Training in progress, step 800, checkpoint

checkpoint-800/config.json

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

checkpoint-800/global_step800/mp_rank_00_model_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7b5835ffa3f642a1d805b97eab146dcdf8fdf71e0269704e18345b444001aa9f
+size 5292979768

checkpoint-800/global_step800/zero_pp_rank_0_mp_rank_00_optim_states.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a562158348c361255d97f488e4f4adf5a427d72a94524121f626fd629f9aa86d
+size 15878620944

checkpoint-800/latest

@@ -0,0 +1 @@
+global_step800

checkpoint-800/model-00001-of-00002.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:86fa3777f8244f56a9056cb329e6714b4753d96e090276392a897f6f9fddbc87
+size 4978208880

checkpoint-800/model-00002-of-00002.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d2bc983a7bf0e6f4fc70443e8f056b56acb5d6d263e035bb819843afe460c360
+size 314703498

checkpoint-800/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-800/rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b08c1a03c6cc74f9794ef7c74d7ddb57c80f6e38283d9395061767f1663897c8
+size 14640

checkpoint-800/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:958f786b31c911d92821e8537c92827d4a57d9d5549221f8f9236f167b6a1b1a
+size 1064

checkpoint-800/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-800/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-800/trainer_state.json

@@ -0,0 +1,618 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 17.112299465240643,
+  "eval_steps": 400,
+  "global_step": 800,
+  "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": 1.9241,
+      "step": 1
+    },
+    {
+      "epoch": 0.21390374331550802,
+      "grad_norm": 0.0,
+      "learning_rate": 0.0,
+      "loss": 1.7796,
+      "step": 10
+    },
+    {
+      "epoch": 0.42780748663101603,
+      "grad_norm": 43.9833984375,
+      "learning_rate": 4e-08,
+      "loss": 1.8211,
+      "step": 20
+    },
+    {
+      "epoch": 0.6417112299465241,
+      "grad_norm": 40.75557327270508,
+      "learning_rate": 4.4e-07,
+      "loss": 1.733,
+      "step": 30
+    },
+    {
+      "epoch": 0.8556149732620321,
+      "grad_norm": 33.69241714477539,
+      "learning_rate": 8.400000000000001e-07,
+      "loss": 1.4004,
+      "step": 40
+    },
+    {
+      "epoch": 1.0695187165775402,
+      "grad_norm": 32.86851501464844,
+      "learning_rate": 1.2400000000000002e-06,
+      "loss": 1.3416,
+      "step": 50
+    },
+    {
+      "epoch": 1.2834224598930482,
+      "grad_norm": 27.803329467773438,
+      "learning_rate": 1.6400000000000002e-06,
+      "loss": 1.081,
+      "step": 60
+    },
+    {
+      "epoch": 1.4973262032085561,
+      "grad_norm": 31.654834747314453,
+      "learning_rate": 2.04e-06,
+      "loss": 1.05,
+      "step": 70
+    },
+    {
+      "epoch": 1.7112299465240641,
+      "grad_norm": 29.529935836791992,
+      "learning_rate": 2.4400000000000004e-06,
+      "loss": 0.9944,
+      "step": 80
+    },
+    {
+      "epoch": 1.9251336898395723,
+      "grad_norm": 29.650171279907227,
+      "learning_rate": 2.84e-06,
+      "loss": 0.8562,
+      "step": 90
+    },
+    {
+      "epoch": 2.1390374331550803,
+      "grad_norm": 19.60784912109375,
+      "learning_rate": 3.2400000000000003e-06,
+      "loss": 0.5071,
+      "step": 100
+    },
+    {
+      "epoch": 2.3529411764705883,
+      "grad_norm": 40.14349365234375,
+      "learning_rate": 3.6400000000000003e-06,
+      "loss": 0.2566,
+      "step": 110
+    },
+    {
+      "epoch": 2.5668449197860963,
+      "grad_norm": 38.05447006225586,
+      "learning_rate": 4.04e-06,
+      "loss": 0.2548,
+      "step": 120
+    },
+    {
+      "epoch": 2.7807486631016043,
+      "grad_norm": 48.719520568847656,
+      "learning_rate": 4.440000000000001e-06,
+      "loss": 0.2113,
+      "step": 130
+    },
+    {
+      "epoch": 2.9946524064171123,
+      "grad_norm": 82.81842803955078,
