Upload folder using huggingface_hub

.gitattributes

@@ -33,3 +33,5 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+processor/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+tokenizer/tokenizer.json filter=lfs diff=lfs merge=lfs -text

.meta.json

@@ -0,0 +1,5 @@
+{
+  "diffusers": "0.40.0.dev0",
+  "torch": "2.13.0+cu126",
+  "transformers": "5.15.0.dev0"
+}

README.md

@@ -0,0 +1,355 @@
+---
+library_name: diffusers
+pipeline_tag: image-text-to-video
+base_model:
+- MiniMaxAI/MiniMax-H3
+---
+
+This tiny model is for debugging. It is randomly initialized with the config adapted from [MiniMaxAI/MiniMax-H3](https://huggingface.co/MiniMaxAI/MiniMax-H3).
+
+File size:
+- ~12MB text_encoder/model.safetensors
+- ~0.5MB transformer/diffusion_pytorch_model.safetensors
+- ~0.5MB transformer_ref/diffusion_pytorch_model.safetensors
+- ~4MB vae/diffusion_pytorch_model.safetensors
+- ~64MB audio_vae/diffusion_pytorch_model.safetensors
+
+Requires `diffusers` from source (`main`) with MiniMax-H3 modular blocks, and a recent `transformers` that ships `Qwen3VLForConditionalGeneration`.
+
+Notes:
+- The conditioner must have `num_hidden_layers > 50` because MiniMax-H3 reads `hidden_states[50]`.
+- Transformer RoPE needs `attention_head_dim >= 6 * rope_freq_dim`.
+- Audio VAE decoder needs `decoder_dim >= 128` for the released 7-stage upsample stack.
+
+| File path | Size |
+|------|------|
+| audio_vae/diffusion_pytorch_model.safetensors | 66.7MB |
+| text_encoder/model.safetensors | 12.1MB |
+| transformer/diffusion_pytorch_model.safetensors | 0.5MB |
+| transformer_ref/diffusion_pytorch_model.safetensors | 0.5MB |
+| vae/diffusion_pytorch_model.safetensors | 4.5MB |
+
+
+### Example usage:
+
+```python
+import torch
+from diffusers import ModularPipeline
+
+model_id = "tiny-random/minimax-h3"
+device = 'cuda' if torch.cuda.is_available() else 'cpu'
+pipe = ModularPipeline.from_pretrained(model_id, workflow='t2va')
+pipe.load_components(dtype=torch.bfloat16)
+if device == 'cuda':
+    pipe.to(device)
+
+outputs = pipe(
+    prompt='A red fox trotting through a snowy pine forest',
+    num_frames=124,
+    height=64,
+    width=64,
+    num_inference_steps=2,
+    generator=torch.Generator(device=device).manual_seed(42),
+    output=['videos', 'audio', 'sampling_rate'],
+)
+print(type(outputs['videos'][0]), getattr(outputs['videos'][0], 'shape', None))
+print(type(outputs['audio'][0]), getattr(outputs['audio'][0], 'shape', None), outputs['sampling_rate'])
+```
+
+### Codes to create this repo:
+
+```python
+import json
+from pathlib import Path
+
+import torch
+from diffusers import (
+    AutoencoderKLMiniMaxH3,
+    AutoencoderKLMiniMaxH3Audio,
+    MiniMaxH3Blocks,
+    MiniMaxH3Scheduler,
+    MiniMaxH3Transformer3DModel,
+)
+from huggingface_hub import hf_hub_download
+from transformers import AutoConfig, AutoProcessor, AutoTokenizer, Qwen3VLForConditionalGeneration
+
+source_model_id = "MiniMaxAI/MiniMax-H3"
+save_folder = "/tmp/tiny-random/minimax-h3"
+
+def save_json(path, obj):
+    Path(path).parent.mkdir(parents=True, exist_ok=True)
+    with open(path, 'w', encoding='utf-8') as f:
+        json.dump(obj, f, indent=2, ensure_ascii=False)
+
+def init_weights(model):
+    torch.manual_seed(42)
+    model = model.cpu()
+    with torch.no_grad():
+        for name, p in sorted(model.named_parameters()):
+            torch.nn.init.normal_(p, 0, 0.1)
+            print(name, p.shape, p.dtype, p.device)
+
+torch.set_default_dtype(torch.bfloat16)
+text_dim = 32
+Path(save_folder).mkdir(parents=True, exist_ok=True)
+
+AutoTokenizer.from_pretrained(source_model_id, subfolder='tokenizer').save_pretrained(
+    f'{save_folder}/tokenizer'
+)
+AutoProcessor.from_pretrained(source_model_id, subfolder='processor').save_pretrained(
+    f'{save_folder}/processor'
