Training in progress, step 2400, checkpoint

checkpoint-2400/config.json

@@ -0,0 +1,32 @@
+{
+  "_name_or_path": "EleutherAI/pythia-2.8b",
+  "architectures": [
+    "GPTNeoForMultipleChoice"
+  ],
+  "attention_bias": true,
+  "attention_dropout": 0.0,
+  "bos_token_id": 0,
+  "classifier_dropout": 0.1,
+  "eos_token_id": 0,
+  "hidden_act": "gelu",
+  "hidden_dropout": 0.0,
+  "hidden_size": 2560,
+  "initializer_range": 0.02,
+  "intermediate_size": 10240,
+  "layer_norm_eps": 1e-05,
+  "max_length": 1024,
+  "max_position_embeddings": 2048,
+  "model_type": "gpt_neox",
+  "num_attention_heads": 32,
+  "num_hidden_layers": 32,
+  "pad_token_id": 0,
+  "rope_scaling": null,
+  "rotary_emb_base": 10000,
+  "rotary_pct": 0.25,
+  "tie_word_embeddings": false,
+  "torch_dtype": "float16",
+  "transformers_version": "4.41.1",
+  "use_cache": true,
+  "use_parallel_residual": true,
+  "vocab_size": 50304
+}

checkpoint-2400/global_step2400/mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f81893638ea1c1e67d094caa0ba64c308c68023cba819caa2873cb17704bd947
+size 5292979768

checkpoint-2400/global_step2400/zero_pp_rank_0_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b68f329a2e423a8295a4a8669728e5b2901b9a4ab1082ad28d8616286880b210
+size 15878620944

checkpoint-2400/latest

@@ -0,0 +1 @@
+global_step2400

checkpoint-2400/model-00001-of-00002.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bd967166b239f4ceb8e4ebef19dc95194614a66ff0560e12cbf462ee1655395a
+size 4978208880

checkpoint-2400/model-00002-of-00002.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8c8a243552a54712d1dae8703ff2cac168f143a541e2d21893010832b26187b6
+size 314703498

checkpoint-2400/model.safetensors.index.json

@@ -0,0 +1,396 @@
+{
+  "metadata": {
+    "total_size": 5292866562
+  },
+  "weight_map": {
+    "classifier.bias": "model-00002-of-00002.safetensors",
+    "classifier.weight": "model-00002-of-00002.safetensors",
+    "gpt_neox.embed_in.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.final_layer_norm.bias": "model-00002-of-00002.safetensors",
+    "gpt_neox.final_layer_norm.weight": "model-00002-of-00002.safetensors",
+    "gpt_neox.layers.0.attention.dense.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.0.attention.dense.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.0.attention.query_key_value.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.0.attention.query_key_value.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.0.input_layernorm.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.0.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.0.mlp.dense_4h_to_h.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.0.mlp.dense_4h_to_h.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.0.mlp.dense_h_to_4h.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.0.mlp.dense_h_to_4h.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.0.post_attention_layernorm.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.0.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.1.attention.dense.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.1.attention.dense.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.1.attention.query_key_value.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.1.attention.query_key_value.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.1.input_layernorm.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.1.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.1.mlp.dense_4h_to_h.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.1.mlp.dense_4h_to_h.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.1.mlp.dense_h_to_4h.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.1.mlp.dense_h_to_4h.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.1.post_attention_layernorm.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.1.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.10.attention.dense.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.10.attention.dense.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.10.attention.query_key_value.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.10.attention.query_key_value.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.10.input_layernorm.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.10.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.10.mlp.dense_4h_to_h.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.10.mlp.dense_4h_to_h.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.10.mlp.dense_h_to_4h.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.10.mlp.dense_h_to_4h.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.10.post_attention_layernorm.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.10.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.11.attention.dense.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.11.attention.dense.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.11.attention.query_key_value.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.11.attention.query_key_value.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.11.input_layernorm.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.11.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.11.mlp.dense_4h_to_h.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.11.mlp.dense_4h_to_h.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.11.mlp.dense_h_to_4h.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.11.mlp.dense_h_to_4h.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.11.post_attention_layernorm.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.11.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.12.attention.dense.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.12.attention.dense.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.12.attention.query_key_value.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.12.attention.query_key_value.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.12.input_layernorm.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.12.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.12.mlp.dense_4h_to_h.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.12.mlp.dense_4h_to_h.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.12.mlp.dense_h_to_4h.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.12.mlp.dense_h_to_4h.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.12.post_attention_layernorm.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.12.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.13.attention.dense.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.13.attention.dense.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.13.attention.query_key_value.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.13.attention.query_key_value.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.13.input_layernorm.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.13.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.13.mlp.dense_4h_to_h.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.13.mlp.dense_4h_to_h.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.13.mlp.dense_h_to_4h.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.13.mlp.dense_h_to_4h.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.13.post_attention_layernorm.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.13.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.14.attention.dense.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.14.attention.dense.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.14.attention.query_key_value.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.14.attention.query_key_value.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.14.input_layernorm.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.14.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.14.mlp.dense_4h_to_h.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.14.mlp.dense_4h_to_h.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.14.mlp.dense_h_to_4h.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.14.mlp.dense_h_to_4h.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.14.post_attention_layernorm.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.14.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.15.attention.dense.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.15.attention.dense.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.15.attention.query_key_value.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.15.attention.query_key_value.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.15.input_layernorm.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.15.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.15.mlp.dense_4h_to_h.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.15.mlp.dense_4h_to_h.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.15.mlp.dense_h_to_4h.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.15.mlp.dense_h_to_4h.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.15.post_attention_layernorm.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.15.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.16.attention.dense.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.16.attention.dense.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.16.attention.query_key_value.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.16.attention.query_key_value.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.16.input_layernorm.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.16.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.16.mlp.dense_4h_to_h.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.16.mlp.dense_4h_to_h.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.16.mlp.dense_h_to_4h.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.16.mlp.dense_h_to_4h.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.16.post_attention_layernorm.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.16.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.17.attention.dense.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.17.attention.dense.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.17.attention.query_key_value.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.17.attention.query_key_value.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.17.input_layernorm.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.17.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.17.mlp.dense_4h_to_h.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.17.mlp.dense_4h_to_h.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.17.mlp.dense_h_to_4h.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.17.mlp.dense_h_to_4h.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.17.post_attention_layernorm.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.17.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.18.attention.dense.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.18.attention.dense.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.18.attention.query_key_value.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.18.attention.query_key_value.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.18.input_layernorm.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.18.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.18.mlp.dense_4h_to_h.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.18.mlp.dense_4h_to_h.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.18.mlp.dense_h_to_4h.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.18.mlp.dense_h_to_4h.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.18.post_attention_layernorm.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.18.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.19.attention.dense.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.19.attention.dense.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.19.attention.query_key_value.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.19.attention.query_key_value.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.19.input_layernorm.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.19.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.19.mlp.dense_4h_to_h.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.19.mlp.dense_4h_to_h.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.19.mlp.dense_h_to_4h.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.19.mlp.dense_h_to_4h.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.19.post_attention_layernorm.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.19.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.2.attention.dense.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.2.attention.dense.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.2.attention.query_key_value.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.2.attention.query_key_value.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.2.input_layernorm.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.2.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.2.mlp.dense_4h_to_h.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.2.mlp.dense_4h_to_h.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.2.mlp.dense_h_to_4h.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.2.mlp.dense_h_to_4h.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.2.post_attention_layernorm.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.2.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.20.attention.dense.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.20.attention.dense.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.20.attention.query_key_value.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.20.attention.query_key_value.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.20.input_layernorm.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.20.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.20.mlp.dense_4h_to_h.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.20.mlp.dense_4h_to_h.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.20.mlp.dense_h_to_4h.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.20.mlp.dense_h_to_4h.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.20.post_attention_layernorm.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.20.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.21.attention.dense.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.21.attention.dense.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.21.attention.query_key_value.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.21.attention.query_key_value.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.21.input_layernorm.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.21.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.21.mlp.dense_4h_to_h.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.21.mlp.dense_4h_to_h.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.21.mlp.dense_h_to_4h.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.21.mlp.dense_h_to_4h.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.21.post_attention_layernorm.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.21.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.22.attention.dense.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.22.attention.dense.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.22.attention.query_key_value.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.22.attention.query_key_value.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.22.input_layernorm.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.22.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.22.mlp.dense_4h_to_h.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.22.mlp.dense_4h_to_h.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.22.mlp.dense_h_to_4h.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.22.mlp.dense_h_to_4h.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.22.post_attention_layernorm.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.22.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.23.attention.dense.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.23.attention.dense.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.23.attention.query_key_value.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.23.attention.query_key_value.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.23.input_layernorm.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.23.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.23.mlp.dense_4h_to_h.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.23.mlp.dense_4h_to_h.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.23.mlp.dense_h_to_4h.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.23.mlp.dense_h_to_4h.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.23.post_attention_layernorm.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.23.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.24.attention.dense.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.24.attention.dense.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.24.attention.query_key_value.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.24.attention.query_key_value.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.24.input_layernorm.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.24.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.24.mlp.dense_4h_to_h.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.24.mlp.dense_4h_to_h.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.24.mlp.dense_h_to_4h.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.24.mlp.dense_h_to_4h.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.24.post_attention_layernorm.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.24.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.25.attention.dense.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.25.attention.dense.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.25.attention.query_key_value.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.25.attention.query_key_value.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.25.input_layernorm.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.25.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.25.mlp.dense_4h_to_h.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.25.mlp.dense_4h_to_h.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.25.mlp.dense_h_to_4h.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.25.mlp.dense_h_to_4h.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.25.post_attention_layernorm.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.25.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.26.attention.dense.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.26.attention.dense.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.26.attention.query_key_value.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.26.attention.query_key_value.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.26.input_layernorm.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.26.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.26.mlp.dense_4h_to_h.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.26.mlp.dense_4h_to_h.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.26.mlp.dense_h_to_4h.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.26.mlp.dense_h_to_4h.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.26.post_attention_layernorm.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.26.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.27.attention.dense.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.27.attention.dense.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.27.attention.query_key_value.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.27.attention.query_key_value.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.27.input_layernorm.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.27.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.27.mlp.dense_4h_to_h.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.27.mlp.dense_4h_to_h.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.27.mlp.dense_h_to_4h.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.27.mlp.dense_h_to_4h.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.27.post_attention_layernorm.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.27.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.28.attention.dense.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.28.attention.dense.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.28.attention.query_key_value.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.28.attention.query_key_value.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.28.input_layernorm.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.28.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.28.mlp.dense_4h_to_h.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.28.mlp.dense_4h_to_h.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.28.mlp.dense_h_to_4h.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.28.mlp.dense_h_to_4h.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.28.post_attention_layernorm.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.28.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.29.attention.dense.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.29.attention.dense.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.29.attention.query_key_value.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.29.attention.query_key_value.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.29.input_layernorm.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.29.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.29.mlp.dense_4h_to_h.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.29.mlp.dense_4h_to_h.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.29.mlp.dense_h_to_4h.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.29.mlp.dense_h_to_4h.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.29.post_attention_layernorm.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.29.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.3.attention.dense.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.3.attention.dense.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.3.attention.query_key_value.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.3.attention.query_key_value.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.3.input_layernorm.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.3.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.3.mlp.dense_4h_to_h.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.3.mlp.dense_4h_to_h.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.3.mlp.dense_h_to_4h.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.3.mlp.dense_h_to_4h.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.3.post_attention_layernorm.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.3.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.30.attention.dense.bias": "model-00002-of-00002.safetensors",
+    "gpt_neox.layers.30.attention.dense.weight": "model-00002-of-00002.safetensors",
+    "gpt_neox.layers.30.attention.query_key_value.bias": "model-00002-of-00002.safetensors",
+    "gpt_neox.layers.30.attention.query_key_value.weight": "model-00002-of-00002.safetensors",
+    "gpt_neox.layers.30.input_layernorm.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.30.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.30.mlp.dense_4h_to_h.bias": "model-00002-of-00002.safetensors",
+    "gpt_neox.layers.30.mlp.dense_4h_to_h.weight": "model-00002-of-00002.safetensors",
+    "gpt_neox.layers.30.mlp.dense_h_to_4h.bias": "model-00002-of-00002.safetensors",
+    "gpt_neox.layers.30.mlp.dense_h_to_4h.weight": "model-00002-of-00002.safetensors",
+    "gpt_neox.layers.30.post_attention_layernorm.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.30.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.31.attention.dense.bias": "model-00002-of-00002.safetensors",
+    "gpt_neox.layers.31.attention.dense.weight": "model-00002-of-00002.safetensors",
+    "gpt_neox.layers.31.attention.query_key_value.bias": "model-00002-of-00002.safetensors",
+    "gpt_neox.layers.31.attention.query_key_value.weight": "model-00002-of-00002.safetensors",
+    "gpt_neox.layers.31.input_layernorm.bias": "model-00002-of-00002.safetensors",
+    "gpt_neox.layers.31.input_layernorm.weight": "model-00002-of-00002.safetensors",
+    "gpt_neox.layers.31.mlp.dense_4h_to_h.bias": "model-00002-of-00002.safetensors",
+    "gpt_neox.layers.31.mlp.dense_4h_to_h.weight": "model-00002-of-00002.safetensors",
+    "gpt_neox.layers.31.mlp.dense_h_to_4h.bias": "model-00002-of-00002.safetensors",
+    "gpt_neox.layers.31.mlp.dense_h_to_4h.weight": "model-00002-of-00002.safetensors",
+    "gpt_neox.layers.31.post_attention_layernorm.bias": "model-00002-of-00002.safetensors",
+    "gpt_neox.layers.31.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
+    "gpt_neox.layers.4.attention.dense.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.4.attention.dense.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.4.attention.query_key_value.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.4.attention.query_key_value.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.4.input_layernorm.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.4.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.4.mlp.dense_4h_to_h.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.4.mlp.dense_4h_to_h.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.4.mlp.dense_h_to_4h.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.4.mlp.dense_h_to_4h.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.4.post_attention_layernorm.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.4.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.5.attention.dense.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.5.attention.dense.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.5.attention.query_key_value.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.5.attention.query_key_value.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.5.input_layernorm.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.5.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.5.mlp.dense_4h_to_h.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.5.mlp.dense_4h_to_h.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.5.mlp.dense_h_to_4h.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.5.mlp.dense_h_to_4h.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.5.post_attention_layernorm.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.5.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.6.attention.dense.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.6.attention.dense.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.6.attention.query_key_value.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.6.attention.query_key_value.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.6.input_layernorm.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.6.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.6.mlp.dense_4h_to_h.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.6.mlp.dense_4h_to_h.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.6.mlp.dense_h_to_4h.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.6.mlp.dense_h_to_4h.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.6.post_attention_layernorm.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.6.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.7.attention.dense.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.7.attention.dense.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.7.attention.query_key_value.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.7.attention.query_key_value.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.7.input_layernorm.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.7.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.7.mlp.dense_4h_to_h.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.7.mlp.dense_4h_to_h.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.7.mlp.dense_h_to_4h.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.7.mlp.dense_h_to_4h.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.7.post_attention_layernorm.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.7.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.8.attention.dense.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.8.attention.dense.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.8.attention.query_key_value.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.8.attention.query_key_value.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.8.input_layernorm.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.8.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.8.mlp.dense_4h_to_h.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.8.mlp.dense_4h_to_h.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.8.mlp.dense_h_to_4h.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.8.mlp.dense_h_to_4h.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.8.post_attention_layernorm.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.8.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.9.attention.dense.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.9.attention.dense.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.9.attention.query_key_value.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.9.attention.query_key_value.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.9.input_layernorm.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.9.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.9.mlp.dense_4h_to_h.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.9.mlp.dense_4h_to_h.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.9.mlp.dense_h_to_4h.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.9.mlp.dense_h_to_4h.weight": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.9.post_attention_layernorm.bias": "model-00001-of-00002.safetensors",
+    "gpt_neox.layers.9.post_attention_layernorm.weight": "model-00001-of-00002.safetensors"
+  }
+}