+      "learning_rate": 4.84e-06,
+      "loss": 0.3946,
+      "step": 140
+    },
+    {
+      "epoch": 3.2085561497326203,
+      "grad_norm": 20.48181915283203,
+      "learning_rate": 5.240000000000001e-06,
+      "loss": 0.09,
+      "step": 150
+    },
+    {
+      "epoch": 3.4224598930481283,
+      "grad_norm": 5.467124938964844,
+      "learning_rate": 5.64e-06,
+      "loss": 0.07,
+      "step": 160
+    },
+    {
+      "epoch": 3.6363636363636362,
+      "grad_norm": 27.81447410583496,
+      "learning_rate": 6.040000000000001e-06,
+      "loss": 0.0837,
+      "step": 170
+    },
+    {
+      "epoch": 3.8502673796791442,
+      "grad_norm": 23.266380310058594,
+      "learning_rate": 6.440000000000001e-06,
+      "loss": 0.1036,
+      "step": 180
+    },
+    {
+      "epoch": 4.064171122994653,
+      "grad_norm": 34.769287109375,
+      "learning_rate": 6.8400000000000014e-06,
+      "loss": 0.0438,
+      "step": 190
+    },
+    {
+      "epoch": 4.278074866310161,
+      "grad_norm": 25.370189666748047,
+      "learning_rate": 7.24e-06,
+      "loss": 0.0541,
+      "step": 200
+    },
+    {
+      "epoch": 4.491978609625669,
+      "grad_norm": 32.902244567871094,
+      "learning_rate": 7.640000000000001e-06,
+      "loss": 0.1416,
+      "step": 210
+    },
+    {
+      "epoch": 4.705882352941177,
+      "grad_norm": 47.98198699951172,
+      "learning_rate": 8.040000000000001e-06,
+      "loss": 0.0939,
+      "step": 220
+    },
+    {
+      "epoch": 4.919786096256685,
+      "grad_norm": 16.503026962280273,
+      "learning_rate": 8.44e-06,
+      "loss": 0.1332,
+      "step": 230
+    },
+    {
+      "epoch": 5.133689839572193,
+      "grad_norm": 6.1169633865356445,
+      "learning_rate": 8.84e-06,
+      "loss": 0.0508,
+      "step": 240
+    },
+    {
+      "epoch": 5.347593582887701,
+      "grad_norm": 6.8135552406311035,
+      "learning_rate": 9.240000000000001e-06,
+      "loss": 0.0915,
+      "step": 250
+    },
+    {
+      "epoch": 5.561497326203209,
+      "grad_norm": 42.288963317871094,
+      "learning_rate": 9.640000000000001e-06,
+      "loss": 0.1364,
+      "step": 260
+    },
+    {
+      "epoch": 5.775401069518717,
+      "grad_norm": 68.40443420410156,
+      "learning_rate": 1e-05,
+      "loss": 0.183,
+      "step": 270
+    },
+    {
+      "epoch": 5.989304812834225,
+      "grad_norm": 18.801176071166992,
+      "learning_rate": 1e-05,
+      "loss": 0.1234,
+      "step": 280
+    },
+    {
+      "epoch": 6.2032085561497325,
+      "grad_norm": 17.309553146362305,
+      "learning_rate": 1e-05,
+      "loss": 0.0899,
+      "step": 290
+    },
+    {
+      "epoch": 6.4171122994652405,
+      "grad_norm": 39.96405792236328,
+      "learning_rate": 1e-05,
+      "loss": 0.1682,
+      "step": 300
+    },
+    {
+      "epoch": 6.6310160427807485,
+      "grad_norm": 37.08085250854492,
+      "learning_rate": 1e-05,
+      "loss": 0.1817,
+      "step": 310
+    },
+    {
+      "epoch": 6.8449197860962565,
+      "grad_norm": 41.293277740478516,
+      "learning_rate": 1e-05,
+      "loss": 0.1853,
+      "step": 320
+    },
+    {
+      "epoch": 7.0588235294117645,
+      "grad_norm": 19.117578506469727,
+      "learning_rate": 1e-05,
+      "loss": 0.084,
+      "step": 330
+    },
+    {
+      "epoch": 7.2727272727272725,
+      "grad_norm": 3.196685552597046,
+      "learning_rate": 1e-05,
+      "loss": 0.0796,
+      "step": 340
+    },
+    {
+      "epoch": 7.4866310160427805,
+      "grad_norm": 34.18360900878906,
+      "learning_rate": 1e-05,
+      "loss": 0.0638,
+      "step": 350
+    },
+    {
+      "epoch": 7.7005347593582885,
+      "grad_norm": 34.174346923828125,
+      "learning_rate": 1e-05,
+      "loss": 0.0499,
+      "step": 360
+    },
+    {
+      "epoch": 7.9144385026737964,
+      "grad_norm": 30.022125244140625,
+      "learning_rate": 1e-05,
+      "loss": 0.0511,
