[improvment] add user friendly method to generate the prompt (#221)

* use the jinja language to generate prompt

demo/cpo/cpo_case_1.yaml

@@ -4,7 +4,8 @@ dispatcher:
 datasets:
   - name: "cpo_data"
     data: "demo/data.json"
-    prompt: "demo/preference_template.json"
+    prompt: "demo/preference_prompt.yaml"
+    prompt_type: "preference"
     preprocess: "default"
 adapters:
   - name: "lora_cpo"

demo/dpo/dpo_case_1.yaml

@@ -4,7 +4,8 @@ dispatcher:
 datasets:
   - name: "dpo_data"
     data: "demo/data.json"
-    prompt: "demo/preference_template.json"
+    prompt: "demo/preference_prompt.yaml"
+    prompt_type: "preference"
     preprocess: "default"
 adapters:
   - name: "lora_dpo"

demo/dpo/dpo_case_2.yaml

@@ -4,7 +4,8 @@ dispatcher:
 datasets:
   - name: "dpo_data"
     data: "demo/data.json"
-    prompt: "demo/preference_template.json"
+    prompt: "demo/preference_prompt.yaml"
+    prompt_type: "preference"
     preprocess: "default"
 adapters:
   - name: "lora_dpo"

demo/dpo/dpo_case_3.yaml

@@ -4,7 +4,8 @@ dispatcher:
 datasets:
   - name: "dpo_data"
     data: "demo/data.json"
-    prompt: "demo/preference_template.json"
+    prompt: "demo/preference_prompt.yaml"
+    prompt_type: "preference"
     preprocess: "default"
 adapters:
   - name: "lora_dpo"

demo/lora/lora_case_1.yaml

@@ -4,7 +4,8 @@ dispatcher:
 datasets:
   - name: "data"
     data: "demo/data.json"
-    prompt: "demo/template.json"
+    prompt: "demo/prompt.yaml"
+    prompt_type: "instruction"
     preprocess: "shuffle"
 adapters:
   - name: "lora_0"

demo/loraplus/loraplus_case_1.yaml

@@ -4,7 +4,8 @@ dispatcher:
 datasets:
   - name: "data"
     data: "demo/data.json"
-    prompt: "demo/template.json"
+    prompt: "demo/prompt.yaml"
+    prompt_type: "instruction"
     preprocess: "shuffle"
 adapters:
   - name: "lora_0"

demo/preference_prompt.yaml

@@ -0,0 +1,13 @@
+template: |
+  ### Instruction:
+  {{ data_point['instruction'] + '\n'}}
+  {% if 'input' in data_point %}
+  ### Input:
+  {{ data_point['input'] + '\n'}}
+  {% endif %}
+  ### Output:
+  {% if is_chosen %}
+  {{ data_point['chosen'] + '\n'}}
+  {% else %}
+  {{ data_point['reject'] + '\n'}}
+  {% endif %}

demo/prompt.yaml

@@ -0,0 +1,9 @@
+template: |
+  ### Instruction:
+  {{ data_point['instruction'] + '\n'}}
+  {% if 'input' in data_point %}
+  ### Input:
+  {{ data_point['input'] + '\n'}}
+  {% endif %}
+  ### Output:
+  {{ data_point['chosen'] + '\n'}}

mlora/cli/dataset.py

@@ -1,6 +1,6 @@
 import json
 import requests
-from InquirerPy import inquirer
+from InquirerPy import inquirer, separator
 from rich import print
 from rich.table import Table
 from rich.box import ASCII
@@ -14,17 +14,21 @@ def list_dataset(obj):
     ret = json.loads(ret.text)
 
     table = Table(show_header=True, show_lines=True, box=ASCII)
-    table.add_column("name", justify="left")
-    table.add_column("train", justify="center")
-    table.add_column("prompt", justify="center")
+    table.add_column("name", justify="center")
+    table.add_column("train data name", justify="center")
+    table.add_column("prompt data name", justify="center")
+    table.add_column("prompter", justify="center")
     table.add_column("preprocess", justify="center")
 
     obj.ret_ = []
 
     for item in ret:
         item = json.loads(item)
-        table.add_row(item["name"], item["train"],
-                      item["prompt"], item["preprocess"])
+        table.add_row(item["name"],
+                      item["data_name"],
+                      item["prompt_name"],
+                      item["prompt_type"],
+                      item["preprocess"])
         obj.ret_.append(item["name"])
 
     obj.pret_ = table
@@ -34,9 +38,9 @@ def create_dataset(obj):
     name = inquirer.text(
         message="name:").execute()
 
