Training in progress, step 1600, checkpoint

checkpoint-1600/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": "4.41.0",
+  "use_cache": true,
+  "use_parallel_residual": true,
+  "vocab_size": 50304
+}

checkpoint-1600/global_step1600/mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a3a18f8457daad3296d78cd1d2c8a6bbfbfe9aa207841b1ea3221fce66e0d337
+size 5292979768

checkpoint-1600/global_step1600/zero_pp_rank_0_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0d5c94aea006745923b7afe4cfe79ff1b0aafd7692381c82b2340cdbfd16d694
+size 15878620944

checkpoint-1600/latest

@@ -0,0 +1 @@
+global_step1600

checkpoint-1600/model-00001-of-00002.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3dac333c37fb63e2340b7e8f0700c86aeed42ce4a3b3d85b06143538dec5d480
+size 4978208880

checkpoint-1600/model-00002-of-00002.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dac76d4391cff89dd0041efcdacfdd2cfd48784ec411e43f277badb144699aaf
+size 314703498

checkpoint-1600/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-1600/rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e636c72ee90e96ecec1452948a8323c12d3a323f8fff7911fa06ddd701df731b
+size 15152

checkpoint-1600/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fcb446744e871ca19d409ac31d13593345b713f31d27ca0c058327eff7db7e10
+size 1064

checkpoint-1600/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-1600/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-1600/trainer_state.json

@@ -0,0 +1,1196 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 34.224598930481285,
+  "eval_steps": 400,
+  "global_step": 1600,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0213903743315508,
+      "grad_norm": 0.0,
+      "learning_rate": 0.0,
+      "loss": 1.8558,
+      "step": 1
+    },
+    {
+      "epoch": 0.21390374331550802,
+      "grad_norm": 0.0,
+      "learning_rate": 0.0,
+      "loss": 2.0957,
+      "step": 10
+    },
+    {
+      "epoch": 0.42780748663101603,
+      "grad_norm": 63.96761703491211,
+      "learning_rate": 8e-08,
+      "loss": 2.1274,
+      "step": 20
+    },
+    {
+      "epoch": 0.6417112299465241,
+      "grad_norm": 56.0494270324707,
+      "learning_rate": 4.4e-07,
+      "loss": 2.0513,
+      "step": 30
+    },
+    {
+      "epoch": 0.8556149732620321,
+      "grad_norm": 31.82117462158203,
+      "learning_rate": 8.400000000000001e-07,
+      "loss": 1.6275,
+      "step": 40
+    },
+    {
+      "epoch": 1.0695187165775402,
+      "grad_norm": 27.546764373779297,
+      "learning_rate": 1.2400000000000002e-06,
+      "loss": 1.3922,
+      "step": 50
+    },
+    {
+      "epoch": 1.2834224598930482,
+      "grad_norm": 27.145185470581055,
+      "learning_rate": 1.6400000000000002e-06,
+      "loss": 1.0424,
+      "step": 60
+    },
+    {
+      "epoch": 1.4973262032085561,
+      "grad_norm": 33.02227020263672,
+      "learning_rate": 2.04e-06,
+      "loss": 1.0434,
+      "step": 70
+    },
+    {
+      "epoch": 1.7112299465240641,
+      "grad_norm": 23.329740524291992,
+      "learning_rate": 2.4400000000000004e-06,
+      "loss": 1.0585,
+      "step": 80
+    },
+    {
+      "epoch": 1.9251336898395723,
+      "grad_norm": 26.408071517944336,
+      "learning_rate": 2.84e-06,
+      "loss": 0.8849,
+      "step": 90
+    },
+    {
+      "epoch": 2.1390374331550803,
+      "grad_norm": 17.790225982666016,
+      "learning_rate": 3.2400000000000003e-06,
+      "loss": 0.5427,
+      "step": 100
+    },
+    {
+      "epoch": 2.3529411764705883,
+      "grad_norm": 29.846942901611328,
+      "learning_rate": 3.6400000000000003e-06,
+      "loss": 0.2784,
+      "step": 110
+    },
+    {
+      "epoch": 2.5668449197860963,
