a

.gitattributes

@@ -33,3 +33,14 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zip filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
+tokenizer_config.json filter=lfs diff=lfs merge=lfs -text
+training_args.bin filter=lfs diff=lfs merge=lfs -text
+added_tokens.json filter=lfs diff=lfs merge=lfs -text
+checkpoint-160 filter=lfs diff=lfs merge=lfs -text
+checkpoint-170 filter=lfs diff=lfs merge=lfs -text
+config.json filter=lfs diff=lfs merge=lfs -text
+special_tokens_map.json filter=lfs diff=lfs merge=lfs -text
+tokenizer.model filter=lfs diff=lfs merge=lfs -text
+checkpoint-220 filter=lfs diff=lfs merge=lfs -text
+generation_config.json filter=lfs diff=lfs merge=lfs -text
+pytorch_model.bin filter=lfs diff=lfs merge=lfs -text

added_tokens.json

@@ -1,11 +1,3 @@
-{
-  "</s>": 2,
-  "</talking>": 32005,
-  "</thinking>": 32003,
-  "<s>": 1,
-  "<talking>": 32004,
-  "<thinking>": 32002,
-  "<unk>": 0,
-  "<|im_end|>": 32000,
-  "<|im_start|>": 32001
-}
+version https://git-lfs.github.com/spec/v1
+oid sha256:ca73f03b622625f22a4d181eca666378c8711c56bb0895aa0d30d19bb674c5ea
+size 182

checkpoint-160/config.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:44e972929ad46908648b92c2feb180fb201dab6bdc232a5bc5776a74d47e0e0a
+size 620

checkpoint-160/generation_config.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:afcf2a1da583dbeb66bb367a759713901e23ce19ed83ed8f5a21b0e877addd5b
+size 116

checkpoint-160/pytorch_model-00001-of-00002.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5161eefe24c7b021b618fe48ae2788dc3ff9daada9430fac69d7d3a541d9e496
+size 9943077196

checkpoint-160/pytorch_model-00002-of-00002.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9a559bb2452356d73b1ff0d2f48b354cb7656babf27ec0d186ca2ad20a15e3fe
+size 4540584799