checkpoint-2400/rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0566e93b83b74c99311859806a46cd4d937efd7086b7a022e5de527f8dfadeb9
+size 14640

checkpoint-2400/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:33abd1e9a7281c9ab9215f203bb1a32f1afed08f58c331e615fef002d43b478f
+size 1064

checkpoint-2400/special_tokens_map.json

@@ -0,0 +1,24 @@
+{
+  "bos_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "eos_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": "<|endoftext|>",
+  "unk_token": {
+    "content": "<|endoftext|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}

checkpoint-2400/tokenizer_config.json

@@ -0,0 +1,214 @@
+{
+  "add_bos_token": false,
+  "add_eos_token": false,
+  "add_prefix_space": false,
+  "added_tokens_decoder": {
+    "0": {
+      "content": "<|endoftext|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "1": {
+      "content": "<|padding|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "50254": {
+      "content": "                        ",
+      "lstrip": false,
+      "normalized": true,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "50255": {
+      "content": "                       ",
+      "lstrip": false,
+      "normalized": true,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "50256": {
+      "content": "                      ",
+      "lstrip": false,
+      "normalized": true,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "50257": {
+      "content": "                     ",
+      "lstrip": false,
+      "normalized": true,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "50258": {
+      "content": "                    ",
+      "lstrip": false,
+      "normalized": true,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "50259": {
+      "content": "                   ",
+      "lstrip": false,
+      "normalized": true,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "50260": {
+      "content": "                  ",
+      "lstrip": false,
+      "normalized": true,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "50261": {
+      "content": "                 ",
+      "lstrip": false,
+      "normalized": true,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "50262": {
+      "content": "                ",
+      "lstrip": false,
+      "normalized": true,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "50263": {
+      "content": "               ",
+      "lstrip": false,
+      "normalized": true,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "50264": {
+      "content": "              ",
+      "lstrip": false,
+      "normalized": true,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "50265": {
+      "content": "             ",
+      "lstrip": false,
+      "normalized": true,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "50266": {
+      "content": "            ",
+      "lstrip": false,
+      "normalized": true,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "50267": {
+      "content": "           ",
+      "lstrip": false,
+      "normalized": true,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "50268": {
+      "content": "          ",
+      "lstrip": false,
+      "normalized": true,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "50269": {
+      "content": "         ",
+      "lstrip": false,
+      "normalized": true,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "50270": {
+      "content": "        ",
+      "lstrip": false,
+      "normalized": true,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "50271": {
+      "content": "       ",
+      "lstrip": false,
+      "normalized": true,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "50272": {
+      "content": "      ",
+      "lstrip": false,
+      "normalized": true,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "50273": {
+      "content": "     ",
+      "lstrip": false,
+      "normalized": true,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "50274": {
+      "content": "    ",
+      "lstrip": false,
+      "normalized": true,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "50275": {
+      "content": "   ",
+      "lstrip": false,
+      "normalized": true,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "50276": {
+      "content": "  ",
+      "lstrip": false,
+      "normalized": true,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    }
+  },
+  "bos_token": "<|endoftext|>",
+  "clean_up_tokenization_spaces": true,
+  "eos_token": "<|endoftext|>",
+  "model_max_length": 1000000000000000019884624838656,
+  "pad_token": "<|endoftext|>",
+  "tokenizer_class": "GPTNeoXTokenizer",
+  "unk_token": "<|endoftext|>"
+}