+      "grad_norm": 20.77630043029785,
+      "learning_rate": 4.04e-06,
+      "loss": 0.2569,
+      "step": 120
+    },
+    {
+      "epoch": 2.7807486631016043,
+      "grad_norm": 33.25271987915039,
+      "learning_rate": 4.440000000000001e-06,
+      "loss": 0.1836,
+      "step": 130
+    },
+    {
+      "epoch": 2.9946524064171123,
+      "grad_norm": 66.23503875732422,
+      "learning_rate": 4.84e-06,
+      "loss": 0.3042,
+      "step": 140
+    },
+    {
+      "epoch": 3.2085561497326203,
+      "grad_norm": 4.618898868560791,
+      "learning_rate": 5.240000000000001e-06,
+      "loss": 0.0455,
+      "step": 150
+    },
+    {
+      "epoch": 3.4224598930481283,
+      "grad_norm": 5.856906890869141,
+      "learning_rate": 5.64e-06,
+      "loss": 0.0558,
+      "step": 160
+    },
+    {
+      "epoch": 3.6363636363636362,
+      "grad_norm": 25.529407501220703,
+      "learning_rate": 6.040000000000001e-06,
+      "loss": 0.0835,
+      "step": 170
+    },
+    {
+      "epoch": 3.8502673796791442,
+      "grad_norm": 16.971477508544922,
+      "learning_rate": 6.440000000000001e-06,
+      "loss": 0.1126,
+      "step": 180
+    },
+    {
+      "epoch": 4.064171122994653,
+      "grad_norm": 9.28547191619873,
+      "learning_rate": 6.8400000000000014e-06,
+      "loss": 0.0838,
+      "step": 190
+    },
+    {
+      "epoch": 4.278074866310161,
+      "grad_norm": 17.296798706054688,
+      "learning_rate": 7.24e-06,
+      "loss": 0.0518,
+      "step": 200
+    },
+    {
+      "epoch": 4.491978609625669,
+      "grad_norm": 12.538177490234375,
+      "learning_rate": 7.640000000000001e-06,
+      "loss": 0.0532,
+      "step": 210
+    },
+    {
+      "epoch": 4.705882352941177,
+      "grad_norm": 24.446659088134766,
+      "learning_rate": 8.040000000000001e-06,
+      "loss": 0.1583,
+      "step": 220
+    },
+    {
+      "epoch": 4.919786096256685,
+      "grad_norm": 25.139781951904297,
+      "learning_rate": 8.44e-06,
+      "loss": 0.0894,
+      "step": 230
+    },
+    {
+      "epoch": 5.133689839572193,
+      "grad_norm": 2.361607074737549,
+      "learning_rate": 8.84e-06,
+      "loss": 0.0531,
+      "step": 240
+    },
+    {
+      "epoch": 5.347593582887701,
+      "grad_norm": 4.768627643585205,
+      "learning_rate": 9.240000000000001e-06,
+      "loss": 0.0664,
+      "step": 250
+    },
+    {
+      "epoch": 5.561497326203209,
+      "grad_norm": 37.02785873413086,
+      "learning_rate": 9.640000000000001e-06,
+      "loss": 0.1502,
+      "step": 260
+    },
+    {
+      "epoch": 5.775401069518717,
+      "grad_norm": 48.693763732910156,
+      "learning_rate": 1e-05,
+      "loss": 0.1953,
+      "step": 270
+    },
+    {
+      "epoch": 5.989304812834225,
+      "grad_norm": 37.84030532836914,
+      "learning_rate": 1e-05,
+      "loss": 0.1806,
+      "step": 280
+    },
+    {
+      "epoch": 6.2032085561497325,
+      "grad_norm": 39.74201965332031,
+      "learning_rate": 1e-05,
+      "loss": 0.0866,
+      "step": 290
+    },
+    {
+      "epoch": 6.4171122994652405,
+      "grad_norm": 7.8511834144592285,
+      "learning_rate": 1e-05,
+      "loss": 0.0994,
+      "step": 300
+    },
+    {
+      "epoch": 6.6310160427807485,
+      "grad_norm": 54.63271713256836,
+      "learning_rate": 1e-05,
+      "loss": 0.1712,
+      "step": 310
+    },
+    {
+      "epoch": 6.8449197860962565,
+      "grad_norm": 44.596153259277344,
+      "learning_rate": 1e-05,
+      "loss": 0.1442,
+      "step": 320
+    },
+    {
+      "epoch": 7.0588235294117645,
+      "grad_norm": 5.094964504241943,
+      "learning_rate": 1e-05,
+      "loss": 0.0526,
+      "step": 330
+    },
+    {
+      "epoch": 7.2727272727272725,
