Upload folder using huggingface_hub

.gitattributes

@@ -46,3 +46,31 @@ checkpoint-118000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
 checkpoint-120000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
 checkpoint-128000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
 checkpoint-130000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
+checkpoint-138000/coco_steps-40_cfg-5/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-138000/coco_steps-40_cfg-5/coco_10_A[[:space:]]horse[[:space:]]grazing[[:space:]]on[[:space:]]top[[:space:]]of[[:space:]]a[[:space:]]lush[[:space:]]green[[:space:]]forest._1.png filter=lfs diff=lfs merge=lfs -text
+checkpoint-138000/coco_steps-40_cfg-5/coco_13_A[[:space:]]woman[[:space:]]eats[[:space:]]a[[:space:]]slice[[:space:]]of[[:space:]]cheese[[:space:]]pizza._1.png filter=lfs diff=lfs merge=lfs -text
+checkpoint-138000/coco_steps-40_cfg-5/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-138000/coco_steps-40_cfg-5/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-138000/coco_steps-40_cfg-5/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-138000/coco_steps-40_cfg-5/coco_17_A[[:space:]]red[[:space:]]fire[[:space:]]hydrant[[:space:]]is[[:space:]]set[[:space:]]up[[:space:]]in[[:space:]]a[[:space:]]grassy[[:space:]]clearing._1.png filter=lfs diff=lfs merge=lfs -text
+checkpoint-138000/coco_steps-40_cfg-5/coco_18_Two[[:space:]]people[[:space:]]are[[:space:]]crossing[[:space:]]the[[:space:]]street[[:space:]]holding[[:space:]]umbrellas._1.png filter=lfs diff=lfs merge=lfs -text
+checkpoint-138000/coco_steps-40_cfg-5/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-138000/coco_steps-40_cfg-5/coco_20_MAN[[:space:]]WORKING[[:space:]]ON[[:space:]]A[[:space:]]MOTORCYCLE[[:space:]]IN[[:space:]]A[[:space:]]GARAGE_1.png filter=lfs diff=lfs merge=lfs -text
+checkpoint-138000/coco_steps-40_cfg-5/coco_22_A[[:space:]]Japan[[:space:]]Airlines[[:space:]]plane[[:space:]]sits[[:space:]]parked[[:space:]]at[[:space:]]an[[:space:]]airport._1.png filter=lfs diff=lfs merge=lfs -text
+checkpoint-138000/coco_steps-40_cfg-5/coco_23_A[[:space:]]young[[:space:]]woman[[:space:]]poses[[:space:]]comically[[:space:]]with[[:space:]]a[[:space:]]piece[[:space:]]of[[:space:]]pizza[[:space:]]in[[:space:]]her[[:space:]]mouth._1.png filter=lfs diff=lfs merge=lfs -text
+checkpoint-138000/coco_steps-40_cfg-5/coco_24_A[[:space:]]black[[:space:]]bear[[:space:]]in[[:space:]]a[[:space:]]tree[[:space:]]at[[:space:]]a[[:space:]]zoo[[:space:]]setting._1.png filter=lfs diff=lfs merge=lfs -text
+checkpoint-138000/coco_steps-40_cfg-5/coco_25_Two[[:space:]]guys[[:space:]]are[[:space:]]on[[:space:]]a[[:space:]]boat,[[:space:]]with[[:space:]]one's[[:space:]]crack[[:space:]]being[[:space:]]exposed.[[:space:]]_1.png filter=lfs diff=lfs merge=lfs -text
+checkpoint-138000/coco_steps-40_cfg-5/coco_27_Two[[:space:]]different[[:space:]]pictures[[:space:]]of[[:space:]]a[[:space:]]giraffe[[:space:]]at[[:space:]]the[[:space:]]zoo._1.png filter=lfs diff=lfs merge=lfs -text
+checkpoint-138000/coco_steps-40_cfg-5/coco_28_A[[:space:]]bus[[:space:]]that[[:space:]]sign[[:space:]]reads[[:space:]]"Crosstown".[[:space:]]It[[:space:]]is[[:space:]]a[[:space:]]metro[[:space:]]bus._1.png filter=lfs diff=lfs merge=lfs -text
+checkpoint-138000/coco_steps-40_cfg-5/coco_2_A[[:space:]]yellow[[:space:]]commuter[[:space:]]train[[:space:]]traveling[[:space:]]past[[:space:]]some[[:space:]]houses._1.png filter=lfs diff=lfs merge=lfs -text
+checkpoint-138000/coco_steps-40_cfg-5/coco_30_A[[:space:]]banana[[:space:]]tree[[:space:]]filled[[:space:]]with[[:space:]]lots[[:space:]]of[[:space:]]unripe[[:space:]]bananas._1.png filter=lfs diff=lfs merge=lfs -text
+checkpoint-138000/coco_steps-40_cfg-5/coco_31_A[[:space:]]red[[:space:]]fire[[:space:]]hydrant[[:space:]]with[[:space:]]weeds[[:space:]]around[[:space:]]it[[:space:]]next[[:space:]]to[[:space:]]a[[:space:]]cement[[:space:]]ball._1.png filter=lfs diff=lfs merge=lfs -text
+checkpoint-138000/coco_steps-40_cfg-5/coco_32_A[[:space:]]single[[:space:]]giraffe[[:space:]]standing[[:space:]]in[[:space:]]the[[:space:]]middle[[:space:]]of[[:space:]]tall[[:space:]]grass_1.png filter=lfs diff=lfs merge=lfs -text
+checkpoint-138000/coco_steps-40_cfg-5/coco_33_the[[:space:]]men[[:space:]]play[[:space:]]soccer[[:space:]]on[[:space:]]the[[:space:]]beach[[:space:]]with[[:space:]]no[[:space:]]shoes_1.png filter=lfs diff=lfs merge=lfs -text
+checkpoint-138000/coco_steps-40_cfg-5/coco_3_An[[:space:]]old[[:space:]]fashioned[[:space:]]oxitue[[:space:]]worth[[:space:]]old[[:space:]]cars[[:space:]]on[[:space:]]street[[:space:]]of[[:space:]]town_1.png filter=lfs diff=lfs merge=lfs -text
+checkpoint-138000/coco_steps-40_cfg-5/coco_4_Two[[:space:]]men[[:space:]]are[[:space:]]in[[:space:]]a[[:space:]]building[[:space:]]with[[:space:]]brick[[:space:]]walls._1.png filter=lfs diff=lfs merge=lfs -text
+checkpoint-138000/coco_steps-40_cfg-5/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-138000/coco_steps-40_cfg-5/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-138000/coco_steps-40_cfg-5/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-138000/trainer_state.json filter=lfs diff=lfs merge=lfs -text
+checkpoint-140000/trainer_state.json filter=lfs diff=lfs merge=lfs -text

