Upload folder using huggingface_hub

config.json

@@ -0,0 +1,56 @@
+{
+  "_name_or_path": "openai-community/gpt2-xl",
+  "activation_function": "gelu_new",
+  "architectures": [
+    "GPT2ForSequenceClassification"
+  ],
+  "attn_pdrop": 0.1,
+  "bos_token_id": 50256,
+  "embd_pdrop": 0.1,
+  "eos_token_id": 50256,
+  "id2label": {
+    "0": "LABEL_0",
+    "1": "LABEL_1",
+    "2": "LABEL_2",
+    "3": "LABEL_3",
+    "4": "LABEL_4"
+  },
+  "initializer_range": 0.02,
+  "label2id": {
+    "LABEL_0": 0,
+    "LABEL_1": 1,
+    "LABEL_2": 2,
+    "LABEL_3": 3,
+    "LABEL_4": 4
+  },
+  "layer_norm_epsilon": 1e-05,
+  "model_type": "gpt2",
+  "n_ctx": 1024,
+  "n_embd": 1600,
+  "n_head": 25,
+  "n_inner": null,
+  "n_layer": 48,
+  "n_positions": 1024,
+  "output_past": true,
+  "pad_token_id": 50256,
+  "problem_type": "single_label_classification",
+  "reorder_and_upcast_attn": false,
+  "resid_pdrop": 0.1,
+  "scale_attn_by_inverse_layer_idx": false,
+  "scale_attn_weights": true,
+  "summary_activation": null,
+  "summary_first_dropout": 0.1,
+  "summary_proj_to_labels": true,
+  "summary_type": "cls_index",
+  "summary_use_proj": true,
+  "task_specific_params": {
+    "text-generation": {
+      "do_sample": true,
+      "max_length": 50
+    }
+  },
+  "torch_dtype": "float32",
+  "transformers_version": "4.38.2",
+  "use_cache": true,
+  "vocab_size": 50257
+}

model-00001-of-00002.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2bb12de2e06ed578137127438112c48579d190e5169e37ecdf903144f66b499e
+size 4959881464

model-00002-of-00002.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:db4627a6cb14a2fc32d2deac782a89d7bd9bbd11623a6695f0b6a6a0e65216d5
+size 1270656176