+      "grad_norm": 12.109665870666504,
+      "learning_rate": 1e-05,
+      "loss": 0.03,
+      "step": 340
+    },
+    {
+      "epoch": 7.4866310160427805,
+      "grad_norm": 34.3682975769043,
+      "learning_rate": 1e-05,
+      "loss": 0.0759,
+      "step": 350
+    },
+    {
+      "epoch": 7.7005347593582885,
+      "grad_norm": 2.3397929668426514,
+      "learning_rate": 1e-05,
+      "loss": 0.0685,
+      "step": 360
+    },
+    {
+      "epoch": 7.9144385026737964,
+      "grad_norm": 21.006542205810547,
+      "learning_rate": 1e-05,
+      "loss": 0.0637,
+      "step": 370
+    },
+    {
+      "epoch": 8.128342245989305,
+      "grad_norm": 2.5418567657470703,
+      "learning_rate": 1e-05,
+      "loss": 0.057,
+      "step": 380
+    },
+    {
+      "epoch": 8.342245989304812,
+      "grad_norm": 18.59604263305664,
+      "learning_rate": 1e-05,
+      "loss": 0.1096,
+      "step": 390
+    },
+    {
+      "epoch": 8.556149732620321,
+      "grad_norm": 52.85743713378906,
+      "learning_rate": 1e-05,
+      "loss": 0.1422,
+      "step": 400
+    },
+    {
+      "epoch": 8.556149732620321,
+      "eval_accuracy": 0.4,
+      "eval_loss": 4.6875,
+      "eval_runtime": 0.8724,
+      "eval_samples_per_second": 11.462,
+      "eval_steps_per_second": 1.146,
+      "step": 400
+    },
+    {
+      "epoch": 8.770053475935828,
+      "grad_norm": 36.897674560546875,
+      "learning_rate": 1e-05,
+      "loss": 0.1633,
+      "step": 410
+    },
+    {
+      "epoch": 8.983957219251337,
+      "grad_norm": 16.528366088867188,
+      "learning_rate": 1e-05,
+      "loss": 0.0336,
+      "step": 420
+    },
+    {
+      "epoch": 9.197860962566844,
+      "grad_norm": 87.09658813476562,
+      "learning_rate": 1e-05,
+      "loss": 0.1071,
+      "step": 430
+    },
+    {
+      "epoch": 9.411764705882353,
+      "grad_norm": 4.741786479949951,
+      "learning_rate": 1e-05,
+      "loss": 0.0701,
+      "step": 440
+    },
+    {
+      "epoch": 9.62566844919786,
+      "grad_norm": 12.09588623046875,
+      "learning_rate": 1e-05,
+      "loss": 0.0746,
+      "step": 450
+    },
+    {
+      "epoch": 9.83957219251337,
+      "grad_norm": 32.604068756103516,
+      "learning_rate": 1e-05,
+      "loss": 0.0872,
+      "step": 460
+    },
+    {
+      "epoch": 10.053475935828876,
+      "grad_norm": 4.408775806427002,
+      "learning_rate": 1e-05,
+      "loss": 0.1084,
+      "step": 470
+    },
+    {
+      "epoch": 10.267379679144385,
+      "grad_norm": 22.274890899658203,
+      "learning_rate": 1e-05,
+      "loss": 0.0297,
+      "step": 480
+    },
+    {
+      "epoch": 10.481283422459892,
+      "grad_norm": 58.223323822021484,
+      "learning_rate": 1e-05,
+      "loss": 0.0637,
+      "step": 490
+    },
+    {
+      "epoch": 10.695187165775401,
+      "grad_norm": 2.7027640342712402,
+      "learning_rate": 1e-05,
+      "loss": 0.1206,
+      "step": 500
+    },
+    {
+      "epoch": 10.909090909090908,
+      "grad_norm": 43.47382736206055,
+      "learning_rate": 1e-05,
+      "loss": 0.045,
+      "step": 510
+    },
+    {
+      "epoch": 11.122994652406417,
+      "grad_norm": 1.6665784120559692,
+      "learning_rate": 1e-05,
+      "loss": 0.045,
+      "step": 520
+    },
+    {
+      "epoch": 11.336898395721924,
+      "grad_norm": 0.5738474130630493,
+      "learning_rate": 1e-05,
+      "loss": 0.008,
+      "step": 530
+    },
+    {
+      "epoch": 11.550802139037433,
+      "grad_norm": 16.410573959350586,
+      "learning_rate": 1e-05,
+      "loss": 0.0995,
+      "step": 540
+    },
+    {
+      "epoch": 11.764705882352942,
+      "grad_norm": 0.7015154361724854,