-    list_file(obj, "train")
-    all_train = [item["name"] for item in obj.ret_]
-    if len(all_train) == 0:
+    list_file(obj, "data")
+    all_train_data = [item["name"] for item in obj.ret_]
+    if len(all_train_data) == 0:
         print("no train data, please upload one")
         return
 
@@ -47,28 +51,57 @@ def create_dataset(obj):
         return
 
     use_train = inquirer.select(
-        message="train data file:", choices=all_train).execute()
+        message="train data file:", choices=[separator.Separator(),
+                                             *all_train_data,
+                                             separator.Separator()]).execute()
     use_prompt = inquirer.select(
-        message="prompt template file:", choices=all_prompt).execute()
+        message="prompt template file:", choices=[separator.Separator(),
+                                                  *all_prompt,
+                                                  separator.Separator()]).execute()
     use_preprocess = inquirer.select(
-        message="data preprocessing:", choices=["default", "shuffle", "sort"]).execute()
+        message="data preprocessing:", choices=[separator.Separator(),
+                                                "default",
+                                                "shuffle",
+                                                "sort",
+                                                separator.Separator()]).execute()
 
     ret = requests.post(url() + "/dataset", json={
         "name": name,
-        "train": use_train,
-        "prompt": use_prompt,
+        "data_name": use_train,
+        "prompt_name": use_prompt,
         "preprocess": use_preprocess
     })
 
     print(json.loads(ret.text))
 
 
+def showcase_dataset(obj):
+    list_dataset(obj)
+    all_dataset = obj.ret_
+
+    if len(all_dataset) == 0:
+        print("no dataset, please create one")
+        return
+
+    use_dataset = inquirer.select(
+        message="dataset name:", choices=[separator.Separator(),
+                                          *all_dataset,
+                                          separator.Separator()]).execute()
+
+    ret = requests.get(url() + f"/showcase?name={use_dataset}")
+    ret = json.loads(ret.text)
+
+    print(ret)
+
+
 def help_dataset(_):
     print("Usage of dataset:")
     print("  ls")
     print("    list all the dataset.")
     print("  create")
     print("    create a new dataset.")
+    print("  showcase")
+    print("    display training data composed of prompt and dataset.")
 
 
 def do_dataset(obj, args):
@@ -79,5 +112,7 @@ def do_dataset(obj, args):
         return print(obj.pret_)
     elif args[0] == "create":
         return create_dataset(obj)
+    elif args[0] == "showcase":
+        return showcase_dataset(obj)
 
     help_dataset(None)

mlora/cli/file.py

@@ -1,13 +1,18 @@
 import json
 import requests
-from InquirerPy import inquirer, validator
+from InquirerPy import inquirer, validator, separator
 from rich import print
 from rich.table import Table
 from rich.box import ASCII
 
 
 from .setting import url
 
+g_file_type_map = {
+    "train data": "data",
+    "prompt data": "prompt"
+}
+
 
 def list_file(obj, file_type: str):
     ret = requests.get(url() + f"/{file_type}")
@@ -16,29 +21,47 @@ def list_file(obj, file_type: str):
     table = Table(show_header=True, show_lines=True, box=ASCII)
     table.add_column("name", justify="center")
     table.add_column("file", justify="center")
+    if file_type == "prompt":
+        table.add_column("prompter", justify="center")
 
     for item in ret:
-        table.add_row(item["name"], item["file"])
+        row_data = [item["name"], item["file"]["file_path"]]
+        if file_type == "prompt":
+            row_data.append(item["file"]["prompt_type"])
+        table.add_row(*row_data)
 
     obj.ret_ = ret
     obj.pret_ = table
 
 
 def upload_file():
-    file_type = inquirer.select(
-        message="type:", choices=["train", "prompt"]).execute()
     name = inquirer.text(
         message="name:",
         validate=validator.EmptyInputValidator("name should not be empty")).execute()
-    path = inquirer.filepath(
-        message="file path:",
-        default="/",
-        validate=validator.PathValidator(
-            is_file=True, message="input is not a file"),
-        only_files=True).execute()
 
-    ret = requests.post(
-        url() + f"/{file_type}?name={name}", files={"data_file": open(path, "rb")})
+    file_type = inquirer.select(message="file type:",
+                                choices=[separator.Separator(),
+                                         *g_file_type_map.keys(),
+                                         separator.Separator()]).execute()
+    file_type = g_file_type_map[file_type]
+
+    post_url = url() + f"/{file_type}?name={name}"
+
+    if file_type == "prompt":
+        prompt_type = inquirer.select(message="prompter type:",
+                                      choices=[separator.Separator(),
+                                               "instruction",
+                                               "preference",
+                                               separator.Separator()]).execute()
+        post_url += f"&prompt_type={prompt_type}"
+
+    path = inquirer.filepath(message="file path:",
+                             default="/",
+                             validate=validator.PathValidator(is_file=True,
+                                                              message="input is not a file"),
+                             only_files=True).execute()
+
+    ret = requests.post(post_url, files={"data_file": open(path, "rb")})
 
     print(json.loads(ret.text))
 
