From a294322ff40c44e1c02db9b41bf15d21160af8c4 Mon Sep 17 00:00:00 2001
From: Hoang Ha <64120343+hahuyhoang411@users.noreply.github.com>
Date: Fri, 1 Nov 2024 10:39:59 +0100
Subject: [PATCH] Chore: Fix (#50)

* fix typo

* fix workflow

* fix

* test

* test

* workflow
---
 .github/workflows/create-model-yml.yml | 42 ++++++++++++--------------
 README.md                              |  2 +-
 2 files changed, 20 insertions(+), 24 deletions(-)

diff --git a/.github/workflows/create-model-yml.yml b/.github/workflows/create-model-yml.yml
index 0137649..4554ba4 100644
--- a/.github/workflows/create-model-yml.yml
+++ b/.github/workflows/create-model-yml.yml
@@ -7,11 +7,21 @@ on:
         description: "Name of the model to create (will be used in repo name and files)"
         required: true
         type: string
+      model_variant:
+        description: "Model variant (e.g., 8b, 13b) - will be used in default field"
+        required: true
+        type: string
+        default: "8b"
       prompt_template:
         description: "Prompt template for the model"
         required: true
         type: string
-        default: "<|im_start|>system\n{system_message}<|im_end|>\n<|im_start|>user\n{prompt}<|im_end|>\n<|im_start|>assistant"
+        default: |
+          <|im_start|>system
+          {system_message}<|im_end|>
+          <|im_start|>user
+          {prompt}<|im_end|>
+          <|im_start|>assistant
       stop_tokens:
         description: "Stop tokens for the model (comma-separated, e.g., <|im_end|>,</s>)"
         required: true
@@ -22,9 +32,11 @@ on:
         required: true
         type: string
         default: "llama-cpp"
+
 env:
   USER_NAME: cortexso
   MODEL_NAME: ${{ inputs.model_name }}
+  MODEL_VARIANT: ${{ inputs.model_variant }}
   PROMPT_TEMPLATE: ${{ inputs.prompt_template }}
   STOP_TOKENS: ${{ inputs.stop_tokens }}
   ENGINE: ${{ inputs.engine }}
@@ -69,8 +81,8 @@ jobs:
               pass
 
           def custom_str_representer(dumper, data):
-              if isinstance(data, str):
-                  return dumper.represent_scalar('tag:yaml.org,2002:str', data, style='')
+              if isinstance(data, str) and '\n' in data:
+                  return dumper.represent_scalar('tag:yaml.org,2002:str', data, style='|')
               return SafeRepresenter.represent_str(dumper, data)
 
           CustomDumper.add_representer(str, custom_str_representer)
@@ -91,24 +103,8 @@ jobs:
               if subsection:
                   file.write(f"# BEGIN {subsection}\n")
               
-              # Special handling for prompt_template to ensure it's a single line with preserved \n
-              if 'prompt_template' in content:
-                  # Create a copy of the content to modify
-                  modified_content = content.copy()
-                  # Convert the prompt template to a literal string that preserves \n
-                  prompt_template = modified_content['prompt_template']
-                  # Remove the prompt_template from the content to handle it separately
-                  del modified_content['prompt_template']
-                  
-                  # Dump the rest of the content
-                  yaml_content = yaml.dump(modified_content, Dumper=CustomDumper, default_flow_style=False, sort_keys=False)
-                  file.write(yaml_content)
-                  
-                  # Write the prompt_template separately with proper escaping
-                  file.write(f'prompt_template: "{prompt_template}"\n')
-              else:
-                  yaml_content = yaml.dump(content, Dumper=CustomDumper, default_flow_style=False, sort_keys=False)
-                  file.write(yaml_content)
+              yaml_content = yaml.dump(content, Dumper=CustomDumper, default_flow_style=False, sort_keys=False)
+              file.write(yaml_content)
                         
               if subsection:
                   file.write(f"# END {subsection}\n")
@@ -173,11 +169,11 @@ jobs:
               write_section(f, required_load, '', 'REQUIRED')
               f.write("# END MODEL LOAD PARAMETERS\n")
 
-          # Create metadata.yml
+          # Create metadata.yml with dynamic default field
           metadata_content = {
               'version': 1,
               'name': os.environ['MODEL_NAME'],
-              'default': '8b-gguf-q4-km'
+              'default': f"{os.environ['MODEL_VARIANT']}-gguf-q4-km"
           }
 
           with open('metadata.yml', 'w') as f:
diff --git a/README.md b/README.md
index 45433b4..0f7949d 100644
--- a/README.md
+++ b/README.md
@@ -34,7 +34,7 @@ This step will create a model repository on [Cortexso's Hugging Face account](ht
 3. Click `Run workflow` dropdown
 4. Fill in the required information:
 - **`source_model_id`**: The source HuggingFace model ID (e.g., meta-llama/Meta-Llama-3.1-8B-Instruct)
-- **`**source_model_size`**: The model size (e.g., 8b)
+- **`source_model_size`**: The model size (e.g., 8b)
 - **`target_model_id`**: The target HuggingFace model ID
 - **`quantization_level`**: Quantization level (e.g., 'q4-km') or 'all' for all levels
 5. Click `Run workflow` button to start the conversion