Dev: add Qwen2.5 and Llama3.2 models

andriydruk · Oct 1, 2024 · ee48f84 · ee48f84
1 parent 1078ee5
commit ee48f84
Show file tree

Hide file tree

Showing 5 changed files with 58 additions and 24 deletions.
diff --git a/README.md b/README.md
@@ -10,12 +10,15 @@ LM Playground is designed as a universal platform for experimenting with various
 ![preview](art/Preview.png)
 
 ## Currently supported models
-* Qwen2 0.5B
+* Qwen2.5 0.5B
+* Qwen2.5 1.5B
+* Llama3.2 1B
+* Llama3.2 3B
 * Google Gemma2 2B
-* Microsoft Phi 3.1B
-* Meta Llama 3 7B
+* Microsoft Phi3.5 mini
+* Llama3.1 7B
 * Mistral 7B
-* Google Gemma2 2B
+* Google Gemma2 9B
 
 ## Install
 If you're just looking to install LM Playground, you can find it on [Google Play](https://play.google.com/store/apps/details?id=com.druk.lmplayground). If you're a developer wanting to contribute, read on.

diff --git a/app/src/main/cpp/LlamaCpp.h b/app/src/main/cpp/LlamaCpp.h
@@ -82,6 +82,7 @@ class LlamaModel {
                    std::string input_prefix,
                    std::string input_suffix,
                    std::vector<std::string> antiprompt,
+                   int32_t n_ctx,
                    int32_t n_gpu_layers,
                    llama_progress_callback progress_callback,
                    void* progress_callback_user_data);

diff --git a/app/src/main/cpp/LlamaModel.cpp b/app/src/main/cpp/LlamaModel.cpp
@@ -35,11 +35,13 @@ void LlamaModel::loadModel(const gpt_params& params_arg,
                            std::string input_prefix,
                            std::string input_suffix,
                            std::vector<std::string> antiprompt,
+                           int32_t n_ctx,
                            int32_t n_gpu_layers,
                            llama_progress_callback progress_callback,
                            void * progress_callback_user_data) {
     params = params_arg;
     params.interactive = true;
+    params.n_ctx = 2048;
     params.interactive_first = true;
     params.input_prefix = std::move(input_prefix);
     params.input_suffix = std::move(input_suffix);

diff --git a/app/src/main/cpp/native-lib.cpp b/app/src/main/cpp/native-lib.cpp
@@ -104,6 +104,7 @@ Java_com_druk_llamacpp_LlamaCpp_loadModel(JNIEnv *env,
                      std::string(inputPrefixCStr),
                      std::string(inputSuffixCStr),
                      antiprompt_vector,
+                     2048,
                      -1,
                      [](float progress, void *ctx) -> bool {
                             auto* context = static_cast<CallbackContext*>(ctx);

