init

adapter_config.json

@@ -4,11 +4,11 @@
   "fan_in_fan_out": false,
   "inference_mode": true,
   "init_lora_weights": true,
-  "lora_alpha": 1024,
+  "lora_alpha": 256,
   "lora_dropout": 0.05,
   "modules_to_save": null,
   "peft_type": "LORA",
-  "r": 512,
+  "r": 128,
   "target_modules": [
     "q_proj",
     "v_proj"

adapter_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2a42ec582bb934b176d722fb2ac82c0f4b45ee9a9829da398e0d04eeed92cf03
+size 419488077

training_parameters.json

@@ -1 +1 @@
-{"lora_name": "charluv-lora", "always_override": true, "save_steps": 0.0, "micro_batch_size": 4, "batch_size": 128, "epochs": 12.0, "learning_rate": "3e-5", "lr_scheduler_type": "cosine_with_restarts", "lora_rank": 512, "lora_alpha": 1024, "lora_dropout": 0.05, "cutoff_len": 256, "dataset": "None", "eval_dataset": "None", "format": "None", "eval_steps": 100.0, "raw_text_file": "aiva", "overlap_len": 128, "newline_favor_len": 128, "do_shuffle": true, "higher_rank_limit": false, "warmup_steps": 100.0, "optimizer": "sgd"}
+{"lora_name": "charluv-lora-v2", "always_override": true, "save_steps": 0.0, "micro_batch_size": 4, "batch_size": 128, "epochs": 3.0, "learning_rate": "3e-4", "lr_scheduler_type": "cosine_with_restarts", "lora_rank": 128, "lora_alpha": 256, "lora_dropout": 0.05, "cutoff_len": 256, "dataset": "None", "eval_dataset": "None", "format": "None", "eval_steps": 100.0, "raw_text_file": "aivo", "overlap_len": 128, "newline_favor_len": 128, "higher_rank_limit": false, "warmup_steps": 100.0, "optimizer": "adamw_torch", "hard_cut_string": "\\n\\n\\n", "train_only_after": ""}