+)
+
+with open(hf_hub_download(source_model_id, filename='text_encoder/config.json', repo_type='model'), 'r', encoding='utf-8') as f:
+    config = json.load(f)
+# MiniMax-H3 conditions on hidden_states[50], so keep >50 layers with a tiny width.
+config['text_config'].update({
+    'head_dim': 8,
+    'hidden_size': text_dim,
+    'intermediate_size': 64,
+    'num_attention_heads': 4,
+    'num_key_value_heads': 2,
+    'num_hidden_layers': 51,
+    'tie_word_embeddings': True,
+})
+config['text_config']['rope_scaling']['mrope_section'] = [2, 1, 1]
+config['vision_config'].update({
+    'depth': 4,
+    'hidden_size': 64,
+    'intermediate_size': 128,
+    'num_heads': 4,
+    'out_hidden_size': text_dim,
+    'deepstack_visual_indexes': [1, 2, 3],
+})
+config['tie_word_embeddings'] = True
+save_json(f'{save_folder}/text_encoder/config.json', config)
+text_encoder = Qwen3VLForConditionalGeneration(
+    AutoConfig.from_pretrained(f'{save_folder}/text_encoder')
+).to(torch.bfloat16)
+init_weights(text_encoder)
+text_encoder.save_pretrained(f'{save_folder}/text_encoder')
+
+# attention_head_dim must cover 2 * 3 * rope_freq_dim rotary channels.
+transformer_kwargs = dict(
+    num_attention_heads=2,
+    attention_head_dim=32,
+    hidden_size=64,
+    num_layers=2,
+    num_refiner_layers=1,
+    ffn_dim=128,
+    in_channels=8,
+    audio_in_channels=8,
+    patch_size=(1, 2, 2),
+    text_dim=text_dim,
+    freq_dim=64,
+    time_embed_hidden_dim=64,
+    time_embed_dim=32,
+    rope_freq_dim=4,
+)
+for subfolder in ('transformer', 'transformer_ref'):
+    transformer = MiniMaxH3Transformer3DModel(**transformer_kwargs)
+    init_weights(transformer)
+    transformer.save_pretrained(f'{save_folder}/{subfolder}')
+
+with open(hf_hub_download(source_model_id, filename='vae/config.json', repo_type='model'), 'r', encoding='utf-8') as f:
+    vae_config = json.load(f)
+vae_config.update({
+    'latent_channels': 8,
+    'block_out_channels': [32, 32, 32, 64, 64, 64],
+    'layers_per_block': 1,
+    'spatial_downsample_factors': [2, 2, 2, 2, 1, 1],
+    'temporal_downsample_factors': [1, 2, 2, 1, 1, 1],
+    'norm_num_groups': 8,
+    'decoder_num_layers': 2,
+    'decoder_num_attention_heads': 2,
+    'decoder_attention_head_dim': 16,
+    'decoder_num_register_tokens': 2,
+    'decoder_ffn_mult': 2,
+    'latents_mean': [0.0] * 8,
+    'latents_std': [1.0] * 8,
+})
+save_json(f'{save_folder}/vae/config.json', vae_config)
+vae = AutoencoderKLMiniMaxH3.from_config(
+    AutoencoderKLMiniMaxH3.load_config(f'{save_folder}/vae')
+)
+init_weights(vae)
+vae.save_pretrained(f'{save_folder}/vae')
+
+# Keep hop length 800 (=32000/40Hz). decoder_dim must stay >= 128 for 7 upsample stages.
+with open(hf_hub_download(source_model_id, filename='audio_vae/config.json', repo_type='model'), 'r', encoding='utf-8') as f:
+    audio_config = json.load(f)
+audio_config.update({
+    'encoder_dim': 32,
+    'latent_dim': 128,
+    'latent_channels': 8,
+    'num_attention_heads': 4,
+    'decoder_dim': 128,
+    'latents_mean': [0.0] * 8,
+    'latents_std': [1.0] * 8,
+})
+save_json(f'{save_folder}/audio_vae/config.json', audio_config)
+audio_vae = AutoencoderKLMiniMaxH3Audio.from_config(
+    AutoencoderKLMiniMaxH3Audio.load_config(f'{save_folder}/audio_vae')
+)
+init_weights(audio_vae)
+audio_vae.save_pretrained(f'{save_folder}/audio_vae')
+
+MiniMaxH3Scheduler.from_pretrained(source_model_id, subfolder='scheduler').save_pretrained(
+    f'{save_folder}/scheduler'
+)
+MiniMaxH3Scheduler.from_pretrained(source_model_id, subfolder='audio_scheduler').save_pretrained(
+    f'{save_folder}/audio_scheduler'
+)
+
+for index_name in ('model_index.json', 'modular_model_index.json'):
+    index = json.load(open(hf_hub_download(source_model_id, filename=index_name, repo_type='model'), encoding='utf-8'))
