15997
|
checkpoint-221000/rng_state_14.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:0a71e225c57abc1b5f2aa39b9c41422e5661a2e80cc50c3ec4025cdbfa0dc640
|
| 3 |
+
size 15997
|
checkpoint-221000/rng_state_15.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:eb4ea3c8c95fe4a7571683c1f9241d9f43b877b05c0be6182f37ace3b401d1a6
|
| 3 |
+
size 15997
|
checkpoint-221000/rng_state_2.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:192fc0f2d04222ca29a401bb806d8f7e39b35e2955cab97ce3a4867f3c158db6
|
| 3 |
+
size 15984
|
checkpoint-221000/rng_state_3.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:7d44d26f473d5039826b7cea3c4d00e74a36d7d9f9813e5b54965d4879fb90c6
|
| 3 |
+
size 15984
|
checkpoint-221000/rng_state_4.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:1e8113bb448eb2c5c9f05a70a37f131d2f7a92fe1a88507a83bbecb25f37b72d
|
| 3 |
+
size 15984
|
checkpoint-221000/rng_state_5.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:3ab88432ab958bc6cb821107a12b7b65147b83e145734fcb89151929c0012b67
|
| 3 |
+
size 15984
|
checkpoint-221000/rng_state_6.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:8459740cce679f478b4dd516c8f27a1a05642e087572f27fb94ecafbf6696117
|
| 3 |
+
size 15984
|
checkpoint-221000/rng_state_7.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:148f1c3752ebddef6d4c87a7df70be665590204560b828bc4837764bc999a4f6
|
| 3 |
+
size 15984
|
checkpoint-221000/rng_state_8.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:b5c1ebb16e080a49e07f68d5babb0b885c02c99bc7b3635f5411535603d4af8f
|
| 3 |
+
size 15984
|
checkpoint-221000/rng_state_9.pth
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:61da20bce2784dced4cadc8db9703ebb700f0f99f51b4a2a6331f8ef3fe102d3
|
| 3 |
+
size 15984
|
checkpoint-221000/scheduler.pt
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:73f3501dd7179f6a2cf9b70bc808e4c3371f404759b1821c1e74e0f78b2317dd
|
| 3 |
+
size 1064
|
checkpoint-221000/trainer_state.json
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:59bf8399739c73ea5cfeed313212e5f12c5ac8981c4f7db6834bb905bceeb309
|
| 3 |
+
size 33794667
|
checkpoint-221000/training_args.bin
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:d43680bcd6a1ea2032e9c5db0154b2b3f4d960479c5e06e3ae15710cc11c65e2
|
| 3 |
+
size 5944
|