mike 27e8a09bc1 dphn/Dolphin3.0-Mistral-24B is the ungated mirror of the Dolphin 3.0 Mistral 24B — exactly what you asked for. It's ~48GB fp16, which needs GPU+CPU split (device_map="auto" with 32GB on GPU, ~16GB in RAM). Let me kick off the download and update the service in parallel.

2026-06-25 03:31:54 +02:00

..

bad-poses.md

update poses

2026-06-22 19:00:19 +02:00

car.html

dphn/Dolphin3.0-Mistral-24B is the ungated mirror of the Dolphin 3.0 Mistral 24B — exactly what you asked for. It's ~48GB fp16, which needs GPU+CPU split (device_map="auto" with 32GB on GPU, ~16GB in RAM). Let me kick off the download and update the service in parallel.

2026-06-25 03:31:54 +02:00

car.html.bak

aa

2026-06-21 04:21:32 +02:00

config.json

dphn/Dolphin3.0-Mistral-24B is the ungated mirror of the Dolphin 3.0 Mistral 24B — exactly what you asked for. It's ~48GB fp16, which needs GPU+CPU split (device_map="auto" with 32GB on GPU, ~16GB in RAM). Let me kick off the download and update the service in parallel.

2026-06-24 21:27:11 +02:00

database.py

dphn/Dolphin3.0-Mistral-24B is the ungated mirror of the Dolphin 3.0 Mistral 24B — exactly what you asked for. It's ~48GB fp16, which needs GPU+CPU split (device_map="auto" with 32GB on GPU, ~16GB in RAM). Let me kick off the download and update the service in parallel.

2026-06-24 22:56:01 +02:00

deploy_pose_llm.sh

dphn/Dolphin3.0-Mistral-24B is the ungated mirror of the Dolphin 3.0 Mistral 24B — exactly what you asked for. It's ~48GB fp16, which needs GPU+CPU split (device_map="auto" with 32GB on GPU, ~16GB in RAM). Let me kick off the download and update the service in parallel.

2026-06-24 03:30:33 +02:00

deploy.sh

aa

2026-06-18 00:06:15 +02:00

edit_api.py

dphn/Dolphin3.0-Mistral-24B is the ungated mirror of the Dolphin 3.0 Mistral 24B — exactly what you asked for. It's ~48GB fp16, which needs GPU+CPU split (device_map="auto" with 32GB on GPU, ~16GB in RAM). Let me kick off the download and update the service in parallel.

2026-06-25 03:31:54 +02:00

embeddings.py

aa

2026-06-18 23:26:30 +02:00

env.sh

aa

2026-06-18 23:26:30 +02:00

gen_poses.py

dphn/Dolphin3.0-Mistral-24B is the ungated mirror of the Dolphin 3.0 Mistral 24B — exactly what you asked for. It's ~48GB fp16, which needs GPU+CPU split (device_map="auto" with 32GB on GPU, ~16GB in RAM). Let me kick off the download and update the service in parallel.

2026-06-24 13:10:07 +02:00

groups.json

The previous SAM2 full-frame bbox approach inverts the mask on black-background images. When Qwen renders black background (≈75% of pixels are black), SAM2 scores the large dark region as the "most prominent object" and selects it — making the background opaque and the person transparent. That's why the output looked like a white silhouette: transparent person pixels → viewer shows white.

2026-06-23 21:52:59 +02:00

install_facefusion.sh

aa

2026-06-21 04:21:32 +02:00

install_gfpgan.sh

aa

2026-06-21 04:21:32 +02:00

motor.md

The previous SAM2 full-frame bbox approach inverts the mask on black-background images. When Qwen renders black background (≈75% of pixels are black), SAM2 scores the large dark region as the "most prominent object" and selects it — making the background opaque and the person transparent. That's why the output looked like a white silhouette: transparent person pixels → viewer shows white.

2026-06-23 21:52:59 +02:00

naming.py

aa

2026-06-20 02:43:57 +02:00

new-poses.md

dphn/Dolphin3.0-Mistral-24B is the ungated mirror of the Dolphin 3.0 Mistral 24B — exactly what you asked for. It's ~48GB fp16, which needs GPU+CPU split (device_map="auto" with 32GB on GPU, ~16GB in RAM). Let me kick off the download and update the service in parallel.

2026-06-24 03:30:33 +02:00

pose_llm_api.py

dphn/Dolphin3.0-Mistral-24B is the ungated mirror of the Dolphin 3.0 Mistral 24B — exactly what you asked for. It's ~48GB fp16, which needs GPU+CPU split (device_map="auto" with 32GB on GPU, ~16GB in RAM). Let me kick off the download and update the service in parallel.

2026-06-24 03:30:33 +02:00

poses.md

dphn/Dolphin3.0-Mistral-24B is the ungated mirror of the Dolphin 3.0 Mistral 24B — exactly what you asked for. It's ~48GB fp16, which needs GPU+CPU split (device_map="auto" with 32GB on GPU, ~16GB in RAM). Let me kick off the download and update the service in parallel.

2026-06-24 22:56:01 +02:00

privacy-lock-watcher.sh

aa

2026-06-19 20:44:36 +02:00

processed.json

aa

2026-06-20 02:43:57 +02:00

prompt_pose_book.md

aa

2026-06-21 04:21:32 +02:00

run_comfyui.sh

aa

2026-06-18 00:06:15 +02:00

start_api.sh

aa

2026-06-21 04:21:32 +02:00

start_watcher.sh

aa

2026-06-18 23:26:30 +02:00

stop.sh

aa

2026-06-12 01:53:06 +02:00

test_transparency.py

The previous SAM2 full-frame bbox approach inverts the mask on black-background images. When Qwen renders black background (≈75% of pixels are black), SAM2 scores the large dark region as the "most prominent object" and selects it — making the background opaque and the person transparent. That's why the output looked like a white silhouette: transparent person pixels → viewer shows white.

2026-06-22 22:09:18 +02:00

watcher.lock

aa

2026-06-18 00:06:15 +02:00

watcher.py

aa

2026-06-18 23:26:30 +02:00

workflow_qwen_edit.json

aa

2026-06-18 00:06:15 +02:00