+      "learning_rate": 1e-05,
+      "loss": 0.0313,
+      "step": 550
+    },
+    {
+      "epoch": 11.97860962566845,
+      "grad_norm": 11.851508140563965,
+      "learning_rate": 1e-05,
+      "loss": 0.0724,
+      "step": 560
+    },
+    {
+      "epoch": 12.192513368983958,
+      "grad_norm": 16.754575729370117,
+      "learning_rate": 1e-05,
+      "loss": 0.0272,
+      "step": 570
+    },
+    {
+      "epoch": 12.406417112299465,
+      "grad_norm": 15.92441177368164,
+      "learning_rate": 1e-05,
+      "loss": 0.0241,
+      "step": 580
+    },
+    {
+      "epoch": 12.620320855614974,
+      "grad_norm": 0.05043609067797661,
+      "learning_rate": 1e-05,
+      "loss": 0.042,
+      "step": 590
+    },
+    {
+      "epoch": 12.834224598930481,
+      "grad_norm": 0.3494648337364197,
+      "learning_rate": 1e-05,
+      "loss": 0.0083,
+      "step": 600
+    },
+    {
+      "epoch": 13.04812834224599,
+      "grad_norm": 23.260435104370117,
+      "learning_rate": 1e-05,
+      "loss": 0.026,
+      "step": 610
+    },
+    {
+      "epoch": 13.262032085561497,
+      "grad_norm": 5.789357662200928,
+      "learning_rate": 1e-05,
+      "loss": 0.0224,
+      "step": 620
+    },
+    {
+      "epoch": 13.475935828877006,
+      "grad_norm": 0.056954510509967804,
+      "learning_rate": 1e-05,
+      "loss": 0.0642,
+      "step": 630
+    },
+    {
+      "epoch": 13.689839572192513,
+      "grad_norm": 72.28430938720703,
+      "learning_rate": 1e-05,
+      "loss": 0.0366,
+      "step": 640
+    },
+    {
+      "epoch": 13.903743315508022,
+      "grad_norm": 0.5189899802207947,
+      "learning_rate": 1e-05,
+      "loss": 0.033,
+      "step": 650
+    },
+    {
+      "epoch": 14.117647058823529,
+      "grad_norm": 0.07448354363441467,
+      "learning_rate": 1e-05,
+      "loss": 0.0088,
+      "step": 660
+    },
+    {
+      "epoch": 14.331550802139038,
+      "grad_norm": 0.05018957331776619,
+      "learning_rate": 1e-05,
+      "loss": 0.0082,
+      "step": 670
+    },
+    {
+      "epoch": 14.545454545454545,
+      "grad_norm": 33.61125183105469,
+      "learning_rate": 1e-05,
+      "loss": 0.0753,
+      "step": 680
+    },
+    {
+      "epoch": 14.759358288770054,
+      "grad_norm": 0.023922119289636612,
+      "learning_rate": 1e-05,
+      "loss": 0.0267,
+      "step": 690
+    },
+    {
+      "epoch": 14.973262032085561,
+      "grad_norm": 0.04713653773069382,
+      "learning_rate": 1e-05,
+      "loss": 0.096,
+      "step": 700
+    },
+    {
+      "epoch": 15.18716577540107,
+      "grad_norm": 0.0457034632563591,
+      "learning_rate": 1e-05,
+      "loss": 0.0083,
+      "step": 710
+    },
+    {
+      "epoch": 15.401069518716577,
+      "grad_norm": 0.011708029545843601,
+      "learning_rate": 1e-05,
+      "loss": 0.072,
+      "step": 720
+    },
+    {
+      "epoch": 15.614973262032086,
+      "grad_norm": 0.3412806987762451,
+      "learning_rate": 1e-05,
+      "loss": 0.0032,
+      "step": 730
+    },
+    {
+      "epoch": 15.828877005347593,
+      "grad_norm": 1.667181372642517,
+      "learning_rate": 1e-05,
+      "loss": 0.0512,
+      "step": 740
+    },
+    {
+      "epoch": 16.0427807486631,
+      "grad_norm": 0.004597651772201061,
+      "learning_rate": 1e-05,
+      "loss": 0.0555,
+      "step": 750
+    },
+    {
+      "epoch": 16.25668449197861,
+      "grad_norm": 0.01679002121090889,
+      "learning_rate": 1e-05,
+      "loss": 0.0294,
+      "step": 760
+    },
+    {
+      "epoch": 16.470588235294116,
+      "grad_norm": 24.51487922668457,
+      "learning_rate": 1e-05,
+      "loss": 0.0979,
+      "step": 770
+    },
+    {
+      "epoch": 16.684491978609625,
