Merge pull request #507 from nerdalert/osx-train

Get OSX support functioning for generate/train

Author: vishnoianil

api-server/.gitignore

@@ -23,6 +23,7 @@ go.work.sum
 
 # binary
 api-server
+ilab-api-server
 
 # app specific
 logs/

api-server/README.md

@@ -21,24 +21,34 @@ go mod download
 
 ### Run the Server
 
+### Build
+
+```bash
+go build -o ilab-api-server
+```
+
 #### For macOS with Metal (MPS)
 
 ```bash
-go run main.go --base-dir /path/to/base-dir --taxonomy-path /path/to/taxonomy --osx
+./ilab-api-server --base-dir /path/to/base-dir --taxonomy-path /path/to/taxonomy --osx  --pipeline simple
+# Or the full pipeline
+./ilab-api-server --base-dir /path/to/base-dir --taxonomy-path /path/to/taxonomy --osx  --pipeline full
 ```
 
 #### For CUDA-enabled environments
 
+Since the device type is cuda, only the accelerated pipeline option is available and set as the default.
+
 ```bash
-go run main.go --base-dir /path/to/base-dir --taxonomy-path /path/to/taxonomy --cuda
+./ilab-api-server --base-dir /path/to/base-dir --taxonomy-path /path/to/taxonomy --cuda
 ```
 
 #### For a RHEL AI machine
 
-- If you're operating on a Red Hat Enterprise Linux AI (RHEL AI) machine, and the ilab binary is already available in your $PATH, you don't need to specify the --base-dir. Additionally, pass CUDA support with `--cuda`.
+- If you're operating on a Red Hat Enterprise Linux AI (RHEL AI) machine, and the ilab binary is already available in your $PATH, you don't need to specify the --base-dir. Additionally, pass CUDA support with `--cuda`. The `accelerated` pipeline is the only option here and also  the default.
 
 ```bash
-go run main.go --taxonomy-path ~/.local/share/instructlab/taxonomy/ --rhelai --cuda
+./ilab-api-server --taxonomy-path ~/.local/share/instructlab/taxonomy/ --rhelai --cuda
 ```
 
 The `--rhelai` flag indicates that the ilab binary is available in the system's $PATH and does not require a virtual environment.
@@ -66,10 +76,10 @@ After this, we can cleanup our temp directory as it is no longer required: `rm -
 
 ### Example command with paths
 
-Here's an example command for running the server on a macOS machine with Metal support:
+Here's an example command for running the server on a macOS machine with Metal support and debugging enabled:
 
 ```bash
-go run main.go --base-dir /Users/user/code/instructlab --taxonomy-path ~/.local/share/instructlab/taxonomy/ --osx
+./ilab-api-server --base-dir /Users/<USERNAME>/<PATH_TO_ILAB>/instructlab/ --taxonomy-path ~/.local/share/instructlab/taxonomy/ --pipeline simple --osx --debug
 ```
 
 ## API Documentation

api-server/jobs.db

@@ -509,40 +509,20 @@ func (srv *ILabServer) startTrainJob(modelName, branchName string, epochs *int)
 		srv.log.Info("No epochs specified; using default number of epochs.")
 	}
 
-	// Additional logic if pipelineType == "simple" (and not rhelai)
 	if srv.pipelineType == "simple" && !srv.rhelai {
-		homeDir, err := os.UserHomeDir()
-		if err != nil {
-			return "", fmt.Errorf("failed to get user home directory: %v", err)
-		}
-		datasetDir := filepath.Join(homeDir, ".local", "share", "instructlab", "datasets")
-
-		// Copy the latest knowledge_train_msgs_*.jsonl => train_gen.jsonl
-		latestTrainFile, err := srv.findLatestFileWithPrefix(datasetDir, "knowledge_train_msgs_")
-		if err != nil {
-			return "", fmt.Errorf("failed to find knowledge_train_msgs_*.jsonl file: %v", err)
-		}
-		trainGenPath := filepath.Join(datasetDir, "train_gen.jsonl")
-		if err := srv.overwriteCopy(latestTrainFile, trainGenPath); err != nil {
-			return "", fmt.Errorf("failed to copy %s to %s: %v", latestTrainFile, trainGenPath, err)
-		}
+		// TODO: Works on RHEL not from ilab main. --model-path seems to only accept the repo/name here and not the full path. Commenting for now.
+		//homeDir, err := os.UserHomeDir()
+		//if err != nil {
+		//  return "", fmt.Errorf("failed to get user home directory: %v", err)
+		//}
+		//datasetDir := filepath.Join(homeDir, ".local", "share", "instructlab", "datasets")
 
-		// Copy the latest test_ggml-model-*.jsonl => test_gen.jsonl
-		latestTestFile, err := srv.findLatestFileWithPrefix(datasetDir, "test_ggml-model")
-		if err != nil {
-			return "", fmt.Errorf("failed to find test_ggml-model*.jsonl file: %v", err)
-		}
-		testGenPath := filepath.Join(datasetDir, "test_gen.jsonl")
-		if err := srv.overwriteCopy(latestTestFile, testGenPath); err != nil {
-			return "", fmt.Errorf("failed to copy %s to %s: %v", latestTestFile, testGenPath, err)
-		}
-
-		// Reset cmdArgs to a simpler set
 		cmdArgs = []string{
 			"model", "train",
 			"--pipeline", srv.pipelineType,
-			fmt.Sprintf("--data-path=%s", datasetDir),
-			fmt.Sprintf("--model-path=%s", fullModelPath),
+			"--optimize-memory",
+			//fmt.Sprintf("--data-path=%s", datasetDir), // Leaving commented out for now until the above todo is resolved.
+			fmt.Sprintf("--model-path=%s", modelName),
 		}
 		if srv.isOSX {
 			cmdArgs = append(cmdArgs, "--device=mps")