model.safetensors.index.json

@@ -0,0 +1,588 @@
+{
+  "metadata": {
+    "total_size": 6230476800
+  },
+  "weight_map": {
+    "score.weight": "model-00002-of-00002.safetensors",
+    "transformer.h.0.attn.c_attn.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.0.attn.c_attn.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.0.attn.c_proj.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.0.attn.c_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.0.ln_1.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.0.ln_1.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.0.ln_2.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.0.ln_2.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.0.mlp.c_fc.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.0.mlp.c_fc.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.0.mlp.c_proj.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.0.mlp.c_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.1.attn.c_attn.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.1.attn.c_attn.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.1.attn.c_proj.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.1.attn.c_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.1.ln_1.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.1.ln_1.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.1.ln_2.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.1.ln_2.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.1.mlp.c_fc.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.1.mlp.c_fc.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.1.mlp.c_proj.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.1.mlp.c_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.10.attn.c_attn.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.10.attn.c_attn.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.10.attn.c_proj.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.10.attn.c_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.10.ln_1.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.10.ln_1.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.10.ln_2.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.10.ln_2.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.10.mlp.c_fc.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.10.mlp.c_fc.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.10.mlp.c_proj.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.10.mlp.c_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.11.attn.c_attn.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.11.attn.c_attn.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.11.attn.c_proj.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.11.attn.c_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.11.ln_1.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.11.ln_1.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.11.ln_2.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.11.ln_2.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.11.mlp.c_fc.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.11.mlp.c_fc.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.11.mlp.c_proj.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.11.mlp.c_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.12.attn.c_attn.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.12.attn.c_attn.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.12.attn.c_proj.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.12.attn.c_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.12.ln_1.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.12.ln_1.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.12.ln_2.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.12.ln_2.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.12.mlp.c_fc.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.12.mlp.c_fc.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.12.mlp.c_proj.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.12.mlp.c_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.13.attn.c_attn.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.13.attn.c_attn.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.13.attn.c_proj.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.13.attn.c_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.13.ln_1.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.13.ln_1.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.13.ln_2.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.13.ln_2.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.13.mlp.c_fc.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.13.mlp.c_fc.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.13.mlp.c_proj.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.13.mlp.c_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.14.attn.c_attn.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.14.attn.c_attn.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.14.attn.c_proj.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.14.attn.c_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.14.ln_1.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.14.ln_1.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.14.ln_2.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.14.ln_2.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.14.mlp.c_fc.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.14.mlp.c_fc.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.14.mlp.c_proj.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.14.mlp.c_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.15.attn.c_attn.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.15.attn.c_attn.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.15.attn.c_proj.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.15.attn.c_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.15.ln_1.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.15.ln_1.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.15.ln_2.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.15.ln_2.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.15.mlp.c_fc.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.15.mlp.c_fc.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.15.mlp.c_proj.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.15.mlp.c_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.16.attn.c_attn.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.16.attn.c_attn.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.16.attn.c_proj.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.16.attn.c_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.16.ln_1.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.16.ln_1.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.16.ln_2.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.16.ln_2.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.16.mlp.c_fc.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.16.mlp.c_fc.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.16.mlp.c_proj.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.16.mlp.c_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.17.attn.c_attn.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.17.attn.c_attn.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.17.attn.c_proj.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.17.attn.c_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.17.ln_1.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.17.ln_1.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.17.ln_2.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.17.ln_2.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.17.mlp.c_fc.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.17.mlp.c_fc.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.17.mlp.c_proj.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.17.mlp.c_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.18.attn.c_attn.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.18.attn.c_attn.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.18.attn.c_proj.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.18.attn.c_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.18.ln_1.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.18.ln_1.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.18.ln_2.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.18.ln_2.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.18.mlp.c_fc.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.18.mlp.c_fc.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.18.mlp.c_proj.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.18.mlp.c_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.19.attn.c_attn.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.19.attn.c_attn.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.19.attn.c_proj.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.19.attn.c_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.19.ln_1.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.19.ln_1.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.19.ln_2.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.19.ln_2.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.19.mlp.c_fc.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.19.mlp.c_fc.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.19.mlp.c_proj.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.19.mlp.c_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.2.attn.c_attn.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.2.attn.c_attn.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.2.attn.c_proj.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.2.attn.c_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.2.ln_1.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.2.ln_1.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.2.ln_2.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.2.ln_2.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.2.mlp.c_fc.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.2.mlp.c_fc.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.2.mlp.c_proj.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.2.mlp.c_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.20.attn.c_attn.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.20.attn.c_attn.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.20.attn.c_proj.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.20.attn.c_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.20.ln_1.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.20.ln_1.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.20.ln_2.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.20.ln_2.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.20.mlp.c_fc.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.20.mlp.c_fc.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.20.mlp.c_proj.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.20.mlp.c_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.21.attn.c_attn.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.21.attn.c_attn.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.21.attn.c_proj.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.21.attn.c_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.21.ln_1.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.21.ln_1.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.21.ln_2.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.21.ln_2.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.21.mlp.c_fc.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.21.mlp.c_fc.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.21.mlp.c_proj.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.21.mlp.c_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.22.attn.c_attn.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.22.attn.c_attn.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.22.attn.c_proj.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.22.attn.c_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.22.ln_1.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.22.ln_1.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.22.ln_2.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.22.ln_2.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.22.mlp.c_fc.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.22.mlp.c_fc.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.22.mlp.c_proj.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.22.mlp.c_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.23.attn.c_attn.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.23.attn.c_attn.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.23.attn.c_proj.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.23.attn.c_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.23.ln_1.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.23.ln_1.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.23.ln_2.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.23.ln_2.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.23.mlp.c_fc.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.23.mlp.c_fc.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.23.mlp.c_proj.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.23.mlp.c_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.24.attn.c_attn.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.24.attn.c_attn.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.24.attn.c_proj.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.24.attn.c_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.24.ln_1.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.24.ln_1.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.24.ln_2.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.24.ln_2.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.24.mlp.c_fc.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.24.mlp.c_fc.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.24.mlp.c_proj.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.24.mlp.c_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.25.attn.c_attn.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.25.attn.c_attn.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.25.attn.c_proj.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.25.attn.c_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.25.ln_1.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.25.ln_1.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.25.ln_2.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.25.ln_2.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.25.mlp.c_fc.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.25.mlp.c_fc.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.25.mlp.c_proj.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.25.mlp.c_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.26.attn.c_attn.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.26.attn.c_attn.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.26.attn.c_proj.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.26.attn.c_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.26.ln_1.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.26.ln_1.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.26.ln_2.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.26.ln_2.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.26.mlp.c_fc.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.26.mlp.c_fc.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.26.mlp.c_proj.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.26.mlp.c_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.27.attn.c_attn.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.27.attn.c_attn.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.27.attn.c_proj.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.27.attn.c_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.27.ln_1.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.27.ln_1.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.27.ln_2.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.27.ln_2.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.27.mlp.c_fc.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.27.mlp.c_fc.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.27.mlp.c_proj.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.27.mlp.c_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.28.attn.c_attn.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.28.attn.c_attn.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.28.attn.c_proj.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.28.attn.c_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.28.ln_1.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.28.ln_1.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.28.ln_2.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.28.ln_2.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.28.mlp.c_fc.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.28.mlp.c_fc.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.28.mlp.c_proj.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.28.mlp.c_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.29.attn.c_attn.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.29.attn.c_attn.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.29.attn.c_proj.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.29.attn.c_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.29.ln_1.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.29.ln_1.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.29.ln_2.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.29.ln_2.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.29.mlp.c_fc.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.29.mlp.c_fc.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.29.mlp.c_proj.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.29.mlp.c_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.3.attn.c_attn.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.3.attn.c_attn.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.3.attn.c_proj.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.3.attn.c_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.3.ln_1.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.3.ln_1.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.3.ln_2.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.3.ln_2.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.3.mlp.c_fc.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.3.mlp.c_fc.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.3.mlp.c_proj.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.3.mlp.c_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.30.attn.c_attn.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.30.attn.c_attn.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.30.attn.c_proj.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.30.attn.c_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.30.ln_1.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.30.ln_1.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.30.ln_2.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.30.ln_2.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.30.mlp.c_fc.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.30.mlp.c_fc.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.30.mlp.c_proj.