checkpoint-138000/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-138000/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-138000/model-00001-of-00002.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d9304880be7ea17765bed974cb5e2f15d0c3ef87359de305698e65a71fe30a32
+size 4974360904

checkpoint-138000/model-00002-of-00002.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ee0fc797f690144fdef0ec99c6d621ed7a7e1627b3d9ef0340bb868c279ab6bd
+size 3234250886

checkpoint-138000/optimizer.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d4d30073b751b4e781369baede3ccf8dec66dda5f2455009c2e6aa2e9b82d0f0
+size 7455061885

checkpoint-138000/rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:af92edd6096bfd3772f9dc195ef69b291a4e29b5e326fc502708ea859db8109e
+size 15024

checkpoint-138000/rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a4884d49c322abc8149ded2641dd29f9308bb89b9c5b813626346a4d9daaf6b9
+size 15024

checkpoint-138000/rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dedf957f0943c3b35e3075e7d20d6aeb54681a4c94af88412a6667ab39d16e88
+size 15024

checkpoint-138000/rng_state_3.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a1906d2ca02a19227b7c566f6ac17aa33ead13c56d42d8170a2258b36736f3f7
+size 15024

checkpoint-138000/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:37fb23e1d842e9514301374aeb0eb2537984b09f62c819ccb746d851435feabc
+size 1064

checkpoint-138000/trainer_state.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0d49206a8129037d93322db80fb99d365d9e6fdfa6952e67e5651a27d29d050e
+size 21242544

checkpoint-138000/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d1c49505c345575e51b0123d9223ddd9263d76b83bc110217e53e85a3e8e90b4
+size 6008