Gitignore aider

# Aider chat conversation:

USER: `invokeai/backend/model_manager/convert_ckpt_to_diffusers.py` Im having issues with cpu memory, not gpu memory

ASSISTANT: To address CPU memory issues, we can make the following changes:

1. Use `del` to explicitly delete large objects when they are no longer needed.
2. Use `gc.collect()` to force garbage collection after deleting large objects.

Here are the *SEARCH/REPLACE* blocks:

invokeai/backend/model_manager/convert_ckpt_to_diffusers.py
```python
<<<<<<< SEARCH
    vae = AutoencoderKL(**vae_config)
    with torch.no_grad():
        vae.load_state_dict(converted_vae_checkpoint)
        vae.to(precision)
        torch.cuda.empty_cache()
=======
    vae = AutoencoderKL(**vae_config)
    with torch.no_grad():
        vae.load_state_dict(converted_vae_checkpoint)
        del converted_vae_checkpoint  # Free memory
        import gc
        gc.collect()
        vae.to(precision)
>>>>>>> REPLACE
```

invokeai/backend/model_manager/convert_ckpt_to_diffusers.py
```python
<<<<<<< SEARCH
    pipe = download_from_original_stable_diffusion_ckpt(Path(checkpoint_path).as_posix(), **kwargs)
    with torch.no_grad():
        pipe = pipe.to(precision)
        torch.cuda.empty_cache()
=======
    pipe = download_from_original_stable_diffusion_ckpt(Path(checkpoint_path).as_posix(), **kwargs)
    with torch.no_grad():
        del kwargs  # Free memory
        import gc
        gc.collect()
        pipe = pipe.to(precision)
>>>>>>> REPLACE
```

invokeai/backend/model_manager/convert_ckpt_to_diffusers.py
```python
<<<<<<< SEARCH
    pipe = download_controlnet_from_original_ckpt(checkpoint_path.as_posix(), **kwargs)
    with torch.no_grad():
        pipe = pipe.to(precision)
        torch.cuda.empty_cache()
=======
    pipe = download_controlnet_from_original_ckpt(checkpoint_path.as_posix(), **kwargs)
    with torch.no_grad():
        del kwargs  # Free memory
        import gc
        gc.collect()
        pipe = pipe.to(precision)
>>>>>>> REPLACE
```

.gitignore

@@ -188,3 +188,4 @@ installer/install.sh
 installer/update.bat
 installer/update.sh
 installer/InvokeAI-Installer/
+.aider*

invokeai/backend/model_manager/convert_ckpt_to_diffusers.py

@@ -30,8 +30,12 @@ def convert_ldm_vae_to_diffusers(
     converted_vae_checkpoint = convert_ldm_vae_checkpoint(checkpoint, vae_config)
 
     vae = AutoencoderKL(**vae_config)
-    vae.load_state_dict(converted_vae_checkpoint)
+    with torch.no_grad():
-    vae.to(precision)
+        vae.load_state_dict(converted_vae_checkpoint)
+        del converted_vae_checkpoint  # Free memory
+        import gc
+        gc.collect()
+        vae.to(precision)
 
     if dump_path:
         vae.save_pretrained(dump_path, safe_serialization=True)
@@ -52,7 +56,11 @@ def convert_ckpt_to_diffusers(
     model to be written.
     """
     pipe = download_from_original_stable_diffusion_ckpt(Path(checkpoint_path).as_posix(), **kwargs)
-    pipe = pipe.to(precision)
+    with torch.no_grad():
+        del kwargs  # Free memory
+        import gc
+        gc.collect()
+        pipe = pipe.to(precision)
 
     # TO DO: save correct repo variant
     if dump_path:
@@ -75,7 +83,11 @@ def convert_controlnet_to_diffusers(
     model to be written.
     """
     pipe = download_controlnet_from_original_ckpt(checkpoint_path.as_posix(), **kwargs)
-    pipe = pipe.to(precision)
+    with torch.no_grad():
+        del kwargs  # Free memory
+        import gc
+        gc.collect()
+        pipe = pipe.to(precision)
 
     # TO DO: save correct repo variant
     if dump_path: