epoch=4: train_epoch_loss.item()=5.126230551155686e-09 eval_epoch_loss.item()=7.991040051891218e-10{"valid_regex_ratio": 0.0}

README.md

@@ -1,3 +1,9 @@
 ---
-license: apache-2.0
+library_name: peft
 ---
+## Training procedure
+
+### Framework versions
+
+
+- PEFT 0.5.0

adapter_config.json

@@ -0,0 +1,16 @@
+{
+  "auto_mapping": null,
+  "base_model_name_or_path": "Salesforce/codegen-350M-mono",
+  "inference_mode": true,
+  "num_attention_heads": 16,
+  "num_layers": 20,
+  "num_transformer_submodules": 1,
+  "num_virtual_tokens": 8,
+  "peft_type": "PROMPT_TUNING",
+  "prompt_tuning_init": "TEXT",
+  "prompt_tuning_init_text": "Infer regex that full match the patterns provided. \nNote that:\n1. Each pattern is provided in a new line.\n2. The resulting regex should be wrapped in double quote.\n3. After the resulting regex is presented, the remaining explaination should be provided in a new line.\n\nFor example:\n\nThe patterns are:\n\"1\"\n\"2\"\n\"3\"\nThe regex is:\n\"[1-3]\"\nbecause the patterns are 1~3.\n\nNow start the real inference process:\n        ",
+  "revision": null,
+  "task_type": "CAUSAL_LM",
+  "token_dim": 1024,
+  "tokenizer_name_or_path": "Salesforce/codegen-350M-mono"
+}

adapter_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:57962ac045872bcb9cc9e97e46be8fda00cda28412dfda82966eb1fc61dea077
+size 34042