+    for value in index.values():
+        if isinstance(value, list) and len(value) >= 3 and isinstance(value[2], dict):
+            value[2]['pretrained_model_name_or_path'] = save_folder
+    save_json(f'{save_folder}/{index_name}', index)
+
+pipe = MiniMaxH3Blocks().init_pipeline(save_folder)
+pipe.load_components(dtype=torch.bfloat16)
+pipe.save_pretrained(save_folder, safe_serialization=True, overwrite_modular_index=True)
+torch.set_default_dtype(torch.float32)
+print(pipe)
+```
+
+### Printing the model:
+
+```text
+MiniMaxH3ModularPipeline {
+  "_blocks_class_name": "MiniMaxH3Blocks",
+  "_class_name": "MiniMaxH3ModularPipeline",
+  "_diffusers_version": "0.40.0.dev0",
+  "audio_scheduler": [
+    "diffusers",
+    "MiniMaxH3Scheduler",
+    {
+      "pretrained_model_name_or_path": "./tmp/tiny-random/minimax-h3",
+      "revision": null,
+      "subfolder": "audio_scheduler",
+      "type_hint": [
+        "diffusers",
+        "MiniMaxH3Scheduler"
+      ],
+      "variant": null
+    }
+  ],
+  "audio_vae": [
+    "diffusers",
+    "AutoencoderKLMiniMaxH3Audio",
+    {
+      "pretrained_model_name_or_path": "./tmp/tiny-random/minimax-h3",
+      "revision": null,
+      "subfolder": "audio_vae",
+      "type_hint": [
+        "diffusers",
+        "AutoencoderKLMiniMaxH3Audio"
+      ],
+      "variant": null
+    }
+  ],
+  "canvas_max_pixels": 1032192,
+  "canvas_short_edge": 768,
+  "processor": [
+    "transformers",
+    "Qwen3VLProcessor",
+    {
+      "pretrained_model_name_or_path": "./tmp/tiny-random/minimax-h3",
+      "revision": null,
+      "subfolder": "processor",
+      "type_hint": [
+        "transformers",
+        "Qwen3VLProcessor"
+      ],
+      "variant": null
+    }
+  ],
+  "reference_image_short_edge": 2048,
+  "scheduler": [
+    "diffusers",
+    "MiniMaxH3Scheduler",
+    {
+      "pretrained_model_name_or_path": "./tmp/tiny-random/minimax-h3",
+      "revision": null,
+      "subfolder": "scheduler",
+      "type_hint": [
+        "diffusers",
+        "MiniMaxH3Scheduler"
+      ],
+      "variant": null
+    }
+  ],
+  "text_encoder": [
+    "transformers",
+    "Qwen3VLForConditionalGeneration",
+    {
+      "pretrained_model_name_or_path": "./tmp/tiny-random/minimax-h3",
+      "revision": null,
+      "subfolder": "text_encoder",
+      "type_hint": [
+        "transformers",
+        "Qwen3VLForConditionalGeneration"
+      ],
+      "variant": null
+    }
+  ],
+  "tokenizer": [
+    "transformers",
+    "Qwen2Tokenizer",
+    {
+      "pretrained_model_name_or_path": "./tmp/tiny-random/minimax-h3",
+      "revision": null,
+      "subfolder": "tokenizer",
+      "type_hint": [
+        "transformers",
+        "Qwen2Tokenizer"
+      ],
+      "variant": null
+    }
+  ],
+  "transformer": [
+    "diffusers",
+    "MiniMaxH3Transformer3DModel",
+    {
+      "pretrained_model_name_or_path": "./tmp/tiny-random/minimax-h3",
+      "revision": null,
+      "subfolder": "transformer",
+      "type_hint": [
+        "diffusers",
+        "MiniMaxH3Transformer3DModel"
+      ],
+      "variant": null
+    }
+  ],
+  "transformer_ref": [
+    "diffusers",
+    "MiniMaxH3Transformer3DModel",
+    {
+      "pretrained_model_name_or_path": "./tmp/tiny-random/minimax-h3",
+      "revision": null,
+      "subfolder": "transformer_ref",
+      "type_hint": [
+        "diffusers",
+        "MiniMaxH3Transformer3DModel"
+      ],
+      "variant": null
+    }
+  ],
+  "vae": [
+    "diffusers",
+    "AutoencoderKLMiniMaxH3",
+    {
+      "pretrained_model_name_or_path": "./tmp/tiny-random/minimax-h3",
+      "revision": null,
+      "subfolder": "vae",
+      "type_hint": [
+        "diffusers",
+        "AutoencoderKLMiniMaxH3"
+      ],
+      "variant": null
+    }
+  ]
+}
+```