checkpoint-2400/trainer_state.json

@@ -0,0 +1,1774 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 7.494145199063232,
+  "eval_steps": 400,
+  "global_step": 2400,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.00312256049960968,
+      "grad_norm": 0.0,
+      "learning_rate": 0.0,
+      "loss": 1.8678,
+      "step": 1
+    },
+    {
+      "epoch": 0.0312256049960968,
+      "grad_norm": 0.0,
+      "learning_rate": 0.0,
+      "loss": 1.7236,
+      "step": 10
+    },
+    {
+      "epoch": 0.0624512099921936,
+      "grad_norm": 37.27561569213867,
+      "learning_rate": 8e-08,
+      "loss": 1.805,
+      "step": 20
+    },
+    {
+      "epoch": 0.0936768149882904,
+      "grad_norm": 41.067867279052734,
+      "learning_rate": 4.800000000000001e-07,
+      "loss": 1.6883,
+      "step": 30
+    },
+    {
+      "epoch": 0.1249024199843872,
+      "grad_norm": 43.672298431396484,
+      "learning_rate": 8.400000000000001e-07,
+      "loss": 1.6964,
+      "step": 40
+    },
+    {
+      "epoch": 0.156128024980484,
+      "grad_norm": 43.07832717895508,
+      "learning_rate": 1.2400000000000002e-06,
+      "loss": 1.6138,
+      "step": 50
+    },
+    {
+      "epoch": 0.1873536299765808,
+      "grad_norm": 37.47705841064453,
+      "learning_rate": 1.6400000000000002e-06,
+      "loss": 1.5515,
+      "step": 60
+    },
+    {
+      "epoch": 0.2185792349726776,
+      "grad_norm": 28.83339500427246,
+      "learning_rate": 2.04e-06,
+      "loss": 1.3408,
+      "step": 70
+    },
+    {
+      "epoch": 0.2498048399687744,
+      "grad_norm": 31.222503662109375,
+      "learning_rate": 2.4400000000000004e-06,
+      "loss": 1.2731,
+      "step": 80
+    },
+    {
+      "epoch": 0.2810304449648712,
+      "grad_norm": 23.76290512084961,
+      "learning_rate": 2.84e-06,
+      "loss": 1.2666,
+      "step": 90
+    },
+    {
+      "epoch": 0.312256049960968,
+      "grad_norm": 23.913143157958984,
+      "learning_rate": 3.2400000000000003e-06,
+      "loss": 1.1393,
+      "step": 100
+    },
+    {
+      "epoch": 0.3434816549570648,
+      "grad_norm": 24.92310905456543,
+      "learning_rate": 3.6400000000000003e-06,
+      "loss": 1.1529,
+      "step": 110
+    },
+    {
+      "epoch": 0.3747072599531616,
+      "grad_norm": 20.76234245300293,
+      "learning_rate": 4.04e-06,
+      "loss": 1.0776,
+      "step": 120
+    },
+    {
+      "epoch": 0.4059328649492584,
+      "grad_norm": 30.90992546081543,
+      "learning_rate": 4.440000000000001e-06,
+      "loss": 1.0028,
+      "step": 130
+    },
+    {
+      "epoch": 0.4371584699453552,
+      "grad_norm": 30.7198429107666,
+      "learning_rate": 4.84e-06,
+      "loss": 0.9807,
+      "step": 140
+    },
+    {
+      "epoch": 0.468384074941452,
+      "grad_norm": 22.76320457458496,
+      "learning_rate": 5.240000000000001e-06,
+      "loss": 0.992,
+      "step": 150
+    },
+    {
+      "epoch": 0.4996096799375488,
+      "grad_norm": 24.735822677612305,
+      "learning_rate": 5.64e-06,
+      "loss": 0.8421,
+      "step": 160
+    },
+    {
+      "epoch": 0.5308352849336456,
+      "grad_norm": 27.185937881469727,
+      "learning_rate": 6.040000000000001e-06,
+      "loss": 1.012,
+      "step": 170
+    },
+    {
+      "epoch": 0.5620608899297423,
+      "grad_norm": 16.42388916015625,
+      "learning_rate": 6.440000000000001e-06,
+      "loss": 0.7315,
+      "step": 180
+    },
+    {
+      "epoch": 0.5932864949258392,
+      "grad_norm": 25.17578887939453,
+      "learning_rate": 6.8400000000000014e-06,
+      "loss": 0.6995,
+      "step": 190
+    },
+    {
+      "epoch": 0.624512099921936,
+      "grad_norm": 19.550037384033203,
+      "learning_rate": 7.24e-06,
+      "loss": 0.8551,
+      "step": 200
+    },
+    {
+      "epoch": 0.6557377049180327,
+      "grad_norm": 22.346853256225586,
+      "learning_rate": 7.640000000000001e-06,
+      "loss": 0.726,
+      "step": 210
+    },
+    {
+      "epoch": 0.6869633099141296,
+      "grad_norm": 31.998685836791992,
+      "learning_rate": 8.040000000000001e-06,
+      "loss": 0.8553,
+      "step": 220
+    },
+    {
+      "epoch": 0.7181889149102264,
+      "grad_norm": 23.751340866088867,
+      "learning_rate": 8.44e-06,
+      "loss": 0.7651,
+      "step": 230
+    },
+    {
+      "epoch": 0.7494145199063232,
+      "grad_norm": 33.09165954589844,
+      "learning_rate": 8.8e-06,
+      "loss": 0.8395,
+      "step": 240
+    },
+    {
+      "epoch": 0.78064012490242,
+      "grad_norm": 35.236629486083984,
+      "learning_rate": 9.200000000000002e-06,
+      "loss": 0.7972,
+      "step": 250
+    },
+    {
+      "epoch": 0.8118657298985168,
+      "grad_norm": 32.98189926147461,
+      "learning_rate": 9.600000000000001e-06,
+      "loss": 0.7247,
+      "step": 260
+    },
+    {
+      "epoch": 0.8430913348946136,
+      "grad_norm": 26.376121520996094,
+      "learning_rate": 1e-05,
+      "loss": 0.7572,
+      "step": 270
+    },
+    {
+      "epoch": 0.8743169398907104,
+      "grad_norm": 40.748741149902344,
+      "learning_rate": 1e-05,
+      "loss": 0.6943,
+      "step": 280
+    },
+    {
+      "epoch": 0.9055425448868072,
+      "grad_norm": 55.08994674682617,
+      "learning_rate": 1e-05,
+      "loss": 0.8476,
+      "step": 290
+    },
+    {
+      "epoch": 0.936768149882904,
+      "grad_norm": 40.200077056884766,
+      "learning_rate": 1e-05,
+      "loss": 0.7114,
+      "step": 300
+    },
+    {
+      "epoch": 0.9679937548790007,
+      "grad_norm": 24.698932647705078,
+      "learning_rate": 1e-05,
+      "loss": 0.7889,
+      "step": 310
+    },
+    {
+      "epoch": 0.9992193598750976,
+      "grad_norm": 20.618940353393555,
+      "learning_rate": 1e-05,
+      "loss": 0.7606,
+      "step": 320
+    },
+    {
+      "epoch": 1.0304449648711944,
+      "grad_norm": 24.90777587890625,
+      "learning_rate": 1e-05,
+      "loss": 0.4925,
+      "step": 330
+    },
+    {
+      "epoch": 1.0616705698672912,
+      "grad_norm": 28.75925636291504,
+      "learning_rate": 1e-05,
+      "loss": 0.4349,
+      "step": 340
+    },
+    {
+      "epoch": 1.092896174863388,
+      "grad_norm": 306.2433166503906,
+      "learning_rate": 1e-05,
+      "loss": 0.4855,
+      "step": 350
+    },
+    {
+      "epoch": 1.1241217798594847,
+      "grad_norm": 30.801406860351562,
+      "learning_rate": 1e-05,
+      "loss": 0.4829,
+      "step": 360
+    },
+    {
+      "epoch": 1.1553473848555815,
+      "grad_norm": 20.874588012695312,
+      "learning_rate": 1e-05,
+      "loss": 0.4967,
+      "step": 370
+    },
+    {
+      "epoch": 1.1865729898516784,
+      "grad_norm": 15.966379165649414,
+      "learning_rate": 1e-05,
+      "loss": 0.4283,
+      "step": 380
+    },
+    {
+      "epoch": 1.2177985948477752,
+      "grad_norm": 82.65829467773438,
+      "learning_rate": 1e-05,
+      "loss": 0.4268,
+      "step": 390
+    },
+    {
+      "epoch": 1.249024199843872,
+      "grad_norm": 32.251461029052734,
+      "learning_rate": 1e-05,
+      "loss": 0.5603,
+      "step": 400
+    },
+    {
+      "epoch": 1.249024199843872,
+      "eval_accuracy": 0.7,
+      "eval_loss": 0.378662109375,
+      "eval_runtime": 0.8734,
+      "eval_samples_per_second": 11.449,
+      "eval_steps_per_second": 1.145,
+      "step": 400
+    },
+    {
+      "epoch": 1.2802498048399689,
+      "grad_norm": 16.248600006103516,
+      "learning_rate": 1e-05,
+      "loss": 0.4496,
+      "step": 410
+    },
+    {
+      "epoch": 1.3114754098360657,
+      "grad_norm": 26.644573211669922,
+      "learning_rate": 1e-05,
+      "loss": 0.45,
+      "step": 420
+    },
+    {
+      "epoch": 1.3427010148321623,
+      "grad_norm": 31.046363830566406,
+      "learning_rate": 1e-05,
+      "loss": 0.4094,
+      "step": 430
+    },
+    {
+      "epoch": 1.3739266198282591,
+      "grad_norm": 25.93197250366211,
+      "learning_rate": 1e-05,
+      "loss": 0.3649,
+      "step": 440
+    },
+    {
+      "epoch": 1.405152224824356,
+      "grad_norm": 19.997283935546875,
+      "learning_rate": 1e-05,
+      "loss": 0.5174,
+      "step": 450
+    },
+    {
+      "epoch": 1.4363778298204528,
+      "grad_norm": 20.04343032836914,
+      "learning_rate": 1e-05,
+      "loss": 0.4514,
+      "step": 460
+    },
+    {
+      "epoch": 1.4676034348165496,
+      "grad_norm": 18.52043914794922,
+      "learning_rate": 1e-05,
+      "loss": 0.3747,
+      "step": 470
+    },
+    {
+      "epoch": 1.4988290398126463,
+      "grad_norm": 74.7401123046875,
+      "learning_rate": 1e-05,
+      "loss": 0.4383,
+      "step": 480
+    },
+    {
+      "epoch": 1.530054644808743,
+      "grad_norm": 114.52285766601562,
+      "learning_rate": 1e-05,
+      "loss": 0.461,
+      "step": 490
+    },
+    {
+      "epoch": 1.56128024980484,
+      "grad_norm": 122.9369125366211,
+      "learning_rate": 1e-05,
+      "loss": 0.6252,
+      "step": 500
+    },
+    {
+      "epoch": 1.5925058548009368,
+      "grad_norm": 44.502681732177734,
+      "learning_rate": 1e-05,
+      "loss": 0.7419,
+      "step": 510
+    },
+    {
+      "epoch": 1.6237314597970336,
+      "grad_norm": 48.50262451171875,
+      "learning_rate": 1e-05,
+      "loss": 0.6756,
+      "step": 520
+    },
+    {
+      "epoch": 1.6549570647931304,
+      "grad_norm": 39.29521942138672,
+      "learning_rate": 1e-05,
+      "loss": 0.6941,
+      "step": 530
+    },
+    {
+      "epoch": 1.6861826697892273,
+      "grad_norm": 33.0960807800293,
+      "learning_rate": 1e-05,
+      "loss": 0.6813,
+      "step": 540
+    },
+    {
+      "epoch": 1.717408274785324,
+      "grad_norm": 25.355117797851562,
+      "learning_rate": 1e-05,
+      "loss": 0.7615,
+      "step": 550
+    },
+    {
+      "epoch": 1.748633879781421,
+      "grad_norm": 20.417200088500977,
+      "learning_rate": 1e-05,
+      "loss": 0.6087,
+      "step": 560
+    },
+    {
+      "epoch": 1.7798594847775175,
+      "grad_norm": 33.266746520996094,
+      "learning_rate": 1e-05,
+      "loss": 0.7996,
+      "step": 570
+    },
+    {
+      "epoch": 1.8110850897736144,
+      "grad_norm": 13.53630542755127,
+      "learning_rate": 1e-05,
+      "loss": 0.6292,
+      "step": 580
+    },
+    {
+      "epoch": 1.8423106947697112,
+      "grad_norm": 39.0125732421875,