checkpoint-160/pytorch_model.bin.index.json

@@ -0,0 +1,298 @@
+{
+  "metadata": {
+    "total_size": 14483562496
+  },
+  "weight_map": {
+    "lm_head.weight": "pytorch_model-00002-of-00002.bin",
+    "model.embed_tokens.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.0.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.0.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.0.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.0.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.0.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.0.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.0.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.0.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.0.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.1.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.1.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.1.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.1.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.1.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.1.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.1.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.1.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.1.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.10.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.10.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.10.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.10.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.10.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.10.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.10.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.10.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.10.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.11.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.11.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.11.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.11.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.11.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.11.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.11.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.11.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.11.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.12.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.12.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.12.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.12.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.12.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.12.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.12.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.12.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.12.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.13.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.13.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.13.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.13.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.13.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.13.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.13.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.13.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.13.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.14.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.14.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.14.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.14.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.14.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.14.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.14.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.14.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.14.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.15.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.15.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.15.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.15.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.15.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.15.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.15.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.15.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.15.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.16.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.16.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.16.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.16.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.16.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.16.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.16.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.16.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.16.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.17.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.17.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.17.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.17.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.17.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.17.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.17.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.17.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.17.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.18.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.18.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.18.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.18.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.18.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.18.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.18.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.18.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.18.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.19.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.19.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.19.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.19.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.19.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.19.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.19.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.19.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.19.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.2.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.2.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.2.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.2.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.2.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.2.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.2.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.2.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.2.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.20.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.20.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.20.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.20.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.20.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.20.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.20.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.20.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.20.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.21.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.21.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.21.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.21.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.21.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.21.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.21.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.21.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.21.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.22.input_layernorm.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.22.mlp.down_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.22.mlp.gate_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.22.mlp.up_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.22.post_attention_layernorm.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.22.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.22.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.22.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.22.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.23.input_layernorm.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.23.mlp.down_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.23.mlp.gate_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.23.mlp.up_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.23.post_attention_layernorm.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.23.self_attn.k_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.23.self_attn.o_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.23.self_attn.q_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.23.self_attn.v_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.24.input_layernorm.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.24.mlp.down_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.24.mlp.gate_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.24.mlp.up_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.24.post_attention_layernorm.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.24.self_attn.k_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.24.self_attn.o_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.24.self_attn.q_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.24.self_attn.v_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.25.input_layernorm.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.25.mlp.down_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.25.mlp.gate_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.25.mlp.up_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.25.post_attention_layernorm.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.25.self_attn.k_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.25.self_attn.o_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.25.self_attn.q_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.25.self_attn.v_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.26.input_layernorm.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.26.mlp.down_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.26.mlp.gate_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.26.mlp.up_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.26.post_attention_layernorm.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.26.self_attn.k_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.26.self_attn.o_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.26.self_attn.q_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.26.self_attn.v_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.27.input_layernorm.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.27.mlp.down_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.27.mlp.gate_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.27.mlp.up_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.27.post_attention_layernorm.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.27.self_attn.k_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.27.self_attn.o_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.27.self_attn.q_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.27.self_attn.v_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.28.input_layernorm.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.28.mlp.down_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.28.mlp.gate_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.28.mlp.up_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.28.post_attention_layernorm.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.28.self_attn.k_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.28.self_attn.o_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.28.self_attn.q_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.28.self_attn.v_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.29.input_layernorm.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.29.mlp.down_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.29.mlp.gate_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.29.mlp.up_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.29.post_attention_layernorm.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.29.self_attn.k_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.29.self_attn.o_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.29.self_attn.q_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.29.self_attn.v_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.3.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.3.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.3.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.3.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.3.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.3.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.3.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.3.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.3.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.30.input_layernorm.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.30.mlp.down_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.30.mlp.gate_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.30.mlp.up_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.30.post_attention_layernorm.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.30.self_attn.k_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.30.self_attn.o_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.30.self_attn.q_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.30.self_attn.v_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.31.input_layernorm.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.31.mlp.down_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.31.mlp.gate_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.31.mlp.up_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.31.post_attention_layernorm.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.31.self_attn.k_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.31.self_attn.o_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.31.self_attn.q_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.31.self_attn.v_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.4.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.4.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.4.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.4.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.4.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.4.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.4.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.4.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.4.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.5.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.5.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.5.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.5.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.5.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.5.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.5.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.5.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.5.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.6.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.6.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.6.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.6.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.6.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.6.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.6.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.6.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.6.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.7.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.7.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.7.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.7.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.7.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.7.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.7.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.7.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.7.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.8.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.8.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.8.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.8.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.8.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.8.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.8.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.8.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.8.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.9.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.9.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.9.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.9.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.9.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.9.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.9.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.9.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.9.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.norm.weight": "pytorch_model-00002-of-00002.bin"
+  }
+}

checkpoint-160/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3f772f6ccbcbbefdc206f1ef8cd2122f95829c41904e09540b951c4e4a2b54d6
+size 5947

checkpoint-170/config.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:44e972929ad46908648b92c2feb180fb201dab6bdc232a5bc5776a74d47e0e0a
+size 620

checkpoint-170/generation_config.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:afcf2a1da583dbeb66bb367a759713901e23ce19ed83ed8f5a21b0e877addd5b
+size 116

checkpoint-170/latest

@@ -0,0 +1 @@
+global_step170

checkpoint-170/pytorch_model-00001-of-00002.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:abfac96d19c169395b86f6cd25ddae0b6ff2087810d03ce501cd0efb17acee19
+size 9943077196

checkpoint-170/pytorch_model-00002-of-00002.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:00c61452ea892344730992c3985c6363656b5171a9a8a88da922d8ce4736e836
+size 4540584799

