eskayML/mistral-7b-plaba-with-context

README.md

@@ -37,15 +37,15 @@ More information needed
 
 The following hyperparameters were used during training:
 - learning_rate: 0.0002
-- train_batch_size: 1
+- train_batch_size: 2
 - eval_batch_size: 8
 - seed: 3407
-- gradient_accumulation_steps: 2
-- total_train_batch_size: 2
+- gradient_accumulation_steps: 8
+- total_train_batch_size: 16
 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
 - lr_scheduler_type: linear
 - lr_scheduler_warmup_steps: 5
-- training_steps: 50
+- training_steps: 70
 - mixed_precision_training: Native AMP
 
 ### Training results

adapter_config.json

@@ -20,12 +20,12 @@
   "rank_pattern": {},
   "revision": null,
   "target_modules": [
-    "k_proj",
-    "o_proj",
+    "v_proj",
     "down_proj",
-    "gate_proj",
+    "o_proj",
     "up_proj",
-    "v_proj",
+    "k_proj",
+    "gate_proj",
     "q_proj"
   ],
   "task_type": "CAUSAL_LM",

adapter_model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:16d4ef93822fa31d6279f91dace055f3a17cf3610aa87577349b99b8774ef56e
+oid sha256:3a69ed07c45f1b0e775ba3377eeb2af2c0ac3809205895543f4da2597e0696cc
 size 167832240

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:98ac43cfdbf19556a93042c8dea609ab8236bbb906076a8cde6df7f9748ad3d7
+oid sha256:7927bdab41cadc85a03f2e9644c66f4365ac234737b13515387f90d84705a084
 size 5368