init

.gitattributes

@@ -33,3 +33,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+tokenizer.json filter=lfs diff=lfs merge=lfs -text

README.md

@@ -0,0 +1,136 @@
+# DPO Chinese Error Correction Model
+使用DPO訓練之中文糾錯模型。
+
+### Usage
+```python
+from transformers import AutoTokenizer, AutoModelForCausalLM, LlamaForCausalLM,AddedToken
+import sys
+
+mode_id = "p208p2002/bloom-1b1-zh-error-correction-dpo"
+model: LlamaForCausalLM = AutoModelForCausalLM.from_pretrained("p208p2002/bloom-1b1-zh-error-correction-dpo")
+tokenizer = AutoTokenizer.from_pretrained("p208p2002/bloom-1b1-zh-error-correction-dpo")
+
+test_texts = [
+    "為了潔約能源請隨守關閉沒有使用的電器",
+    "今天新情很好",
+    "你快樂我也很高心",
+    "但不再算再找實習生了",
+    "今天太陽很大要注意篩傷",
+    "你要不要和我依起去台北",
+    "清晨六點終太陽會升起",
+    "傾城六點鐘太陽會升起",
+    "鍋馬路時你應該要注意虹綠燈",
+    "他正在學學彈吉他",
+    "下樓梯請注意階梯",
+    "此信件為系統自動發送之通知",
+    "此信件為系統自動發送知通知",
+    "如為誤傳也請立即刪除本郵件並通知寄件者"
+]
+for text in test_texts:
+    inputs = tokenizer(
+        f"{tokenizer.bos_token}{text} {tokenizer.eos_token}\n {tokenizer.bos_token}",
+        return_tensors="pt",
+        add_special_tokens=False
+    )["input_ids"]
+
+    out = model.generate(
+        inputs,
+        max_new_tokens=20,
+    )
+    decode_out = tokenizer.decode(out[0])
+
+    input_text,output_text = decode_out.split("\n") 
+    input_text = input_text.strip()
+    output_text = output_text.strip()
+
+    print("input :",input_text)
+    print("output:",output_text)
+    print('-'*30)
+```
+```
+input: <s>為了潔約能源請隨守關閉沒有使用的電器 </s>
+output: <s>為了節約能源請隨時關閉沒有使用的電器 </s>
+------------------------------
+input: <s>今天新情很好 </s>
+output: <s>今天心情很好 </s>
+------------------------------
+input: <s>你快樂我也很高心 </s>
+output: <s>你快樂我也很高興 </s>
+------------------------------
+input: <s>但不再算再找實習生了 </s>
+output: <s>但不再去找實習生了 </s>
+------------------------------
+input: <s>今天太陽很大要注意篩傷 </s>
+output: <s>今天太陽很大要注意一下 </s>
+------------------------------
+input: <s>你要不要和我依起去台北 </s>
+output: <s>你要不要和我一起去台北 </s>
+------------------------------
+input: <s>清晨六點終太陽會升起 </s>
+output: <s>清晨六點鐘太陽會升起 </s>
+------------------------------
+input: <s>傾城六點鐘太陽會升起 </s>
+output: <s>凌晨六點鐘太陽會升起 </s>
+------------------------------
+input: <s>鍋馬路時你應該要注意虹綠燈 </s>
+output: <s>過馬路時你應該要注意紅綠燈 </s>
+------------------------------
+input: <s>他正在學學彈吉他 </s>
+output: <s>他正在學習彈吉他 </s>
+------------------------------
+input: <s>下樓梯請注意階梯 </s>
+output: <s>下樓梯請注意階梯 </s>
+------------------------------
+input: <s>此信件為系統自動發送之通知 </s>
+output: <s>此信件為系統自動發送之通知 </s>
+------------------------------
+input: <s>此信件為系統自動發送知通知 </s>
+output: <s>此信件為系統自動發送通知 </s>
+------------------------------
+input: <s>如為誤傳也請立即刪除本郵件並通知寄件者 </s>
+output: <s>如為誤傳也請立即刪除本郵件並通知寄件者 </s>
+------------------------------
+(venv) philip@nca100-3-G1:~/ec-dpo$ python test_model.py dpo_trainer/checkpoint-250
+input : <s>為了潔約能源請隨守關閉沒有使用的電器 </s>
+output: <s>為了節約能源請隨時關閉沒有使用的電器 </s>
+------------------------------
+input : <s>今天新情很好 </s>
+output: <s>今天心情很好 </s>
+------------------------------
+input : <s>你快樂我也很高心 </s>
+output: <s>你快樂我也很高興 </s>
+------------------------------
+input : <s>但不再算再找實習生了 </s>
+output: <s>但不再去找實習生了 </s>
+------------------------------
+input : <s>今天太陽很大要注意篩傷 </s>
+output: <s>今天太陽很大要注意一下 </s>
+------------------------------
+input : <s>你要不要和我依起去台北 </s>
+output: <s>你要不要和我一起去台北 </s>
+------------------------------
+input : <s>清晨六點終太陽會升起 </s>
+output: <s>清晨六點鐘太陽會升起 </s>
+------------------------------
+input : <s>傾城六點鐘太陽會升起 </s>
+output: <s>凌晨六點鐘太陽會升起 </s>
+------------------------------
+input : <s>鍋馬路時你應該要注意虹綠燈 </s>
+output: <s>過馬路時你應該要注意紅綠燈 </s>
+------------------------------
+input : <s>他正在學學彈吉他 </s>
+output: <s>他正在學習彈吉他 </s>
+------------------------------
+input : <s>下樓梯請注意階梯 </s>
+output: <s>下樓梯請注意階梯 </s>
+------------------------------
+input : <s>此信件為系統自動發送之通知 </s>
+output: <s>此信件為系統自動發送之通知 </s>
+------------------------------
+input : <s>此信件為系統自動發送知通知 </s>
+output: <s>此信件為系統自動發送通知 </s>
+------------------------------
+input : <s>如為誤傳也請立即刪除本郵件並通知寄件者 </s>
+output: <s>如為誤傳也請立即刪除本郵件並通知寄件者 </s>
+------------------------------
+```