checkpoint-170/pytorch_model.bin.index.json

@@ -0,0 +1,298 @@
+{
+  "metadata": {
+    "total_size": 14483562496
+  },
+  "weight_map": {
+    "lm_head.weight": "pytorch_model-00002-of-00002.bin",
+    "model.embed_tokens.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.0.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.0.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.0.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.0.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.0.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.0.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.0.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.0.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.0.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.1.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.1.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.1.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.1.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.1.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.1.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.1.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.1.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.1.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.10.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.10.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.10.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.10.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.10.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.10.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.10.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.10.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.10.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.11.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.11.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.11.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.11.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.11.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.11.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.11.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.11.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.11.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.12.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.12.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.12.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.12.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.12.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.12.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.12.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.12.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.12.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.13.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.13.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.13.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.13.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.13.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.13.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.13.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.13.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.13.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.14.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.14.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.14.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.14.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.14.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.14.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.14.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.14.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.14.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.15.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.15.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.15.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.15.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.15.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.15.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.15.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.15.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.15.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.16.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.16.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.16.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.16.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.16.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.16.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.16.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.16.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.16.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.17.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.17.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.17.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.17.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.17.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.17.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.17.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.17.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.17.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.18.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.18.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.18.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.18.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.18.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.18.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.18.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.18.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.18.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.19.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.19.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.19.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.19.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.19.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.19.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.19.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.19.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.19.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.2.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.2.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.2.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.2.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.2.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.2.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.2.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.2.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.2.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.20.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.20.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.20.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.20.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.20.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.20.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.20.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.20.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.20.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.21.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.21.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.21.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.21.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.21.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.21.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.21.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.21.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.21.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.22.input_layernorm.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.22.mlp.down_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.22.mlp.gate_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.22.mlp.up_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.22.post_attention_layernorm.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.22.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.22.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.22.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.22.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.23.input_layernorm.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.23.mlp.down_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.23.mlp.gate_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.23.mlp.up_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.23.post_attention_layernorm.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.23.self_attn.k_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.23.self_attn.o_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.23.self_attn.q_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.23.self_attn.v_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.24.input_layernorm.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.24.mlp.down_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.24.mlp.gate_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.24.mlp.up_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.24.post_attention_layernorm.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.24.self_attn.k_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.24.self_attn.o_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.24.self_attn.q_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.24.self_attn.v_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.25.input_layernorm.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.25.mlp.down_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.25.mlp.gate_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.25.mlp.up_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.25.post_attention_layernorm.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.25.self_attn.k_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.25.self_attn.o_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.25.self_attn.q_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.25.self_attn.v_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.26.input_layernorm.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.26.mlp.down_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.26.mlp.gate_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.26.mlp.up_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.26.post_attention_layernorm.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.26.self_attn.k_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.26.self_attn.o_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.26.self_attn.q_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.26.self_attn.v_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.27.input_layernorm.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.27.mlp.down_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.27.mlp.gate_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.27.mlp.up_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.27.post_attention_layernorm.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.27.self_attn.k_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.27.self_attn.o_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.27.self_attn.q_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.27.self_attn.v_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.28.input_layernorm.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.28.mlp.down_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.28.mlp.gate_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.28.mlp.up_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.28.post_attention_layernorm.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.28.self_attn.k_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.28.self_attn.o_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.28.self_attn.q_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.28.self_attn.v_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.29.input_layernorm.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.29.mlp.down_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.29.mlp.gate_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.29.mlp.up_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.29.post_attention_layernorm.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.29.self_attn.k_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.29.self_attn.o_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.29.self_attn.q_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.29.self_attn.v_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.3.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.3.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.3.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.3.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.3.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.3.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.3.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.3.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.3.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.30.input_layernorm.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.30.mlp.down_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.30.mlp.gate_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.30.mlp.up_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.30.post_attention_layernorm.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.30.self_attn.k_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.30.self_attn.o_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.30.self_attn.q_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.30.self_attn.v_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.31.input_layernorm.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.31.mlp.down_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.31.mlp.gate_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.31.mlp.up_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.31.post_attention_layernorm.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.31.self_attn.k_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.31.self_attn.o_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.31.self_attn.q_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.31.self_attn.v_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.4.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.4.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.4.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.4.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.4.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.4.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.4.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.4.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.4.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.5.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.5.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.5.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.5.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.5.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.5.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.5.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.5.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.5.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.6.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.6.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.6.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.6.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.6.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.6.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.6.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.6.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.6.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.7.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.7.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.7.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.7.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.7.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.7.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.7.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.7.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.7.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.8.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.8.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.8.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.8.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.8.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.8.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.8.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.8.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.8.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.9.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.9.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.9.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.9.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.9.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.9.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.9.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.9.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.9.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.norm.weight": "pytorch_model-00002-of-00002.bin"
+  }
+}

checkpoint-170/rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:db779cf9587a6d610fb5a77992eaff5b7023434a619f30e062d99dea1ff2918c
+size 17655

checkpoint-170/rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6e811b297222dd684a3ed489577251e81e4ee09e43d649c972861259cbc5b070
+size 17655

checkpoint-170/rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ae7f943d9c364bdc8d9a868e3c9a8a1d71d15aeaa438fb1461daa8ad5833d4ee
+size 17655