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.30.mlp.c_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.31.attn.c_attn.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.31.attn.c_attn.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.31.attn.c_proj.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.31.attn.c_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.31.ln_1.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.31.ln_1.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.31.ln_2.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.31.ln_2.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.31.mlp.c_fc.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.31.mlp.c_fc.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.31.mlp.c_proj.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.31.mlp.c_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.32.attn.c_attn.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.32.attn.c_attn.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.32.attn.c_proj.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.32.attn.c_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.32.ln_1.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.32.ln_1.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.32.ln_2.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.32.ln_2.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.32.mlp.c_fc.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.32.mlp.c_fc.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.32.mlp.c_proj.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.32.mlp.c_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.33.attn.c_attn.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.33.attn.c_attn.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.33.attn.c_proj.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.33.attn.c_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.33.ln_1.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.33.ln_1.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.33.ln_2.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.33.ln_2.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.33.mlp.c_fc.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.33.mlp.c_fc.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.33.mlp.c_proj.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.33.mlp.c_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.34.attn.c_attn.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.34.attn.c_attn.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.34.attn.c_proj.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.34.attn.c_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.34.ln_1.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.34.ln_1.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.34.ln_2.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.34.ln_2.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.34.mlp.c_fc.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.34.mlp.c_fc.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.34.mlp.c_proj.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.34.mlp.c_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.35.attn.c_attn.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.35.attn.c_attn.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.35.attn.c_proj.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.35.attn.c_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.35.ln_1.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.35.ln_1.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.35.ln_2.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.35.ln_2.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.35.mlp.c_fc.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.35.mlp.c_fc.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.35.mlp.c_proj.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.35.mlp.c_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.36.attn.c_attn.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.36.attn.c_attn.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.36.attn.c_proj.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.36.attn.c_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.36.ln_1.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.36.ln_1.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.36.ln_2.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.36.ln_2.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.36.mlp.c_fc.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.36.mlp.c_fc.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.36.mlp.c_proj.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.36.mlp.c_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.37.attn.c_attn.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.37.attn.c_attn.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.37.attn.c_proj.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.37.attn.c_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.37.ln_1.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.37.ln_1.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.37.ln_2.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.37.ln_2.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.37.mlp.c_fc.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.37.mlp.c_fc.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.37.mlp.c_proj.bias": "model-00002-of-00002.safetensors",
+    "transformer.h.37.mlp.c_proj.weight": "model-00002-of-00002.safetensors",
+    "transformer.h.38.attn.c_attn.bias": "model-00002-of-00002.safetensors",
+    "transformer.h.38.attn.c_attn.weight": "model-00002-of-00002.safetensors",
+    "transformer.h.38.attn.c_proj.bias": "model-00002-of-00002.safetensors",
+    "transformer.h.38.attn.c_proj.weight": "model-00002-of-00002.safetensors",
+    "transformer.h.38.ln_1.bias": "model-00002-of-00002.safetensors",
+    "transformer.h.38.ln_1.weight": "model-00002-of-00002.safetensors",
+    "transformer.h.38.ln_2.bias": "model-00002-of-00002.safetensors",
+    "transformer.h.38.ln_2.weight": "model-00002-of-00002.safetensors",
+    "transformer.h.38.mlp.c_fc.bias": "model-00002-of-00002.safetensors",
+    "transformer.h.38.mlp.c_fc.weight": "model-00002-of-00002.safetensors",
+    "transformer.h.38.mlp.c_proj.bias": "model-00002-of-00002.safetensors",
+    "transformer.h.38.mlp.c_proj.weight": "model-00002-of-00002.safetensors",
+    "transformer.h.39.attn.c_attn.bias": "model-00002-of-00002.safetensors",
+    "transformer.h.39.attn.c_attn.weight": "model-00002-of-00002.safetensors",
+    "transformer.h.39.attn.c_proj.bias": "model-00002-of-00002.safetensors",
+    "transformer.h.39.attn.c_proj.weight": "model-00002-of-00002.safetensors",
+    "transformer.h.39.ln_1.bias": "model-00002-of-00002.safetensors",
+    "transformer.h.39.ln_1.weight": "model-00002-of-00002.safetensors",
+    "transformer.h.39.ln_2.bias": "model-00002-of-00002.safetensors",
+    "transformer.h.39.ln_2.weight": "model-00002-of-00002.safetensors",
+    "transformer.h.39.mlp.c_fc.bias": "model-00002-of-00002.safetensors",
+    "transformer.h.39.mlp.c_fc.weight": "model-00002-of-00002.safetensors",
+    "transformer.h.39.mlp.c_proj.bias": "model-00002-of-00002.safetensors",
+    "transformer.h.39.mlp.c_proj.weight": "model-00002-of-00002.safetensors",
+    "transformer.h.4.attn.c_attn.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.4.attn.c_attn.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.4.attn.c_proj.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.4.attn.c_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.4.ln_1.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.4.ln_1.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.4.ln_2.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.4.ln_2.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.4.mlp.c_fc.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.4.mlp.c_fc.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.4.mlp.c_proj.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.4.mlp.c_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.40.attn.c_attn.bias": "model-00002-of-00002.safetensors",
+    "transformer.h.40.attn.c_attn.weight": "model-00002-of-00002.safetensors",
+    "transformer.h.40.attn.c_proj.bias": "model-00002-of-00002.safetensors",
+    "transformer.h.40.attn.c_proj.weight": "model-00002-of-00002.safetensors",
+    "transformer.h.40.ln_1.bias": "model-00002-of-00002.safetensors",
+    "transformer.h.40.ln_1.weight": "model-00002-of-00002.safetensors",
+    "transformer.h.40.ln_2.bias": "model-00002-of-00002.safetensors",
+    "transformer.h.40.ln_2.weight": "model-00002-of-00002.safetensors",
+    "transformer.h.40.mlp.c_fc.bias": "model-00002-of-00002.safetensors",
+    "transformer.h.40.mlp.c_fc.weight": "model-00002-of-00002.safetensors",
+    "transformer.h.40.mlp.c_proj.bias": "model-00002-of-00002.safetensors",
+    "transformer.h.40.mlp.c_proj.weight": "model-00002-of-00002.safetensors",
+    "transformer.h.41.attn.c_attn.bias": "model-00002-of-00002.safetensors",
+    "transformer.h.41.attn.c_attn.weight": "model-00002-of-00002.safetensors",
+    "transformer.h.41.attn.c_proj.bias": "model-00002-of-00002.safetensors",
+    "transformer.h.41.attn.c_proj.weight": "model-00002-of-00002.safetensors",
+    "transformer.h.41.ln_1.bias": "model-00002-of-00002.safetensors",
+    "transformer.h.41.ln_1.weight": "model-00002-of-00002.safetensors",
+    "transformer.h.41.ln_2.bias": "model-00002-of-00002.safetensors",
+    "transformer.h.41.ln_2.weight": "model-00002-of-00002.safetensors",
+    "transformer.h.41.mlp.c_fc.bias": "model-00002-of-00002.safetensors",
+    "transformer.h.41.mlp.c_fc.weight": "model-00002-of-00002.safetensors",
+    "transformer.h.41.mlp.c_proj.bias": "model-00002-of-00002.safetensors",
+    "transformer.h.41.mlp.c_proj.weight": "model-00002-of-00002.safetensors",
+    "transformer.h.42.attn.c_attn.bias": "model-00002-of-00002.safetensors",
+    "transformer.h.42.attn.c_attn.weight": "model-00002-of-00002.safetensors",
+    "transformer.h.42.attn.c_proj.bias": "model-00002-of-00002.safetensors",
+    "transformer.h.42.attn.c_proj.weight": "model-00002-of-00002.safetensors",
+    "transformer.h.42.ln_1.bias": "model-00002-of-00002.safetensors",
+    "transformer.h.42.ln_1.weight": "model-00002-of-00002.safetensors",
+    "transformer.h.42.ln_2.bias": "model-00002-of-00002.safetensors",
+    "transformer.h.42.ln_2.weight": "model-00002-of-00002.safetensors",
+    "transformer.h.42.mlp.c_fc.bias": "model-00002-of-00002.safetensors",
+    "transformer.h.42.mlp.c_fc.weight": "model-00002-of-00002.safetensors",
+    "transformer.h.42.mlp.c_proj.bias": "model-00002-of-00002.safetensors",
+    "transformer.h.42.mlp.c_proj.weight": "model-00002-of-00002.safetensors",
+    "transformer.h.43.attn.c_attn.bias": "model-00002-of-00002.safetensors",
+    "transformer.h.43.attn.c_attn.weight": "model-00002-of-00002.safetensors",
+    "transformer.h.43.attn.c_proj.bias": "model-00002-of-00002.safetensors",
+    "transformer.h.43.attn.c_proj.weight": "model-00002-of-00002.safetensors",
+    "transformer.h.43.ln_1.bias": "model-00002-of-00002.safetensors",
+    "transformer.h.43.ln_1.weight": "model-00002-of-00002.safetensors",
+    "transformer.h.43.ln_2.bias": "model-00002-of-00002.safetensors",
+    "transformer.h.43.ln_2.weight": "model-00002-of-00002.safetensors",
+    "transformer.h.43.mlp.c_fc.bias": "model-00002-of-00002.safetensors",
+    "transformer.h.43.mlp.c_fc.weight": "model-00002-of-00002.safetensors",
+    "transformer.h.43.mlp.c_proj.bias": "model-00002-of-00002.safetensors",
+    "transformer.h.43.mlp.c_proj.weight": "model-00002-of-00002.safetensors",
+    "transformer.h.44.attn.c_attn.bias": "model-00002-of-00002.safetensors",
+    "transformer.h.44.attn.c_attn.weight": "model-00002-of-00002.safetensors",
+    "transformer.h.44.attn.c_proj.bias": "model-00002-of-00002.safetensors",
+    "transformer.h.44.attn.c_proj.weight": "model-00002-of-00002.safetensors",
+    "transformer.h.44.ln_1.bias": "model-00002-of-00002.safetensors",
+    "transformer.h.44.ln_1.weight": "model-00002-of-00002.safetensors",
+    "transformer.h.44.ln_2.bias": "model-00002-of-00002.safetensors",
+    "transformer.h.44.ln_2.weight": "model-00002-of-00002.safetensors",
+    "transformer.h.44.mlp.c_fc.bias": "model-00002-of-00002.safetensors",
+    "transformer.h.44.mlp.c_fc.weight": "model-00002-of-00002.safetensors",
+    "transformer.h.44.mlp.c_proj.bias": "model-00002-of-00002.safetensors",
+    "transformer.h.44.mlp.c_proj.weight": "model-00002-of-00002.safetensors",
+    "transformer.h.45.attn.c_attn.bias": "model-00002-of-00002.safetensors",
+    "transformer.h.45.attn.c_attn.weight": "model-00002-of-00002.safetensors",
+    "transformer.h.45.attn.c_proj.bias": "model-00002-of-00002.safetensors",
+    "transformer.h.45.attn.c_proj.weight": "model-00002-of-00002.safetensors",
+    "transformer.h.45.ln_1.bias": "model-00002-of-00002.safetensors",
+    "transformer.h.45.ln_1.weight": "model-00002-of-00002.safetensors",
+    "transformer.h.45.ln_2.bias": "model-00002-of-00002.safetensors",
+    "transformer.h.45.ln_2.weight": "model-00002-of-00002.safetensors",
+    "transformer.h.45.mlp.c_fc.bias": "model-00002-of-00002.safetensors",
+    "transformer.h.45.mlp.c_fc.weight": "model-00002-of-00002.safetensors",
+    "transformer.h.45.mlp.c_proj.bias": "model-00002-of-00002.safetensors",
+    "transformer.h.45.mlp.c_proj.weight": "model-00002-of-00002.safetensors",
+    "transformer.h.46.attn.c_attn.bias": "model-00002-of-00002.safetensors",
+    "transformer.h.46.attn.c_attn.weight": "model-00002-of-00002.safetensors",
+    "transformer.h.46.attn.c_proj.bias": "model-00002-of-00002.safetensors",
+    "transformer.h.46.attn.c_proj.weight": "model-00002-of-00002.safetensors",
+    "transformer.h.46.ln_1.bias": "model-00002-of-00002.safetensors",
+    "transformer.h.46.ln_1.weight": "model-00002-of-00002.safetensors",
+    "transformer.h.46.ln_2.bias": "model-00002-of-00002.safetensors",
+    "transformer.h.46.ln_2.weight": "model-00002-of-00002.safetensors",
+    "transformer.h.46.mlp.c_fc.bias": "model-00002-of-00002.safetensors",
+    "transformer.h.46.mlp.c_fc.weight": "model-00002-of-00002.safetensors",
+    "transformer.h.46.mlp.c_proj.bias": "model-00002-of-00002.safetensors",
+    "transformer.h.46.mlp.c_proj.weight": "model-00002-of-00002.safetensors",
+    "transformer.h.47.attn.c_attn.bias": "model-00002-of-00002.safetensors",
+    "transformer.h.47.attn.c_attn.weight": "model-00002-of-00002.safetensors",
+    "transformer.h.47.attn.c_proj.bias": "model-00002-of-00002.safetensors",
+    "transformer.h.47.attn.c_proj.weight": "model-00002-of-00002.safetensors",
+    "transformer.h.47.ln_1.bias": "model-00002-of-00002.safetensors",
+    "transformer.h.47.ln_1.weight": "model-00002-of-00002.safetensors",
+    "transformer.h.47.ln_2.bias": "model-00002-of-00002.safetensors",
+    "transformer.h.47.ln_2.weight": "model-00002-of-00002.safetensors",
+    "transformer.h.47.mlp.c_fc.bias": "model-00002-of-00002.safetensors",
+    "transformer.h.47.mlp.c_fc.weight": "model-00002-of-00002.safetensors",
+    "transformer.h.47.mlp.c_proj.bias": "model-00002-of-00002.safetensors",
+    "transformer.h.47.mlp.c_proj.weight": "model-00002-of-00002.safetensors",
+    "transformer.h.5.attn.c_attn.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.5.attn.c_attn.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.5.attn.c_proj.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.5.attn.c_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.5.ln_1.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.5.ln_1.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.5.ln_2.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.5.ln_2.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.5.mlp.c_fc.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.5.mlp.c_fc.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.5.mlp.c_proj.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.5.mlp.c_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.6.attn.c_attn.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.6.attn.c_attn.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.6.attn.c_proj.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.6.attn.c_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.6.ln_1.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.6.ln_1.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.6.ln_2.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.6.ln_2.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.6.mlp.c_fc.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.6.mlp.c_fc.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.6.mlp.c_proj.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.6.mlp.c_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.7.attn.c_attn.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.7.attn.c_attn.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.7.attn.c_proj.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.7.attn.c_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.7.ln_1.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.7.ln_1.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.7.ln_2.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.7.ln_2.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.7.mlp.c_fc.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.7.mlp.c_fc.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.7.mlp.c_proj.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.7.mlp.c_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.8.attn.c_attn.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.8.attn.c_attn.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.8.attn.c_proj.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.8.attn.c_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.8.ln_1.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.8.ln_1.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.8.ln_2.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.8.ln_2.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.8.mlp.c_fc.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.8.mlp.c_fc.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.8.mlp.c_proj.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.8.mlp.c_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.9.attn.c_attn.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.9.attn.c_attn.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.9.attn.c_proj.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.9.attn.c_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.9.ln_1.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.9.ln_1.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.9.ln_2.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.9.ln_2.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.9.mlp.c_fc.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.9.mlp.c_fc.weight": "model-00001-of-00002.safetensors",
+    "transformer.h.9.mlp.c_proj.bias": "model-00001-of-00002.safetensors",
+    "transformer.h.9.mlp.c_proj.weight": "model-00001-of-00002.safetensors",
+    "transformer.ln_f.bias": "model-00002-of-00002.safetensors",
+    "transformer.ln_f.weight": "model-00002-of-00002.safetensors",
+    "transformer.wpe.weight": "model-00001-of-00002.safetensors",
+    "transformer.wte.weight": "model-00001-of-00002.safetensors"
+  }
+}