+      "learning_rate": 1e-05,
+      "loss": 0.591,
+      "step": 590
+    },
+    {
+      "epoch": 1.8735362997658078,
+      "grad_norm": 24.019407272338867,
+      "learning_rate": 1e-05,
+      "loss": 0.6722,
+      "step": 600
+    },
+    {
+      "epoch": 1.9047619047619047,
+      "grad_norm": 27.3595027923584,
+      "learning_rate": 1e-05,
+      "loss": 0.5955,
+      "step": 610
+    },
+    {
+      "epoch": 1.9359875097580015,
+      "grad_norm": 22.498308181762695,
+      "learning_rate": 1e-05,
+      "loss": 0.5076,
+      "step": 620
+    },
+    {
+      "epoch": 1.9672131147540983,
+      "grad_norm": 18.389278411865234,
+      "learning_rate": 1e-05,
+      "loss": 0.6773,
+      "step": 630
+    },
+    {
+      "epoch": 1.9984387197501952,
+      "grad_norm": 17.433815002441406,
+      "learning_rate": 1e-05,
+      "loss": 0.5944,
+      "step": 640
+    },
+    {
+      "epoch": 2.029664324746292,
+      "grad_norm": 11.7727632522583,
+      "learning_rate": 1e-05,
+      "loss": 0.1184,
+      "step": 650
+    },
+    {
+      "epoch": 2.060889929742389,
+      "grad_norm": 44.985408782958984,
+      "learning_rate": 1e-05,
+      "loss": 0.4219,
+      "step": 660
+    },
+    {
+      "epoch": 2.0921155347384857,
+      "grad_norm": 27.04376220703125,
+      "learning_rate": 1e-05,
+      "loss": 0.1695,
+      "step": 670
+    },
+    {
+      "epoch": 2.1233411397345825,
+      "grad_norm": 29.073190689086914,
+      "learning_rate": 1e-05,
+      "loss": 0.2694,
+      "step": 680
+    },
+    {
+      "epoch": 2.1545667447306793,
+      "grad_norm": 30.895280838012695,
+      "learning_rate": 1e-05,
+      "loss": 0.2046,
+      "step": 690
+    },
+    {
+      "epoch": 2.185792349726776,
+      "grad_norm": 10.022652626037598,
+      "learning_rate": 1e-05,
+      "loss": 0.1136,
+      "step": 700
+    },
+    {
+      "epoch": 2.2170179547228726,
+      "grad_norm": 26.809078216552734,
+      "learning_rate": 1e-05,
+      "loss": 0.1925,
+      "step": 710
+    },
+    {
+      "epoch": 2.2482435597189694,
+      "grad_norm": 36.76298141479492,
+      "learning_rate": 1e-05,
+      "loss": 0.2269,
+      "step": 720
+    },
+    {
+      "epoch": 2.279469164715066,
+      "grad_norm": 15.884474754333496,
+      "learning_rate": 1e-05,
+      "loss": 0.2236,
+      "step": 730
+    },
+    {
+      "epoch": 2.310694769711163,
+      "grad_norm": 48.100120544433594,
+      "learning_rate": 1e-05,
+      "loss": 0.2063,
+      "step": 740
+    },
+    {
+      "epoch": 2.34192037470726,
+      "grad_norm": 7.69113302230835,
+      "learning_rate": 1e-05,
+      "loss": 0.1649,
+      "step": 750
+    },
+    {
+      "epoch": 2.3731459797033567,
+      "grad_norm": 37.846527099609375,
+      "learning_rate": 1e-05,
+      "loss": 0.1523,
+      "step": 760
+    },
+    {
+      "epoch": 2.4043715846994536,
+      "grad_norm": 17.19913101196289,
+      "learning_rate": 1e-05,
+      "loss": 0.2338,
+      "step": 770
+    },
+    {
+      "epoch": 2.4355971896955504,
+      "grad_norm": 42.62053298950195,
+      "learning_rate": 1e-05,
+      "loss": 0.4299,
+      "step": 780
+    },
+    {
+      "epoch": 2.4668227946916472,
+      "grad_norm": 14.81313705444336,
+      "learning_rate": 1e-05,
+      "loss": 0.2679,
+      "step": 790
+    },
+    {
+      "epoch": 2.498048399687744,
+      "grad_norm": 16.247289657592773,
+      "learning_rate": 1e-05,
+      "loss": 0.2645,
+      "step": 800
+    },
+    {
+      "epoch": 2.498048399687744,
+      "eval_accuracy": 0.7,
+      "eval_loss": 0.490234375,
+      "eval_runtime": 0.8679,
+      "eval_samples_per_second": 11.522,
+      "eval_steps_per_second": 1.152,
+      "step": 800
+    },
+    {
+      "epoch": 2.529274004683841,
+      "grad_norm": 26.519615173339844,
+      "learning_rate": 1e-05,
+      "loss": 0.2979,
+      "step": 810
+    },
+    {
+      "epoch": 2.5604996096799377,
+      "grad_norm": 35.26914596557617,
+      "learning_rate": 1e-05,
+      "loss": 0.2336,
+      "step": 820
+    },
+    {
+      "epoch": 2.5917252146760346,
+      "grad_norm": 21.243257522583008,
+      "learning_rate": 1e-05,
+      "loss": 0.2344,
+      "step": 830
+    },
+    {
+      "epoch": 2.6229508196721314,
+      "grad_norm": 59.89961624145508,
+      "learning_rate": 1e-05,
+      "loss": 0.2617,
+      "step": 840
+    },
+    {
+      "epoch": 2.654176424668228,
+      "grad_norm": 19.667827606201172,
+      "learning_rate": 1e-05,
+      "loss": 0.197,
+      "step": 850
+    },
+    {
+      "epoch": 2.6854020296643246,
+      "grad_norm": 27.412151336669922,
+      "learning_rate": 1e-05,
+      "loss": 0.1607,
+      "step": 860
+    },
+    {
+      "epoch": 2.7166276346604215,
+      "grad_norm": 10.426700592041016,
+      "learning_rate": 1e-05,
+      "loss": 0.2341,
+      "step": 870
+    },
+    {
+      "epoch": 2.7478532396565183,
+      "grad_norm": 25.850656509399414,
+      "learning_rate": 1e-05,
+      "loss": 0.1947,
+      "step": 880
+    },
+    {
+      "epoch": 2.779078844652615,
+      "grad_norm": 33.998863220214844,
+      "learning_rate": 1e-05,
+      "loss": 0.2047,
+      "step": 890
+    },
+    {
+      "epoch": 2.810304449648712,
+      "grad_norm": 17.702449798583984,
+      "learning_rate": 1e-05,
+      "loss": 0.238,
+      "step": 900
+    },
+    {
+      "epoch": 2.841530054644809,
+      "grad_norm": 3.9858572483062744,
+      "learning_rate": 1e-05,
+      "loss": 0.2327,
+      "step": 910
+    },
+    {
+      "epoch": 2.8727556596409056,
+      "grad_norm": 35.145668029785156,
+      "learning_rate": 1e-05,
+      "loss": 0.1995,
+      "step": 920
+    },
+    {
+      "epoch": 2.9039812646370025,
+      "grad_norm": 46.61024856567383,
+      "learning_rate": 1e-05,
+      "loss": 0.1658,
+      "step": 930
+    },
+    {
+      "epoch": 2.9352068696330993,
+      "grad_norm": 23.774057388305664,
+      "learning_rate": 1e-05,
+      "loss": 0.2819,
+      "step": 940
+    },
+    {
+      "epoch": 2.9664324746291957,
+      "grad_norm": 15.349525451660156,
+      "learning_rate": 1e-05,
+      "loss": 0.1376,
+      "step": 950
+    },
+    {
+      "epoch": 2.9976580796252925,
+      "grad_norm": 13.426594734191895,
+      "learning_rate": 1e-05,
+      "loss": 0.2882,
+      "step": 960
+    },
+    {
+      "epoch": 3.0288836846213893,
+      "grad_norm": 6.281402587890625,
+      "learning_rate": 1e-05,
+      "loss": 0.0894,
+      "step": 970
+    },
+    {
+      "epoch": 3.060109289617486,
+      "grad_norm": 2.655089855194092,
+      "learning_rate": 1e-05,
+      "loss": 0.0745,
+      "step": 980
+    },
+    {
+      "epoch": 3.091334894613583,
+      "grad_norm": 3.948760986328125,
+      "learning_rate": 1e-05,
+      "loss": 0.07,
+      "step": 990
+    },
+    {
+      "epoch": 3.12256049960968,
+      "grad_norm": 20.85759735107422,
+      "learning_rate": 1e-05,
+      "loss": 0.0706,
+      "step": 1000
+    },
+    {
+      "epoch": 3.1537861046057767,
+      "grad_norm": 17.535884857177734,
+      "learning_rate": 1e-05,
+      "loss": 0.2045,
+      "step": 1010
+    },
+    {
+      "epoch": 3.1850117096018735,
+      "grad_norm": 21.014545440673828,
+      "learning_rate": 1e-05,
+      "loss": 0.1267,
+      "step": 1020
+    },
+    {
+      "epoch": 3.2162373145979704,
+      "grad_norm": 6.366164207458496,
+      "learning_rate": 1e-05,
+      "loss": 0.1616,
+      "step": 1030
+    },
+    {
+      "epoch": 3.247462919594067,
+      "grad_norm": 20.15192222595215,
+      "learning_rate": 1e-05,
+      "loss": 0.0979,
+      "step": 1040
+    },
+    {
+      "epoch": 3.278688524590164,
+      "grad_norm": 0.4769607186317444,
+      "learning_rate": 1e-05,
+      "loss": 0.1403,
+      "step": 1050
+    },
+    {
+      "epoch": 3.309914129586261,
+      "grad_norm": 9.628069877624512,
+      "learning_rate": 1e-05,
+      "loss": 0.0501,
+      "step": 1060
+    },
+    {
+      "epoch": 3.3411397345823577,
+      "grad_norm": 52.10974884033203,
+      "learning_rate": 1e-05,
+      "loss": 0.1405,
+      "step": 1070
+    },
+    {
+      "epoch": 3.3723653395784545,
+      "grad_norm": 23.110986709594727,
+      "learning_rate": 1e-05,
+      "loss": 0.1177,
+      "step": 1080
+    },
+    {
+      "epoch": 3.4035909445745514,
+      "grad_norm": 19.135101318359375,
+      "learning_rate": 1e-05,
+      "loss": 0.1152,
+      "step": 1090
+    },
+    {
+      "epoch": 3.4348165495706477,
+      "grad_norm": 10.451769828796387,
+      "learning_rate": 1e-05,
+      "loss": 0.1005,
+      "step": 1100
+    },
+    {
+      "epoch": 3.4660421545667446,
+      "grad_norm": 15.533573150634766,
+      "learning_rate": 1e-05,
+      "loss": 0.0585,
+      "step": 1110
+    },
+    {
+      "epoch": 3.4972677595628414,
+      "grad_norm": 8.309584617614746,
+      "learning_rate": 1e-05,
+      "loss": 0.1311,
+      "step": 1120
+    },
+    {
+      "epoch": 3.5284933645589383,
+      "grad_norm": 3.8508894443511963,
+      "learning_rate": 1e-05,
+      "loss": 0.0971,
+      "step": 1130
+    },
+    {
+      "epoch": 3.559718969555035,
+      "grad_norm": 16.79774284362793,
+      "learning_rate": 1e-05,
+      "loss": 0.1476,
+      "step": 1140
+    },
+    {
+      "epoch": 3.590944574551132,
+      "grad_norm": 1.4701294898986816,
+      "learning_rate": 1e-05,
+      "loss": 0.1692,
+      "step": 1150
+    },
+    {
+      "epoch": 3.6221701795472288,
+      "grad_norm": 13.413945198059082,
+      "learning_rate": 1e-05,
+      "loss": 0.1472,
+      "step": 1160
+    },
+    {
+      "epoch": 3.6533957845433256,
+      "grad_norm": 27.413959503173828,
+      "learning_rate": 1e-05,
+      "loss": 0.1762,
+      "step": 1170
+    },
+    {
+      "epoch": 3.6846213895394224,
+      "grad_norm": 32.048553466796875,
+      "learning_rate": 1e-05,
+      "loss": 0.1562,
+      "step": 1180
+    },
+    {
+      "epoch": 3.7158469945355193,
+      "grad_norm": 31.58294677734375,
+      "learning_rate": 1e-05,
+      "loss": 0.1347,
+      "step": 1190
+    },
+    {
+      "epoch": 3.747072599531616,
+      "grad_norm": 17.824254989624023,
+      "learning_rate": 1e-05,
+      "loss": 0.1158,
+      "step": 1200
+    },
+    {
+      "epoch": 3.747072599531616,
+      "eval_accuracy": 0.6,
+      "eval_loss": 1.248046875,