config.json

@@ -0,0 +1,32 @@
+{
+  "_name_or_path": "sft_trainer/checkpoint-4500/",
+  "apply_residual_connection_post_layernorm": false,
+  "architectures": [
+    "BloomForCausalLM"
+  ],
+  "attention_dropout": 0.0,
+  "attention_softmax_in_fp32": true,
+  "bias_dropout_fusion": true,
+  "bos_token_id": 1,
+  "eos_token_id": 2,
+  "hidden_dropout": 0.0,
+  "hidden_size": 1536,
+  "initializer_range": 0.02,
+  "layer_norm_epsilon": 1e-05,
+  "masked_softmax_fusion": true,
+  "model_type": "bloom",
+  "n_head": 16,
+  "n_inner": null,
+  "n_layer": 24,
+  "offset_alibi": 100,
+  "pad_token_id": 3,
+  "pretraining_tp": 1,
+  "skip_bias_add": true,
+  "skip_bias_add_qkv": false,
+  "slow_but_exact": false,
+  "torch_dtype": "float32",
+  "transformers_version": "4.37.2",
+  "unk_token_id": 0,
+  "use_cache": true,
+  "vocab_size": 250880
+}

generation_config.json

@@ -0,0 +1,7 @@
+{
+  "_from_model_config": true,
+  "bos_token_id": 1,
+  "eos_token_id": 2,
+  "pad_token_id": 3,
+  "transformers_version": "4.37.2"
+}

model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2a4d934b57e15c85fabeee1c80fc1ba3fb58d9bd959865a102d1fedd35b0ebcd
+size 4261291440

optimizer.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2e82e1971b8b37f9437ead50ede64293d81ecf954e006d50246065f3b12a49f5
+size 8522768386

rng_state.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1ff264f99d31b522cc7e2a4eac9d38606d0c58a34c0adc74d71e0ca8b371dc36
+size 14244

scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6f3ec4f70580d870f44b786edc3a8bc0395e2f10d51f478622a7a57d30160892
+size 1064

special_tokens_map.json

@@ -0,0 +1,24 @@
+{
+  "bos_token": {
+    "content": "<s>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "eos_token": {
+    "content": "</s>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": "</s>",
+  "unk_token": {
+    "content": "<unk>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}

tokenizer.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:17a208233d2ee8d8c83b23bc214df737c44806a1919f444e89b31e586cd956ba
+size 14500471

tokenizer_config.json

@@ -0,0 +1,49 @@
+{
+  "add_prefix_space": false,
+  "added_tokens_decoder": {
+    "0": {
+      "content": "<unk>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "1": {
+      "content": "<s>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "2": {
+      "content": "</s>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "3": {
+      "content": "<pad>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    }
+  },
+  "bos_token": "<s>",
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "</s>",
+  "max_length": 256,
+  "model_max_length": 1000000000000000019884624838656,
+  "pad_token": "</s>",
+  "padding_side": "right",
+  "stride": 0,
+  "tokenizer_class": "BloomTokenizer",
+  "truncation_side": "right",
+  "truncation_strategy": "longest_first",
+  "unk_token": "<unk>"
+}

training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9299d7ea4fb442144a1ab68d137cae8b85e61eaf3c86b5bdbffc30c723e505cf
+size 4664