Upload 10 files

all_results.json

@@ -0,0 +1,14 @@
+{
+    "epoch": 1.0,
+    "eval_loss": 0.6936763525009155,
+    "eval_runtime": 4.0578,
+    "eval_samples": 932,
+    "eval_samples_per_second": 229.682,
+    "eval_steps_per_second": 7.393,
+    "perplexity": 2.0010586239542834,
+    "train_loss": 1.0507700066817434,
+    "train_runtime": 1517.4958,
+    "train_samples": 92214,
+    "train_samples_per_second": 60.767,
+    "train_steps_per_second": 0.238
+}

config.json

@@ -19,7 +19,7 @@
   "tie_word_embeddings": false,
   "torch_dtype": "float32",
   "transformers_version": "4.24.0",
-  "use_cache": true,
+  "use_cache": false,
   "use_parallel_residual": true,
-  "vocab_size": 50304
+  "vocab_size": 50277
 }

eval_results.json

@@ -0,0 +1,9 @@
+{
+    "epoch": 1.0,
+    "eval_loss": 0.6936763525009155,
+    "eval_runtime": 4.0578,
+    "eval_samples": 932,
+    "eval_samples_per_second": 229.682,
+    "eval_steps_per_second": 7.393,
+    "perplexity": 2.0010586239542834
+}

pytorch_model.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:695991fbaa6ecfda1bb18ae0bb89e8f43cda4502f82ff93c3421856c4fe0f8c4
-size 1722122365
+oid sha256:433863b0500502a7ba7d2313fb826b51aab489a26fbe66b354bdce4d4f4c7c71
+size 1721842252

special_tokens_map.json

@@ -1,5 +1,5 @@
 {
-  "bos_token": "<s>",
-  "eos_token": "</s>",
+  "bos_token": "<|endoftext|>",
+  "eos_token": "<|endoftext|>",
   "unk_token": "<|endoftext|>"
 }

train_results.json

@@ -0,0 +1,8 @@
+{
+    "epoch": 1.0,
+    "train_loss": 1.0507700066817434,
+    "train_runtime": 1517.4958,
+    "train_samples": 92214,
+    "train_samples_per_second": 60.767,
+    "train_steps_per_second": 0.238
+}

trainer_state.json

@@ -0,0 +1,33 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 1.0,
+  "global_step": 361,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.55,
+      "eval_loss": 0.7341781854629517,
+      "eval_runtime": 4.0764,
+      "eval_samples_per_second": 228.635,
+      "eval_steps_per_second": 7.359,
+      "step": 200
+    },
+    {
+      "epoch": 1.0,
+      "step": 361,
+      "total_flos": 5.481095099737702e+16,
+      "train_loss": 1.0507700066817434,
+      "train_runtime": 1517.4958,
+      "train_samples_per_second": 60.767,
+      "train_steps_per_second": 0.238
+    }
+  ],
+  "max_steps": 361,
+  "num_train_epochs": 1,
+  "total_flos": 5.481095099737702e+16,
+  "trial_name": null,
+  "trial_params": null
+}

training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cc13da0a8126df1f5b54f1121574bc60d44d4487278697423e787b7dbb7253ec
+size 4411