checkpoint-170/rng_state_3.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d72265887dfe6344a4ca37a0c48b3dfa9066a8410d2cb192dc651b3ad539316e
+size 17655

checkpoint-170/trainer_state.json

@@ -0,0 +1,1039 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.0357142857142856,
+  "eval_steps": 250,
+  "global_step": 170,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0,
+      "loss": 1.3708,
+      "step": 1
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 6.000000000000001e-07,
+      "loss": 1.0095,
+      "step": 2
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.2000000000000002e-06,
+      "loss": 1.033,
+      "step": 3
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.8e-06,
+      "loss": 0.919,
+      "step": 4
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.4000000000000003e-06,
+      "loss": 0.8426,
+      "step": 5
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 3e-06,
+      "loss": 0.7499,
+      "step": 6
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.6e-06,
+      "loss": 0.7977,
+      "step": 7
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.2e-06,
+      "loss": 0.7178,
+      "step": 8
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.800000000000001e-06,
+      "loss": 0.721,
+      "step": 9
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 5.4e-06,
+      "loss": 0.686,
+      "step": 10
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 6e-06,
+      "loss": 0.7123,
+      "step": 11
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 5.9719626168224305e-06,
+      "loss": 0.6778,
+      "step": 12
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 5.94392523364486e-06,
+      "loss": 0.6724,
+      "step": 13
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 5.9158878504672904e-06,
+      "loss": 0.6661,
+      "step": 14
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 5.88785046728972e-06,
+      "loss": 0.6647,
+      "step": 15
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 5.85981308411215e-06,
+      "loss": 0.6411,
+      "step": 16
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 5.83177570093458e-06,
+      "loss": 0.6361,
+      "step": 17
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 5.8037383177570094e-06,
+      "loss": 0.6205,
+      "step": 18
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 5.775700934579439e-06,
+      "loss": 0.6062,
+      "step": 19
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 5.747663551401869e-06,
+      "loss": 0.6578,
+      "step": 20
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 5.719626168224299e-06,
+      "loss": 0.6056,
+      "step": 21
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 5.691588785046729e-06,
+      "loss": 0.6033,
+      "step": 22
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 5.663551401869159e-06,
+      "loss": 0.618,
+      "step": 23
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 5.635514018691589e-06,
+      "loss": 0.6132,
+      "step": 24
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.607476635514019e-06,
+      "loss": 0.6046,
+      "step": 25
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.579439252336449e-06,
+      "loss": 0.6043,
+      "step": 26
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.551401869158879e-06,
+      "loss": 0.552,
+      "step": 27
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.523364485981309e-06,
+      "loss": 0.5772,
+      "step": 28
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5.4953271028037385e-06,
+      "loss": 0.5818,
+      "step": 29
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 5.467289719626169e-06,
+      "loss": 0.5933,
+      "step": 30
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 5.4392523364485985e-06,
+      "loss": 0.5727,
+      "step": 31
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 5.411214953271029e-06,
+      "loss": 0.5911,
+      "step": 32
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 5.383177570093458e-06,
+      "loss": 0.5568,
+      "step": 33
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 5.355140186915888e-06,
+      "loss": 0.5238,
+      "step": 34
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 5.3271028037383174e-06,
+      "loss": 0.5666,
+      "step": 35
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.299065420560748e-06,
+      "loss": 0.5509,
+      "step": 36
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.271028037383177e-06,
+      "loss": 0.5887,
+      "step": 37
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.242990654205608e-06,
+      "loss": 0.5499,
+      "step": 38
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 5.214953271028037e-06,
+      "loss": 0.5745,
+      "step": 39
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.186915887850468e-06,
+      "loss": 0.5301,
+      "step": 40
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.158878504672897e-06,
+      "loss": 0.556,
+      "step": 41
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.1308411214953276e-06,
+      "loss": 0.5634,
+      "step": 42
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.102803738317757e-06,
+      "loss": 0.5457,
+      "step": 43
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.0747663551401875e-06,
+      "loss": 0.5901,
+      "step": 44
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.046728971962617e-06,
+      "loss": 0.5568,
+      "step": 45
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 5.018691588785047e-06,
+      "loss": 0.5588,
+      "step": 46
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 4.990654205607477e-06,
+      "loss": 0.5605,
+      "step": 47
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.962616822429907e-06,
+      "loss": 0.5529,
+      "step": 48
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.934579439252337e-06,
+      "loss": 0.5553,
+      "step": 49
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 4.906542056074766e-06,