+      "grad_norm": 0.010406852699816227,
+      "learning_rate": 1e-05,
+      "loss": 0.0361,
+      "step": 780
+    },
+    {
+      "epoch": 16.898395721925134,
+      "grad_norm": 4.567187786102295,
+      "learning_rate": 1e-05,
+      "loss": 0.1229,
+      "step": 790
+    },
+    {
+      "epoch": 17.112299465240643,
+      "grad_norm": 1.7842605113983154,
+      "learning_rate": 1e-05,
+      "loss": 0.0121,
+      "step": 800
+    },
+    {
+      "epoch": 17.112299465240643,
+      "eval_accuracy": 0.5,
+      "eval_loss": 6.6640625,
+      "eval_runtime": 0.8689,
+      "eval_samples_per_second": 11.509,
+      "eval_steps_per_second": 1.151,
+      "step": 800
+    },
+    {
+      "epoch": 17.32620320855615,
+      "grad_norm": 0.061683282256126404,
+      "learning_rate": 1e-05,
+      "loss": 0.0079,
+      "step": 810
+    },
+    {
+      "epoch": 17.540106951871657,
+      "grad_norm": 0.11749367415904999,
+      "learning_rate": 1e-05,
+      "loss": 0.0857,
+      "step": 820
+    },
+    {
+      "epoch": 17.754010695187166,
+      "grad_norm": 0.02787230722606182,
+      "learning_rate": 1e-05,
+      "loss": 0.0051,
+      "step": 830
+    },
+    {
+      "epoch": 17.967914438502675,
+      "grad_norm": 0.015304004773497581,
+      "learning_rate": 1e-05,
+      "loss": 0.0882,
+      "step": 840
+    },
+    {
+      "epoch": 18.181818181818183,
+      "grad_norm": 0.41026026010513306,
+      "learning_rate": 1e-05,
+      "loss": 0.0002,
+      "step": 850
+    },
+    {
+      "epoch": 18.39572192513369,
+      "grad_norm": 0.012662127614021301,
+      "learning_rate": 1e-05,
+      "loss": 0.0013,
+      "step": 860
+    },
+    {
+      "epoch": 18.609625668449198,
+      "grad_norm": 0.10844732820987701,
+      "learning_rate": 1e-05,
+      "loss": 0.0062,
+      "step": 870
+    },
+    {
+      "epoch": 18.823529411764707,
+      "grad_norm": 2.6394922733306885,
+      "learning_rate": 1e-05,
+      "loss": 0.0335,
+      "step": 880
+    },
+    {
+      "epoch": 19.037433155080215,
+      "grad_norm": 0.28504815697669983,
+      "learning_rate": 1e-05,
+      "loss": 0.0092,
+      "step": 890
+    },
+    {
+      "epoch": 19.25133689839572,
+      "grad_norm": 0.03371066227555275,
+      "learning_rate": 1e-05,
+      "loss": 0.0001,
+      "step": 900
+    },
+    {
+      "epoch": 19.46524064171123,
+      "grad_norm": 0.01855628192424774,
+      "learning_rate": 1e-05,
+      "loss": 0.0046,
+      "step": 910
+    },
+    {
+      "epoch": 19.67914438502674,
+      "grad_norm": 0.008938683196902275,
+      "learning_rate": 1e-05,
+      "loss": 0.0014,
+      "step": 920
+    },
+    {
+      "epoch": 19.893048128342247,
+      "grad_norm": 0.0025397331919521093,
+      "learning_rate": 1e-05,
+      "loss": 0.0023,
+      "step": 930
+    },
+    {
+      "epoch": 20.106951871657753,
+      "grad_norm": 0.0024771266616880894,
+      "learning_rate": 1e-05,
+      "loss": 0.001,
+      "step": 940
+    },
+    {
+      "epoch": 20.32085561497326,
+      "grad_norm": 56.46254348754883,
+      "learning_rate": 1e-05,
+      "loss": 0.0533,
+      "step": 950
+    },
+    {
+      "epoch": 20.53475935828877,
+      "grad_norm": 0.046234406530857086,
+      "learning_rate": 1e-05,
+      "loss": 0.0113,
+      "step": 960
+    },
+    {
+      "epoch": 20.74866310160428,
+      "grad_norm": 0.03888735547661781,
+      "learning_rate": 1e-05,
+      "loss": 0.0001,
+      "step": 970
+    },
+    {
+      "epoch": 20.962566844919785,
+      "grad_norm": 4.142932415008545,
+      "learning_rate": 1e-05,
+      "loss": 0.0078,
+      "step": 980
+    },
+    {
+      "epoch": 21.176470588235293,
