first

conversation.py

@@ -0,0 +1,86 @@
+from fastchat.conversation import Conversation, SeparatorStyle
+
+# 找tokenizer_config.json
+
+CONVS = {
+    'Qwen1.5': Conversation(
+        name="qwen",
+        system_template="<|im_start|>system\n{system_message}",
+        system_message="You are a helpful assistant.",
+        roles=("<|im_start|>user", "<|im_start|>assistant"),
+        sep_style=SeparatorStyle.CHATML,
+        sep="<|im_end|>",
+        stop_token_ids=[
+            151643,
+            151644,
+            151645,
+            14582,
+        ],  # "<|endoftext|>", "<|im_start|>", "<|im_end|>"
+        stop_str="<|endoftext|>",
+    ),
+    'Qwen1.5-Chat': Conversation(
+        name="qwen",
+        system_template="<|im_start|>system\n{system_message}",
+        system_message="You are a helpful assistant.",
+        roles=("<|im_start|>user", "<|im_start|>assistant"),
+        sep_style=SeparatorStyle.CHATML,
+        sep="<|im_end|>",
+        stop_token_ids=[
+            151643,
+            151644,
+            151645,
+            14582,
+        ],  # "<|endoftext|>", "<|im_start|>", "<|im_end|>"
+        stop_str="<|endoftext|>",
+    ),
+    'Qwen': Conversation(
+        name="qwen",
+        system_template="<|im_start|>system\n{system_message}",
+        system_message="You are a helpful assistant.",
+        roles=("<|im_start|>user", "<|im_start|>assistant"),
+        sep_style=SeparatorStyle.CHATML,
+        sep="<|im_end|>",
+        stop_token_ids=[
+            151643,
+            151644,
+            151645,
+            14582,
+        ],  # "<|endoftext|>", "<|im_start|>", "<|im_end|>"
+        stop_str="<|endoftext|>",
+    ),
+    'Llama-2': Conversation(
+        name="llama-2",
+        system_template="[INST] <<SYS>>\n{system_message}\n<</SYS>>\n\n",
+        roles=("[INST]", "[/INST]"),
+        sep_style=SeparatorStyle.LLAMA2,
+        sep=" ",
+        sep2=" </s><s>",
+    )
+}
+
+
+def generate_inputs(standard_conv, prompt_q, tokenizer):
+    # print(standard_conv.name)
+    if standard_conv.name == 'llama-2':
+        conv = standard_conv.copy()
+        conv.set_system_message("You will write beautiful compliments according to needs")
+        # conv.append_message("<|user|>", prompt_q)
+        # conv.append_message("<|assistant|>", None)
+        conv.append_message("[INST]", prompt_q)
+        conv.append_message("[/INST]", None)
+        inputs = tokenizer(
+            conv.get_prompt(),
+            return_tensors='pt'
+        )["input_ids"]
+    elif standard_conv.name == 'qwen':
+        conv = standard_conv.copy()
+        conv.set_system_message("You will write beautiful compliments according to needs")
+        conv.append_message("<|im_start|>user", prompt_q)
+        conv.append_message("<|im_start|>assistant", None)
+        inputs = tokenizer(
+            conv.get_prompt(),
+            return_tensors='pt'
+        )["input_ids"]
+    else:
+        return None
+    return inputs

data/MMLU-STEM/dataset_dict.json

@@ -0,0 +1 @@
+{"splits": ["train"]}

data/MMLU-STEM/train/dataset_info.json

@@ -0,0 +1,53 @@
+{
+  "builder_name": "json",
+  "citation": "",
+  "config_name": "default",
+  "dataset_name": "mmlu-stem",
+  "dataset_size": 1000011,
+  "description": "",
+  "download_checksums": {
+    "hf://datasets/TIGER-Lab/MMLU-STEM@299b1a95bf25a953b7af2f71dacb39fa7a5299c5/stem.json": {
+      "num_bytes": 1291805,
+      "checksum": null
+    }
+  },
+  "download_size": 1291805,
+  "features": {
+    "question": {
+      "dtype": "string",
+      "_type": "Value"
+    },
+    "subject": {
+      "dtype": "string",
+      "_type": "Value"
+    },
+    "choices": {
+      "feature": {
+        "dtype": "string",
+        "_type": "Value"
+      },
+      "_type": "Sequence"
+    },
+    "answer": {
+      "dtype": "int64",
+      "_type": "Value"
+    }
+  },
+  "homepage": "",
+  "license": "",
+  "size_in_bytes": 2291816,
+  "splits": {
+    "train": {
+      "name": "train",
+      "num_bytes": 1000011,
+      "num_examples": 3134,
+      "dataset_name": "mmlu-stem"
+    }
+  },
+  "version": {
+    "version_str": "0.0.0",
+    "major": 0,
+    "minor": 0,
+    "patch": 0
+  }
+}

