NumExample_75000Finetune_Epoch_3

README.md

@@ -6,16 +6,16 @@ library_name: peft
 
 The following `bitsandbytes` quantization config was used during training:
 - quant_method: bitsandbytes
-- load_in_8bit: True
-- load_in_4bit: False
+- load_in_8bit: False
+- load_in_4bit: True
 - llm_int8_threshold: 6.0
 - llm_int8_skip_modules: None
 - llm_int8_enable_fp32_cpu_offload: False
 - llm_int8_has_fp16_weight: False
-- bnb_4bit_quant_type: fp4
-- bnb_4bit_use_double_quant: False
-- bnb_4bit_compute_dtype: float32
+- bnb_4bit_quant_type: nf4
+- bnb_4bit_use_double_quant: True
+- bnb_4bit_compute_dtype: bfloat16
 ### Framework versions
 
 
-- PEFT 0.5.0.dev0
+- PEFT 0.6.0.dev0

adapter_config.json

@@ -1,14 +1,14 @@
 {
   "auto_mapping": null,
-  "base_model_name_or_path": "SKT27182/flan_t5_large_fine_tuned_head",
+  "base_model_name_or_path": "google/flan-t5-large",
   "bias": "none",
   "fan_in_fan_out": false,
   "inference_mode": true,
   "init_lora_weights": true,
   "layers_pattern": null,
   "layers_to_transform": null,
-  "lora_alpha": 16,
-  "lora_dropout": 0.05,
+  "lora_alpha": 512,
+  "lora_dropout": 0.01,
   "modules_to_save": null,
   "peft_type": "LORA",
   "r": 8,

adapter_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:0c02b673cb84b2cae7784a9fd3aafcd7999ff6917f73295ff0696ec5ba6dcb9b
+oid sha256:5717e6ceb65601c11660ffde54a9a336d2190ef9b1b4f4f92be505bd7f3ffed9
 size 9543245