+      "grad_norm": 0.01926487125456333,
+      "learning_rate": 1e-05,
+      "loss": 0.0015,
+      "step": 990
+    },
+    {
+      "epoch": 21.390374331550802,
+      "grad_norm": 0.03864801302552223,
+      "learning_rate": 1e-05,
+      "loss": 0.0032,
+      "step": 1000
+    },
+    {
+      "epoch": 21.60427807486631,
+      "grad_norm": 0.40588998794555664,
+      "learning_rate": 1e-05,
+      "loss": 0.0003,
+      "step": 1010
+    },
+    {
+      "epoch": 21.818181818181817,
+      "grad_norm": 0.2934487760066986,
+      "learning_rate": 1e-05,
+      "loss": 0.0081,
+      "step": 1020
+    },
+    {
+      "epoch": 22.032085561497325,
+      "grad_norm": 0.004641254432499409,
+      "learning_rate": 1e-05,
+      "loss": 0.0,
+      "step": 1030
+    },
+    {
+      "epoch": 22.245989304812834,
+      "grad_norm": 14.084956169128418,
+      "learning_rate": 1e-05,
+      "loss": 0.0018,
+      "step": 1040
+    },
+    {
+      "epoch": 22.459893048128343,
+      "grad_norm": 1.7572118043899536,
+      "learning_rate": 1e-05,
+      "loss": 0.001,
+      "step": 1050
+    },
+    {
+      "epoch": 22.67379679144385,
+      "grad_norm": 2.7290031909942627,
+      "learning_rate": 1e-05,
+      "loss": 0.0002,
+      "step": 1060
+    },
+    {
+      "epoch": 22.887700534759357,
+      "grad_norm": 0.005051525309681892,
+      "learning_rate": 1e-05,
+      "loss": 0.0007,
+      "step": 1070
+    },
+    {
+      "epoch": 23.101604278074866,
+      "grad_norm": 0.002047200920060277,
+      "learning_rate": 1e-05,
+      "loss": 0.0,
+      "step": 1080
+    },
+    {
+      "epoch": 23.315508021390375,
+      "grad_norm": 0.0035107170697301626,
+      "learning_rate": 1e-05,
+      "loss": 0.0494,
+      "step": 1090
+    },
+    {
+      "epoch": 23.529411764705884,
+      "grad_norm": 0.004472868517041206,
+      "learning_rate": 1e-05,
+      "loss": 0.0034,
+      "step": 1100
+    },
+    {
+      "epoch": 23.74331550802139,
+      "grad_norm": 0.08594793826341629,
+      "learning_rate": 1e-05,
+      "loss": 0.0063,
+      "step": 1110
+    },
+    {
+      "epoch": 23.9572192513369,
+      "grad_norm": 0.752574622631073,
+      "learning_rate": 1e-05,
+      "loss": 0.0002,
+      "step": 1120
+    },
+    {
+      "epoch": 24.171122994652407,
+      "grad_norm": 0.021934740245342255,
+      "learning_rate": 1e-05,
+      "loss": 0.0019,
+      "step": 1130
+    },
+    {
+      "epoch": 24.385026737967916,
+      "grad_norm": 0.007998108863830566,
+      "learning_rate": 1e-05,
+      "loss": 0.0,
+      "step": 1140
+    },
+    {
+      "epoch": 24.59893048128342,
+      "grad_norm": 0.010390130802989006,
+      "learning_rate": 1e-05,
+      "loss": 0.0072,
+      "step": 1150
+    },
+    {
+      "epoch": 24.81283422459893,
+      "grad_norm": 0.002071298658847809,
+      "learning_rate": 1e-05,
+      "loss": 0.0013,
+      "step": 1160
+    },
+    {
+      "epoch": 25.02673796791444,
+      "grad_norm": 0.0045625255443155766,
+      "learning_rate": 1e-05,
+      "loss": 0.002,
+      "step": 1170
+    },
+    {
+      "epoch": 25.240641711229948,
+      "grad_norm": 0.017212701961398125,
+      "learning_rate": 1e-05,
+      "loss": 0.0048,
+      "step": 1180
+    },
+    {
+      "epoch": 25.454545454545453,
+      "grad_norm": 12.734994888305664,
+      "learning_rate": 1e-05,
+      "loss": 0.0018,
+      "step": 1190
+    },
+    {
+      "epoch": 25.668449197860962,
+      "grad_norm": 0.0035849984269589186,
+      "learning_rate": 1e-05,
+      "loss": 0.0,
+      "step": 1200
+    },
+    {
+      "epoch": 25.668449197860962,
+      "eval_accuracy": 0.8,
+      "eval_loss": 4.71484375,
+      "eval_runtime": 0.8712,