optimizer.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d6fa81ba98739d7e4d636f96a25e273984133de1c10c50fa2595ca20d70aa18b
+size 12461449840

rng_state.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:da5e6f302edbb9f0258fc10b547b7cb03de97694d574f43fe2eb34f4dcf774cd
+size 14575

scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:54aeff90d18da8078e79567e715fb5689447dcbff149fb0ee15c76b534d4eb83
+size 627

special_tokens_map.json

@@ -0,0 +1,6 @@
+{
+  "bos_token": "<|endoftext|>",
+  "eos_token": "<|endoftext|>",
+  "pad_token": "<|endoftext|>",
+  "unk_token": "<|endoftext|>"
+}

tokenizer_config.json

@@ -0,0 +1,20 @@
+{
+  "add_prefix_space": false,
+  "added_tokens_decoder": {
+    "50256": {
+      "content": "<|endoftext|>",
+      "lstrip": false,
+      "normalized": true,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    }
+  },
+  "bos_token": "<|endoftext|>",
+  "clean_up_tokenization_spaces": true,
+  "eos_token": "<|endoftext|>",
+  "model_max_length": 1024,
+  "pad_token": "<|endoftext|>",
+  "tokenizer_class": "GPT2Tokenizer",
+  "unk_token": "<|endoftext|>"
+}