audio_scheduler/scheduler_config.json

@@ -0,0 +1,5 @@
+{
+  "_class_name": "MiniMaxH3Scheduler",
+  "_diffusers_version": "0.40.0.dev0",
+  "shift": 3.0
+}

audio_vae/config.json

@@ -0,0 +1,78 @@
+{
+  "_class_name": "AutoencoderKLMiniMaxH3Audio",
+  "_diffusers_version": "0.40.0.dev0",
+  "_name_or_path": "./tmp/tiny-random/minimax-h3",
+  "decoder_dim": 128,
+  "decoder_kernel_sizes": [
+    9,
+    9,
+    4,
+    4,
+    4,
+    4,
+    4
+  ],
+  "decoder_rates": [
+    5,
+    5,
+    2,
+    2,
+    2,
+    2,
+    2
+  ],
+  "encoder_dim": 32,
+  "encoder_rates": [
+    2,
+    4,
+    4,
+    5,
+    5
+  ],
+  "latent_channels": 8,
+  "latent_dim": 128,
+  "latents_mean": [
+    0.0,
+    0.0,
+    0.0,
+    0.0,
+    0.0,
+    0.0,
+    0.0,
+    0.0
+  ],
+  "latents_std": [
+    1.0,
+    1.0,
+    1.0,
+    1.0,
+    1.0,
+    1.0,
+    1.0,
+    1.0
+  ],
+  "num_attention_heads": 4,
+  "resblock_dilation_sizes": [
+    [
+      1,
+      3,
+      5
+    ],
+    [
+      1,
+      3,
+      5
+    ],
+    [
+      1,
+      3,
+      5
+    ]
+  ],
+  "resblock_kernel_sizes": [
+    3,
+    7,
+    11
+  ],
+  "sampling_rate": 32000
+}

audio_vae/diffusion_pytorch_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5a9928676863a1a8c2a9ec549e5cc85f7ad986587c5ec1b7597aaa5db2c59a3c
+size 66721604

model_index.json

@@ -0,0 +1,131 @@
+{
+  "_class_name": "MiniMaxH3ModularPipeline",
+  "_diffusers_version": "0.36.0.dev0",
+  "_blocks_class_name": "MiniMaxH3Blocks",
+  "text_encoder": [
+    "transformers",
+    "Qwen3VLForConditionalGeneration",
+    {
+      "type_hint": [
+        "transformers",
+        "Qwen3VLForConditionalGeneration"
+      ],
+      "pretrained_model_name_or_path": "./tmp/tiny-random/minimax-h3",
+      "subfolder": "text_encoder",
+      "variant": null,
+      "revision": null
+    }
+  ],
+  "tokenizer": [
+    "transformers",
+    "Qwen2TokenizerFast",
+    {
+      "type_hint": [
+        "transformers",
+        "Qwen2TokenizerFast"
+      ],
+      "pretrained_model_name_or_path": "./tmp/tiny-random/minimax-h3",
+      "subfolder": "tokenizer",
+      "variant": null,
+      "revision": null
+    }
+  ],
+  "processor": [
+    "transformers",
+    "Qwen3VLProcessor",
+    {
+      "type_hint": [
+        "transformers",
+        "Qwen3VLProcessor"
+      ],
+      "pretrained_model_name_or_path": "./tmp/tiny-random/minimax-h3",
+      "subfolder": "processor",
+      "variant": null,
+      "revision": null
+    }
+  ],
+  "vae": [
+    "diffusers",
+    "AutoencoderKLMiniMaxH3",
+    {
+      "type_hint": [
+        "diffusers",
+        "AutoencoderKLMiniMaxH3"
+      ],
+      "pretrained_model_name_or_path": "./tmp/tiny-random/minimax-h3",
+      "subfolder": "vae",
+      "variant": null,
+      "revision": null
+    }
+  ],
+  "audio_vae": [
+    "diffusers",
+    "AutoencoderKLMiniMaxH3Audio",
+    {
+      "type_hint": [
+        "diffusers",
+        "AutoencoderKLMiniMaxH3Audio"
+      ],
+      "pretrained_model_name_or_path": "./tmp/tiny-random/minimax-h3",
+      "subfolder": "audio_vae",
+      "variant": null,
+      "revision": null
+    }
+  ],
+  "transformer": [
+    "diffusers",
+    "MiniMaxH3Transformer3DModel",
+    {
+      "type_hint": [
+        "diffusers",
+        "MiniMaxH3Transformer3DModel"
+      ],
+      "pretrained_model_name_or_path": "./tmp/tiny-random/minimax-h3",
+      "subfolder": "transformer",
+      "variant": null,
+      "revision": null
+    }
+  ],
+  "transformer_ref": [
+    "diffusers",
+    "MiniMaxH3Transformer3DModel",
+    {
+      "type_hint": [
+        "diffusers",
+        "MiniMaxH3Transformer3DModel"
+      ],
+      "pretrained_model_name_or_path": "./tmp/tiny-random/minimax-h3",
+      "subfolder": "transformer_ref",
+      "variant": null,
+      "revision": null
+    }
+  ],
+  "scheduler": [
+    "diffusers",
+    "MiniMaxH3Scheduler",
+    {
+      "type_hint": [
+        "diffusers",
+        "MiniMaxH3Scheduler"
+      ],
+      "pretrained_model_name_or_path": "./tmp/tiny-random/minimax-h3",
+      "subfolder": "scheduler",
+      "variant": null,
+      "revision": null
+    }
+  ],
+  "audio_scheduler": [
+    "diffusers",
+    "MiniMaxH3Scheduler",
+    {
+      "type_hint": [
+        "diffusers",
+        "MiniMaxH3Scheduler"
+      ],
+      "pretrained_model_name_or_path": "./tmp/tiny-random/minimax-h3",
+      "subfolder": "audio_scheduler",
+      "variant": null,
+      "revision": null
+    }
+  ]
+}

modular_model_index.json

@@ -0,0 +1,134 @@
+{
+  "_blocks_class_name": "MiniMaxH3Blocks",
+  "_class_name": "MiniMaxH3ModularPipeline",
+  "_diffusers_version": "0.40.0.dev0",
+  "audio_scheduler": [
+    "diffusers",
+    "MiniMaxH3Scheduler",
+    {
+      "pretrained_model_name_or_path": "./tmp/tiny-random/minimax-h3",
+      "revision": null,
+      "subfolder": "audio_scheduler",
+      "type_hint": [
+        "diffusers",
+        "MiniMaxH3Scheduler"
+      ],
+      "variant": null
+    }
+  ],
+  "audio_vae": [
+    "diffusers",
+    "AutoencoderKLMiniMaxH3Audio",
+    {
+      "pretrained_model_name_or_path": "./tmp/tiny-random/minimax-h3",
+      "revision": null,
+      "subfolder": "audio_vae",
+      "type_hint": [
+        "diffusers",
+        "AutoencoderKLMiniMaxH3Audio"
+      ],
+      "variant": null
+    }
+  ],
+  "canvas_max_pixels": 1032192,
+  "canvas_short_edge": 768,
+  "processor": [
+    "transformers",
+    "Qwen3VLProcessor",
+    {
+      "pretrained_model_name_or_path": "./tmp/tiny-random/minimax-h3",
+      "revision": null,
+      "subfolder": "processor",
+      "type_hint": [
+        "transformers",
+        "Qwen3VLProcessor"
+      ],
+      "variant": null
+    }
+  ],
+  "reference_image_short_edge": 2048,
+  "scheduler": [
+    "diffusers",
+    "MiniMaxH3Scheduler",
+    {
+      "pretrained_model_name_or_path": "./tmp/tiny-random/minimax-h3",
+      "revision": null,
+      "subfolder": "scheduler",
+      "type_hint": [
+        "diffusers",
+        "MiniMaxH3Scheduler"
+      ],
+      "variant": null
+    }
+  ],
+  "text_encoder": [
+    "transformers",
+    "Qwen3VLForConditionalGeneration",
+    {
+      "pretrained_model_name_or_path": "./tmp/tiny-random/minimax-h3",
+      "revision": null,
+      "subfolder": "text_encoder",
+      "type_hint": [
+        "transformers",
+        "Qwen3VLForConditionalGeneration"
+      ],
+      "variant": null
+    }
+  ],
+  "tokenizer": [
+    "transformers",
+    "Qwen2Tokenizer",
+    {
+      "pretrained_model_name_or_path": "./tmp/tiny-random/minimax-h3",
+      "revision": null,
+      "subfolder": "tokenizer",
+      "type_hint": [
+        "transformers",
+        "Qwen2Tokenizer"
+      ],
+      "variant": null
+    }
+  ],
+  "transformer": [
+    "diffusers",
+    "MiniMaxH3Transformer3DModel",
+    {
+      "pretrained_model_name_or_path": "./tmp/tiny-random/minimax-h3",
+      "revision": null,
+      "subfolder": "transformer",
+      "type_hint": [
+        "diffusers",
+        "MiniMaxH3Transformer3DModel"
+      ],
+      "variant": null
+    }
+  ],
+  "transformer_ref": [
+    "diffusers",
+    "MiniMaxH3Transformer3DModel",
+    {
+      "pretrained_model_name_or_path": "./tmp/tiny-random/minimax-h3",
+      "revision": null,
+      "subfolder": "transformer_ref",
+      "type_hint": [
+        "diffusers",
+        "MiniMaxH3Transformer3DModel"
+      ],
+      "variant": null
+    }
+  ],
+  "vae": [
+    "diffusers",
+    "AutoencoderKLMiniMaxH3",
+    {
+      "pretrained_model_name_or_path": "./tmp/tiny-random/minimax-h3",
+      "revision": null,
+      "subfolder": "vae",
+      "type_hint": [
+        "diffusers",
+        "AutoencoderKLMiniMaxH3"
+      ],
+      "variant": null
+    }
+  ]
+}

