From ceebcfccbe72839be185a68d7a09e36259d3288c Mon Sep 17 00:00:00 2001
From: Vladimir Mandic <mandic00@live.com>
Date: Mon, 23 Jan 2023 16:44:42 -0500
Subject: [PATCH] update train defaults

---
 cli/train.py | 8 ++++----
 config.json  | 9 +++++++--
 train.sh     | 9 +++++++++
 3 files changed, 20 insertions(+), 6 deletions(-)
 create mode 100755 train.sh

diff --git a/cli/train.py b/cli/train.py
index a0422b559..d6a7f45b5 100755
--- a/cli/train.py
+++ b/cli/train.py
@@ -138,7 +138,7 @@ async def preprocess(params):
     elif os.path.isdir(params.src):
         if params.overwrite:
             preprocess_cleanup(params)
-        else:
+        elif os.path.isdir(params.dst):
             log.error({ 'preprocess output folder already exists': params.dst })
             return 0
 
@@ -342,7 +342,7 @@ async def pipeline(params):
 
     await plotloss(params)
 
-    create_preview(params.name, params.init)
+    # create_preview(params.name, params.init)
 
     log.debug({ 'pipeline end' })
     return
@@ -415,11 +415,11 @@ async def main():
     parser.add_argument("--init", type = str, default = "person", required = False, help = "initialization class, default: %(default)s")
     parser.add_argument("--dst", type = str, default = "/tmp", required = False, help = "destination image folder for processed images, default: %(default)s")
     parser.add_argument("--steps", type = int, default = -1, required = False, help = "training steps, default: %(default)s")
-    parser.add_argument("--maxsteps", type = int, default = 5000, required = False, help = "max training steps used when dynamic gradient is active, default: %(default)s")
+    parser.add_argument("--maxsteps", type = int, default = 4000, required = False, help = "max training steps used when dynamic gradient is active, default: %(default)s")
     parser.add_argument("--vectors", type = int, default = -1, required = False, help = "number of vectors per token, default: dynamic based on number of input images")
     parser.add_argument("--batch", type = int, default = 1, required = False, help = "batch size, default: %(default)s")
     parser.add_argument("--rate", type = str, default = "", required = False, help = "learn rate, default: dynamic")
-    parser.add_argument("--rstart", type = float, default = 0.01, required = False, help = "starting learn rate if using dynamic rate, default: %(default)s")
+    parser.add_argument("--rstart", type = float, default = 0.02, required = False, help = "starting learn rate if using dynamic rate, default: %(default)s")
     parser.add_argument("--rend", type = float, default = 0.0005, required = False, help = "ending learn rate if using dynamic rate, default: %(default)s")
     parser.add_argument("--rdescend", type = float, default = 2, required = False, help = "learn rate descend power when using dynamic rate, default: %(default)s")
     parser.add_argument("--grad", type = int, default = -1, required = False, help = "accumulate gradient over n images, default: : %(default)s")
diff --git a/config.json b/config.json
index 55d5b359a..9324ac72a 100644
--- a/config.json
+++ b/config.json
@@ -87,7 +87,7 @@
     "interrogate_clip_num_beams": 1,
     "interrogate_clip_min_length": 32,
     "interrogate_clip_max_length": 128,
-    "interrogate_clip_dict_limit": 0.0,
+    "interrogate_clip_dict_limit": 2048.0,
     "interrogate_deepbooru_score_threshold": 0.65,
     "deepbooru_sort_alpha": false,
     "deepbooru_use_spaces": false,
@@ -165,5 +165,10 @@
     "extra_networks_default_multiplier": 1,
     "keyedit_precision_attention": 0.1,
     "keyedit_precision_extra": 0.05,
-    "ui_extra_networks_tab_reorder": ""
+    "ui_extra_networks_tab_reorder": "",
+    "interrogate_clip_skip_categories": [
+        "artists",
+        "movements",
+        "flavors"
+    ]
 }
\ No newline at end of file
diff --git a/train.sh b/train.sh
new file mode 100755
index 000000000..015542bf0
--- /dev/null
+++ b/train.sh
@@ -0,0 +1,9 @@
+#/bin/env bash
+echo "Running without optimizations"
+
+export TF_CPP_MIN_LOG_LEVEL=2
+export FORCE_CUDA="1"
+export ATTN_PRECISION=fp16
+export PYTORCH_CUDA_ALLOC_CONF=garbage_collection_threshold:0.9,max_split_size_mb:512
+python launch.py --api --disable-opt-split-attention --disable-console-progressbars "$@"
+# --opt-channelslast