trainer_state.json

@@ -0,0 +1,1542 @@
+{
+  "best_metric": 0.750106155872345,
+  "best_model_checkpoint": "sean_test_out_large/checkpoint-108334",
+  "epoch": 1.0,
+  "eval_steps": 500,
+  "global_step": 108334,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0,
+      "grad_norm": 1387778.25,
+      "learning_rate": 5.972307862720845e-05,
+      "loss": 1.1897,
+      "step": 500
+    },
+    {
+      "epoch": 0.01,
+      "grad_norm": 664302.625,
+      "learning_rate": 5.94461572544169e-05,
+      "loss": 1.0759,
+      "step": 1000
+    },
+    {
+      "epoch": 0.01,
+      "grad_norm": 428486.84375,
+      "learning_rate": 5.9169235881625344e-05,
+      "loss": 1.0728,
+      "step": 1500
+    },
+    {
+      "epoch": 0.02,
+      "grad_norm": 299683.625,
+      "learning_rate": 5.8892314508833796e-05,
+      "loss": 1.0409,
+      "step": 2000
+    },
+    {
+      "epoch": 0.02,
+      "grad_norm": 503517.40625,
+      "learning_rate": 5.861539313604224e-05,
+      "loss": 1.0335,
+      "step": 2500
+    },
+    {
+      "epoch": 0.03,
+      "grad_norm": 657503.8125,
+      "learning_rate": 5.833847176325069e-05,
+      "loss": 0.9533,
+      "step": 3000
+    },
+    {
+      "epoch": 0.03,
+      "grad_norm": 533787.3125,
+      "learning_rate": 5.806155039045914e-05,
+      "loss": 0.9992,
+      "step": 3500
+    },
+    {
+      "epoch": 0.04,
+      "grad_norm": 913784.6875,
+      "learning_rate": 5.7784629017667584e-05,
+      "loss": 0.9657,
+      "step": 4000
+    },
+    {
+      "epoch": 0.04,
+      "grad_norm": 310152.21875,
+      "learning_rate": 5.7507707644876036e-05,
+      "loss": 0.9652,
+      "step": 4500
+    },
+    {
+      "epoch": 0.05,
+      "grad_norm": 309690.34375,
+      "learning_rate": 5.723078627208448e-05,
+      "loss": 0.969,
+      "step": 5000
+    },
+    {
+      "epoch": 0.05,
+      "grad_norm": 211456.5,
+      "learning_rate": 5.695386489929293e-05,
+      "loss": 0.9515,
+      "step": 5500
+    },
+    {
+      "epoch": 0.06,
+      "grad_norm": 391416.65625,
+      "learning_rate": 5.667694352650138e-05,
+      "loss": 0.9221,
+      "step": 6000
+    },
+    {
+      "epoch": 0.06,
+      "grad_norm": 221054.703125,
+      "learning_rate": 5.6400022153709824e-05,
+      "loss": 0.9679,
+      "step": 6500
+    },
+    {
+      "epoch": 0.06,
+      "grad_norm": 472076.15625,
+      "learning_rate": 5.612310078091827e-05,
+      "loss": 0.9519,
+      "step": 7000
+    },
+    {
+      "epoch": 0.07,
+      "grad_norm": 988913.375,
+      "learning_rate": 5.584617940812672e-05,
+      "loss": 0.9447,
+      "step": 7500
+    },
+    {
+      "epoch": 0.07,
+      "grad_norm": 661127.75,
+      "learning_rate": 5.556925803533517e-05,
+      "loss": 0.9009,
+      "step": 8000
+    },
+    {
+      "epoch": 0.08,
+      "grad_norm": 212504.640625,
+      "learning_rate": 5.529233666254362e-05,
+      "loss": 0.9456,
+      "step": 8500
+    },
+    {
+      "epoch": 0.08,
+      "grad_norm": 473405.96875,
+      "learning_rate": 5.5015415289752064e-05,
+      "loss": 0.9385,
+      "step": 9000
+    },
+    {
+      "epoch": 0.09,
+      "grad_norm": 266084.78125,
+      "learning_rate": 5.473849391696051e-05,
+      "loss": 1.0131,
+      "step": 9500
+    },
+    {
+      "epoch": 0.09,
+      "grad_norm": 491548.53125,
+      "learning_rate": 5.446157254416896e-05,
+      "loss": 0.9254,
+      "step": 10000
+    },
+    {
+      "epoch": 0.1,
+      "grad_norm": 267938.4375,
+      "learning_rate": 5.418465117137741e-05,
+      "loss": 0.9101,
+      "step": 10500
+    },
+    {
+      "epoch": 0.1,
+      "grad_norm": 258540.84375,
+      "learning_rate": 5.390772979858586e-05,
+      "loss": 0.9234,
+      "step": 11000
+    },
+    {
+      "epoch": 0.11,
+      "grad_norm": 200250.609375,
+      "learning_rate": 5.3630808425794305e-05,
+      "loss": 0.9157,
+      "step": 11500
+    },
+    {
+      "epoch": 0.11,
+      "grad_norm": 480926.25,
+      "learning_rate": 5.335388705300275e-05,
+      "loss": 0.9532,
+      "step": 12000
+    },
+    {
+      "epoch": 0.12,
+      "grad_norm": 593979.4375,
+      "learning_rate": 5.30769656802112e-05,
+      "loss": 0.9184,
+      "step": 12500
+    },
+    {
+      "epoch": 0.12,
+      "grad_norm": 686144.1875,
+      "learning_rate": 5.280004430741965e-05,
+      "loss": 0.9215,
+      "step": 13000
+    },
+    {
+      "epoch": 0.12,
+      "grad_norm": 582782.0,
+      "learning_rate": 5.25231229346281e-05,
+      "loss": 0.9151,
+      "step": 13500
+    },
+    {
+      "epoch": 0.13,
+      "grad_norm": 554729.875,
+      "learning_rate": 5.2246201561836545e-05,
+      "loss": 0.9042,
+      "step": 14000
+    },
+    {
+      "epoch": 0.13,
+      "grad_norm": 423190.0,
+      "learning_rate": 5.196928018904499e-05,
+      "loss": 0.9114,
+      "step": 14500
+    },
+    {
+      "epoch": 0.14,
+      "grad_norm": 724497.5,
+      "learning_rate": 5.169235881625344e-05,
+      "loss": 0.9123,
+      "step": 15000
+    },
+    {
+      "epoch": 0.14,
+      "grad_norm": 333863.78125,
+      "learning_rate": 5.141543744346189e-05,
+      "loss": 0.9237,
+      "step": 15500
+    },
+    {
+      "epoch": 0.15,
+      "grad_norm": 282522.5,
+      "learning_rate": 5.113851607067033e-05,
+      "loss": 0.8927,
+      "step": 16000
+    },
+    {
+      "epoch": 0.15,
+      "grad_norm": 326572.34375,
+      "learning_rate": 5.0861594697878785e-05,
+      "loss": 0.9256,
+      "step": 16500
+    },
+    {
+      "epoch": 0.16,
+      "grad_norm": 454635.34375,
+      "learning_rate": 5.058467332508723e-05,
+      "loss": 0.8991,
+      "step": 17000
+    },
+    {
+      "epoch": 0.16,
+      "grad_norm": 306963.25,
+      "learning_rate": 5.030775195229568e-05,
+      "loss": 0.8937,
+      "step": 17500
+    },
+    {
+      "epoch": 0.17,
+      "grad_norm": 580224.0,
+      "learning_rate": 5.003083057950413e-05,
+      "loss": 0.9108,
+      "step": 18000
+    },
+    {
+      "epoch": 0.17,
+      "grad_norm": 549220.75,
+      "learning_rate": 4.975390920671257e-05,
+      "loss": 0.9307,
+      "step": 18500
+    },
+    {
+      "epoch": 0.18,
+      "grad_norm": 188370.0,
+      "learning_rate": 4.9476987833921025e-05,
+      "loss": 0.9009,
+      "step": 19000
+    },
+    {
+      "epoch": 0.18,
+      "grad_norm": 535151.8125,
+      "learning_rate": 4.920006646112947e-05,
+      "loss": 0.8773,
+      "step": 19500
+    },
+    {
+      "epoch": 0.18,
+      "grad_norm": 675718.1875,
+      "learning_rate": 4.892314508833792e-05,
+      "loss": 0.8984,
+      "step": 20000
+    },
+    {
+      "epoch": 0.19,
+      "grad_norm": 247562.0,
+      "learning_rate": 4.864622371554637e-05,
+      "loss": 0.8896,
+      "step": 20500
+    },
+    {
+      "epoch": 0.19,
+      "grad_norm": 412286.65625,
+      "learning_rate": 4.836930234275481e-05,
+      "loss": 0.8914,
+      "step": 21000
+    },
+    {
+      "epoch": 0.2,
+      "grad_norm": 362339.53125,
+      "learning_rate": 4.8092380969963265e-05,
+      "loss": 0.8906,
+      "step": 21500
+    },
+    {
+      "epoch": 0.2,
+      "grad_norm": 307259.8125,
+      "learning_rate": 4.781545959717171e-05,
+      "loss": 0.8615,
+      "step": 22000
+    },
+    {
+      "epoch": 0.21,
+      "grad_norm": 264488.625,
+      "learning_rate": 4.753853822438016e-05,
+      "loss": 0.896,
+      "step": 22500
+    },
+    {
+      "epoch": 0.21,
+      "grad_norm": 392361.15625,
+      "learning_rate": 4.726161685158861e-05,
+      "loss": 0.8802,
+      "step": 23000
+    },
+    {
+      "epoch": 0.22,
+      "grad_norm": 844823.375,
+      "learning_rate": 4.698469547879705e-05,
+      "loss": 0.8927,
+      "step": 23500
+    },
+    {
+      "epoch": 0.22,
+      "grad_norm": 252408.484375,
+      "learning_rate": 4.6707774106005505e-05,
+      "loss": 0.8942,
+      "step": 24000
+    },
+    {
+      "epoch": 0.23,
+      "grad_norm": 591983.0625,
+      "learning_rate": 4.643085273321395e-05,
+      "loss": 0.8859,
+      "step": 24500
+    },
+    {
+      "epoch": 0.23,
+      "grad_norm": 284245.46875,
+      "learning_rate": 4.6153931360422396e-05,
+      "loss": 0.8953,
+      "step": 25000
+    },
+    {
+      "epoch": 0.24,
+      "grad_norm": 288328.15625,
+      "learning_rate": 4.587700998763085e-05,
+      "loss": 0.8731,
+      "step": 25500
+    },
+    {
+      "epoch": 0.24,
+      "grad_norm": 422104.84375,
+      "learning_rate": 4.560008861483929e-05,
+      "loss": 0.8695,
+      "step": 26000
+    },
+    {
+      "epoch": 0.24,
+      "grad_norm": 450527.375,
+      "learning_rate": 4.5323167242047745e-05,
+      "loss": 0.8616,
+      "step": 26500
+    },
+    {
+      "epoch": 0.25,
+      "grad_norm": 440466.5,
+      "learning_rate": 4.504624586925619e-05,
+      "loss": 0.8871,
+      "step": 27000
+    },
+    {
+      "epoch": 0.25,
