comfy-aimdo 0.2 - Improved pytorch allocator integration (#12557)

Integrate comfy-aimdo 0.2 which takes a different approach to
installing the memory allocator hook. Instead of using the complicated
and buggy pytorch MemPool+CudaPluggableAlloctor, cuda is directly hooked
making the process much more transparent to both comfy and pytorch. As
far as pytorch knows, aimdo doesnt exist anymore, and just operates
behind the scenes.

Remove all the mempool setup stuff for dynamic_vram and bump the
comfy-aimdo version. Remove the allocator object from memory_management
and demote its use as an enablment check to a boolean flag.

Comfy-aimdo 0.2 also support the pytorch cuda async allocator, so
remove the dynamic_vram based force disablement of cuda_malloc and
just go back to the old settings of allocators based on command line
input.

execution.py

@@ -9,7 +9,6 @@ import traceback
 from enum import Enum
 from typing import List, Literal, NamedTuple, Optional, Union
 import asyncio
-from contextlib import nullcontext
 
 import torch
 
@@ -521,19 +520,14 @@ async def execute(server, dynprompt, caches, current_item, extra_data, executed,
                 # TODO - How to handle this with async functions without contextvars (which requires Python 3.12)?
                 GraphBuilder.set_default_prefix(unique_id, call_index, 0)
 
-            #Do comfy_aimdo mempool chunking here on the per-node level. Multi-model workflows
-            #will cause all sorts of incompatible memory shapes to fragment the pytorch alloc
-            #that we just want to cull out each model run.
-            allocator = comfy.memory_management.aimdo_allocator
-            with nullcontext() if allocator is None else torch.cuda.use_mem_pool(torch.cuda.MemPool(allocator.allocator())):
-                try:
-                    output_data, output_ui, has_subgraph, has_pending_tasks = await get_output_data(prompt_id, unique_id, obj, input_data_all, execution_block_cb=execution_block_cb, pre_execute_cb=pre_execute_cb, v3_data=v3_data)
-                finally:
-                    if allocator is not None:
-                        if args.verbose == "DEBUG":
-                            comfy_aimdo.model_vbar.vbars_analyze()
-                        comfy.model_management.reset_cast_buffers()
-                        comfy_aimdo.model_vbar.vbars_reset_watermark_limits()
+            try:
+                output_data, output_ui, has_subgraph, has_pending_tasks = await get_output_data(prompt_id, unique_id, obj, input_data_all, execution_block_cb=execution_block_cb, pre_execute_cb=pre_execute_cb, v3_data=v3_data)
+            finally:
+                if comfy.memory_management.aimdo_enabled:
+                    if args.verbose == "DEBUG":
+                        comfy_aimdo.control.analyze()
+                    comfy.model_management.reset_cast_buffers()
+                    comfy_aimdo.model_vbar.vbars_reset_watermark_limits()
 
             if has_pending_tasks:
                 pending_async_nodes[unique_id] = output_data