processor/chat_template.jinja

@@ -0,0 +1,120 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {%- if messages[0].content is string %}
+            {{- messages[0].content }}
+        {%- else %}
+            {%- for content in messages[0].content %}
+                {%- if 'text' in content %}
+                    {{- content.text }}
+                {%- endif %}
+            {%- endfor %}
+        {%- endif %}
+        {{- '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' }}
+        {%- if messages[0].content is string %}
+            {{- messages[0].content }}
+        {%- else %}
+            {%- for content in messages[0].content %}
+                {%- if 'text' in content %}
+                    {{- content.text }}
+                {%- endif %}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- for message in messages %}
+    {%- if message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' }}
+        {%- if message.content is string %}
+            {{- message.content }}
+        {%- else %}
+            {%- for content in message.content %}
+                {%- if content.type == 'image' or 'image' in content or 'image_url' in content %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                    {%- if add_vision_id %}Picture {{ image_count.value }}: {% endif -%}
+                    <|vision_start|><|image_pad|><|vision_end|>
+                {%- elif content.type == 'video' or 'video' in content %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                    {%- if add_vision_id %}Video {{ video_count.value }}: {% endif -%}
+                    <|vision_start|><|video_pad|><|vision_end|>
+                {%- elif 'text' in content %}
+                    {{- content.text }}
+                {%- endif %}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "assistant" %}
+        {{- '<|im_start|>' + message.role + '\n' }}
+        {%- if message.content is string %}
+            {{- message.content }}
+        {%- else %}
+            {%- for content_item in message.content %}
+                {%- if 'text' in content_item %}
+                    {{- content_item.text }}
+                {%- endif %}
+            {%- endfor %}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and message.content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {%- if message.content is string %}
+            {{- message.content }}
+        {%- else %}
+            {%- for content in message.content %}
+                {%- if content.type == 'image' or 'image' in content or 'image_url' in content %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                    {%- if add_vision_id %}Picture {{ image_count.value }}: {% endif -%}
+                    <|vision_start|><|image_pad|><|vision_end|>
+                {%- elif content.type == 'video' or 'video' in content %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                    {%- if add_vision_id %}Video {{ video_count.value }}: {% endif -%}
+                    <|vision_start|><|video_pad|><|vision_end|>
+                {%- elif 'text' in content %}
+                    {{- content.text }}
+                {%- endif %}
+            {%- endfor %}
+        {%- endif %}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+{%- endif %}

processor/processor_config.json

@@ -0,0 +1,60 @@
+{
+  "image_processor": {
+    "do_convert_rgb": true,
+    "do_normalize": true,
+    "do_rescale": true,
+    "do_resize": true,
+    "image_mean": [
+      0.5,
+      0.5,
+      0.5
+    ],
+    "image_processor_type": "Qwen2VLImageProcessor",
+    "image_std": [
+      0.5,
+      0.5,
+      0.5
+    ],
+    "merge_size": 2,
+    "patch_size": 16,
+    "resample": 3,
+    "rescale_factor": 0.00392156862745098,
+    "size": {
+      "longest_edge": 16777216,
+      "shortest_edge": 65536
+    },
+    "temporal_patch_size": 2
+  },
+  "processor_class": "Qwen3VLProcessor",
+  "video_processor": {
+    "do_convert_rgb": true,
+    "do_normalize": true,
+    "do_rescale": true,
+    "do_resize": true,
+    "do_sample_frames": true,
+    "fps": 2,
+    "image_mean": [
+      0.5,
+      0.5,
+      0.5
+    ],
+    "image_std": [
+      0.5,
+      0.5,
+      0.5
+    ],
+    "max_frames": 768,
+    "merge_size": 2,
+    "min_frames": 4,
+    "patch_size": 16,
+    "resample": 3,
+    "rescale_factor": 0.00392156862745098,
+    "return_metadata": false,
+    "size": {
+      "longest_edge": 25165824,
+      "shortest_edge": 4096
+    },
+    "temporal_patch_size": 2,
+    "video_processor_type": "Qwen3VLVideoProcessor"
+  }
+}

processor/tokenizer.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:170c542379436ef9d4574ad325af3bf4c8c3ef295951fdc060a319a905bb7ff2
+size 11423968

processor/tokenizer_config.json

@@ -0,0 +1,39 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|im_end|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>",
+    "<d>",
+    "</d>",
+    "<|cutoff|>",
+    "<|lyrics_start|>",
+    "<|lyrics_end|>",
+    "<|caption_start|>",
+    "<|caption_end|>"
+  ],
+  "is_local": true,
+  "local_files_only": false,
+  "model_max_length": 262144,
+  "pad_token": "<|endoftext|>",
+  "processor_class": "Qwen3VLProcessor",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null,
+  "variant": null
+}

scheduler/scheduler_config.json

@@ -0,0 +1,5 @@
+{
+  "_class_name": "MiniMaxH3Scheduler",
+  "_diffusers_version": "0.40.0.dev0",
+  "shift": 12.0
+}