+      "grad_norm": 386183.34375,
+      "learning_rate": 4.4769324496464636e-05,
+      "loss": 0.8645,
+      "step": 27500
+    },
+    {
+      "epoch": 0.26,
+      "grad_norm": 542583.0,
+      "learning_rate": 4.449240312367309e-05,
+      "loss": 0.8755,
+      "step": 28000
+    },
+    {
+      "epoch": 0.26,
+      "grad_norm": 497213.90625,
+      "learning_rate": 4.421548175088153e-05,
+      "loss": 0.8663,
+      "step": 28500
+    },
+    {
+      "epoch": 0.27,
+      "grad_norm": 1084880.375,
+      "learning_rate": 4.3938560378089985e-05,
+      "loss": 0.8371,
+      "step": 29000
+    },
+    {
+      "epoch": 0.27,
+      "grad_norm": 365434.125,
+      "learning_rate": 4.366163900529843e-05,
+      "loss": 0.8791,
+      "step": 29500
+    },
+    {
+      "epoch": 0.28,
+      "grad_norm": 310639.5,
+      "learning_rate": 4.3384717632506876e-05,
+      "loss": 0.8298,
+      "step": 30000
+    },
+    {
+      "epoch": 0.28,
+      "grad_norm": 1322288.0,
+      "learning_rate": 4.310779625971533e-05,
+      "loss": 0.8658,
+      "step": 30500
+    },
+    {
+      "epoch": 0.29,
+      "grad_norm": 488033.0625,
+      "learning_rate": 4.283087488692377e-05,
+      "loss": 0.8556,
+      "step": 31000
+    },
+    {
+      "epoch": 0.29,
+      "grad_norm": 404639.3125,
+      "learning_rate": 4.2553953514132225e-05,
+      "loss": 0.8547,
+      "step": 31500
+    },
+    {
+      "epoch": 0.3,
+      "grad_norm": 324229.59375,
+      "learning_rate": 4.227703214134067e-05,
+      "loss": 0.8853,
+      "step": 32000
+    },
+    {
+      "epoch": 0.3,
+      "grad_norm": 391828.6875,
+      "learning_rate": 4.2000110768549116e-05,
+      "loss": 0.8725,
+      "step": 32500
+    },
+    {
+      "epoch": 0.3,
+      "grad_norm": 481346.0,
+      "learning_rate": 4.172318939575757e-05,
+      "loss": 0.8484,
+      "step": 33000
+    },
+    {
+      "epoch": 0.31,
+      "grad_norm": 500705.90625,
+      "learning_rate": 4.144626802296601e-05,
+      "loss": 0.8415,
+      "step": 33500
+    },
+    {
+      "epoch": 0.31,
+      "grad_norm": 484916.0625,
+      "learning_rate": 4.116934665017446e-05,
+      "loss": 0.8464,
+      "step": 34000
+    },
+    {
+      "epoch": 0.32,
+      "grad_norm": 305320.15625,
+      "learning_rate": 4.089242527738291e-05,
+      "loss": 0.8439,
+      "step": 34500
+    },
+    {
+      "epoch": 0.32,
+      "grad_norm": 275214.6875,
+      "learning_rate": 4.0615503904591356e-05,
+      "loss": 0.8257,
+      "step": 35000
+    },
+    {
+      "epoch": 0.33,
+      "grad_norm": 364166.0,
+      "learning_rate": 4.033858253179981e-05,
+      "loss": 0.8763,
+      "step": 35500
+    },
+    {
+      "epoch": 0.33,
+      "grad_norm": 285496.46875,
+      "learning_rate": 4.006166115900825e-05,
+      "loss": 0.856,
+      "step": 36000
+    },
+    {
+      "epoch": 0.34,
+      "grad_norm": 959005.0625,
+      "learning_rate": 3.97847397862167e-05,
+      "loss": 0.8306,
+      "step": 36500
+    },
+    {
+      "epoch": 0.34,
+      "grad_norm": 357090.28125,
+      "learning_rate": 3.950781841342515e-05,
+      "loss": 0.8564,
+      "step": 37000
+    },
+    {
+      "epoch": 0.35,
+      "grad_norm": 605763.8125,
+      "learning_rate": 3.9230897040633596e-05,
+      "loss": 0.8476,
+      "step": 37500
+    },
+    {
+      "epoch": 0.35,
+      "grad_norm": 418555.625,
+      "learning_rate": 3.895397566784205e-05,
+      "loss": 0.857,
+      "step": 38000
+    },
+    {
+      "epoch": 0.36,
+      "grad_norm": 726787.125,
+      "learning_rate": 3.8677054295050493e-05,
+      "loss": 0.8627,
+      "step": 38500
+    },
+    {
+      "epoch": 0.36,
+      "grad_norm": 212226.21875,
+      "learning_rate": 3.840013292225894e-05,
+      "loss": 0.8172,
+      "step": 39000
+    },
+    {
+      "epoch": 0.36,
+      "grad_norm": 219878.765625,
+      "learning_rate": 3.812321154946739e-05,
+      "loss": 0.8476,
+      "step": 39500
+    },
+    {
+      "epoch": 0.37,
+      "grad_norm": 520633.34375,
+      "learning_rate": 3.7846290176675836e-05,
+      "loss": 0.8226,
+      "step": 40000
+    },
+    {
+      "epoch": 0.37,
+      "grad_norm": 317160.71875,
+      "learning_rate": 3.756936880388429e-05,
+      "loss": 0.8235,
+      "step": 40500
+    },
+    {
+      "epoch": 0.38,
+      "grad_norm": 409949.8125,
+      "learning_rate": 3.7292447431092734e-05,
+      "loss": 0.8112,
+      "step": 41000
+    },
+    {
+      "epoch": 0.38,
+      "grad_norm": 287649.84375,
+      "learning_rate": 3.701552605830118e-05,
+      "loss": 0.8357,
+      "step": 41500
+    },
+    {
+      "epoch": 0.39,
+      "grad_norm": 386562.09375,
+      "learning_rate": 3.673860468550963e-05,
+      "loss": 0.8425,
+      "step": 42000
+    },
+    {
+      "epoch": 0.39,
+      "grad_norm": 454850.125,
+      "learning_rate": 3.6461683312718076e-05,
+      "loss": 0.8347,
+      "step": 42500
+    },
+    {
+      "epoch": 0.4,
+      "grad_norm": 239857.5,
+      "learning_rate": 3.618476193992653e-05,
+      "loss": 0.8619,
+      "step": 43000
+    },
+    {
+      "epoch": 0.4,
+      "grad_norm": 227077.171875,
+      "learning_rate": 3.5907840567134974e-05,
+      "loss": 0.847,
+      "step": 43500
+    },
+    {
+      "epoch": 0.41,
+      "grad_norm": 465131.4375,
+      "learning_rate": 3.563091919434342e-05,
+      "loss": 0.8333,
+      "step": 44000
+    },
+    {
+      "epoch": 0.41,
+      "grad_norm": 329121.21875,
+      "learning_rate": 3.535399782155187e-05,
+      "loss": 0.8466,
+      "step": 44500
+    },
+    {
+      "epoch": 0.42,
+      "grad_norm": 362261.40625,
+      "learning_rate": 3.5077076448760316e-05,
+      "loss": 0.8076,
+      "step": 45000
+    },
+    {
+      "epoch": 0.42,
+      "grad_norm": 729279.75,
+      "learning_rate": 3.480015507596876e-05,
+      "loss": 0.8015,
+      "step": 45500
+    },
+    {
+      "epoch": 0.42,
+      "grad_norm": 311664.15625,
+      "learning_rate": 3.4523233703177214e-05,
+      "loss": 0.838,
+      "step": 46000
+    },
+    {
+      "epoch": 0.43,
+      "grad_norm": 267956.34375,
+      "learning_rate": 3.424631233038566e-05,
+      "loss": 0.8229,
+      "step": 46500
+    },
+    {
+      "epoch": 0.43,
+      "grad_norm": 393154.1875,
+      "learning_rate": 3.396939095759411e-05,
+      "loss": 0.7973,
+      "step": 47000
+    },
+    {
+      "epoch": 0.44,
+      "grad_norm": 298007.96875,
+      "learning_rate": 3.3692469584802556e-05,
+      "loss": 0.8309,
+      "step": 47500
+    },
+    {
+      "epoch": 0.44,
+      "grad_norm": 235954.09375,
+      "learning_rate": 3.3415548212011e-05,
+      "loss": 0.808,
+      "step": 48000
+    },
+    {
+      "epoch": 0.45,
+      "grad_norm": 565394.3125,
+      "learning_rate": 3.3138626839219454e-05,
+      "loss": 0.8123,
+      "step": 48500
+    },
+    {
+      "epoch": 0.45,
+      "grad_norm": 228396.78125,
+      "learning_rate": 3.28617054664279e-05,
+      "loss": 0.8456,
+      "step": 49000
+    },
+    {
+      "epoch": 0.46,
+      "grad_norm": 314704.09375,
+      "learning_rate": 3.258478409363635e-05,
+      "loss": 0.8116,
+      "step": 49500
+    },
+    {
+      "epoch": 0.46,
+      "grad_norm": 250648.609375,
+      "learning_rate": 3.2307862720844797e-05,
+      "loss": 0.8284,
+      "step": 50000
+    },
+    {
+      "epoch": 0.47,
+      "grad_norm": 409734.96875,
+      "learning_rate": 3.203094134805324e-05,
+      "loss": 0.8015,
+      "step": 50500
+    },
+    {
+      "epoch": 0.47,
+      "grad_norm": 171156.265625,
+      "learning_rate": 3.1754019975261694e-05,
+      "loss": 0.8322,
+      "step": 51000
+    },
+    {
+      "epoch": 0.48,
+      "grad_norm": 300950.65625,
+      "learning_rate": 3.147709860247014e-05,
+      "loss": 0.8215,
+      "step": 51500
+    },
+    {
+      "epoch": 0.48,
+      "grad_norm": 223661.734375,
+      "learning_rate": 3.120017722967859e-05,
+      "loss": 0.8004,
+      "step": 52000
+    },
+    {
+      "epoch": 0.48,
+      "grad_norm": 303242.25,
+      "learning_rate": 3.0923255856887037e-05,
+      "loss": 0.8109,
+      "step": 52500
+    },
+    {
+      "epoch": 0.49,
+      "grad_norm": 190445.625,
+      "learning_rate": 3.064633448409548e-05,
+      "loss": 0.8126,
+      "step": 53000
+    },
+    {
+      "epoch": 0.49,
+      "grad_norm": 312735.21875,
+      "learning_rate": 3.036941311130393e-05,
+      "loss": 0.8186,
+      "step": 53500
+    },
+    {
+      "epoch": 0.5,
+      "grad_norm": 492028.34375,
+      "learning_rate": 3.009249173851238e-05,
+      "loss": 0.7993,
+      "step": 54000
+    },
+    {
+      "epoch": 0.5,
+      "grad_norm": 159610.984375,
+      "learning_rate": 2.9815570365720828e-05,
+      "loss": 0.8209,
+      "step": 54500
+    },
+    {
+      "epoch": 0.51,
+      "grad_norm": 361054.71875,
