From 3797126de40bee0e3ccd2c353f1d35abcafde458 Mon Sep 17 00:00:00 2001
From: Carl-Robert Linnupuu <carlrobertoh@gmail.com>
Date: Thu, 23 Nov 2023 17:22:48 +0200
Subject: [PATCH] Add deepseek coder instruct models (1-33B)

---
 .../codegpt/completions/HuggingFaceModel.java   | 10 ++++++++++
 .../codegpt/completions/llama/LlamaModel.java   | 17 +++++++++++++++++
 2 files changed, 27 insertions(+)

diff --git a/src/main/java/ee/carlrobert/codegpt/completions/HuggingFaceModel.java b/src/main/java/ee/carlrobert/codegpt/completions/HuggingFaceModel.java
index a3a8784e..e95c44f1 100644
--- a/src/main/java/ee/carlrobert/codegpt/completions/HuggingFaceModel.java
+++ b/src/main/java/ee/carlrobert/codegpt/completions/HuggingFaceModel.java
@@ -21,6 +21,16 @@ public enum HuggingFaceModel {
   CODE_BOOGA_34B_Q4(34, 4, "CodeBooga-34B-v0.1-GGUF"),
   CODE_BOOGA_34B_Q5(34, 5, "CodeBooga-34B-v0.1-GGUF"),
 
+  DEEPSEEK_CODER_1_3B_Q3(1, 3, "deepseek-coder-1.3b-instruct-GGUF"),
+  DEEPSEEK_CODER_1_3B_Q4(1, 4, "deepseek-coder-1.3b-instruct-GGUF"),
+  DEEPSEEK_CODER_1_3B_Q5(1, 5, "deepseek-coder-1.3b-instruct-GGUF"),
+  DEEPSEEK_CODER_6_7B_Q3(7, 3, "deepseek-coder-6.7b-instruct-GGUF"),
+  DEEPSEEK_CODER_6_7B_Q4(7, 4, "deepseek-coder-6.7b-instruct-GGUF"),
+  DEEPSEEK_CODER_6_7B_Q5(7, 5, "deepseek-coder-6.7b-instruct-GGUF"),
+  DEEPSEEK_CODER_33B_Q3(33, 3, "deepseek-coder-33b-instruct-GGUF"),
+  DEEPSEEK_CODER_33B_Q4(33, 4, "deepseek-coder-33b-instruct-GGUF"),
+  DEEPSEEK_CODER_33B_Q5(33, 5, "deepseek-coder-33b-instruct-GGUF"),
+
   PHIND_CODE_LLAMA_34B_Q3(34, 3, "Phind-CodeLlama-34B-v2-GGUF"),
   PHIND_CODE_LLAMA_34B_Q4(34, 4, "Phind-CodeLlama-34B-v2-GGUF"),
   PHIND_CODE_LLAMA_34B_Q5(34, 5, "Phind-CodeLlama-34B-v2-GGUF"),
diff --git a/src/main/java/ee/carlrobert/codegpt/completions/llama/LlamaModel.java b/src/main/java/ee/carlrobert/codegpt/completions/llama/LlamaModel.java
index 8cbacc6f..4e0afd49 100644
--- a/src/main/java/ee/carlrobert/codegpt/completions/llama/LlamaModel.java
+++ b/src/main/java/ee/carlrobert/codegpt/completions/llama/LlamaModel.java
@@ -38,6 +38,23 @@ public enum LlamaModel {
           HuggingFaceModel.CODE_BOOGA_34B_Q3,
           HuggingFaceModel.CODE_BOOGA_34B_Q4,
           HuggingFaceModel.CODE_BOOGA_34B_Q5)),
+  DEEPSEEK_CODER(
+      "Deepseek Coder",
+      "Deepseek Coder is composed of a series of code language models, each trained "
+          + "from scratch on 2T tokens, with a composition of 87% code and 13% natural language "
+          + "in both English and Chinese. It achieves state-of-the-art performance among "
+          + "open-source code models on multiple programming languages and various benchmarks.",
+      PromptTemplate.ALPACA,
+      List.of(
+          HuggingFaceModel.DEEPSEEK_CODER_1_3B_Q3,
+          HuggingFaceModel.DEEPSEEK_CODER_1_3B_Q4,
+          HuggingFaceModel.DEEPSEEK_CODER_1_3B_Q5,
+          HuggingFaceModel.DEEPSEEK_CODER_6_7B_Q3,
+          HuggingFaceModel.DEEPSEEK_CODER_6_7B_Q4,
+          HuggingFaceModel.DEEPSEEK_CODER_6_7B_Q5,
+          HuggingFaceModel.DEEPSEEK_CODER_33B_Q3,
+          HuggingFaceModel.DEEPSEEK_CODER_33B_Q4,
+          HuggingFaceModel.DEEPSEEK_CODER_33B_Q5)),
   PHIND_CODE_LLAMA(
       "Phind Code Llama",
       "This model is fine-tuned from Phind-CodeLlama-34B-v1 on an additional 1.5B tokens "