Megnis/saiga2-7b-vkr-analis

README.md

@@ -36,8 +36,9 @@ The following hyperparameters were used during training:
 - eval_batch_size: 8
 - seed: 42
 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
-- lr_scheduler_type: constant
+- lr_scheduler_type: linear
 - lr_scheduler_warmup_ratio: 0.03
+- lr_scheduler_warmup_steps: 10
 - num_epochs: 1
 
 ### Training results

adapter_config.json

@@ -1,21 +1,19 @@
 {
   "auto_mapping": null,
   "base_model_name_or_path": "pieken/saiga2_7b_lora_merged",
-  "bias": "none",
   "fan_in_fan_out": false,
+  "feedforward_modules": [
+    "down_proj"
+  ],
   "inference_mode": true,
-  "init_lora_weights": true,
-  "layers_pattern": null,
-  "layers_to_transform": null,
-  "lora_alpha": 16,
-  "lora_dropout": 0.1,
+  "init_ia3_weights": true,
   "modules_to_save": null,
-  "peft_type": "LORA",
-  "r": 8,
+  "peft_type": "IA3",
   "revision": null,
   "target_modules": [
-    "q_proj",
-    "v_proj"
+    "k_proj",
+    "v_proj",
+    "down_proj"
   ],
   "task_type": "CAUSAL_LM"
 }

adapter_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:f10d0cebf74c7e0c55e7c8751c11c62c743d731d5b6c927e2d039606461ad502
-size 16823434
+oid sha256:01105901817db08bed9e323f171edc53a18b9cc42337e6d81eef352a460e849a
+size 2491406

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:caf343bc4faa38a42fd4295254057bc7881ee89eac5ad128c9420e88645c06e8
+oid sha256:89709411c0f3350dcaa2dca754e3e9c8dc96e389902e97cbc7a3f4e91bf3155c
 size 4472