Upload folder using huggingface_hub

.gitattributes

@@ -84,3 +84,20 @@ checkpoint-178000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
 checkpoint-180000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
 checkpoint-188000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
 checkpoint-190000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
+checkpoint-198000/coco_steps-40_cfg-5_0/coco_0_A[[:space:]]group[[:space:]]of[[:space:]]zebras[[:space:]]grazing[[:space:]]in[[:space:]]the[[:space:]]grass._1.png filter=lfs diff=lfs merge=lfs -text
+checkpoint-198000/coco_steps-40_cfg-5_0/coco_10_A[[:space:]]traditional[[:space:]]Japanese[[:space:]]street[[:space:]]in[[:space:]]spring,[[:space:]]with[[:space:]]cherry[[:space:]]blossoms[[:space:]]drifting[[:space:]]through[[:space:]]the[[:space:]]air[[:space:]]and[[:space:]]petals[[:space:]]scattered[[:space:]]across[[:space:]]stone[[:space:]]steps.[[:space:]]A[[:space:]]small[[:space:]]Shiba[[:space:]]Inu[[:space:]]dog[[:space:]]_1.png filter=lfs diff=lfs merge=lfs -text
+checkpoint-198000/coco_steps-40_cfg-5_0/coco_12_A[[:space:]]wintry[[:space:]]New[[:space:]]York[[:space:]]Central[[:space:]]Park[[:space:]]morning,[[:space:]]with[[:space:]]joggers[[:space:]]bundled[[:space:]]in[[:space:]]scarves,[[:space:]]dog[[:space:]]walkers[[:space:]]guiding[[:space:]]all[[:space:]]kinds[[:space:]]of[[:space:]]breeds[[:space:]]past[[:space:]]snow-dusted[[:space:]]benches,[[:space:]]and[[:space:]]a[[:space:]]flock_1.png filter=lfs diff=lfs merge=lfs -text
+checkpoint-198000/coco_steps-40_cfg-5_0/coco_14_A[[:space:]]colorful[[:space:]]umbrella[[:space:]]falling[[:space:]]from[[:space:]]a[[:space:]]tall[[:space:]]building._1.png filter=lfs diff=lfs merge=lfs -text
+checkpoint-198000/coco_steps-40_cfg-5_0/coco_15_Group[[:space:]]of[[:space:]]people[[:space:]]standing[[:space:]]near[[:space:]]or[[:space:]]riding[[:space:]]a[[:space:]]very[[:space:]]small[[:space:]]train[[:space:]]engine.[[:space:]]_1.png filter=lfs diff=lfs merge=lfs -text
+checkpoint-198000/coco_steps-40_cfg-5_0/coco_16_A[[:space:]]green[[:space:]]train[[:space:]]parked,[[:space:]]sitting[[:space:]]in[[:space:]]the[[:space:]]snow_1.png filter=lfs diff=lfs merge=lfs -text
+checkpoint-198000/coco_steps-40_cfg-5_0/coco_17_A[[:space:]]vintage[[:space:]]green[[:space:]]train,[[:space:]]its[[:space:]]exterior[[:space:]]softly[[:space:]]gleaming[[:space:]]through[[:space:]]patches[[:space:]]of[[:space:]]frost,[[:space:]]sits[[:space:]]quietly[[:space:]]on[[:space:]]snow-blanketed[[:space:]]tracks[[:space:]]at[[:space:]]a[[:space:]]quaint[[:space:]]rural[[:space:]]station.[[:space:]]Snow[[:space:]]ge_1.png filter=lfs diff=lfs merge=lfs -text
