Merge branch 'max_memory' of https://github.com/elwolf6/text-generation-webui into elwolf6-max_memory

modules/models.py

@@ -96,7 +96,12 @@ def load_model(model_name):
             params.append("load_in_8bit=True" if shared.args.load_in_8bit else "torch_dtype=torch.bfloat16" if shared.args.bf16 else "torch_dtype=torch.float16")
 
             if shared.args.gpu_memory:
-                params.append(f"max_memory={{0: '{shared.args.gpu_memory or '99'}GiB', 'cpu': '{shared.args.cpu_memory or '99'}GiB'}}")
+                memory_map = shared.args.gpu_memory.split(",")
+                max_memory = f"max_memory={{0: '{memory_map[0]}GiB'"
+                for i in range(1,len(memory_map)):
+                    max_memory+=(f", {i}: '{memory_map[i]}GiB'")
+                max_memory+=(f", 'cpu': '{shared.args.cpu_memory or '99'}GiB'" + "}")
+                params.append(max_memory)
             elif not shared.args.load_in_8bit:
                 total_mem = (torch.cuda.get_device_properties(0).total_memory/(1024*1024))
                 suggestion = round((total_mem-1000)/1000)*1000

modules/shared.py

@@ -44,7 +44,7 @@ parser.add_argument('--bf16', action='store_true', help='Load the model with bfl
 parser.add_argument('--auto-devices', action='store_true', help='Automatically split the model across the available GPU(s) and CPU.')
 parser.add_argument('--disk', action='store_true', help='If the model is too large for your GPU(s) and CPU combined, send the remaining layers to the disk.')
 parser.add_argument('--disk-cache-dir', type=str, default="cache", help='Directory to save the disk cache to. Defaults to "cache".')
-parser.add_argument('--gpu-memory', type=int, help='Maximum GPU memory in GiB to allocate. This is useful if you get out of memory errors while trying to generate text. Must be an integer number.')
+parser.add_argument('--gpu-memory', type=str, help='Maxmimum GPU memory in GiB to be allocated per GPU. Ex to allocate 11 GiB on device 0 and 5 GiB on device 1, use --gpu-memory 11,5')
 parser.add_argument('--cpu-memory', type=int, help='Maximum CPU memory in GiB to allocate for offloaded weights. Must be an integer number. Defaults to 99.')
 parser.add_argument('--flexgen', action='store_true', help='Enable the use of FlexGen offloading.')
 parser.add_argument('--percent', nargs="+", type=int, default=[0, 100, 100, 0, 100, 0], help='FlexGen: allocation percentages. Must be 6 numbers separated by spaces (default: 0, 100, 100, 0, 100, 0).')