+      "eval_samples_per_second": 11.478,
+      "eval_steps_per_second": 1.148,
+      "step": 1200
+    },
+    {
+      "epoch": 25.88235294117647,
+      "grad_norm": 0.005425190087407827,
+      "learning_rate": 1e-05,
+      "loss": 0.0,
+      "step": 1210
+    },
+    {
+      "epoch": 26.09625668449198,
+      "grad_norm": 0.007795536890625954,
+      "learning_rate": 1e-05,
+      "loss": 0.0034,
+      "step": 1220
+    },
+    {
+      "epoch": 26.310160427807485,
+      "grad_norm": 0.00858351681381464,
+      "learning_rate": 1e-05,
+      "loss": 0.0032,
+      "step": 1230
+    },
+    {
+      "epoch": 26.524064171122994,
+      "grad_norm": 0.0035222836304455996,
+      "learning_rate": 1e-05,
+      "loss": 0.0103,
+      "step": 1240
+    },
+    {
+      "epoch": 26.737967914438503,
+      "grad_norm": 0.004781092517077923,
+      "learning_rate": 1e-05,
+      "loss": 0.0002,
+      "step": 1250
+    },
+    {
+      "epoch": 26.951871657754012,
+      "grad_norm": 0.0031953530851751566,
+      "learning_rate": 1e-05,
+      "loss": 0.0013,
+      "step": 1260
+    },
+    {
+      "epoch": 27.165775401069517,
+      "grad_norm": 0.00516524026170373,
+      "learning_rate": 1e-05,
+      "loss": 0.0007,
+      "step": 1270
+    },
+    {
+      "epoch": 27.379679144385026,
+      "grad_norm": 0.004281027242541313,
+      "learning_rate": 1e-05,
+      "loss": 0.0106,
+      "step": 1280
+    },
+    {
+      "epoch": 27.593582887700535,
+      "grad_norm": 0.0015023744199424982,
+      "learning_rate": 1e-05,
+      "loss": 0.0,
+      "step": 1290
+    },
+    {
+      "epoch": 27.807486631016044,
+      "grad_norm": 0.005484332330524921,
+      "learning_rate": 1e-05,
+      "loss": 0.0023,
+      "step": 1300
+    },
+    {
+      "epoch": 28.02139037433155,
+      "grad_norm": 0.0035302124451845884,
+      "learning_rate": 1e-05,
+      "loss": 0.0001,
+      "step": 1310
+    },
+    {
+      "epoch": 28.235294117647058,
+      "grad_norm": 0.0033441612031310797,
+      "learning_rate": 1e-05,
+      "loss": 0.0025,
+      "step": 1320
+    },
+    {
+      "epoch": 28.449197860962567,
+      "grad_norm": 0.0035446221008896828,
+      "learning_rate": 1e-05,
+      "loss": 0.0014,
+      "step": 1330
+    },
+    {
+      "epoch": 28.663101604278076,
+      "grad_norm": 0.010898416861891747,
+      "learning_rate": 1e-05,
+      "loss": 0.002,
+      "step": 1340
+    },
+    {
+      "epoch": 28.87700534759358,
+      "grad_norm": 0.008727981708943844,
+      "learning_rate": 1e-05,
+      "loss": 0.1828,
+      "step": 1350
+    },
+    {
+      "epoch": 29.09090909090909,
+      "grad_norm": 0.004256559070199728,
+      "learning_rate": 1e-05,
+      "loss": 0.0431,
+      "step": 1360
+    },
+    {
+      "epoch": 29.3048128342246,
+      "grad_norm": 0.21178306639194489,
+      "learning_rate": 1e-05,
+      "loss": 0.0099,
+      "step": 1370
+    },
+    {
+      "epoch": 29.518716577540108,
+      "grad_norm": 0.2352713793516159,
+      "learning_rate": 1e-05,
+      "loss": 0.0835,
+      "step": 1380
+    },
+    {
+      "epoch": 29.732620320855617,
+      "grad_norm": 0.03791838884353638,
+      "learning_rate": 1e-05,
+      "loss": 0.0522,
+      "step": 1390
+    },
+    {
+      "epoch": 29.946524064171122,
+      "grad_norm": 0.04343949630856514,
+      "learning_rate": 1e-05,
+      "loss": 0.0004,
+      "step": 1400
+    },
+    {
+      "epoch": 30.16042780748663,
+      "grad_norm": 2.7350242137908936,
+      "learning_rate": 1e-05,
+      "loss": 0.003,
+      "step": 1410
+    },
+    {
+      "epoch": 30.37433155080214,
+      "grad_norm": 0.030035167932510376,
+      "learning_rate": 1e-05,