data/MMLU-STEM/train/state.json

@@ -0,0 +1,13 @@
+{
+  "_data_files": [
+    {
+      "filename": "data-00000-of-00001.arrow"
+    }
+  ],
+  "_fingerprint": "d699cf3943feae0b",
+  "_format_columns": null,
+  "_format_kwargs": {},
+  "_format_type": null,
+  "_output_all_columns": false,
+  "_split": "train"
+}

data/gsm8k/dataset_dict.json

@@ -0,0 +1 @@
+{"splits": ["train", "test"]}

data/gsm8k/test/dataset_info.json

@@ -0,0 +1,52 @@
+{
+  "builder_name": "parquet",
+  "citation": "",
+  "config_name": "main",
+  "dataset_name": "gsm8k",
+  "dataset_size": 4676934,
+  "description": "",
+  "download_checksums": {
+    "hf://datasets/gsm8k@e53f048856ff4f594e959d75785d2c2d37b678ee/main/train-00000-of-00001.parquet": {
+      "num_bytes": 2306545,
+      "checksum": null
+    },
+    "hf://datasets/gsm8k@e53f048856ff4f594e959d75785d2c2d37b678ee/main/test-00000-of-00001.parquet": {
+      "num_bytes": 419088,
+      "checksum": null
+    }
+  },
+  "download_size": 2725633,
+  "features": {
+    "question": {
+      "dtype": "string",
+      "_type": "Value"
+    },
+    "answer": {
+      "dtype": "string",
+      "_type": "Value"
+    }
+  },
+  "homepage": "",
+  "license": "",
+  "size_in_bytes": 7402567,
+  "splits": {
+    "train": {
+      "name": "train",
+      "num_bytes": 3963202,
+      "num_examples": 7473,
+      "dataset_name": "gsm8k"
+    },
+    "test": {
+      "name": "test",
+      "num_bytes": 713732,
+      "num_examples": 1319,
+      "dataset_name": "gsm8k"
+    }
+  },
+  "version": {
+    "version_str": "0.0.0",
+    "major": 0,
+    "minor": 0,
+    "patch": 0
+  }
+}

data/gsm8k/test/state.json

@@ -0,0 +1,13 @@
+{
+  "_data_files": [
+    {
+      "filename": "data-00000-of-00001.arrow"
+    }
+  ],
+  "_fingerprint": "77dafb2c804c8978",
+  "_format_columns": null,
+  "_format_kwargs": {},
+  "_format_type": null,
+  "_output_all_columns": false,
+  "_split": "test"
+}

data/gsm8k/train/dataset_info.json

@@ -0,0 +1,52 @@
+{
+  "builder_name": "parquet",
+  "citation": "",
+  "config_name": "main",
+  "dataset_name": "gsm8k",
+  "dataset_size": 4676934,
+  "description": "",
+  "download_checksums": {
+    "hf://datasets/gsm8k@e53f048856ff4f594e959d75785d2c2d37b678ee/main/train-00000-of-00001.parquet": {
+      "num_bytes": 2306545,
+      "checksum": null
+    },
+    "hf://datasets/gsm8k@e53f048856ff4f594e959d75785d2c2d37b678ee/main/test-00000-of-00001.parquet": {
+      "num_bytes": 419088,
+      "checksum": null
+    }
+  },
+  "download_size": 2725633,
+  "features": {
+    "question": {
+      "dtype": "string",
+      "_type": "Value"
+    },
+    "answer": {
+      "dtype": "string",
+      "_type": "Value"
+    }
+  },
+  "homepage": "",
+  "license": "",
+  "size_in_bytes": 7402567,
+  "splits": {
+    "train": {
+      "name": "train",
+      "num_bytes": 3963202,
+      "num_examples": 7473,
+      "dataset_name": "gsm8k"
+    },
+    "test": {
+      "name": "test",
+      "num_bytes": 713732,
+      "num_examples": 1319,
+      "dataset_name": "gsm8k"
+    }
+  },
+  "version": {
+    "version_str": "0.0.0",
+    "major": 0,
+    "minor": 0,
+    "patch": 0
+  }
+}

data/gsm8k/train/state.json

@@ -0,0 +1,13 @@
+{
+  "_data_files": [
+    {
+      "filename": "data-00000-of-00001.arrow"
+    }
+  ],
+  "_fingerprint": "6c618aa7490feacd",
+  "_format_columns": null,
+  "_format_kwargs": {},
+  "_format_type": null,
+  "_output_all_columns": false,
+  "_split": "train"
+}