ModelPrésentéExterne

Depth LoRA Checkpoints Diode Diode Outdoor

> Base model: metric3dvitgiant2 > loaded via torch.hub.load("yvanyin/metric3d", "metric3dvitgiant2")

Sourcehuggingface.co/igzi/depth-lora-checkpoints_diode-diode_outdoor/tree/1a21a1a94d3d50a27f56627c09f05e3e2ce4e248

TypeLoRAVersionadapter_modelBClasse

Source du modèle

Type: LoRA
Taille du fichier: 22,7 MB
Version: vadapter_model
Mis à jour: 20 oct. 2026
Téléchargements: 4
Source: Hugging Face

Description de la source

Metric3D ViT-giant2 — LoRA adapter (DIODE Outdoor)

Base model: metric3d_vit_giant2 loaded via torch.hub.load("yvanyin/metric3d", "metric3d_vit_giant2")

LoRA fine-tuning of Metric3D v2 (metric3d_vit_giant2) on the DIODE outdoor split (metric depth in metres).

Training details


Base model	`metric3d_vit_giant2` (torch.hub `yvanyin/metric3d`)
Dataset	the DIODE outdoor split (metric depth in metres)
Loss	Direct metric-depth L1 + gradient loss
LoRA rank / alpha	16 / 32
LoRA targets	`qkv`, `proj`

Usage

import torch
from peft import PeftModel
import torch.nn as nn
import torch.utils.checkpoint as torch_checkpoint

# Load base model
model = torch.hub.load("yvanyin/metric3d", "metric3d_vit_giant2",
                       pretrain=True, trust_repo=True)

# Apply the same gradient-checkpointing wrapper used during training
# (needed so PEFT key names match the saved adapter)
def enable_gradient_checkpointing(model):
    try:
        encoder = model.depth_model.encoder
    except AttributeError:
        encoder = model.base_model.model.depth_model.encoder

    class _CheckpointedBlock(nn.Module):
        def __init__(self, block):
            super().__init__()
            self.block = block
        def forward(self, x):
            return torch_checkpoint.checkpoint(self.block, x, use_reentrant=False)

    for blk_group in encoder.blocks:
        for key in list(blk_group._modules.keys()):
            blk_group._modules[key] = _CheckpointedBlock(blk_group._modules[key])

enable_gradient_checkpointing(model)
model = PeftModel.from_pretrained(model, "igzi/depth-lora-checkpoints_diode-diode_outdoor")
model.eval()

# Inference: input pixel_values shape (B, 3, 616, 1064), values normalised
# with mean=[123.675, 116.28, 103.53], std=[58.395, 57.12, 57.375]
pred_canonical, _, _ = model({"input": pixel_values})
# De-canonicalise: pred_metric = pred_canonical * (fx_scaled / 1000)

Sources

1 source

Hugging FacePrincipalVersion 1a21a1a94d3d50a27f56627c09f05e3e2ce4e248

huggingface.co/igzi/depth-lora-checkpoints_diode-diode_outdoor/tree/1a21a1a94d3d50a27f56627c09f05e3e2ce4e248

Vérifié 20 oct.

Artefacts du modèle

1 artefact

adapter_model.safetensors

safetensors · 22,7 MB · SHA-256 a98d5e83b7c2…4394 · Hugging Face

Download file

Extraits de sources

3 extraits

Readme3 aoûthuggingface.co/igzi/depth-lora-checkpoints_diode-diode_outdoor

--- library_name: peft tags: - depth-estimation - lora - peft - metric3d - diode - outdoor license: apache-2.0 --- # Metric3D ViT-giant2 — LoRA adapter (DIODE Outdoor) > **Base model:** [`metric3d_vit_giant2`](https://github.com/YvanYin/Metric3D) > loaded via `torch.hub.load("yvanyin/metric3d", "metric3d_vit_giant2")` LoRA fine-tuning of [Metric3D v2](https://github.com/YvanYin/Metric3D) (`metric3d_vit_giant2`) on the [DIODE](https://diode-dataset.org/) outdoor split (metric depth in metres). ## Training details | | | |---|---| | Base model | `metric3d_vit_giant2` (torch.hub `yvanyin/metric3d`) | | Dataset | the [DIODE](https://diode-dataset.org/) outdoor split (metric depth in metres) | | Loss | Direct metric-depth L1 + gradient loss | | LoRA rank / alpha | 16 / 32 | | LoRA targets | `qkv`, `proj` | ## Usage ```python import torch from peft import PeftModel import torch.nn as nn import torch.utils.checkpoint as torch_checkpoint # Load base model model = torch.hub.load("yvanyin/metric3d", "metric3d_vit_giant2", pretrain=True, trust_repo=True) # Apply the same gradient-checkpointing wrapper used during training # (needed so PEFT key names match the saved adapter) def enable_gradient_checkpointing(model): try: encoder = model.depth_model.encoder except AttributeError: encoder = model.base_model.model.depth_model.encoder class _CheckpointedBlock(nn.Module): def __init__(self, block): super().__init__() self.block = block def forward(self, x): return torch_checkpoint.checkpoint(self.block, x, use_reentrant=False) for blk_group in encoder.blocks: for key in list(blk_group._modules.keys()): blk_group._modules[key] = _CheckpointedBlock(blk_group._modules[key]) enable_gradient_checkpointing(model) model = PeftModel.from_pretrained(model, "igzi/depth-lora-checkpoints_diode-diode_outdoor") model.eval() # Inference: input pixel_values shape (B, 3, 616, 1064), values normalised # with mean=[123.675, 116.28, 103.53], std=[58.395, 57.12, 57.375] pred_canonical, _, _ = model({"input": pixel_values}) # De-canonicalise: pred_metric = pred_canonical * (fx_scaled / 1000) ```

Note20 oct.huggingface.co/igzi/depth-lora-checkpoints_diode-diode_outdoor

Source context: 4 downloads · 0 likes · Pipeline depth-estimation · Library peft · Repo igzi/depth-lora-checkpoints_diode-diode_outdoor

Note3 aoûthuggingface.co/igzi/depth-lora-checkpoints_diode-diode_outdoor

Source context: 3 downloads · 0 likes · Pipeline depth-estimation · Library peft · Repo igzi/depth-lora-checkpoints_diode-diode_outdoor