+      "learning_rate": 2.9538648992929277e-05,
+      "loss": 0.8034,
+      "step": 55000
+    },
+    {
+      "epoch": 0.51,
+      "grad_norm": 333324.875,
+      "learning_rate": 2.9261727620137725e-05,
+      "loss": 0.7747,
+      "step": 55500
+    },
+    {
+      "epoch": 0.52,
+      "grad_norm": 220997.15625,
+      "learning_rate": 2.898480624734617e-05,
+      "loss": 0.7898,
+      "step": 56000
+    },
+    {
+      "epoch": 0.52,
+      "grad_norm": 427477.28125,
+      "learning_rate": 2.870788487455462e-05,
+      "loss": 0.8487,
+      "step": 56500
+    },
+    {
+      "epoch": 0.53,
+      "grad_norm": 494515.15625,
+      "learning_rate": 2.8430963501763068e-05,
+      "loss": 0.7867,
+      "step": 57000
+    },
+    {
+      "epoch": 0.53,
+      "grad_norm": 131962.8125,
+      "learning_rate": 2.8154042128971517e-05,
+      "loss": 0.8257,
+      "step": 57500
+    },
+    {
+      "epoch": 0.54,
+      "grad_norm": 360923.75,
+      "learning_rate": 2.7877120756179962e-05,
+      "loss": 0.824,
+      "step": 58000
+    },
+    {
+      "epoch": 0.54,
+      "grad_norm": 537102.3125,
+      "learning_rate": 2.760019938338841e-05,
+      "loss": 0.7947,
+      "step": 58500
+    },
+    {
+      "epoch": 0.54,
+      "grad_norm": 185466.890625,
+      "learning_rate": 2.732327801059686e-05,
+      "loss": 0.7965,
+      "step": 59000
+    },
+    {
+      "epoch": 0.55,
+      "grad_norm": 449036.0,
+      "learning_rate": 2.7046356637805308e-05,
+      "loss": 0.8359,
+      "step": 59500
+    },
+    {
+      "epoch": 0.55,
+      "grad_norm": 302378.875,
+      "learning_rate": 2.6769435265013757e-05,
+      "loss": 0.791,
+      "step": 60000
+    },
+    {
+      "epoch": 0.56,
+      "grad_norm": 170428.640625,
+      "learning_rate": 2.6492513892222202e-05,
+      "loss": 0.804,
+      "step": 60500
+    },
+    {
+      "epoch": 0.56,
+      "grad_norm": 486227.0,
+      "learning_rate": 2.621559251943065e-05,
+      "loss": 0.7807,
+      "step": 61000
+    },
+    {
+      "epoch": 0.57,
+      "grad_norm": 657373.0,
+      "learning_rate": 2.59386711466391e-05,
+      "loss": 0.8004,
+      "step": 61500
+    },
+    {
+      "epoch": 0.57,
+      "grad_norm": 386137.0625,
+      "learning_rate": 2.5661749773847548e-05,
+      "loss": 0.7669,
+      "step": 62000
+    },
+    {
+      "epoch": 0.58,
+      "grad_norm": 451425.8125,
+      "learning_rate": 2.5384828401055994e-05,
+      "loss": 0.7948,
+      "step": 62500
+    },
+    {
+      "epoch": 0.58,
+      "grad_norm": 188410.265625,
+      "learning_rate": 2.5107907028264442e-05,
+      "loss": 0.7988,
+      "step": 63000
+    },
+    {
+      "epoch": 0.59,
+      "grad_norm": 543696.4375,
+      "learning_rate": 2.483098565547289e-05,
+      "loss": 0.7927,
+      "step": 63500
+    },
+    {
+      "epoch": 0.59,
+      "grad_norm": 317845.90625,
+      "learning_rate": 2.455406428268134e-05,
+      "loss": 0.8021,
+      "step": 64000
+    },
+    {
+      "epoch": 0.6,
+      "grad_norm": 567647.5,
+      "learning_rate": 2.427714290988979e-05,
+      "loss": 0.7853,
+      "step": 64500
+    },
+    {
+      "epoch": 0.6,
+      "grad_norm": 413894.4375,
+      "learning_rate": 2.4000221537098234e-05,
+      "loss": 0.7973,
+      "step": 65000
+    },
+    {
+      "epoch": 0.6,
+      "grad_norm": 519043.9375,
+      "learning_rate": 2.3723300164306682e-05,
+      "loss": 0.7516,
+      "step": 65500
+    },
+    {
+      "epoch": 0.61,
+      "grad_norm": 129269.5234375,
+      "learning_rate": 2.344637879151513e-05,
+      "loss": 0.8054,
+      "step": 66000
+    },
+    {
+      "epoch": 0.61,
+      "grad_norm": 272380.0625,
+      "learning_rate": 2.316945741872358e-05,
+      "loss": 0.8112,
+      "step": 66500
+    },
+    {
+      "epoch": 0.62,
+      "grad_norm": 299374.875,
+      "learning_rate": 2.2892536045932025e-05,
+      "loss": 0.7687,
+      "step": 67000
+    },
+    {
+      "epoch": 0.62,
+      "grad_norm": 597694.5,
+      "learning_rate": 2.2615614673140474e-05,
+      "loss": 0.8072,
+      "step": 67500
+    },
+    {
+      "epoch": 0.63,
+      "grad_norm": 273690.375,
+      "learning_rate": 2.2338693300348922e-05,
+      "loss": 0.7895,
+      "step": 68000
+    },
+    {
+      "epoch": 0.63,
+      "grad_norm": 237268.21875,
+      "learning_rate": 2.206177192755737e-05,
+      "loss": 0.7802,
+      "step": 68500
+    },
+    {
+      "epoch": 0.64,
+      "grad_norm": 347646.09375,
+      "learning_rate": 2.178485055476582e-05,
+      "loss": 0.7991,
+      "step": 69000
+    },
+    {
+      "epoch": 0.64,
+      "grad_norm": 266270.28125,
+      "learning_rate": 2.1507929181974265e-05,
+      "loss": 0.7979,
+      "step": 69500
+    },
+    {
+      "epoch": 0.65,
+      "grad_norm": 371650.84375,
+      "learning_rate": 2.1231007809182714e-05,
+      "loss": 0.7616,
+      "step": 70000
+    },
+    {
+      "epoch": 0.65,
+      "grad_norm": 389250.75,
+      "learning_rate": 2.0954086436391163e-05,
+      "loss": 0.8056,
+      "step": 70500
+    },
+    {
+      "epoch": 0.66,
+      "grad_norm": 472425.71875,
+      "learning_rate": 2.067716506359961e-05,
+      "loss": 0.7632,
+      "step": 71000
+    },
+    {
+      "epoch": 0.66,
+      "grad_norm": 341248.3125,
+      "learning_rate": 2.0400243690808057e-05,
+      "loss": 0.7648,
+      "step": 71500
+    },
+    {
+      "epoch": 0.66,
+      "grad_norm": 470803.96875,
+      "learning_rate": 2.0123322318016505e-05,
+      "loss": 0.7699,
+      "step": 72000
+    },
+    {
+      "epoch": 0.67,
+      "grad_norm": 363643.71875,
+      "learning_rate": 1.9846400945224954e-05,
+      "loss": 0.7929,
+      "step": 72500
+    },
+    {
+      "epoch": 0.67,
+      "grad_norm": 380603.65625,
+      "learning_rate": 1.9569479572433403e-05,
+      "loss": 0.7614,
+      "step": 73000
+    },
+    {
+      "epoch": 0.68,
+      "grad_norm": 282191.78125,
+      "learning_rate": 1.929255819964185e-05,
+      "loss": 0.7508,
+      "step": 73500
+    },
+    {
+      "epoch": 0.68,
+      "grad_norm": 270537.5625,
+      "learning_rate": 1.9015636826850297e-05,
+      "loss": 0.7724,
+      "step": 74000
+    },
+    {
+      "epoch": 0.69,
+      "grad_norm": 786935.9375,
+      "learning_rate": 1.8738715454058745e-05,
+      "loss": 0.7903,
+      "step": 74500
+    },
+    {
+      "epoch": 0.69,
+      "grad_norm": 276354.875,
+      "learning_rate": 1.8461794081267194e-05,
+      "loss": 0.8034,
+      "step": 75000
+    },
+    {
+      "epoch": 0.7,
+      "grad_norm": 488051.25,
+      "learning_rate": 1.8184872708475643e-05,
+      "loss": 0.7988,
+      "step": 75500
+    },
+    {
+      "epoch": 0.7,
+      "grad_norm": 268592.53125,
+      "learning_rate": 1.7907951335684088e-05,
+      "loss": 0.771,
+      "step": 76000
+    },
+    {
+      "epoch": 0.71,
+      "grad_norm": 286408.90625,
+      "learning_rate": 1.7631029962892537e-05,
+      "loss": 0.7732,
+      "step": 76500
+    },
+    {
+      "epoch": 0.71,
+      "grad_norm": 287447.375,
+      "learning_rate": 1.7354108590100985e-05,
+      "loss": 0.7788,
+      "step": 77000
+    },
+    {
+      "epoch": 0.72,
+      "grad_norm": 427488.15625,
+      "learning_rate": 1.7077187217309434e-05,
+      "loss": 0.7763,
+      "step": 77500
+    },
+    {
+      "epoch": 0.72,
+      "grad_norm": 327204.96875,
+      "learning_rate": 1.6800265844517883e-05,
+      "loss": 0.7814,
+      "step": 78000
+    },
+    {
+      "epoch": 0.72,
+      "grad_norm": 446818.0,
+      "learning_rate": 1.6523344471726328e-05,
+      "loss": 0.773,
+      "step": 78500
+    },
+    {
+      "epoch": 0.73,
+      "grad_norm": 367178.46875,
+      "learning_rate": 1.6246423098934777e-05,
+      "loss": 0.7784,
+      "step": 79000
+    },
+    {
+      "epoch": 0.73,
+      "grad_norm": 564747.75,
+      "learning_rate": 1.5969501726143226e-05,
+      "loss": 0.7751,
+      "step": 79500
+    },
+    {
+      "epoch": 0.74,
+      "grad_norm": 373075.34375,
+      "learning_rate": 1.5692580353351674e-05,
+      "loss": 0.7701,
+      "step": 80000
+    },
+    {
+      "epoch": 0.74,
+      "grad_norm": 220253.890625,
+      "learning_rate": 1.541565898056012e-05,
+      "loss": 0.7717,
+      "step": 80500
+    },
+    {
+      "epoch": 0.75,
+      "grad_norm": 442242.46875,
+      "learning_rate": 1.513873760776857e-05,
+      "loss": 0.7699,
+      "step": 81000
+    },
+    {
+      "epoch": 0.75,
+      "grad_norm": 281603.1875,
+      "learning_rate": 1.4861816234977015e-05,
+      "loss": 0.7626,
+      "step": 81500
+    },
+    {
+      "epoch": 0.76,
+      "grad_norm": 420246.21875,
+      "learning_rate": 1.4584894862185464e-05,
+      "loss": 0.7719,
+      "step": 82000
+    },
+    {
+      "epoch": 0.76,
+      "grad_norm": 347264.375,
+      "learning_rate": 1.4307973489393911e-05,
+      "loss": 0.764,
