From 7a8500f6facb9bafd576f7f7d261a10ad88eec08 Mon Sep 17 00:00:00 2001
From: Frank <frank@dankware>
Date: Tue, 25 Aug 2026 02:30:25 -0700
Subject: [PATCH] Tune frank-default-large params

- frank-default-large.env: CTX_SIZE 262144 -> 196608 (192K context),
  BATCH_SIZE/UBATCH_SIZE 4096 -> 8192, CTX_CHECKPOINTS 16 -> 32, add
  preserve_thinking:true to chat_template_kwargs
- kingjones-strix-mtp.md: sync launch excerpt + note on preserve_thinking
  and the peg-native/deepseek chat_format pairing
---
 docs/LLM/kingjones-strix-mtp.md |  8 +++++++-
 scripts/frank-default-large.env | 18 ++++++++++--------
 2 files changed, 17 insertions(+), 9 deletions(-)

diff --git a/docs/LLM/kingjones-strix-mtp.md b/docs/LLM/kingjones-strix-mtp.md
index 7e66399..8ade2ac 100644
--- a/docs/LLM/kingjones-strix-mtp.md
+++ b/docs/LLM/kingjones-strix-mtp.md
@@ -68,7 +68,9 @@ llama-server \
   -m Qwen3.8-27B-Uncensored-Q6_0_ROCMFPX_AGENT.gguf \
   --mmproj mmproj-Qwen3.8-27B-Q8_0.gguf \
   --chat-template-file chat-template.jinja \
-  -ngl 999 -dev ROCm0 -c 262144 -np 1 \
+  --chat-template-kwargs '{"enable_thinking":true,"preserve_thinking":true}' \
+  -ngl 999 -dev ROCm0 -c 196608 -np 1 \
+  -b 8192 -ub 8192 -ctxcp 32 \
   --spec-type draft-mtp --model-draft mtp-Qwen3.8-27B-Q4_0.gguf \
   --spec-draft-ngl 999 --spec-draft-n-max 4 \
   -fa on --no-mmap --cache-type-k f16 --cache-type-v f16
@@ -82,6 +84,10 @@ llama-server \
    `enable_thinking: false` instead.
 4. Small `max_tokens` with thinking on returns empty `content`. The Sharp
    template drives thinking on by default (effort medium).
+5. `preserve_thinking:true` (in `chat_template_kwargs`) keeps `<think>`
+   blocks in recorded assistant messages. `preserve_thinking:false` strips
+   them. See [[deployment]] for the chat_format/reasoning_format pairing
+   (`peg-native` + `deepseek`) the Pi agent expects from this endpoint.
 
 ## See Also
 
diff --git a/scripts/frank-default-large.env b/scripts/frank-default-large.env
index 763c112..19409b2 100644
--- a/scripts/frank-default-large.env
+++ b/scripts/frank-default-large.env
@@ -29,8 +29,10 @@ MMPROJ_FILE=mmproj-Qwen3.8-27B-Q8_0.gguf
 
 # Qwen-Sharp chat template (tracked in the repo, staged on the host).
 CHAT_TEMPLATE_FILE=/home/dank/serverAppConfigs/scripts/chat-template.jinja
-# enable_thinking on by default in the Sharp template; medium effort default.
-CHAT_TEMPLATE_KWARGS='{"enable_thinking":true}'
+# enable_thinking + preserve_thinking on by default in the Sharp template;
+# medium effort default. preserve_thinking keeps <think> blocks in recorded
+# assistant messages (set to false to strip them, true to always keep them).
+CHAT_TEMPLATE_KWARGS='{"enable_thinking":true,"preserve_thinking":true}'
 
 # ROCm 7.15 SDK paths (inside the retired ling3-ciru venv).
 ROCM_DEVEL=/home/dank/ling3-ciru/Ling-3.0-Flash-CIRU-int4-Strix-native/.venv/lib/python3.12/site-packages/_rocm_sdk_devel
@@ -41,16 +43,16 @@ ALIAS=frank-default-large
 BIND_ADDR=0.0.0.0
 LISTEN_PORT=7770
 
-# Serving: full native 262144 context, 1 slot. f16 KV, context checkpoints,
+# Serving: 196608 (192K) context, 1 slot. f16 KV, context checkpoints,
 # prompt cache, draft-mtp with standalone Q4_0 head, n_max 4 (the knee for
 # the dense model). Unified memory; rocm-smi reports the 64 GiB budget.
-CTX_SIZE=262144
+CTX_SIZE=196608
 PARALLEL=1
-BATCH_SIZE=4096
-UBATCH_SIZE=4096
+BATCH_SIZE=8192
+UBATCH_SIZE=8192
 CACHE_TYPE_K=f16
 CACHE_TYPE_V=f16
-CTX_CHECKPOINTS=16
+CTX_CHECKPOINTS=32
 CACHE_RAM=8192
 
 # Dense-model MTP draft (standalone Q4_0 head, n_max 4). No PromptForge .pfs
@@ -70,4 +72,4 @@ TOP_P=0.95
 REASONING=on
 REASONING_FORMAT=deepseek
 
-VERBOSITY=4
\ No newline at end of file
+VERBOSITY=4