+      "step": 370
+    },
+    {
+      "epoch": 8.128342245989305,
+      "grad_norm": 5.687129020690918,
+      "learning_rate": 1e-05,
+      "loss": 0.0381,
+      "step": 380
+    },
+    {
+      "epoch": 8.342245989304812,
+      "grad_norm": 8.238314628601074,
+      "learning_rate": 1e-05,
+      "loss": 0.0699,
+      "step": 390
+    },
+    {
+      "epoch": 8.556149732620321,
+      "grad_norm": 48.88713836669922,
+      "learning_rate": 1e-05,
+      "loss": 0.1932,
+      "step": 400
+    },
+    {
+      "epoch": 8.556149732620321,
+      "eval_accuracy": 0.6,
+      "eval_loss": 4.55859375,
+      "eval_runtime": 0.8621,
+      "eval_samples_per_second": 11.599,
+      "eval_steps_per_second": 1.16,
+      "step": 400
+    },
+    {
+      "epoch": 8.770053475935828,
+      "grad_norm": 1.5831031799316406,
+      "learning_rate": 1e-05,
+      "loss": 0.1437,
+      "step": 410
+    },
+    {
+      "epoch": 8.983957219251337,
+      "grad_norm": 1.2749309539794922,
+      "learning_rate": 1e-05,
+      "loss": 0.0762,
+      "step": 420
+    },
+    {
+      "epoch": 9.197860962566844,
+      "grad_norm": 56.0820426940918,
+      "learning_rate": 1e-05,
+      "loss": 0.056,
+      "step": 430
+    },
+    {
+      "epoch": 9.411764705882353,
+      "grad_norm": 21.727741241455078,
+      "learning_rate": 1e-05,
+      "loss": 0.0458,
+      "step": 440
+    },
+    {
+      "epoch": 9.62566844919786,
+      "grad_norm": 3.6833608150482178,
+      "learning_rate": 1e-05,
+      "loss": 0.044,
+      "step": 450
+    },
+    {
+      "epoch": 9.83957219251337,
+      "grad_norm": 43.264564514160156,
+      "learning_rate": 1e-05,
+      "loss": 0.0981,
+      "step": 460
+    },
+    {
+      "epoch": 10.053475935828876,
+      "grad_norm": 0.6142730712890625,
+      "learning_rate": 1e-05,
+      "loss": 0.0384,
+      "step": 470
+    },
+    {
+      "epoch": 10.267379679144385,
+      "grad_norm": 1.119858980178833,
+      "learning_rate": 1e-05,
+      "loss": 0.0681,
+      "step": 480
+    },
+    {
+      "epoch": 10.481283422459892,
+      "grad_norm": 0.2561619281768799,
+      "learning_rate": 1e-05,
+      "loss": 0.0087,
+      "step": 490
+    },
+    {
+      "epoch": 10.695187165775401,
+      "grad_norm": 37.55344009399414,
+      "learning_rate": 1e-05,
+      "loss": 0.0371,
+      "step": 500
+    },
+    {
+      "epoch": 10.909090909090908,
+      "grad_norm": 25.553625106811523,
+      "learning_rate": 1e-05,
+      "loss": 0.1983,
+      "step": 510
+    },
+    {
+      "epoch": 11.122994652406417,
+      "grad_norm": 3.476731777191162,
+      "learning_rate": 1e-05,
+      "loss": 0.0475,
+      "step": 520
+    },
+    {
+      "epoch": 11.336898395721924,
+      "grad_norm": 32.92385482788086,
+      "learning_rate": 1e-05,
+      "loss": 0.0108,
+      "step": 530
+    },
+    {
+      "epoch": 11.550802139037433,
+      "grad_norm": 14.769713401794434,
+      "learning_rate": 1e-05,
+      "loss": 0.0174,
+      "step": 540
+    },
+    {
+      "epoch": 11.764705882352942,
+      "grad_norm": 1.3393691778182983,
+      "learning_rate": 1e-05,
+      "loss": 0.0503,
+      "step": 550
+    },
+    {
+      "epoch": 11.97860962566845,
+      "grad_norm": 0.19340232014656067,
+      "learning_rate": 1e-05,
+      "loss": 0.0393,
+      "step": 560
+    },
+    {
+      "epoch": 12.192513368983958,
+      "grad_norm": 2.4283690452575684,
+      "learning_rate": 1e-05,
+      "loss": 0.0659,
+      "step": 570
+    },
+    {
+      "epoch": 12.406417112299465,
+      "grad_norm": 0.5373936295509338,
+      "learning_rate": 1e-05,
+      "loss": 0.0014,
+      "step": 580
+    },
+    {
+      "epoch": 12.620320855614974,
+      "grad_norm": 1.1644189357757568,
+      "learning_rate": 1e-05,
+      "loss": 0.0651,
+      "step": 590
+    },
+    {
+      "epoch": 12.834224598930481,
+      "grad_norm": 0.6392247676849365,
+      "learning_rate": 1e-05,