+      "loss": 0.006,
+      "step": 1420
+    },
+    {
+      "epoch": 30.58823529411765,
+      "grad_norm": 3.1150763034820557,
+      "learning_rate": 1e-05,
+      "loss": 0.0016,
+      "step": 1430
+    },
+    {
+      "epoch": 30.802139037433154,
+      "grad_norm": 0.028651399537920952,
+      "learning_rate": 1e-05,
+      "loss": 0.0016,
+      "step": 1440
+    },
+    {
+      "epoch": 31.016042780748663,
+      "grad_norm": 37.155609130859375,
+      "learning_rate": 1e-05,
+      "loss": 0.0128,
+      "step": 1450
+    },
+    {
+      "epoch": 31.22994652406417,
+      "grad_norm": 0.002968219807371497,
+      "learning_rate": 1e-05,
+      "loss": 0.0141,
+      "step": 1460
+    },
+    {
+      "epoch": 31.44385026737968,
+      "grad_norm": 5.24705696105957,
+      "learning_rate": 1e-05,
+      "loss": 0.0121,
+      "step": 1470
+    },
+    {
+      "epoch": 31.657754010695186,
+      "grad_norm": 0.0024348751176148653,
+      "learning_rate": 1e-05,
+      "loss": 0.1083,
+      "step": 1480
+    },
+    {
+      "epoch": 31.871657754010695,
+      "grad_norm": 0.5254661440849304,
+      "learning_rate": 1e-05,
+      "loss": 0.1179,
+      "step": 1490
+    },
+    {
+      "epoch": 32.0855614973262,
+      "grad_norm": 72.65499114990234,
+      "learning_rate": 1e-05,
+      "loss": 0.1361,
+      "step": 1500
+    },
+    {
+      "epoch": 32.29946524064171,
+      "grad_norm": 14.370610237121582,
+      "learning_rate": 1e-05,
+      "loss": 0.0346,
+      "step": 1510
+    },
+    {
+      "epoch": 32.51336898395722,
+      "grad_norm": 20.988916397094727,
+      "learning_rate": 1e-05,
+      "loss": 0.0025,
+      "step": 1520
+    },
+    {
+      "epoch": 32.72727272727273,
+      "grad_norm": 0.002341994782909751,
+      "learning_rate": 1e-05,
+      "loss": 0.0004,
+      "step": 1530
+    },
+    {
+      "epoch": 32.94117647058823,
+      "grad_norm": 2.404449701309204,
+      "learning_rate": 1e-05,
+      "loss": 0.013,
+      "step": 1540
+    },
+    {
+      "epoch": 33.155080213903744,
+      "grad_norm": 0.005042242351919413,
+      "learning_rate": 1e-05,
+      "loss": 0.0011,
+      "step": 1550
+    },
+    {
+      "epoch": 33.36898395721925,
+      "grad_norm": 0.00044189911568537354,
+      "learning_rate": 1e-05,
+      "loss": 0.0139,
+      "step": 1560
+    },
+    {
+      "epoch": 33.58288770053476,
+      "grad_norm": 0.0014409746509045362,
+      "learning_rate": 1e-05,
+      "loss": 0.0284,
+      "step": 1570
+    },
+    {
+      "epoch": 33.79679144385027,
+      "grad_norm": 2.608125686645508,
+      "learning_rate": 1e-05,
+      "loss": 0.015,
+      "step": 1580
+    },
+    {
+      "epoch": 34.01069518716577,
+      "grad_norm": 0.0005788679700344801,
+      "learning_rate": 1e-05,
+      "loss": 0.0201,
+      "step": 1590
+    },
+    {
+      "epoch": 34.224598930481285,
+      "grad_norm": 75.40068817138672,
+      "learning_rate": 1e-05,
+      "loss": 0.028,
+      "step": 1600
+    },
+    {
+      "epoch": 34.224598930481285,
+      "eval_accuracy": 0.7,
+      "eval_loss": 6.69921875,
+      "eval_runtime": 0.8667,
+      "eval_samples_per_second": 11.537,
+      "eval_steps_per_second": 1.154,
+      "step": 1600
+    }
+  ],
+  "logging_steps": 10,
+  "max_steps": 2500,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 55,
+  "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": 1.3208690665078129e+18,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}

checkpoint-1600/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7ea5f51f632ffed5e26cc809a24c0a64055c95fd6a4bba189c05a309473a06bf
+size 6136

checkpoint-1600/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)