+checkpoint-198000/coco_steps-40_cfg-5_0/coco_1_a[[:space:]]number[[:space:]]of[[:space:]]people[[:space:]]standing[[:space:]]around[[:space:]]a[[:space:]]large[[:space:]]group[[:space:]]of[[:space:]]luggage[[:space:]]bags_1.png filter=lfs diff=lfs merge=lfs -text
+checkpoint-198000/coco_steps-40_cfg-5_0/coco_2_A[[:space:]]snowy[[:space:]]mountain[[:space:]]cabin[[:space:]]scene[[:space:]]at[[:space:]]dusk,[[:space:]]where[[:space:]]a[[:space:]]cheerful[[:space:]]Saint[[:space:]]Bernard[[:space:]]dog[[:space:]]bounds[[:space:]]through[[:space:]]deep[[:space:]]snow,[[:space:]]leaving[[:space:]]a[[:space:]]trail[[:space:]]of[[:space:]]paw[[:space:]]prints.[[:space:]]Icicles[[:space:]]hang[[:space:]]from[[:space:]]th_1.png filter=lfs diff=lfs merge=lfs -text
+checkpoint-198000/coco_steps-40_cfg-5_0/coco_3_A[[:space:]]whimsical[[:space:]]artist’s[[:space:]]studio[[:space:]]overlooking[[:space:]]the[[:space:]]ocean,[[:space:]]with[[:space:]]tall[[:space:]]windows[[:space:]]framing[[:space:]]crashing[[:space:]]waves[[:space:]]and[[:space:]]soaring[[:space:]]seagulls.[[:space:]]On[[:space:]]the[[:space:]]floor,[[:space:]]a[[:space:]]curious[[:space:]]rabbit[[:space:]]sniff_1.png filter=lfs diff=lfs merge=lfs -text
+checkpoint-198000/coco_steps-40_cfg-5_0/coco_4_A[[:space:]]lively[[:space:]]beach[[:space:]]at[[:space:]]sunset[[:space:]]with[[:space:]]families[[:space:]]building[[:space:]]sandcastles,[[:space:]]dogs[[:space:]]playing[[:space:]]fetch[[:space:]]near[[:space:]]the[[:space:]]water’s[[:space:]]edge,[[:space:]]and[[:space:]]children[[:space:]]flying[[:space:]]colorful[[:space:]]kites[[:space:]]shaped[[:space:]]like[[:space:]]_1.png filter=lfs diff=lfs merge=lfs -text
+checkpoint-198000/coco_steps-40_cfg-5_0/coco_5_A[[:space:]]woman[[:space:]]baking[[:space:]]a[[:space:]]pizza[[:space:]]pie[[:space:]]in[[:space:]]the[[:space:]]kitchen_1.png filter=lfs diff=lfs merge=lfs -text
+checkpoint-198000/coco_steps-40_cfg-5_0/coco_6_A[[:space:]]couple[[:space:]]of[[:space:]]men[[:space:]]standing[[:space:]]on[[:space:]]a[[:space:]]field[[:space:]]playing[[:space:]]baseball._1.png filter=lfs diff=lfs merge=lfs -text
+checkpoint-198000/coco_steps-40_cfg-5_0/coco_8_LOOKS[[:space:]]LIKE[[:space:]]A[[:space:]]PERSON[[:space:]]DROPPING[[:space:]]THERE[[:space:]]TENNIS[[:space:]]RACK_1.png filter=lfs diff=lfs merge=lfs -text
+checkpoint-198000/coco_steps-40_cfg-5_0/coco_9_Three[[:space:]]people[[:space:]]are[[:space:]]walking[[:space:]]down[[:space:]]the[[:space:]]road[[:space:]]with[[:space:]]three[[:space:]]horses._1.png filter=lfs diff=lfs merge=lfs -text
+checkpoint-198000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
+checkpoint-200000/trainer_state.json filter=lfs diff=lfs merge=lfs -text