text_encoder/config.json

@@ -0,0 +1,66 @@
+{
+  "architectures": [
+    "Qwen3VLForConditionalGeneration"
+  ],
+  "dtype": "bfloat16",
+  "image_token_id": 151655,
+  "model_type": "qwen3_vl",
+  "text_config": {
+    "attention_bias": false,
+    "attention_dropout": 0.0,
+    "bos_token_id": 151643,
+    "dtype": "bfloat16",
+    "eos_token_id": 151645,
+    "head_dim": 8,
+    "hidden_act": "silu",
+    "hidden_size": 32,
+    "initializer_range": 0.02,
+    "intermediate_size": 64,
+    "max_position_embeddings": 262144,
+    "model_type": "qwen3_vl_text",
+    "num_attention_heads": 4,
+    "num_hidden_layers": 51,
+    "num_key_value_heads": 2,
+    "pad_token_id": null,
+    "rms_norm_eps": 1e-06,
+    "rope_parameters": {
+      "mrope_interleaved": true,
+      "mrope_section": [
+        2,
+        1,
+        1
+      ],
+      "rope_theta": 5000000,
+      "rope_type": "default"
+    },
+    "tie_word_embeddings": true,
+    "use_cache": true,
+    "vocab_size": 151936
+  },
+  "tie_word_embeddings": true,
+  "transformers_version": "5.15.0.dev0",
+  "video_token_id": 151656,
+  "vision_config": {
+    "deepstack_visual_indexes": [
+      1,
+      2,
+      3
+    ],
+    "depth": 4,
+    "dtype": "bfloat16",
+    "hidden_act": "gelu_pytorch_tanh",
+    "hidden_size": 64,
+    "in_channels": 3,
+    "initializer_range": 0.02,
+    "intermediate_size": 128,
+    "model_type": "qwen3_vl_vision",
+    "num_heads": 4,
+    "num_position_embeddings": 2304,
+    "out_hidden_size": 32,
+    "patch_size": 16,
+    "spatial_merge_size": 2,
+    "temporal_patch_size": 2
+  },
+  "vision_end_token_id": 151653,
+  "vision_start_token_id": 151652
+}

text_encoder/generation_config.json

@@ -0,0 +1,9 @@
+{
+  "_from_model_config": true,
+  "bos_token_id": 151643,
+  "eos_token_id": 151645,
+  "output_attentions": false,
+  "output_hidden_states": false,
+  "transformers_version": "5.15.0.dev0",
+  "use_cache": true
+}

text_encoder/model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:975ada26892c830c648bbe1322371a5aa646ad93d923720fbdef0d6023b0966e
+size 12104296

tokenizer/chat_template.jinja

@@ -0,0 +1,120 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {%- if messages[0].content is string %}
+            {{- messages[0].content }}
+        {%- else %}
+            {%- for content in messages[0].content %}
+                {%- if 'text' in content %}
+                    {{- content.text }}
+                {%- endif %}
+            {%- endfor %}
+        {%- endif %}
+        {{- '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' }}
+        {%- if messages[0].content is string %}
+            {{- messages[0].content }}
+        {%- else %}
+            {%- for content in messages[0].content %}
+                {%- if 'text' in content %}
+                    {{- content.text }}
+                {%- endif %}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- set image_count = namespace(value=0) %}
+{%- set video_count = namespace(value=0) %}
+{%- for message in messages %}
+    {%- if message.role == "user" %}
+        {{- '<|im_start|>' + message.role + '\n' }}
+        {%- if message.content is string %}
+            {{- message.content }}
+        {%- else %}
+            {%- for content in message.content %}
+                {%- if content.type == 'image' or 'image' in content or 'image_url' in content %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                    {%- if add_vision_id %}Picture {{ image_count.value }}: {% endif -%}
+                    <|vision_start|><|image_pad|><|vision_end|>
+                {%- elif content.type == 'video' or 'video' in content %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                    {%- if add_vision_id %}Video {{ video_count.value }}: {% endif -%}
+                    <|vision_start|><|video_pad|><|vision_end|>
+                {%- elif 'text' in content %}
+                    {{- content.text }}
+                {%- endif %}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "assistant" %}
+        {{- '<|im_start|>' + message.role + '\n' }}
+        {%- if message.content is string %}
+            {{- message.content }}
+        {%- else %}
+            {%- for content_item in message.content %}
+                {%- if 'text' in content_item %}
+                    {{- content_item.text }}
+                {%- endif %}
+            {%- endfor %}
+        {%- endif %}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and message.content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {%- if message.content is string %}
+            {{- message.content }}
+        {%- else %}
+            {%- for content in message.content %}
+                {%- if content.type == 'image' or 'image' in content or 'image_url' in content %}
+                    {%- set image_count.value = image_count.value + 1 %}
+                    {%- if add_vision_id %}Picture {{ image_count.value }}: {% endif -%}
+                    <|vision_start|><|image_pad|><|vision_end|>
+                {%- elif content.type == 'video' or 'video' in content %}
+                    {%- set video_count.value = video_count.value + 1 %}
+                    {%- if add_vision_id %}Video {{ video_count.value }}: {% endif -%}
+                    <|vision_start|><|video_pad|><|vision_end|>
+                {%- elif 'text' in content %}
+                    {{- content.text }}
+                {%- endif %}
+            {%- endfor %}
+        {%- endif %}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+{%- endif %}

