From 695738a89bfd433c0506fa5e6a0e5a12d800fa87 Mon Sep 17 00:00:00 2001
From: Linchen Xiao <xxllcc1993@gmail.com>
Date: Fri, 1 Nov 2024 15:34:23 +0800
Subject: [PATCH] [Update] Add lmdeploy DeepSeek configs (#1656)

* [Update] Add lmdeploy DeepSeek configs

* update max out length
---
 .../models/deepseek/lmdeploy_deepseek_v2.py   | 18 +++++++++++++++++
 .../models/deepseek/lmdeploy_deepseek_v2_5.py | 20 +++++++++++++++++++
 2 files changed, 38 insertions(+)
 create mode 100644 opencompass/configs/models/deepseek/lmdeploy_deepseek_v2.py
 create mode 100644 opencompass/configs/models/deepseek/lmdeploy_deepseek_v2_5.py

diff --git a/opencompass/configs/models/deepseek/lmdeploy_deepseek_v2.py b/opencompass/configs/models/deepseek/lmdeploy_deepseek_v2.py
new file mode 100644
index 00000000..a535003e
--- /dev/null
+++ b/opencompass/configs/models/deepseek/lmdeploy_deepseek_v2.py
@@ -0,0 +1,18 @@
+# flake8: noqa
+from mmengine.config import read_base
+from opencompass.models import (
+    TurboMindModel,
+)
+lmdeploy_deepseek_v2_model = [
+    dict(
+        type=TurboMindModel,
+        abbr='deepseek-v2-turbomind',
+        path='deepseek-ai/DeepSeek-V2',
+        engine_config=dict(session_len=7168, max_batch_size=4, tp=8, cache_max_entry_count=0.7),
+        gen_config=dict(top_k=1, temperature=1e-6, top_p=0.9),
+        max_seq_len=7168,
+        max_out_len=2048,
+        batch_size=4,
+        run_cfg=dict(num_gpus=8),
+    )
+]
\ No newline at end of file
diff --git a/opencompass/configs/models/deepseek/lmdeploy_deepseek_v2_5.py b/opencompass/configs/models/deepseek/lmdeploy_deepseek_v2_5.py
new file mode 100644
index 00000000..34574eff
--- /dev/null
+++ b/opencompass/configs/models/deepseek/lmdeploy_deepseek_v2_5.py
@@ -0,0 +1,20 @@
+from opencompass.models import TurboMindModelwithChatTemplate
+
+models = [
+    dict(
+        type=TurboMindModelwithChatTemplate,
+        abbr='deepseek-v2_5-turbomind',
+        path='deepseek-ai/DeepSeek-V2.5',
+        engine_config=dict(
+            session_len=7168,
+            max_batch_size=4,
+            tp=8,
+            cache_max_entry_count=0.7,
+        ),
+        gen_config=dict(top_k=1, temperature=1e-6, top_p=0.9),
+        max_seq_len=7168,
+        max_out_len=2048,
+        batch_size=4,
+        run_cfg=dict(num_gpus=8),
+    )
+]