+      "eval_runtime": 0.8648,
+      "eval_samples_per_second": 11.563,
+      "eval_steps_per_second": 1.156,
+      "step": 1200
+    },
+    {
+      "epoch": 3.7782982045277125,
+      "grad_norm": 46.47492599487305,
+      "learning_rate": 1e-05,
+      "loss": 0.1508,
+      "step": 1210
+    },
+    {
+      "epoch": 3.8095238095238093,
+      "grad_norm": 13.830499649047852,
+      "learning_rate": 1e-05,
+      "loss": 0.0936,
+      "step": 1220
+    },
+    {
+      "epoch": 3.840749414519906,
+      "grad_norm": 19.533958435058594,
+      "learning_rate": 1e-05,
+      "loss": 0.063,
+      "step": 1230
+    },
+    {
+      "epoch": 3.871975019516003,
+      "grad_norm": 43.4871940612793,
+      "learning_rate": 1e-05,
+      "loss": 0.1794,
+      "step": 1240
+    },
+    {
+      "epoch": 3.9032006245121,
+      "grad_norm": 17.626535415649414,
+      "learning_rate": 1e-05,
+      "loss": 0.1324,
+      "step": 1250
+    },
+    {
+      "epoch": 3.9344262295081966,
+      "grad_norm": 18.589401245117188,
+      "learning_rate": 1e-05,
+      "loss": 0.1517,
+      "step": 1260
+    },
+    {
+      "epoch": 3.9656518345042935,
+      "grad_norm": 8.064416885375977,
+      "learning_rate": 1e-05,
+      "loss": 0.108,
+      "step": 1270
+    },
+    {
+      "epoch": 3.9968774395003903,
+      "grad_norm": 3.094780206680298,
+      "learning_rate": 1e-05,
+      "loss": 0.1716,
+      "step": 1280
+    },
+    {
+      "epoch": 4.028103044496487,
+      "grad_norm": 9.602354049682617,
+      "learning_rate": 1e-05,
+      "loss": 0.0586,
+      "step": 1290
+    },
+    {
+      "epoch": 4.059328649492584,
+      "grad_norm": 17.06719207763672,
+      "learning_rate": 1e-05,
+      "loss": 0.0568,
+      "step": 1300
+    },
+    {
+      "epoch": 4.090554254488681,
+      "grad_norm": 23.80466079711914,
+      "learning_rate": 1e-05,
+      "loss": 0.0135,
+      "step": 1310
+    },
+    {
+      "epoch": 4.121779859484778,
+      "grad_norm": 1.7121708393096924,
+      "learning_rate": 1e-05,
+      "loss": 0.0382,
+      "step": 1320
+    },
+    {
+      "epoch": 4.1530054644808745,
+      "grad_norm": 0.5317578315734863,
+      "learning_rate": 1e-05,
+      "loss": 0.086,
+      "step": 1330
+    },
+    {
+      "epoch": 4.184231069476971,
+      "grad_norm": 46.14189147949219,
+      "learning_rate": 1e-05,
+      "loss": 0.2723,
+      "step": 1340
+    },
+    {
+      "epoch": 4.215456674473068,
+      "grad_norm": 14.067253112792969,
+      "learning_rate": 1e-05,
+      "loss": 0.1464,
+      "step": 1350
+    },
+    {
+      "epoch": 4.246682279469165,
+      "grad_norm": 5.362925052642822,
+      "learning_rate": 1e-05,
+      "loss": 0.0283,
+      "step": 1360
+    },
+    {
+      "epoch": 4.277907884465262,
+      "grad_norm": 6.1237874031066895,
+      "learning_rate": 1e-05,
+      "loss": 0.0601,
+      "step": 1370
+    },
+    {
+      "epoch": 4.309133489461359,
+      "grad_norm": 1.5201495885849,
+      "learning_rate": 1e-05,
+      "loss": 0.0139,
+      "step": 1380
+    },
+    {
+      "epoch": 4.3403590944574555,
+      "grad_norm": 12.532272338867188,
+      "learning_rate": 1e-05,
+      "loss": 0.0499,
+      "step": 1390
+    },
+    {
+      "epoch": 4.371584699453552,
+      "grad_norm": 6.465614318847656,
+      "learning_rate": 1e-05,
+      "loss": 0.1156,
+      "step": 1400
+    },
+    {
+      "epoch": 4.402810304449648,
+      "grad_norm": 32.81221389770508,
+      "learning_rate": 1e-05,
+      "loss": 0.0678,
+      "step": 1410
+    },
+    {
+      "epoch": 4.434035909445745,
+      "grad_norm": 0.24042364954948425,
+      "learning_rate": 1e-05,
+      "loss": 0.1699,
+      "step": 1420
+    },
+    {
+      "epoch": 4.465261514441842,
+      "grad_norm": 50.17581558227539,
+      "learning_rate": 1e-05,
+      "loss": 0.09,
+      "step": 1430
+    },
+    {
+      "epoch": 4.496487119437939,
+      "grad_norm": 3.710916519165039,
+      "learning_rate": 1e-05,
+      "loss": 0.1545,
+      "step": 1440
+    },
+    {
+      "epoch": 4.527712724434036,
+      "grad_norm": 7.061243534088135,
+      "learning_rate": 1e-05,
+      "loss": 0.2035,
+      "step": 1450
+    },
+    {
+      "epoch": 4.558938329430132,
+      "grad_norm": 13.808802604675293,
+      "learning_rate": 1e-05,
+      "loss": 0.0959,
+      "step": 1460
+    },
+    {
+      "epoch": 4.590163934426229,
+      "grad_norm": 7.443483352661133,
+      "learning_rate": 1e-05,
+      "loss": 0.0549,
+      "step": 1470
+    },
+    {
+      "epoch": 4.621389539422326,
+      "grad_norm": 1.2829999923706055,
+      "learning_rate": 1e-05,
+      "loss": 0.1526,
+      "step": 1480
+    },
+    {
+      "epoch": 4.652615144418423,
+      "grad_norm": 26.241554260253906,
+      "learning_rate": 1e-05,
+      "loss": 0.0783,
+      "step": 1490
+    },
+    {
+      "epoch": 4.68384074941452,
+      "grad_norm": 43.98433303833008,
+      "learning_rate": 1e-05,
+      "loss": 0.0907,
+      "step": 1500
+    },
+    {
+      "epoch": 4.715066354410617,
+      "grad_norm": 1.828418254852295,
+      "learning_rate": 1e-05,
+      "loss": 0.1057,
+      "step": 1510
+    },
+    {
+      "epoch": 4.7462919594067134,
+      "grad_norm": 19.284440994262695,
+      "learning_rate": 1e-05,
+      "loss": 0.0701,
+      "step": 1520
+    },
+    {
+      "epoch": 4.77751756440281,
+      "grad_norm": 18.53413963317871,
+      "learning_rate": 1e-05,
+      "loss": 0.1294,
+      "step": 1530
+    },
+    {
+      "epoch": 4.808743169398907,
+      "grad_norm": 2.0131237506866455,
+      "learning_rate": 1e-05,
+      "loss": 0.1589,
+      "step": 1540
+    },
+    {
+      "epoch": 4.839968774395004,
+      "grad_norm": 7.335690021514893,
+      "learning_rate": 1e-05,
+      "loss": 0.1426,
+      "step": 1550
+    },
+    {
+      "epoch": 4.871194379391101,
+      "grad_norm": 28.594770431518555,
+      "learning_rate": 1e-05,
+      "loss": 0.113,
+      "step": 1560
+    },
+    {
+      "epoch": 4.902419984387198,
+      "grad_norm": 4.218417644500732,
+      "learning_rate": 1e-05,
+      "loss": 0.1795,
+      "step": 1570
+    },
+    {
+      "epoch": 4.9336455893832944,
+      "grad_norm": 37.12601089477539,
+      "learning_rate": 1e-05,
+      "loss": 0.1044,
+      "step": 1580
+    },
+    {
+      "epoch": 4.964871194379391,
+      "grad_norm": 28.900989532470703,
+      "learning_rate": 1e-05,
+      "loss": 0.1998,
+      "step": 1590
+    },
+    {
+      "epoch": 4.996096799375488,
+      "grad_norm": 15.175968170166016,
+      "learning_rate": 1e-05,
+      "loss": 0.0844,
+      "step": 1600
+    },
+    {
+      "epoch": 4.996096799375488,
+      "eval_accuracy": 0.8,
+      "eval_loss": 1.69921875,
+      "eval_runtime": 0.8704,
+      "eval_samples_per_second": 11.49,
+      "eval_steps_per_second": 1.149,
+      "step": 1600
+    },
+    {
+      "epoch": 5.027322404371585,
+      "grad_norm": 96.47978973388672,
+      "learning_rate": 1e-05,
+      "loss": 0.1599,
+      "step": 1610
+    },
+    {
+      "epoch": 5.058548009367682,
+      "grad_norm": 5.848822116851807,
+      "learning_rate": 1e-05,
+      "loss": 0.0671,
+      "step": 1620
+    },
+    {
+      "epoch": 5.089773614363779,
+      "grad_norm": 8.831692695617676,
+      "learning_rate": 1e-05,
+      "loss": 0.047,
+      "step": 1630
+    },
+    {
+      "epoch": 5.1209992193598755,
+      "grad_norm": 0.2928885221481323,
+      "learning_rate": 1e-05,
+      "loss": 0.0895,
+      "step": 1640
+    },
+    {
+      "epoch": 5.152224824355972,
+      "grad_norm": 4.588135242462158,
+      "learning_rate": 1e-05,
+      "loss": 0.0109,
+      "step": 1650
+    },
+    {
+      "epoch": 5.183450429352069,
+      "grad_norm": 0.0034015802666544914,
+      "learning_rate": 1e-05,
+      "loss": 0.1328,
+      "step": 1660
+    },
+    {
+      "epoch": 5.214676034348165,
+      "grad_norm": 2.1403472423553467,
+      "learning_rate": 1e-05,
+      "loss": 0.1678,
+      "step": 1670
+    },
+    {
+      "epoch": 5.245901639344262,
+      "grad_norm": 38.722293853759766,
+      "learning_rate": 1e-05,
+      "loss": 0.0738,
+      "step": 1680
+    },
+    {
+      "epoch": 5.277127244340359,
+      "grad_norm": 24.931602478027344,
+      "learning_rate": 1e-05,
+      "loss": 0.0893,
+      "step": 1690
+    },
+    {
+      "epoch": 5.308352849336456,
+      "grad_norm": 8.807583808898926,
+      "learning_rate": 1e-05,
+      "loss": 0.0387,
+      "step": 1700
+    },
+    {
+      "epoch": 5.339578454332552,
+      "grad_norm": 56.61589431762695,
+      "learning_rate": 1e-05,
+      "loss": 0.1091,
+      "step": 1710
+    },
+    {
+      "epoch": 5.370804059328649,
+      "grad_norm": 3.9017961025238037,
+      "learning_rate": 1e-05,
+      "loss": 0.1022,
+      "step": 1720
+    },
+    {
+      "epoch": 5.402029664324746,
+      "grad_norm": 13.145605087280273,
+      "learning_rate": 1e-05,
+      "loss": 0.0593,
+      "step": 1730
+    },
+    {
+      "epoch": 5.433255269320843,
+      "grad_norm": 2.734715223312378,
+      "learning_rate": 1e-05,
+      "loss": 0.0412,
+      "step": 1740
+    },
+    {
+      "epoch": 5.46448087431694,
+      "grad_norm": 11.634307861328125,
+      "learning_rate": 1e-05,
+      "loss": 0.0745,
+      "step": 1750
+    },
+    {
+      "epoch": 5.495706479313037,
+      "grad_norm": 32.81011962890625,
+      "learning_rate": 1e-05,
+      "loss": 0.0778,
+      "step": 1760
+    },
+    {
+      "epoch": 5.526932084309133,
+      "grad_norm": 4.1930975914001465,
+      "learning_rate": 1e-05,
+      "loss": 0.1031,
+      "step": 1770
+    },
+    {
+      "epoch": 5.55815768930523,
+      "grad_norm": 1.3936034440994263,
+      "learning_rate": 1e-05,
+      "loss": 0.241,
+      "step": 1780
+    },
+    {
+      "epoch": 5.589383294301327,
+      "grad_norm": 31.164995193481445,
+      "learning_rate": 1e-05,
+      "loss": 0.0586,
+      "step": 1790
+    },
+    {
+      "epoch": 5.620608899297424,
+      "grad_norm": 2.2932653427124023,
+      "learning_rate": 1e-05,
+      "loss": 0.0132,
+      "step": 1800
+    },
+    {
+      "epoch": 5.651834504293521,
+      "grad_norm": 0.4385182857513428,
+      "learning_rate": 1e-05,
+      "loss": 0.1186,
+      "step": 1810
+    },
+    {
+      "epoch": 5.683060109289618,