checkpoint-198000/config.json

@@ -0,0 +1,79 @@
+{
+  "ar_steps": 1,
+  "architectures": [
+    "DiffVLMDiffusion"
+  ],
+  "attention_dropout": 0.0,
+  "bos_token_id": 151643,
+  "condition_layer": -1,
+  "eos_token_id": 151645,
+  "hidden_act": "silu",
+  "hidden_size": 1536,
+  "image_token_id": 151655,
+  "img_cross_attention_dim": 2048,
+  "img_diffuser_depth": 1,
+  "img_ffn_dim_multiplier": null,
+  "img_hidden_size": 1536,
+  "img_multiple_of": 256,
+  "img_norm_eps": 1e-05,
+  "img_num_attention_heads": 12,
+  "img_num_kv_heads": 12,
+  "img_qk_norm": true,
+  "in_channels": 32,
+  "initializer_range": 0.02,
+  "inject_img_diffuser": false,
+  "input_size": 32,
+  "intermediate_size": 8960,
+  "layer_group_size": 7,
+  "layerwise_start_idx": 0,
+  "lora_alpha": 16,
+  "lora_bias": "none",
+  "lora_dropout": 0.05,
+  "lora_enable": false,
+  "lora_r": 64,
+  "max_position_embeddings": 32768,
+  "max_window_layers": 28,
+  "model_type": "qwen2_vl",
+  "non_linearity": 1,
+  "norm_elementwise_affine": true,
+  "num_attention_heads": 12,
+  "num_hidden_layers": 28,
+  "num_key_value_heads": 2,
+  "patch_size": 1,
+  "repa_coeff": 0.5,
+  "repa_layers": null,
+  "repa_shared": false,
+  "rms_norm_eps": 1e-06,
+  "rope_scaling": {
+    "mrope_section": [
+      16,
+      24,
+      24
+    ],
+    "rope_type": "default",
+    "type": "default"
+  },
+  "rope_theta": 1000000.0,
+  "sample_size": 128,
+  "sampling_steps": 28,
+  "sliding_window": null,
+  "tie_word_embeddings": true,
+  "torch_dtype": "bfloat16",
+  "transformers_version": "4.47.0",
+  "use_cache": true,
+  "use_repa": false,
+  "use_residual_attn": true,
+  "use_sliding_window": false,
+  "vae_path": "mit-han-lab/dc-ae-f32c32-sana-1.1-diffusers",
+  "video_token_id": 151656,
+  "vision_config": {
+    "hidden_size": 1536,
+    "in_chans": 3,
+    "model_type": "qwen2_vl",
+    "spatial_patch_size": 14
+  },
+  "vision_end_token_id": 151653,
+  "vision_start_token_id": 151652,
+  "vision_token_id": 151654,
+  "vocab_size": 151936
+}

checkpoint-198000/generation_config.json

@@ -0,0 +1,6 @@
+{
+  "_from_model_config": true,
+  "bos_token_id": 151643,
+  "eos_token_id": 151645,
+  "transformers_version": "4.47.0"
+}

checkpoint-198000/model-00001-of-00002.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:abea94db3a193dcb57fbef3334a144f3f35ded5cc0fec92e86671a86094492b7
+size 4974360904

checkpoint-198000/model-00002-of-00002.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:325d119f7be9a285f2c1f1e3cbfedfc983d48cbe81aa4da1595273acae22b7c4
+size 3234250886

checkpoint-198000/optimizer.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dd175cd228e62a2039cb412ccb264d9c767c8fade21ed6c1567be85200861d5a
+size 7455061885

checkpoint-198000/rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:aec1d59f0f0fc320e3e9c3c9759c35c1cc3d5492d1424dc08cda457a901a1972
+size 15024

checkpoint-198000/rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:32acbc145c8522d1a41050b9243d859a51a6082d4305700f4930851402427f07
+size 15024

checkpoint-198000/rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:12075c643084f53cd6ea1345fef1255eeab0fb1fab4aeec770bc94190d56a910
+size 15024

checkpoint-198000/rng_state_3.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f699b0b7260d2b33badc918856ad0e4d73e3ff8ffa329dd520b92b7ff922767d
+size 15024

checkpoint-198000/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:51745d42f170b114de9dcc326bb9e738fc90c57fa247ee1bbfb17dcd22f7d1d9
+size 1064

checkpoint-198000/trainer_state.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:03e6cdb48966f64fa0cd307653dd82bda4d8feb9d6fbbc6332f5f3ff3ccc031d
+size 30505988

checkpoint-198000/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b4c5ab601b530ed40fb76ff6da7f4319bb893e9b4e2295da5da00473f27040c2
+size 6008