+      "step": 82500
+    },
+    {
+      "epoch": 0.77,
+      "grad_norm": 326494.21875,
+      "learning_rate": 1.403105211660236e-05,
+      "loss": 0.7617,
+      "step": 83000
+    },
+    {
+      "epoch": 0.77,
+      "grad_norm": 193308.671875,
+      "learning_rate": 1.3754130743810807e-05,
+      "loss": 0.7599,
+      "step": 83500
+    },
+    {
+      "epoch": 0.78,
+      "grad_norm": 283088.8125,
+      "learning_rate": 1.3477209371019255e-05,
+      "loss": 0.7628,
+      "step": 84000
+    },
+    {
+      "epoch": 0.78,
+      "grad_norm": 539284.0,
+      "learning_rate": 1.3200287998227702e-05,
+      "loss": 0.7358,
+      "step": 84500
+    },
+    {
+      "epoch": 0.78,
+      "grad_norm": 473964.84375,
+      "learning_rate": 1.2923366625436151e-05,
+      "loss": 0.755,
+      "step": 85000
+    },
+    {
+      "epoch": 0.79,
+      "grad_norm": 327867.3125,
+      "learning_rate": 1.26464452526446e-05,
+      "loss": 0.7718,
+      "step": 85500
+    },
+    {
+      "epoch": 0.79,
+      "grad_norm": 263024.28125,
+      "learning_rate": 1.2369523879853047e-05,
+      "loss": 0.7722,
+      "step": 86000
+    },
+    {
+      "epoch": 0.8,
+      "grad_norm": 266528.5625,
+      "learning_rate": 1.2092602507061495e-05,
+      "loss": 0.7635,
+      "step": 86500
+    },
+    {
+      "epoch": 0.8,
+      "grad_norm": 126082.46875,
+      "learning_rate": 1.1815681134269942e-05,
+      "loss": 0.7508,
+      "step": 87000
+    },
+    {
+      "epoch": 0.81,
+      "grad_norm": 300010.3125,
+      "learning_rate": 1.1538759761478391e-05,
+      "loss": 0.7755,
+      "step": 87500
+    },
+    {
+      "epoch": 0.81,
+      "grad_norm": 471967.5625,
+      "learning_rate": 1.1261838388686838e-05,
+      "loss": 0.7416,
+      "step": 88000
+    },
+    {
+      "epoch": 0.82,
+      "grad_norm": 261213.0625,
+      "learning_rate": 1.0984917015895287e-05,
+      "loss": 0.7705,
+      "step": 88500
+    },
+    {
+      "epoch": 0.82,
+      "grad_norm": 192903.46875,
+      "learning_rate": 1.0707995643103734e-05,
+      "loss": 0.7392,
+      "step": 89000
+    },
+    {
+      "epoch": 0.83,
+      "grad_norm": 682632.8125,
+      "learning_rate": 1.0431074270312182e-05,
+      "loss": 0.7577,
+      "step": 89500
+    },
+    {
+      "epoch": 0.83,
+      "grad_norm": 218102.4375,
+      "learning_rate": 1.0154152897520631e-05,
+      "loss": 0.7586,
+      "step": 90000
+    },
+    {
+      "epoch": 0.84,
+      "grad_norm": 212267.5625,
+      "learning_rate": 9.877231524729078e-06,
+      "loss": 0.7472,
+      "step": 90500
+    },
+    {
+      "epoch": 0.84,
+      "grad_norm": 320534.59375,
+      "learning_rate": 9.600310151937527e-06,
+      "loss": 0.7602,
+      "step": 91000
+    },
+    {
+      "epoch": 0.84,
+      "grad_norm": 295994.96875,
+      "learning_rate": 9.323388779145974e-06,
+      "loss": 0.7368,
+      "step": 91500
+    },
+    {
+      "epoch": 0.85,
+      "grad_norm": 221153.6875,
+      "learning_rate": 9.046467406354423e-06,
+      "loss": 0.7492,
+      "step": 92000
+    },
+    {
+      "epoch": 0.85,
+      "grad_norm": 250322.6875,
+      "learning_rate": 8.76954603356287e-06,
+      "loss": 0.7582,
+      "step": 92500
+    },
+    {
+      "epoch": 0.86,
+      "grad_norm": 364772.46875,
+      "learning_rate": 8.492624660771318e-06,
+      "loss": 0.747,
+      "step": 93000
+    },
+    {
+      "epoch": 0.86,
+      "grad_norm": 226077.453125,
+      "learning_rate": 8.215703287979767e-06,
+      "loss": 0.7515,
+      "step": 93500
+    },
+    {
+      "epoch": 0.87,
+      "grad_norm": 260477.265625,
+      "learning_rate": 7.938781915188214e-06,
+      "loss": 0.7402,
+      "step": 94000
+    },
+    {
+      "epoch": 0.87,
+      "grad_norm": 380977.46875,
+      "learning_rate": 7.661860542396663e-06,
+      "loss": 0.7512,
+      "step": 94500
+    },
+    {
+      "epoch": 0.88,
+      "grad_norm": 1052145.875,
+      "learning_rate": 7.38493916960511e-06,
+      "loss": 0.7589,
+      "step": 95000
+    },
+    {
+      "epoch": 0.88,
+      "grad_norm": 213601.171875,
+      "learning_rate": 7.1080177968135575e-06,
+      "loss": 0.7581,
+      "step": 95500
+    },
+    {
+      "epoch": 0.89,
+      "grad_norm": 467877.71875,
+      "learning_rate": 6.831096424022006e-06,
+      "loss": 0.7464,
+      "step": 96000
+    },
+    {
+      "epoch": 0.89,
+      "grad_norm": 566204.3125,
+      "learning_rate": 6.554175051230454e-06,
+      "loss": 0.7557,
+      "step": 96500
+    },
+    {
+      "epoch": 0.9,
+      "grad_norm": 330177.5625,
+      "learning_rate": 6.277253678438902e-06,
+      "loss": 0.7304,
+      "step": 97000
+    },
+    {
+      "epoch": 0.9,
+      "grad_norm": 319206.65625,
+      "learning_rate": 6.00033230564735e-06,
+      "loss": 0.736,
+      "step": 97500
+    },
+    {
+      "epoch": 0.9,
+      "grad_norm": 410904.90625,
+      "learning_rate": 5.723410932855798e-06,
+      "loss": 0.752,
+      "step": 98000
+    },
+    {
+      "epoch": 0.91,
+      "grad_norm": 375357.15625,
+      "learning_rate": 5.4464895600642454e-06,
+      "loss": 0.7358,
+      "step": 98500
+    },
+    {
+      "epoch": 0.91,
+      "grad_norm": 307483.375,
+      "learning_rate": 5.169568187272693e-06,
+      "loss": 0.7749,
+      "step": 99000
+    },
+    {
+      "epoch": 0.92,
+      "grad_norm": 434196.59375,
+      "learning_rate": 4.892646814481141e-06,
+      "loss": 0.7324,
+      "step": 99500
+    },
+    {
+      "epoch": 0.92,
+      "grad_norm": 554119.4375,
+      "learning_rate": 4.61572544168959e-06,
+      "loss": 0.7096,
+      "step": 100000
+    },
+    {
+      "epoch": 0.93,
+      "grad_norm": 224006.78125,
+      "learning_rate": 4.338804068898038e-06,
+      "loss": 0.7205,
+      "step": 100500
+    },
+    {
+      "epoch": 0.93,
+      "grad_norm": 215164.78125,
+      "learning_rate": 4.0618826961064855e-06,
+      "loss": 0.732,
+      "step": 101000
+    },
+    {
+      "epoch": 0.94,
+      "grad_norm": 391820.53125,
+      "learning_rate": 3.7849613233149334e-06,
+      "loss": 0.7228,
+      "step": 101500
+    },
+    {
+      "epoch": 0.94,
+      "grad_norm": 306694.15625,
+      "learning_rate": 3.5080399505233816e-06,
+      "loss": 0.7507,
+      "step": 102000
+    },
+    {
+      "epoch": 0.95,
+      "grad_norm": 467995.03125,
+      "learning_rate": 3.2311185777318295e-06,
+      "loss": 0.7433,
+      "step": 102500
+    },
+    {
+      "epoch": 0.95,
+      "grad_norm": 284427.625,
+      "learning_rate": 2.9541972049402773e-06,
+      "loss": 0.7047,
+      "step": 103000
+    },
+    {
+      "epoch": 0.96,
+      "grad_norm": 193486.921875,
+      "learning_rate": 2.6772758321487256e-06,
+      "loss": 0.7646,
+      "step": 103500
+    },
+    {
+      "epoch": 0.96,
+      "grad_norm": 262856.15625,
+      "learning_rate": 2.4003544593571735e-06,
+      "loss": 0.7306,
+      "step": 104000
+    },
+    {
+      "epoch": 0.96,
+      "grad_norm": 267263.5,
+      "learning_rate": 2.1234330865656213e-06,
+      "loss": 0.7631,
+      "step": 104500
+    },
+    {
+      "epoch": 0.97,
+      "grad_norm": 229738.515625,
+      "learning_rate": 1.8465117137740692e-06,
+      "loss": 0.731,
+      "step": 105000
+    },
+    {
+      "epoch": 0.97,
+      "grad_norm": 528997.6875,
+      "learning_rate": 1.569590340982517e-06,
+      "loss": 0.7393,
+      "step": 105500
+    },
+    {
+      "epoch": 0.98,
+      "grad_norm": 346399.28125,
+      "learning_rate": 1.292668968190965e-06,
+      "loss": 0.7304,
+      "step": 106000
+    },
+    {
+      "epoch": 0.98,
+      "grad_norm": 315753.28125,
+      "learning_rate": 1.015747595399413e-06,
+      "loss": 0.7239,
+      "step": 106500
+    },
+    {
+      "epoch": 0.99,
+      "grad_norm": 245733.109375,
+      "learning_rate": 7.388262226078609e-07,
+      "loss": 0.7067,
+      "step": 107000
+    },
+    {
+      "epoch": 0.99,
+      "grad_norm": 423298.15625,
+      "learning_rate": 4.619048498163089e-07,
+      "loss": 0.734,
+      "step": 107500
+    },
+    {
+      "epoch": 1.0,
+      "grad_norm": 361011.8125,
+      "learning_rate": 1.8498347702475678e-07,
+      "loss": 0.7356,
+      "step": 108000
+    },
+    {
+      "epoch": 1.0,
+      "eval_accuracy": 0.68,
+      "eval_loss": 0.750106155872345,
+      "eval_runtime": 570.5256,
+      "eval_samples_per_second": 17.528,
+      "eval_steps_per_second": 2.922,
+      "step": 108334
+    }
+  ],
+  "logging_steps": 500,
+  "max_steps": 108334,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 1,
+  "save_steps": 500,
+  "total_flos": 2.328341808688051e+18,
+  "train_batch_size": 6,
+  "trial_name": null,
+  "trial_params": null
+}

training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:393b55a1a65866dbcb76d7f98fecbc3fc9f9dcd3895a7e49a819a0d1d762db05
+size 4603