+      "grad_norm": 25.183168411254883,
+      "learning_rate": 1e-05,
+      "loss": 0.0207,
+      "step": 1820
+    },
+    {
+      "epoch": 5.714285714285714,
+      "grad_norm": 8.401402473449707,
+      "learning_rate": 1e-05,
+      "loss": 0.0865,
+      "step": 1830
+    },
+    {
+      "epoch": 5.745511319281811,
+      "grad_norm": 6.048158168792725,
+      "learning_rate": 1e-05,
+      "loss": 0.1033,
+      "step": 1840
+    },
+    {
+      "epoch": 5.776736924277908,
+      "grad_norm": 10.991080284118652,
+      "learning_rate": 1e-05,
+      "loss": 0.1009,
+      "step": 1850
+    },
+    {
+      "epoch": 5.807962529274005,
+      "grad_norm": 5.008920669555664,
+      "learning_rate": 1e-05,
+      "loss": 0.1051,
+      "step": 1860
+    },
+    {
+      "epoch": 5.839188134270102,
+      "grad_norm": 59.1823616027832,
+      "learning_rate": 1e-05,
+      "loss": 0.1015,
+      "step": 1870
+    },
+    {
+      "epoch": 5.870413739266199,
+      "grad_norm": 31.044307708740234,
+      "learning_rate": 1e-05,
+      "loss": 0.0661,
+      "step": 1880
+    },
+    {
+      "epoch": 5.901639344262295,
+      "grad_norm": 53.48557662963867,
+      "learning_rate": 1e-05,
+      "loss": 0.0645,
+      "step": 1890
+    },
+    {
+      "epoch": 5.932864949258392,
+      "grad_norm": 85.64656066894531,
+      "learning_rate": 1e-05,
+      "loss": 0.0618,
+      "step": 1900
+    },
+    {
+      "epoch": 5.964090554254488,
+      "grad_norm": 55.22670364379883,
+      "learning_rate": 1e-05,
+      "loss": 0.1957,
+      "step": 1910
+    },
+    {
+      "epoch": 5.995316159250585,
+      "grad_norm": 20.682653427124023,
+      "learning_rate": 1e-05,
+      "loss": 0.0895,
+      "step": 1920
+    },
+    {
+      "epoch": 6.026541764246682,
+      "grad_norm": 20.45547103881836,
+      "learning_rate": 1e-05,
+      "loss": 0.0349,
+      "step": 1930
+    },
+    {
+      "epoch": 6.057767369242779,
+      "grad_norm": 0.7434096336364746,
+      "learning_rate": 1e-05,
+      "loss": 0.0839,
+      "step": 1940
+    },
+    {
+      "epoch": 6.0889929742388755,
+      "grad_norm": 3.747971534729004,
+      "learning_rate": 1e-05,
+      "loss": 0.0582,
+      "step": 1950
+    },
+    {
+      "epoch": 6.120218579234972,
+      "grad_norm": 12.133618354797363,
+      "learning_rate": 1e-05,
+      "loss": 0.1125,
+      "step": 1960
+    },
+    {
+      "epoch": 6.151444184231069,
+      "grad_norm": 0.936773955821991,
+      "learning_rate": 1e-05,
+      "loss": 0.0583,
+      "step": 1970
+    },
+    {
+      "epoch": 6.182669789227166,
+      "grad_norm": 50.38084411621094,
+      "learning_rate": 1e-05,
+      "loss": 0.0407,
+      "step": 1980
+    },
+    {
+      "epoch": 6.213895394223263,
+      "grad_norm": 26.78063201904297,
+      "learning_rate": 1e-05,
+      "loss": 0.0461,
+      "step": 1990
+    },
+    {
+      "epoch": 6.24512099921936,
+      "grad_norm": 0.004929454065859318,
+      "learning_rate": 1e-05,
+      "loss": 0.0454,
+      "step": 2000
+    },
+    {
+      "epoch": 6.24512099921936,
+      "eval_accuracy": 0.8,
+      "eval_loss": 0.7734375,
+      "eval_runtime": 0.8692,
+      "eval_samples_per_second": 11.505,
+      "eval_steps_per_second": 1.151,
+      "step": 2000
+    },
+    {
+      "epoch": 6.2763466042154565,
+      "grad_norm": 0.012857209891080856,
+      "learning_rate": 1e-05,
+      "loss": 0.0506,
+      "step": 2010
+    },
+    {
+      "epoch": 6.307572209211553,
+      "grad_norm": 0.12446296960115433,
+      "learning_rate": 1e-05,
+      "loss": 0.05,
+      "step": 2020
+    },
+    {
+      "epoch": 6.33879781420765,
+      "grad_norm": 0.12134945392608643,
+      "learning_rate": 1e-05,
+      "loss": 0.101,
+      "step": 2030
+    },
+    {
+      "epoch": 6.370023419203747,
+      "grad_norm": 10.001453399658203,
+      "learning_rate": 1e-05,
+      "loss": 0.0425,
+      "step": 2040
+    },
+    {
+      "epoch": 6.401249024199844,
+      "grad_norm": 0.5767087340354919,
+      "learning_rate": 1e-05,
+      "loss": 0.0806,
+      "step": 2050
+    },
+    {
+      "epoch": 6.432474629195941,
+      "grad_norm": 0.0031043330673128366,
+      "learning_rate": 1e-05,
+      "loss": 0.0399,
+      "step": 2060
+    },
+    {
+      "epoch": 6.4637002341920375,
+      "grad_norm": 12.768810272216797,
+      "learning_rate": 1e-05,
+      "loss": 0.0352,
+      "step": 2070
+    },
+    {
+      "epoch": 6.494925839188134,
+      "grad_norm": 26.4639835357666,
+      "learning_rate": 1e-05,
+      "loss": 0.0731,
+      "step": 2080
+    },
+    {
+      "epoch": 6.526151444184231,
+      "grad_norm": 19.031003952026367,
+      "learning_rate": 1e-05,
+      "loss": 0.0712,
+      "step": 2090
+    },
+    {
+      "epoch": 6.557377049180328,
+      "grad_norm": 0.18513961136341095,
+      "learning_rate": 1e-05,
+      "loss": 0.076,
+      "step": 2100
+    },
+    {
+      "epoch": 6.588602654176425,
+      "grad_norm": 0.551861047744751,
+      "learning_rate": 1e-05,
+      "loss": 0.0737,
+      "step": 2110
+    },
+    {
+      "epoch": 6.619828259172522,
+      "grad_norm": 2.401113510131836,
+      "learning_rate": 1e-05,
+      "loss": 0.0674,
+      "step": 2120
+    },
+    {
+      "epoch": 6.6510538641686185,
+      "grad_norm": 14.345379829406738,
+      "learning_rate": 1e-05,
+      "loss": 0.0446,
+      "step": 2130
+    },
+    {
+      "epoch": 6.682279469164715,
+      "grad_norm": 3.8702871799468994,
+      "learning_rate": 1e-05,
+      "loss": 0.0041,
+      "step": 2140
+    },
+    {
+      "epoch": 6.713505074160812,
+      "grad_norm": 73.34154510498047,
+      "learning_rate": 1e-05,
+      "loss": 0.1014,
+      "step": 2150
+    },
+    {
+      "epoch": 6.744730679156909,
+      "grad_norm": 12.140496253967285,
+      "learning_rate": 1e-05,
+      "loss": 0.1052,
+      "step": 2160
+    },
+    {
+      "epoch": 6.775956284153006,
+      "grad_norm": 0.11863163858652115,
+      "learning_rate": 1e-05,
+      "loss": 0.0121,
+      "step": 2170
+    },
+    {
+      "epoch": 6.807181889149103,
+      "grad_norm": 15.587754249572754,
+      "learning_rate": 1e-05,
+      "loss": 0.0602,
+      "step": 2180
+    },
+    {
+      "epoch": 6.8384074941451995,
+      "grad_norm": 49.31621170043945,
+      "learning_rate": 1e-05,
+      "loss": 0.0718,
+      "step": 2190
+    },
+    {
+      "epoch": 6.8696330991412955,
+      "grad_norm": 17.83144187927246,
+      "learning_rate": 1e-05,
+      "loss": 0.0884,
+      "step": 2200
+    },
+    {
+      "epoch": 6.900858704137392,
+      "grad_norm": 36.556373596191406,
+      "learning_rate": 1e-05,
+      "loss": 0.0755,
+      "step": 2210
+    },
+    {
+      "epoch": 6.932084309133489,
+      "grad_norm": 1.8881489038467407,
+      "learning_rate": 1e-05,
+      "loss": 0.0701,
+      "step": 2220
+    },
+    {
+      "epoch": 6.963309914129586,
+      "grad_norm": 0.718502402305603,
+      "learning_rate": 1e-05,
+      "loss": 0.0518,
+      "step": 2230
+    },
+    {
+      "epoch": 6.994535519125683,
+      "grad_norm": 126.71410369873047,
+      "learning_rate": 1e-05,
+      "loss": 0.1361,
+      "step": 2240
+    },
+    {
+      "epoch": 7.02576112412178,
+      "grad_norm": 31.042253494262695,
+      "learning_rate": 1e-05,
+      "loss": 0.018,
+      "step": 2250
+    },
+    {
+      "epoch": 7.0569867291178765,
+      "grad_norm": 3.930874824523926,
+      "learning_rate": 1e-05,
+      "loss": 0.0375,
+      "step": 2260
+    },
+    {
+      "epoch": 7.088212334113973,
+      "grad_norm": 5.475597381591797,
+      "learning_rate": 1e-05,
+      "loss": 0.0812,
+      "step": 2270
+    },
+    {
+      "epoch": 7.11943793911007,
+      "grad_norm": 0.0065327598713338375,
+      "learning_rate": 1e-05,
+      "loss": 0.008,
+      "step": 2280
+    },
+    {
+      "epoch": 7.150663544106167,
+      "grad_norm": 4.2306389808654785,
+      "learning_rate": 1e-05,
+      "loss": 0.0085,
+      "step": 2290
+    },
+    {
+      "epoch": 7.181889149102264,
+      "grad_norm": 0.0050476002506911755,
+      "learning_rate": 1e-05,
+      "loss": 0.0342,
+      "step": 2300
+    },
+    {
+      "epoch": 7.213114754098361,
+      "grad_norm": 83.71935272216797,
+      "learning_rate": 1e-05,
+      "loss": 0.1075,
+      "step": 2310
+    },
+    {
+      "epoch": 7.2443403590944575,
+      "grad_norm": 5.99688720703125,
+      "learning_rate": 1e-05,
+      "loss": 0.032,
+      "step": 2320
+    },
+    {
+      "epoch": 7.275565964090554,
+      "grad_norm": 0.8152950406074524,
+      "learning_rate": 1e-05,
+      "loss": 0.0473,
+      "step": 2330
+    },
+    {
+      "epoch": 7.306791569086651,
+      "grad_norm": 22.056900024414062,
+      "learning_rate": 1e-05,
+      "loss": 0.0473,
+      "step": 2340
+    },
+    {
+      "epoch": 7.338017174082748,
+      "grad_norm": 0.7729480862617493,
+      "learning_rate": 1e-05,
+      "loss": 0.0032,
+      "step": 2350
+    },
+    {
+      "epoch": 7.369242779078845,
+      "grad_norm": 0.011774387210607529,
+      "learning_rate": 1e-05,
+      "loss": 0.0281,
+      "step": 2360
+    },
+    {
+      "epoch": 7.400468384074942,
+      "grad_norm": 0.002286819973960519,
+      "learning_rate": 1e-05,
+      "loss": 0.0241,
+      "step": 2370
+    },
+    {
+      "epoch": 7.4316939890710385,
+      "grad_norm": 3.6069295406341553,
+      "learning_rate": 1e-05,
+      "loss": 0.2117,
+      "step": 2380
+    },
+    {
+      "epoch": 7.462919594067135,
+      "grad_norm": 59.043479919433594,
+      "learning_rate": 1e-05,
+      "loss": 0.1253,
+      "step": 2390
+    },
+    {
+      "epoch": 7.494145199063232,
+      "grad_norm": 23.903953552246094,
+      "learning_rate": 1e-05,
+      "loss": 0.0803,
+      "step": 2400
+    },
+    {
+      "epoch": 7.494145199063232,
+      "eval_accuracy": 0.7,
+      "eval_loss": 1.1728515625,
+      "eval_runtime": 0.8741,
+      "eval_samples_per_second": 11.44,
+      "eval_steps_per_second": 1.144,
+      "step": 2400
+    }
+  ],
+  "logging_steps": 10,
+  "max_steps": 2500,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 8,
+  "save_steps": 400,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 2.199744656109994e+18,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}