checkpoint-200000/config.json

@@ -0,0 +1,79 @@
+{
+  "ar_steps": 1,
+  "architectures": [
+    "DiffVLMDiffusion"
+  ],
+  "attention_dropout": 0.0,
+  "bos_token_id": 151643,
+  "condition_layer": -1,
+  "eos_token_id": 151645,
+  "hidden_act": "silu",
+  "hidden_size": 1536,
+  "image_token_id": 151655,
+  "img_cross_attention_dim": 2048,
+  "img_diffuser_depth": 1,
+  "img_ffn_dim_multiplier": null,
+  "img_hidden_size": 1536,
+  "img_multiple_of": 256,
+  "img_norm_eps": 1e-05,
+  "img_num_attention_heads": 12,
+  "img_num_kv_heads": 12,
+  "img_qk_norm": true,
+  "in_channels": 32,
+  "initializer_range": 0.02,
+  "inject_img_diffuser": false,
+  "input_size": 32,
+  "intermediate_size": 8960,
+  "layer_group_size": 7,
+  "layerwise_start_idx": 0,
+  "lora_alpha": 16,
+  "lora_bias": "none",
+  "lora_dropout": 0.05,
+  "lora_enable": false,
+  "lora_r": 64,
+  "max_position_embeddings": 32768,
+  "max_window_layers": 28,
+  "model_type": "qwen2_vl",
+  "non_linearity": 1,
+  "norm_elementwise_affine": true,
+  "num_attention_heads": 12,
+  "num_hidden_layers": 28,
+  "num_key_value_heads": 2,
+  "patch_size": 1,
+  "repa_coeff": 0.5,
+  "repa_layers": null,
+  "repa_shared": false,
+  "rms_norm_eps": 1e-06,
+  "rope_scaling": {
+    "mrope_section": [
+      16,
+      24,
+      24
+    ],
+    "rope_type": "default",
+    "type": "default"
+  },
+  "rope_theta": 1000000.0,
+  "sample_size": 128,
+  "sampling_steps": 28,
+  "sliding_window": null,
+  "tie_word_embeddings": true,
+  "torch_dtype": "bfloat16",
+  "transformers_version": "4.47.0",
+  "use_cache": true,
+  "use_repa": false,
+  "use_residual_attn": true,
+  "use_sliding_window": false,
+  "vae_path": "mit-han-lab/dc-ae-f32c32-sana-1.1-diffusers",
+  "video_token_id": 151656,
+  "vision_config": {
+    "hidden_size": 1536,
+    "in_chans": 3,
+    "model_type": "qwen2_vl",
+    "spatial_patch_size": 14
+  },
+  "vision_end_token_id": 151653,
+  "vision_start_token_id": 151652,
+  "vision_token_id": 151654,
+  "vocab_size": 151936
+}

checkpoint-200000/generation_config.json

@@ -0,0 +1,6 @@
+{
+  "_from_model_config": true,
+  "bos_token_id": 151643,
+  "eos_token_id": 151645,
+  "transformers_version": "4.47.0"
+}

checkpoint-200000/model-00001-of-00002.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6ee21ec0851b403369416ef8a74904934c4b2476eb9da255523909f35c7a065d
+size 4974360904

checkpoint-200000/model-00002-of-00002.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1d3684655c5929e96acf0d9e115c516d966d4641e5d8b4ac3a08cb39ac6ef96c
+size 3234250886

checkpoint-200000/optimizer.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c272999499a5b606fcb8c198e4401b184ceb924d84889910c8b4ccaaecc0bafb
+size 7455061885

checkpoint-200000/rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:25cb94ad6f6ead0bf0fbb83ed5941f78d5b6e007f3ffa46173f087689a617de9
+size 15024

checkpoint-200000/rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:122676ef3436e99017317e55f3f6f0e90a811adebe498c122734e0a0728bff8a
+size 15024

checkpoint-200000/rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d0b6e94309dd042efafb586992c54794be464d27a0d1b02f7626f7e77c818f56
+size 15024

checkpoint-200000/rng_state_3.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2ec9ad9638fd089eebae80702a27fcb7cc26073608323b5ac0acfe4f05d2331f
+size 15024

checkpoint-200000/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7cb0584af88a439fe3c2644d4b34719d5e34d085616e4fed428123c84c09fd11
+size 1064

checkpoint-200000/trainer_state.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:848c7bca706ace7addb8a62f71b28369c5b5923450e9376f198ac85b7d88a8a8
+size 30814838

checkpoint-200000/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b4c5ab601b530ed40fb76ff6da7f4319bb893e9b4e2295da5da00473f27040c2
+size 6008