@@ -57,7 +80,12 @@ def do_file(obj, args):
     if args[0] == "ls":
         # to chose file type
         file_type = inquirer.select(
-            message="type:", choices=["train", "prompt"]).execute()
+            message="type:",
+            choices=[separator.Separator(),
+                     *g_file_type_map.keys(),
+                     separator.Separator()]
+        ).execute()
+        file_type = g_file_type_map[file_type]
         list_file(obj, file_type)
         return print(obj.pret_)
     elif args[0] == "upload":

mlora/config/dataset.py

@@ -7,12 +7,14 @@ class DatasetConfig(DictConfig):
     name_: str = ""
     data_path_: str = ""
     prompt_path_: str = ""
+    prompt_type_: str = ""
     preprocess_: str = "shuffle"
 
     __params_map: Dict[str, str] = {
         "name_": "name",
         "data_path_": "data",
         "prompt_path_": "prompt",
+        "prompt_type_": "prompt_type",
         "preprocess_": "preprocess",
     }
 

mlora/executor/context/__init__.py

@@ -12,6 +12,7 @@
     "loraplus": InferenceLoRAContext
 }
 
+
 __all__ = [
     "TRAINCONTEXT_CLASS",
     "INFERENCECONTEXT_CLASS",

mlora/executor/task/cpo_task.py

@@ -1,7 +1,5 @@
 from mlora.model.args import LinearInfo, Tokens, MLoRADataConfig
 from mlora.model.tokenizer import Tokenizer
-from mlora.config import CPOTaskConfig
-from mlora.prompter import PreferenceDataPrompter
 
 import torch
 import logging
@@ -15,10 +13,6 @@
 class CPOTask(TrainTask):
     now_epoch_: int = 0
 
-    def __init__(self, config: CPOTaskConfig, llm_name: str) -> None:
-        super().__init__(config, llm_name)
-        self.prompter_ = PreferenceDataPrompter(config.dataset_.prompt_path_)
-
     @override
     def prepare(self, linears_info: OrderedDict[str, LinearInfo], tokenizer: Tokenizer):
         self.tokenizer_ = tokenizer
@@ -42,13 +36,14 @@ def __cpo_loss_hinge(self, logits: torch.Tensor) -> torch.Tensor:
     @override
     def data(self, start_idx: int) -> Tuple[List[Tokens], List[MLoRADataConfig]]:
         logging.info(
-            f'Adapter {self.context_.name_} epoch: {self.now_epoch_}/{self.config_.num_epochs_}'
+            f'Adapter {self.context_.name_} epoch: {
+                self.now_epoch_}/{self.config_.num_epochs_}'
             f' iteration: {self.now_data_idx_}/{len(self.data_)} step: {self.now_step_}')
         data_idx_s = self.now_data_idx_
         data_idx_e = self.now_data_idx_ + self.config_.mini_batch_size_
 
         # get the train raw string
-        batch_str = self.prompter_.generate_prompt_batch(
+        batch_str = self.prompter_.generate_prompt(
             self.data_[data_idx_s:data_idx_e])
 
         # convert the string to tokens

mlora/executor/task/dpo_task.py

@@ -1,7 +1,5 @@
 from mlora.model.tokenizer import Tokenizer
 from mlora.model.modules import AdapterModel
-from mlora.config import DPOTaskConfig
-from mlora.prompter import PreferenceDataPrompter
 from mlora.model.args import Tokens, MLoRADataConfig, LinearInfo
 from mlora.executor.context import TaskContext, INFERENCECONTEXT_CLASS
 
@@ -17,11 +15,6 @@
 class DPOTask(TrainTask):
     ref_context_: TaskContext = None
 
-    def __init__(self, config: DPOTaskConfig, llm_name: str) -> None:
-        super().__init__(config, llm_name)
-
-        self.prompter_ = PreferenceDataPrompter(config.dataset_.prompt_path_)
-
     @override
     def prepare(self, linears_info: OrderedDict[str, LinearInfo], tokenizer: Tokenizer):
         self.tokenizer_ = tokenizer
@@ -88,7 +81,7 @@ def data(self, start_idx: int) -> Tuple[List[Tokens], List[MLoRADataConfig]]:
 
         # 0...mid is chosen data
         # mid.end is reject data
-        batch_str = self.prompter_.generate_prompt_batch(
+        batch_str = self.prompter_.generate_prompt(
             self.data_[data_idx_s:data_idx_e])
 
         assert len(batch_str) % 2 == 0