checkpoint-2400/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d25dce953fe638888ddcd442f993075f5d6a6b6733071c3ba33f827a1cdecae6
+size 6136

checkpoint-2400/zero_to_fp32.py

@@ -0,0 +1,604 @@
+#!/usr/bin/env python
+
+# Copyright (c) Microsoft Corporation.
+# SPDX-License-Identifier: Apache-2.0
+
+# DeepSpeed Team
+
+# This script extracts fp32 consolidated weights from a zero 1, 2 and 3 DeepSpeed checkpoints. It gets
+# copied into the top level checkpoint dir, so the user can easily do the conversion at any point in
+# the future. Once extracted, the weights don't require DeepSpeed and can be used in any
+# application.
+#
+# example: python zero_to_fp32.py . pytorch_model.bin
+
+import argparse
+import torch
+import glob
+import math
+import os
+import re
+from collections import OrderedDict
+from dataclasses import dataclass
+
+# while this script doesn't use deepspeed to recover data, since the checkpoints are pickled with
+# DeepSpeed data structures it has to be available in the current python environment.
+from deepspeed.utils import logger
+from deepspeed.checkpoint.constants import (DS_VERSION, OPTIMIZER_STATE_DICT, SINGLE_PARTITION_OF_FP32_GROUPS,
+                                            FP32_FLAT_GROUPS, ZERO_STAGE, PARTITION_COUNT, PARAM_SHAPES, BUFFER_NAMES,
+                                            FROZEN_PARAM_SHAPES, FROZEN_PARAM_FRAGMENTS)
+
+
+@dataclass
+class zero_model_state:
+    buffers: dict()
+    param_shapes: dict()
+    shared_params: list
+    ds_version: int
+    frozen_param_shapes: dict()
+    frozen_param_fragments: dict()
+
+
+debug = 0
+
+# load to cpu
+device = torch.device('cpu')
+
+
+def atoi(text):
+    return int(text) if text.isdigit() else text
+
+
+def natural_keys(text):
+    '''
+    alist.sort(key=natural_keys) sorts in human order
+    http://nedbatchelder.com/blog/200712/human_sorting.html
+    (See Toothy's implementation in the comments)
+    '''
+    return [atoi(c) for c in re.split(r'(\d+)', text)]
+
+
+def get_model_state_file(checkpoint_dir, zero_stage):
+    if not os.path.isdir(checkpoint_dir):
+        raise FileNotFoundError(f"Directory '{checkpoint_dir}' doesn't exist")
+
+    # there should be only one file
+    if zero_stage <= 2:
+        file = os.path.join(checkpoint_dir, "mp_rank_00_model_states.pt")
+    elif zero_stage == 3:
+        file = os.path.join(checkpoint_dir, "zero_pp_rank_0_mp_rank_00_model_states.pt")
+
+    if not os.path.exists(file):
+        raise FileNotFoundError(f"can't find model states file at '{file}'")
+
+    return file
+
+
+def get_checkpoint_files(checkpoint_dir, glob_pattern):
+    # XXX: need to test that this simple glob rule works for multi-node setup too
+    ckpt_files = sorted(glob.glob(os.path.join(checkpoint_dir, glob_pattern)), key=natural_keys)
+
+    if len(ckpt_files) == 0:
+        raise FileNotFoundError(f"can't find {glob_pattern} files in directory '{checkpoint_dir}'")
+
+    return ckpt_files
+
+
+def get_optim_files(checkpoint_dir):
+    return get_checkpoint_files(checkpoint_dir, "*_optim_states.pt")
+
+
+def get_model_state_files(checkpoint_dir):
+    return get_checkpoint_files(checkpoint_dir, "*_model_states.pt")
+
+
+def parse_model_states(files):
+    zero_model_states = []
+    for file in files:
+        state_dict = torch.load(file, map_location=device)
+
+        if BUFFER_NAMES not in state_dict:
+            raise ValueError(f"{file} is not a model state checkpoint")
+        buffer_names = state_dict[BUFFER_NAMES]
+        if debug:
+            print("Found buffers:", buffer_names)
+
+        # recover just the buffers while restoring them to fp32 if they were saved in fp16
+        buffers = {k: v.float() for k, v in state_dict["module"].items() if k in buffer_names}
+        param_shapes = state_dict[PARAM_SHAPES]
+
+        # collect parameters that are included in param_shapes
+        param_names = []
+        for s in param_shapes:
+            for name in s.keys():
+                param_names.append(name)
+
+        # update with frozen parameters
+        frozen_param_shapes = state_dict.get(FROZEN_PARAM_SHAPES, None)
+        if frozen_param_shapes is not None:
+            if debug:
+                print(f"Found frozen_param_shapes: {frozen_param_shapes}")
+            param_names += list(frozen_param_shapes.keys())
+
+        # handle shared params
+        shared_params = [[k, v] for k, v in state_dict["shared_params"].items()]
+
+        ds_version = state_dict.get(DS_VERSION, None)
+
+        frozen_param_fragments = state_dict.get(FROZEN_PARAM_FRAGMENTS, None)
+
+        z_model_state = zero_model_state(buffers=buffers,
+                                         param_shapes=param_shapes,
+                                         shared_params=shared_params,
+                                         ds_version=ds_version,
+                                         frozen_param_shapes=frozen_param_shapes,
+                                         frozen_param_fragments=frozen_param_fragments)
+        zero_model_states.append(z_model_state)
+
+    return zero_model_states
+
+
+def parse_optim_states(files, ds_checkpoint_dir):
+
+    total_files = len(files)
+    state_dicts = []
+    for f in files:
+        state_dict = torch.load(f, map_location=device)
+        # immediately discard the potentially huge 2 optimizer states as we only care for fp32 master weights
+        # and also handle the case where it was already removed by another helper script
+        state_dict["optimizer_state_dict"].pop("optimizer_state_dict", None)
+        state_dicts.append(state_dict)
+
+    if not ZERO_STAGE in state_dicts[0][OPTIMIZER_STATE_DICT]:
+        raise ValueError(f"{files[0]} is not a zero checkpoint")
+    zero_stage = state_dicts[0][OPTIMIZER_STATE_DICT][ZERO_STAGE]
+    world_size = state_dicts[0][OPTIMIZER_STATE_DICT][PARTITION_COUNT]
+
+    # For ZeRO-2 each param group can have different partition_count as data parallelism for expert
+    # parameters can be different from data parallelism for non-expert parameters. So we can just
+    # use the max of the partition_count to get the dp world_size.
+
+    if type(world_size) is list:
+        world_size = max(world_size)
+
+    if world_size != total_files:
+        raise ValueError(
+            f"Expected {world_size} of '*_optim_states.pt' under '{ds_checkpoint_dir}' but found {total_files} files. "
+            "Possibly due to an overwrite of an old checkpoint, or a checkpoint didn't get saved by one or more processes."
+        )
+
+    # the groups are named differently in each stage
+    if zero_stage <= 2:
+        fp32_groups_key = SINGLE_PARTITION_OF_FP32_GROUPS
+    elif zero_stage == 3:
+        fp32_groups_key = FP32_FLAT_GROUPS
+    else:
+        raise ValueError(f"unknown zero stage {zero_stage}")
+
+    if zero_stage <= 2:
+        fp32_flat_groups = [state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key] for i in range(len(state_dicts))]
+    elif zero_stage == 3:
+        # if there is more than one param group, there will be multiple flattened tensors - one
+        # flattened tensor per group - for simplicity merge them into a single tensor
+        #
+        # XXX: could make the script more memory efficient for when there are multiple groups - it
+        # will require matching the sub-lists of param_shapes for each param group flattened tensor
+
+        fp32_flat_groups = [
+            torch.cat(state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key], 0) for i in range(len(state_dicts))
+        ]
+
+    return zero_stage, world_size, fp32_flat_groups
+
+
+def _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir, exclude_frozen_parameters):
+    """
+    Returns fp32 state_dict reconstructed from ds checkpoint
+
+    Args:
+        - ``ds_checkpoint_dir``: path to the deepspeed checkpoint folder (where the optimizer files are)
+
+    """
+    print(f"Processing zero checkpoint '{ds_checkpoint_dir}'")
+
+    optim_files = get_optim_files(ds_checkpoint_dir)
+    zero_stage, world_size, fp32_flat_groups = parse_optim_states(optim_files, ds_checkpoint_dir)
+    print(f"Detected checkpoint of type zero stage {zero_stage}, world_size: {world_size}")
+
+    model_files = get_model_state_files(ds_checkpoint_dir)
+
+    zero_model_states = parse_model_states(model_files)
+    print(f'Parsing checkpoint created by deepspeed=={zero_model_states[0].ds_version}')
+
+    if zero_stage <= 2:
+        return _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                                          exclude_frozen_parameters)
+    elif zero_stage == 3:
+        return _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                                          exclude_frozen_parameters)
+
+
+def _zero2_merge_frozen_params(state_dict, zero_model_states):
+    if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
+        return
+
+    frozen_param_shapes = zero_model_states[0].frozen_param_shapes
+    frozen_param_fragments = zero_model_states[0].frozen_param_fragments
+
+    if debug:
+        num_elem = sum(s.numel() for s in frozen_param_shapes.values())
+        print(f'rank 0: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
+
+        wanted_params = len(frozen_param_shapes)
+        wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
+        avail_numel = sum([p.numel() for p in frozen_param_fragments.values()])
+        print(f'Frozen params: Have {avail_numel} numels to process.')
+        print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
+
+    total_params = 0
+    total_numel = 0
+    for name, shape in frozen_param_shapes.items():
+        total_params += 1
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+
+        state_dict[name] = frozen_param_fragments[name]
+
+        if debug:
+            print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
+
+    print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _has_callable(obj, fn):
+    attr = getattr(obj, fn, None)
+    return callable(attr)
+
+
+def _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
+    param_shapes = zero_model_states[0].param_shapes
+
+    # Reconstruction protocol:
+    #
+    # XXX: document this
+
+    if debug:
+        for i in range(world_size):
+            for j in range(len(fp32_flat_groups[0])):
+                print(f"{FP32_FLAT_GROUPS}[{i}][{j}].shape={fp32_flat_groups[i][j].shape}")
+
+    # XXX: memory usage doubles here (zero2)
+    num_param_groups = len(fp32_flat_groups[0])
+    merged_single_partition_of_fp32_groups = []
+    for i in range(num_param_groups):
+        merged_partitions = [sd[i] for sd in fp32_flat_groups]
+        full_single_fp32_vector = torch.cat(merged_partitions, 0)
+        merged_single_partition_of_fp32_groups.append(full_single_fp32_vector)
+    avail_numel = sum(
+        [full_single_fp32_vector.numel() for full_single_fp32_vector in merged_single_partition_of_fp32_groups])
+
+    if debug:
+        wanted_params = sum([len(shapes) for shapes in param_shapes])
+        wanted_numel = sum([sum(shape.numel() for shape in shapes.values()) for shapes in param_shapes])
+        # not asserting if there is a mismatch due to possible padding
+        print(f"Have {avail_numel} numels to process.")
+        print(f"Need {wanted_numel} numels in {wanted_params} params.")
+
+    # params
+    # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
+    # out-of-core computing solution
+    total_numel = 0
+    total_params = 0
+    for shapes, full_single_fp32_vector in zip(param_shapes, merged_single_partition_of_fp32_groups):
+        offset = 0
+        avail_numel = full_single_fp32_vector.numel()
+        for name, shape in shapes.items():
+
+            unpartitioned_numel = shape.numel() if _has_callable(shape, 'numel') else math.prod(shape)
+            total_numel += unpartitioned_numel
+            total_params += 1