diff --git a/app/src/main/java/com/druk/lmplayground/models/ModelInfoProvider.kt b/app/src/main/java/com/druk/lmplayground/models/ModelInfoProvider.kt
@@ -10,14 +10,41 @@ object ModelInfoProvider {
         val files = path.listFiles()
         return listOf(
             ModelInfo(
-                name = "Qwen2 0.5B",
-                file = files?.firstOrNull { it.name == "qwen2-0_5b-instruct-q4_k_m.gguf" },
-                remoteUri = Uri.parse("https://huggingface.co/Qwen/Qwen2-0.5B-Instruct-GGUF/resolve/main/qwen2-0_5b-instruct-q4_k_m.gguf"),
+                name = "Qwen2.5 0.5B",
+                file = files?.firstOrNull { it.name == "Qwen2.5-0.5B-Instruct-Q4_K_M.gguf" },
+                remoteUri = Uri.parse("https://huggingface.co/lmstudio-community/Qwen2.5-0.5B-Instruct-GGUF/resolve/main/Qwen2.5-0.5B-Instruct-Q4_K_M.gguf?download=true"),
                 inputPrefix = "<|im_start|>user\n",
                 inputSuffix = "<|im_end|>\n<|im_start|>assistant\n",
                 antiPrompt = arrayOf("<|im_end|>"),
                 description = "0.5 billion parameters language model"
             ),
+            ModelInfo(
+                name = "Qwen2.5 1.5B",
+                file = files?.firstOrNull { it.name == "Qwen2.5-1.5B-Instruct-Q4_K_M.gguf" },
+                remoteUri = Uri.parse("https://huggingface.co/lmstudio-community/Qwen2.5-1.5B-Instruct-GGUF/resolve/main/Qwen2.5-1.5B-Instruct-Q4_K_M.gguf?download=true"),
+                inputPrefix = "<|im_start|>user\n",
+                inputSuffix = "<|im_end|>\n<|im_start|>assistant\n",
+                antiPrompt = arrayOf("<|im_end|>"),
+                description = "1.5 billion parameters language model"
+            ),
+            ModelInfo(
+                name = "Llama3.2 1B",
+                file = files?.firstOrNull { it.name == "Llama-3.2-1B-Instruct-Q4_K_M.gguf" },
+                remoteUri = Uri.parse("https://huggingface.co/lmstudio-community/Llama-3.2-1B-Instruct-GGUF/resolve/main/Llama-3.2-1B-Instruct-Q4_K_M.gguf?download=true"),
+                inputPrefix = "<|start_header_id|>user<|end_header_id|>\n\n",
+                inputSuffix = "<|eot_id|><|start_header_id|>assistant<|end_header_id|>\n\n",
+                antiPrompt = arrayOf("<|eot_id|>"),
+                description = "1 billions parameters language model"
+            ),
+            ModelInfo(
+                name = "Llama3.2 3B",
+                file = files?.firstOrNull { it.name == "Llama-3.2-3B-Instruct-Q4_K_M.gguf" },
+                remoteUri = Uri.parse("https://huggingface.co/lmstudio-community/Llama-3.2-3B-Instruct-GGUF/resolve/main/Llama-3.2-3B-Instruct-Q4_K_M.gguf?download=true"),
+                inputPrefix = "<|start_header_id|>user<|end_header_id|>\n\n",
+                inputSuffix = "<|eot_id|><|start_header_id|>assistant<|end_header_id|>\n\n",
+                antiPrompt = arrayOf("<|eot_id|>"),
+                description = "3 billions parameters language model"
+            ),
             ModelInfo(
                 name = "Gemma2 2B",
                 file = files?.firstOrNull { it.name == "gemma-2b-it-q4_k_m.gguf" },
@@ -28,39 +55,39 @@ object ModelInfoProvider {
                 description = "2 billions parameters language model"
             ),
             ModelInfo(
-                name = "Phi 3.1",
-                file = files?.firstOrNull { it.name == "Phi-3.1-mini-4k-instruct-Q4_K_M.gguf" },
-                remoteUri = Uri.parse("https://huggingface.co/bartowski/Phi-3.1-mini-4k-instruct-GGUF/resolve/main/Phi-3.1-mini-4k-instruct-Q4_K_M.gguf"),
+                name = "Phi3.5 mini",
+                file = files?.firstOrNull { it.name == "Phi-3.5-mini-instruct-Q4_K_M.gguf" },
+                remoteUri = Uri.parse("https://huggingface.co/bartowski/Phi-3.5-mini-instruct-GGUF/resolve/main/Phi-3.5-mini-instruct-Q4_K_M.gguf?download=true"),
                 inputPrefix = "<|user|>\n",
                 inputSuffix = "<|end|>\n<|assistant|>\n",
                 antiPrompt = arrayOf("<|end|>", "<|assistant|>"),
                 description = "3.8 billions parameters language model"
             ),
-            ModelInfo(
-                name = "Gemma2 9B",
-                file = files?.firstOrNull { it.name == "gemma-2-9b-it-Q4_K_M.gguf" },
-                remoteUri = Uri.parse("https://huggingface.co/bartowski/gemma-2-9b-it-GGUF/resolve/main/gemma-2-9b-it-Q4_K_M.gguf"),
-                inputPrefix = "<start_of_turn>user\n",
-                inputSuffix = "<end_of_turn>\n<start_of_turn>model\n",
-                antiPrompt = arrayOf("<start_of_turn>user", "<start_of_turn>model", "<end_of_turn>"),
-                description = "8.5 billions parameters language model"
-            ),
             ModelInfo(
                 name = "Mistral 7B",
-                file = files?.firstOrNull { it.name == "mistral-7b-instruct-v0.2.Q4_K_M.gguf" },
-                remoteUri = Uri.parse("https://huggingface.co/TheBloke/Mistral-7B-Instruct-v0.2-GGUF/resolve/main/mistral-7b-instruct-v0.2.Q4_K_M.gguf"),
+                file = files?.firstOrNull { it.name == "Mistral-7B-Instruct-v0.3-Q4_K_M.gguf" },
+                remoteUri = Uri.parse("https://huggingface.co/lmstudio-community/Mistral-7B-Instruct-v0.3-GGUF/resolve/main/Mistral-7B-Instruct-v0.3-Q4_K_M.gguf?download=true"),
                 inputPrefix = "[INST]",
                 inputSuffix = "[/INST]",
                 description = "7.3 billions parameter language model"
             ),
             ModelInfo(
-                name = "Llama 3 8B",
-                file = files?.firstOrNull { it.name == "Meta-Llama-3-8B-Instruct-Q4_K_M.gguf" },
-                remoteUri = Uri.parse("https://huggingface.co/lmstudio-community/Meta-Llama-3-8B-Instruct-GGUF/resolve/main/Meta-Llama-3-8B-Instruct-Q4_K_M.gguf"),
+                name = "Llama3.1 8B",
+                file = files?.firstOrNull { it.name == "Meta-Llama-3.1-8B-Instruct-Q4_K_M.gguf" },
+                remoteUri = Uri.parse("https://huggingface.co/lmstudio-community/Meta-Llama-3.1-8B-Instruct-GGUF/resolve/main/Meta-Llama-3.1-8B-Instruct-Q4_K_M.gguf?download=true"),
                 inputPrefix = "<|start_header_id|>user<|end_header_id|>\n\n",
                 inputSuffix = "<|eot_id|><|start_header_id|>assistant<|end_header_id|>\n\n",
                 antiPrompt = arrayOf("<|eot_id|>"),
                 description = "8 billions parameters language model"
+            ),
+            ModelInfo(
+                name = "Gemma2 9B",
+                file = files?.firstOrNull { it.name == "gemma-2-9b-it-Q4_K_M.gguf" },
+                remoteUri = Uri.parse("https://huggingface.co/bartowski/gemma-2-9b-it-GGUF/resolve/main/gemma-2-9b-it-Q4_K_M.gguf"),
+                inputPrefix = "<start_of_turn>user\n",
+                inputSuffix = "<end_of_turn>\n<start_of_turn>model\n",
+                antiPrompt = arrayOf("<start_of_turn>user", "<start_of_turn>model", "<end_of_turn>"),
+                description = "8.5 billions parameters language model"
             )
         )
     }