tokenizer/tokenizer.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:170c542379436ef9d4574ad325af3bf4c8c3ef295951fdc060a319a905bb7ff2
+size 11423968

tokenizer/tokenizer_config.json

@@ -0,0 +1,38 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|im_end|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>",
+    "<d>",
+    "</d>",
+    "<|cutoff|>",
+    "<|lyrics_start|>",
+    "<|lyrics_end|>",
+    "<|caption_start|>",
+    "<|caption_end|>"
+  ],
+  "is_local": true,
+  "local_files_only": false,
+  "model_max_length": 262144,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null,
+  "variant": null
+}

transformer/config.json

@@ -0,0 +1,27 @@
+{
+  "_class_name": "MiniMaxH3Transformer3DModel",
+  "_diffusers_version": "0.40.0.dev0",
+  "_name_or_path": "./tmp/tiny-random/minimax-h3",
+  "attention_head_dim": 32,
+  "audio_in_channels": 8,
+  "ffn_dim": 128,
+  "final_norm_eps": 1e-05,
+  "freq_dim": 64,
+  "hidden_size": 64,
+  "in_channels": 8,
+  "norm_eps": 1e-05,
+  "num_attention_heads": 2,
+  "num_layers": 2,
+  "num_refiner_layers": 1,
+  "patch_size": [
+    1,
+    2,
+    2
+  ],
+  "qk_norm_eps": 1e-05,
+  "rope_freq_dim": 4,
+  "rope_theta": 10000.0,
+  "text_dim": 32,
+  "time_embed_dim": 32,
+  "time_embed_hidden_dim": 64
+}

transformer/diffusion_pytorch_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a17afecb6ae5e65d91c46a61ed19fd69e4717cbe6d125f6c6046fbf65d76f6fb
+size 463184

transformer_ref/config.json

@@ -0,0 +1,27 @@
+{
+  "_class_name": "MiniMaxH3Transformer3DModel",
+  "_diffusers_version": "0.40.0.dev0",
+  "_name_or_path": "./tmp/tiny-random/minimax-h3",
+  "attention_head_dim": 32,
+  "audio_in_channels": 8,
+  "ffn_dim": 128,
+  "final_norm_eps": 1e-05,
+  "freq_dim": 64,
+  "hidden_size": 64,
+  "in_channels": 8,
+  "norm_eps": 1e-05,
+  "num_attention_heads": 2,
+  "num_layers": 2,
+  "num_refiner_layers": 1,
+  "patch_size": [
+    1,
+    2,
+    2
+  ],
+  "qk_norm_eps": 1e-05,
+  "rope_freq_dim": 4,
+  "rope_theta": 10000.0,
+  "text_dim": 32,
+  "time_embed_dim": 32,
+  "time_embed_hidden_dim": 64
+}

transformer_ref/diffusion_pytorch_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a17afecb6ae5e65d91c46a61ed19fd69e4717cbe6d125f6c6046fbf65d76f6fb
+size 463184

vae/config.json

@@ -0,0 +1,66 @@
+{
+  "_class_name": "AutoencoderKLMiniMaxH3",
+  "_diffusers_version": "0.40.0.dev0",
+  "_name_or_path": "./tmp/tiny-random/minimax-h3",
+  "block_out_channels": [
+    32,
+    32,
+    32,
+    64,
+    64,
+    64
+  ],
+  "clip_length": 17,
+  "decoder_attention_head_dim": 16,
+  "decoder_ffn_mult": 2,
+  "decoder_norm_eps": 1e-05,
+  "decoder_num_attention_heads": 2,
+  "decoder_num_layers": 2,
+  "decoder_num_register_tokens": 2,
+  "decoder_rope_dim_ratio": 0.75,
+  "decoder_rope_theta": 100.0,
+  "in_channels": 3,
+  "latent_channels": 8,
+  "latents_mean": [
+    0.0,
+    0.0,
+    0.0,
+    0.0,
+    0.0,
+    0.0,
+    0.0,
+    0.0
+  ],
+  "latents_std": [
+    1.0,
+    1.0,
+    1.0,
+    1.0,
+    1.0,
+    1.0,
+    1.0,
+    1.0
+  ],
+  "layers_per_block": 1,
+  "norm_eps": 1e-06,
+  "norm_num_groups": 8,
+  "out_channels": 3,
+  "spatial_downsample_factors": [
+    2,
+    2,
+    2,
+    2,
+    1,
+    1
+  ],
+  "spatial_padding_mode": "reflect",
+  "temporal_downsample_factors": [
+    1,
+    2,
+    2,
+    1,
+    1,
+    1
+  ],
+  "token_drop": 3
+}

vae/diffusion_pytorch_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1d7653f9103b32fe03919fbcca9482fb2cf0ee95dc8423c668baf11a67bb068c
+size 4513304