+
+            if debug:
+                print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
+            state_dict[name] = full_single_fp32_vector.narrow(0, offset, unpartitioned_numel).view(shape)
+            offset += unpartitioned_numel
+
+        # Z2 started to align to 2*world_size to improve nccl performance. Therefore both offset and
+        # avail_numel can differ by anywhere between 0..2*world_size. Due to two unrelated complex
+        # paddings performed in the code it's almost impossible to predict the exact numbers w/o the
+        # live optimizer object, so we are checking that the numbers are within the right range
+        align_to = 2 * world_size
+
+        def zero2_align(x):
+            return align_to * math.ceil(x / align_to)
+
+        if debug:
+            print(f"original offset={offset}, avail_numel={avail_numel}")
+
+        offset = zero2_align(offset)
+        avail_numel = zero2_align(avail_numel)
+
+        if debug:
+            print(f"aligned  offset={offset}, avail_numel={avail_numel}")
+
+        # Sanity check
+        if offset != avail_numel:
+            raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
+
+    print(f"Reconstructed fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                               exclude_frozen_parameters):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    if not exclude_frozen_parameters:
+        _zero2_merge_frozen_params(state_dict, zero_model_states)
+
+    _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
+
+    # recover shared parameters
+    for pair in zero_model_states[0].shared_params:
+        if pair[1] in state_dict:
+            state_dict[pair[0]] = state_dict[pair[1]]
+
+    return state_dict
+
+
+def zero3_partitioned_param_info(unpartitioned_numel, world_size):
+    remainder = unpartitioned_numel % world_size
+    padding_numel = (world_size - remainder) if remainder else 0
+    partitioned_numel = math.ceil(unpartitioned_numel / world_size)
+    return partitioned_numel, padding_numel
+
+
+def _zero3_merge_frozen_params(state_dict, world_size, zero_model_states):
+    if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
+        return
+
+    if debug:
+        for i in range(world_size):
+            num_elem = sum(s.numel() for s in zero_model_states[i].frozen_param_fragments.values())
+            print(f'rank {i}: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
+
+        frozen_param_shapes = zero_model_states[0].frozen_param_shapes
+        wanted_params = len(frozen_param_shapes)
+        wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
+        avail_numel = sum([p.numel() for p in zero_model_states[0].frozen_param_fragments.values()]) * world_size
+        print(f'Frozen params: Have {avail_numel} numels to process.')
+        print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
+
+    total_params = 0
+    total_numel = 0
+    for name, shape in zero_model_states[0].frozen_param_shapes.items():
+        total_params += 1
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+
+        param_frags = tuple(model_state.frozen_param_fragments[name] for model_state in zero_model_states)
+        state_dict[name] = torch.cat(param_frags, 0).narrow(0, 0, unpartitioned_numel).view(shape)
+
+        partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
+
+        if debug:
+            print(
+                f"Frozen params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
+            )
+
+    print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
+    param_shapes = zero_model_states[0].param_shapes
+    avail_numel = fp32_flat_groups[0].numel() * world_size
+    # Reconstruction protocol: For zero3 we need to zip the partitions together at boundary of each
+    # param, re-consolidating each param, while dealing with padding if any
+
+    # merge list of dicts, preserving order
+    param_shapes = {k: v for d in param_shapes for k, v in d.items()}
+
+    if debug:
+        for i in range(world_size):
+            print(f"{FP32_FLAT_GROUPS}[{i}].shape={fp32_flat_groups[i].shape}")
+
+        wanted_params = len(param_shapes)
+        wanted_numel = sum(shape.numel() for shape in param_shapes.values())
+        # not asserting if there is a mismatch due to possible padding
+        avail_numel = fp32_flat_groups[0].numel() * world_size
+        print(f"Trainable params: Have {avail_numel} numels to process.")
+        print(f"Trainable params: Need {wanted_numel} numels in {wanted_params} params.")
+
+    # params
+    # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
+    # out-of-core computing solution
+    offset = 0
+    total_numel = 0
+    total_params = 0
+    for name, shape in param_shapes.items():
+
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+        total_params += 1
+
+        partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
+
+        if debug:
+            print(
+                f"Trainable params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
+            )
+
+        # XXX: memory usage doubles here
+        state_dict[name] = torch.cat(
+            tuple(fp32_flat_groups[i].narrow(0, offset, partitioned_numel) for i in range(world_size)),
+            0).narrow(0, 0, unpartitioned_numel).view(shape)
+        offset += partitioned_numel
+
+    offset *= world_size
+
+    # Sanity check
+    if offset != avail_numel:
+        raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
+
+    print(f"Reconstructed Trainable fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                               exclude_frozen_parameters):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    if not exclude_frozen_parameters:
+        _zero3_merge_frozen_params(state_dict, world_size, zero_model_states)
+
+    _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
+
+    # recover shared parameters
+    for pair in zero_model_states[0].shared_params:
+        if pair[1] in state_dict:
+            state_dict[pair[0]] = state_dict[pair[1]]
+
+    return state_dict
+
+
+def get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag=None, exclude_frozen_parameters=False):
+    """
+    Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated state_dict that can be loaded with
+    ``load_state_dict()`` and used for training without DeepSpeed or shared with others, for example
+    via a model hub.
+
+    Args:
+        - ``checkpoint_dir``: path to the desired checkpoint folder
+        - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in 'latest' file. e.g., ``global_step14``
+        - ``exclude_frozen_parameters``: exclude frozen parameters
+
+    Returns:
+        - pytorch ``state_dict``
+
+    Note: this approach may not work if your application doesn't have sufficient free CPU memory and
+    you may need to use the offline approach using the ``zero_to_fp32.py`` script that is saved with
+    the checkpoint.
+
+    A typical usage might be ::
+
+        from deepspeed.utils.zero_to_fp32 import get_fp32_state_dict_from_zero_checkpoint
+        # do the training and checkpoint saving
+        state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir) # already on cpu
+        model = model.cpu() # move to cpu
+        model.load_state_dict(state_dict)
+        # submit to model hub or save the model to share with others
+
+    In this example the ``model`` will no longer be usable in the deepspeed context of the same
+    application. i.e. you will need to re-initialize the deepspeed engine, since
+    ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
+
+    If you want it all done for you, use ``load_state_dict_from_zero_checkpoint`` instead.
+
+    """
+    if tag is None:
+        latest_path = os.path.join(checkpoint_dir, 'latest')
+        if os.path.isfile(latest_path):
+            with open(latest_path, 'r') as fd:
+                tag = fd.read().strip()
+        else:
+            raise ValueError(f"Unable to find 'latest' file at {latest_path}")
+
+    ds_checkpoint_dir = os.path.join(checkpoint_dir, tag)
+
+    if not os.path.isdir(ds_checkpoint_dir):
+        raise FileNotFoundError(f"Directory '{ds_checkpoint_dir}' doesn't exist")
+
+    return _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir, exclude_frozen_parameters)
+
+
+def convert_zero_checkpoint_to_fp32_state_dict(checkpoint_dir, output_file, tag=None, exclude_frozen_parameters=False):
+    """
+    Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict`` file that can be
+    loaded with ``torch.load(file)`` + ``load_state_dict()`` and used for training without DeepSpeed.
+
+    Args:
+        - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
+        - ``output_file``: path to the pytorch fp32 state_dict output file (e.g. path/pytorch_model.bin)
+        - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
+        - ``exclude_frozen_parameters``: exclude frozen parameters
+    """
+
+    state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag, exclude_frozen_parameters)
+    print(f"Saving fp32 state dict to {output_file}")
+    torch.save(state_dict, output_file)
+
+
+def load_state_dict_from_zero_checkpoint(model, checkpoint_dir, tag=None):
+    """
+    1. Put the provided model to cpu
+    2. Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict``
+    3. Load it into the provided model
+
+    Args:
+        - ``model``: the model object to update
+        - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
+        - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
+
+    Returns:
+        - ``model`: modified model
+
+    Make sure you have plenty of CPU memory available before you call this function. If you don't
+    have enough use the ``zero_to_fp32.py`` utility to do the conversion. You will find it
+    conveniently placed for you in the checkpoint folder.
+
+    A typical usage might be ::
+
+        from deepspeed.utils.zero_to_fp32 import load_state_dict_from_zero_checkpoint
+        model = load_state_dict_from_zero_checkpoint(trainer.model, checkpoint_dir)
+        # submit to model hub or save the model to share with others
+
+    Note, that once this was run, the ``model`` will no longer be usable in the deepspeed context
+    of the same application. i.e. you will need to re-initialize the deepspeed engine, since
+    ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
+
+    """
+    logger.info(f"Extracting fp32 weights")
+    state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag)
+
+    logger.info(f"Overwriting model with fp32 weights")
+    model = model.cpu()
+    model.load_state_dict(state_dict, strict=False)
+
+    return model
+
+
+if __name__ == "__main__":
+
+    parser = argparse.ArgumentParser()
+    parser.add_argument("checkpoint_dir",
+                        type=str,
+                        help="path to the desired checkpoint folder, e.g., path/checkpoint-12")
+    parser.add_argument(
+        "output_file",
+        type=str,
+        help="path to the pytorch fp32 state_dict output file (e.g. path/checkpoint-12/pytorch_model.bin)")
+    parser.add_argument("-t",
+                        "--tag",
+                        type=str,
+                        default=None,
+                        help="checkpoint tag used as a unique identifier for checkpoint. e.g., global_step1")
+    parser.add_argument("--exclude_frozen_parameters", action='store_true', help="exclude frozen parameters")
+    parser.add_argument("-d", "--debug", action='store_true', help="enable debug")
+    args = parser.parse_args()
+
+    debug = args.debug
+
+    convert_zero_checkpoint_to_fp32_state_dict(args.checkpoint_dir,
+                                               args.output_file,
+                                               tag=args.tag,
+                                               exclude_frozen_parameters=args.exclude_frozen_parameters)