+      "loss": 0.013,
+      "step": 600
+    },
+    {
+      "epoch": 13.04812834224599,
+      "grad_norm": 0.13331516087055206,
+      "learning_rate": 1e-05,
+      "loss": 0.0215,
+      "step": 610
+    },
+    {
+      "epoch": 13.262032085561497,
+      "grad_norm": 0.08020053058862686,
+      "learning_rate": 1e-05,
+      "loss": 0.0267,
+      "step": 620
+    },
+    {
+      "epoch": 13.475935828877006,
+      "grad_norm": 6.262119770050049,
+      "learning_rate": 1e-05,
+      "loss": 0.0137,
+      "step": 630
+    },
+    {
+      "epoch": 13.689839572192513,
+      "grad_norm": 0.7786157727241516,
+      "learning_rate": 1e-05,
+      "loss": 0.037,
+      "step": 640
+    },
+    {
+      "epoch": 13.903743315508022,
+      "grad_norm": 6.777099132537842,
+      "learning_rate": 1e-05,
+      "loss": 0.0015,
+      "step": 650
+    },
+    {
+      "epoch": 14.117647058823529,
+      "grad_norm": 0.0033113746903836727,
+      "learning_rate": 1e-05,
+      "loss": 0.0255,
+      "step": 660
+    },
+    {
+      "epoch": 14.331550802139038,
+      "grad_norm": 13.173226356506348,
+      "learning_rate": 1e-05,
+      "loss": 0.0056,
+      "step": 670
+    },
+    {
+      "epoch": 14.545454545454545,
+      "grad_norm": 0.021669812500476837,
+      "learning_rate": 1e-05,
+      "loss": 0.0026,
+      "step": 680
+    },
+    {
+      "epoch": 14.759358288770054,
+      "grad_norm": 0.03204642981290817,
+      "learning_rate": 1e-05,
+      "loss": 0.0046,
+      "step": 690
+    },
+    {
+      "epoch": 14.973262032085561,
+      "grad_norm": 0.27756085991859436,
+      "learning_rate": 1e-05,
+      "loss": 0.0131,
+      "step": 700
+    },
+    {
+      "epoch": 15.18716577540107,
+      "grad_norm": 2.8744680881500244,
+      "learning_rate": 1e-05,
+      "loss": 0.0051,
+      "step": 710
+    },
+    {
+      "epoch": 15.401069518716577,
+      "grad_norm": 6.371706962585449,
+      "learning_rate": 1e-05,
+      "loss": 0.0471,
+      "step": 720
+    },
+    {
+      "epoch": 15.614973262032086,
+      "grad_norm": 0.01101295743137598,
+      "learning_rate": 1e-05,
+      "loss": 0.0057,
+      "step": 730
+    },
+    {
+      "epoch": 15.828877005347593,
+      "grad_norm": 0.021259065717458725,
+      "learning_rate": 1e-05,
+      "loss": 0.0822,
+      "step": 740
+    },
+    {
+      "epoch": 16.0427807486631,
+      "grad_norm": 19.74135398864746,
+      "learning_rate": 1e-05,
+      "loss": 0.0141,
+      "step": 750
+    },
+    {
+      "epoch": 16.25668449197861,
+      "grad_norm": 0.0036215828731656075,
+      "learning_rate": 1e-05,
+      "loss": 0.0158,
+      "step": 760
+    },
+    {
+      "epoch": 16.470588235294116,
+      "grad_norm": 19.020896911621094,
+      "learning_rate": 1e-05,
+      "loss": 0.0107,
+      "step": 770
+    },
+    {
+      "epoch": 16.684491978609625,
+      "grad_norm": 0.029464904218912125,
+      "learning_rate": 1e-05,
+      "loss": 0.0004,
+      "step": 780
+    },
+    {
+      "epoch": 16.898395721925134,
+      "grad_norm": 0.0017767682438716292,
+      "learning_rate": 1e-05,
+      "loss": 0.0006,
+      "step": 790
+    },
+    {
+      "epoch": 17.112299465240643,
+      "grad_norm": 4.318113327026367,
+      "learning_rate": 1e-05,
+      "loss": 0.0061,
+      "step": 800
+    },
+    {
+      "epoch": 17.112299465240643,
+      "eval_accuracy": 0.6,
+      "eval_loss": 5.87890625,
+      "eval_runtime": 0.8622,
+      "eval_samples_per_second": 11.598,
+      "eval_steps_per_second": 1.16,
+      "step": 800
+    }
+  ],
+  "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": 6.603619036121006e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}

checkpoint-800/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5bee4e0aae3883768eae4bdb744484177b7f7ed1fa0052758ccb7b1a9fbc6b83
+size 6136

checkpoint-800/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)