From 07d6dc9694e0ea942683bf6e3e374887d9f5b054 Mon Sep 17 00:00:00 2001 From: huiwenshi Date: Fri, 13 Jun 2025 16:23:40 +0000 Subject: [PATCH] Upload folder using huggingface_hub --- hunyuan3d-dit-v2-1/config.yaml | 82 ++++++++++++++++++++++++++++++ hunyuan3d-dit-v2-1/model.fp16.ckpt | 3 ++ hunyuan3d-vae-v2-1/config.yaml | 19 +++++++ hunyuan3d-vae-v2-1/model.fp16.ckpt | 3 ++ 4 files changed, 107 insertions(+) create mode 100644 hunyuan3d-dit-v2-1/config.yaml create mode 100644 hunyuan3d-dit-v2-1/model.fp16.ckpt create mode 100644 hunyuan3d-vae-v2-1/config.yaml create mode 100644 hunyuan3d-vae-v2-1/model.fp16.ckpt diff --git a/hunyuan3d-dit-v2-1/config.yaml b/hunyuan3d-dit-v2-1/config.yaml new file mode 100644 index 0000000..695b7d5 --- /dev/null +++ b/hunyuan3d-dit-v2-1/config.yaml @@ -0,0 +1,82 @@ +model: + target: hy3dshape.models.denoisers.hunyuandit.HunYuanDiTPlain + params: + input_size: &num_latents 4096 + in_channels: 64 + hidden_size: 2048 + context_dim: 1024 + depth: 21 + num_heads: 16 + qk_norm: true + text_len: 1370 + with_decoupled_ca: false + use_attention_pooling: false + qk_norm_type: 'rms' + qkv_bias: false + use_pos_emb: false + num_moe_layers: 6 + num_experts: 8 + moe_top_k: 2 + +vae: + target: hy3dshape.models.autoencoders.ShapeVAE + params: + num_latents: *num_latents + embed_dim: 64 + num_freqs: 8 + include_pi: false + heads: 16 + width: 1024 + num_encoder_layers: 8 + num_decoder_layers: 16 + qkv_bias: false + qk_norm: true + scale_factor: 1.0039506158752403 + geo_decoder_mlp_expand_ratio: 4 + geo_decoder_downsample_ratio: 1 + geo_decoder_ln_post: true + point_feats: 4 + pc_size: 81920 + pc_sharpedge_size: 0 + +conditioner: + target: hy3dshape.models.conditioner.SingleImageEncoder + params: + main_image_encoder: + type: DinoImageEncoder # dino large + kwargs: + config: + attention_probs_dropout_prob: 0.0 + drop_path_rate: 0.0 + hidden_act: gelu + hidden_dropout_prob: 0.0 + hidden_size: 1024 + image_size: 518 + initializer_range: 0.02 + layer_norm_eps: 1.e-6 + layerscale_value: 1.0 + mlp_ratio: 4 + model_type: dinov2 + num_attention_heads: 16 + num_channels: 3 + num_hidden_layers: 24 + patch_size: 14 + qkv_bias: true + torch_dtype: float32 + use_swiglu_ffn: false + image_size: 518 + use_cls_token: true + +scheduler: + target: hy3dshape.schedulers.FlowMatchEulerDiscreteScheduler + params: + num_train_timesteps: 1000 + +image_processor: + target: hy3dshape.preprocessors.ImageProcessorV2 + params: + size: 512 + border_ratio: 0.15 + +pipeline: + target: hy3dshape.pipelines.Hunyuan3DDiTFlowMatchingPipeline diff --git a/hunyuan3d-dit-v2-1/model.fp16.ckpt b/hunyuan3d-dit-v2-1/model.fp16.ckpt new file mode 100644 index 0000000..3dd05dc --- /dev/null +++ b/hunyuan3d-dit-v2-1/model.fp16.ckpt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:6b519fc7242f78e9b5f47ea4d55668fe3d944a2d27332f4ca68d29a6ff603f5e +size 7366389768 diff --git a/hunyuan3d-vae-v2-1/config.yaml b/hunyuan3d-vae-v2-1/config.yaml new file mode 100644 index 0000000..2421cf0 --- /dev/null +++ b/hunyuan3d-vae-v2-1/config.yaml @@ -0,0 +1,19 @@ +target: hy3dshape.models.ShapeVAE +params: + num_latents: 4096 + embed_dim: 64 + num_freqs: 8 + include_pi: false + heads: 16 + width: 1024 + num_encoder_layers: 8 + num_decoder_layers: 16 + qkv_bias: false + qk_norm: true + scale_factor: 1.0039506158752403 + geo_decoder_mlp_expand_ratio: 4 + geo_decoder_downsample_ratio: 1 + geo_decoder_ln_post: true + point_feats: 4 + pc_size: 81920 + pc_sharpedge_size: 0 \ No newline at end of file diff --git a/hunyuan3d-vae-v2-1/model.fp16.ckpt b/hunyuan3d-vae-v2-1/model.fp16.ckpt new file mode 100644 index 0000000..6ae3b35 --- /dev/null +++ b/hunyuan3d-vae-v2-1/model.fp16.ckpt @@ -0,0 +1,3 @@ +version https://git-lfs.github.com/spec/v1 +oid sha256:5cbe97f25e6e7abd4bccc80ab07524ec0c86d24118486a9ba49bb5dfb070288a +size 655648152