+      "loss": 0.5185,
+      "step": 50
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.878504672897196e-06,
+      "loss": 0.5436,
+      "step": 51
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 4.850467289719626e-06,
+      "loss": 0.529,
+      "step": 52
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 4.822429906542056e-06,
+      "loss": 0.5315,
+      "step": 53
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.794392523364486e-06,
+      "loss": 0.5261,
+      "step": 54
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.766355140186916e-06,
+      "loss": 0.5634,
+      "step": 55
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.738317757009346e-06,
+      "loss": 0.5225,
+      "step": 56
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 4.710280373831776e-06,
+      "loss": 0.4673,
+      "step": 57
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 4.682242990654206e-06,
+      "loss": 0.5062,
+      "step": 58
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 4.6542056074766356e-06,
+      "loss": 0.4767,
+      "step": 59
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 4.626168224299066e-06,
+      "loss": 0.4665,
+      "step": 60
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 4.5981308411214955e-06,
+      "loss": 0.4807,
+      "step": 61
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 4.570093457943926e-06,
+      "loss": 0.4797,
+      "step": 62
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 4.542056074766355e-06,
+      "loss": 0.445,
+      "step": 63
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 4.514018691588785e-06,
+      "loss": 0.4966,
+      "step": 64
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 4.485981308411215e-06,
+      "loss": 0.4892,
+      "step": 65
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 4.457943925233645e-06,
+      "loss": 0.5169,
+      "step": 66
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 4.429906542056074e-06,
+      "loss": 0.4722,
+      "step": 67
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 4.401869158878505e-06,
+      "loss": 0.4431,
+      "step": 68
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 4.373831775700934e-06,
+      "loss": 0.4756,
+      "step": 69
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 4.345794392523365e-06,
+      "loss": 0.4779,
+      "step": 70
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 4.317757009345794e-06,
+      "loss": 0.4775,
+      "step": 71
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 4.289719626168225e-06,
+      "loss": 0.4802,
+      "step": 72
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 4.261682242990654e-06,
+      "loss": 0.4698,
+      "step": 73
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 4.2336448598130845e-06,
+      "loss": 0.4514,
+      "step": 74
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 4.205607476635514e-06,
+      "loss": 0.4423,
+      "step": 75
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.177570093457944e-06,
+      "loss": 0.4484,
+      "step": 76
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.149532710280374e-06,
+      "loss": 0.4898,
+      "step": 77
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.121495327102804e-06,
+      "loss": 0.4627,
+      "step": 78
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.093457943925234e-06,
+      "loss": 0.4528,
+      "step": 79
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 4.065420560747663e-06,
+      "loss": 0.4737,
+      "step": 80
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 4.037383177570093e-06,
+      "loss": 0.4742,
+      "step": 81
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 4.009345794392523e-06,
+      "loss": 0.4621,
+      "step": 82
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.981308411214953e-06,
+      "loss": 0.4683,
+      "step": 83
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.953271028037383e-06,
+      "loss": 0.4616,
+      "step": 84
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 3.925233644859813e-06,
+      "loss": 0.4376,
+      "step": 85
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 3.897196261682243e-06,
+      "loss": 0.4539,
+      "step": 86
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 3.869158878504673e-06,
+      "loss": 0.4366,
+      "step": 87
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 3.841121495327103e-06,
+      "loss": 0.434,
+      "step": 88
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 3.8130841121495326e-06,
+      "loss": 0.4848,
+      "step": 89
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 3.785046728971963e-06,
+      "loss": 0.4676,
+      "step": 90
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.7570093457943925e-06,
+      "loss": 0.4743,
+      "step": 91
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.7289719626168225e-06,
+      "loss": 0.4555,
+      "step": 92
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.700934579439252e-06,
+      "loss": 0.4616,
+      "step": 93
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.6728971962616824e-06,
+      "loss": 0.4527,
+      "step": 94
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.644859813084112e-06,
+      "loss": 0.4281,
+      "step": 95
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 3.6168224299065423e-06,
+      "loss": 0.4759,
+      "step": 96
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 3.588785046728972e-06,
+      "loss": 0.4906,
+      "step": 97
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 3.560747663551402e-06,
+      "loss": 0.485,
+      "step": 98
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 3.5327102803738317e-06,
+      "loss": 0.4704,
+      "step": 99
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 3.5046728971962617e-06,
+      "loss": 0.4604,
+      "step": 100
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 3.4766355140186912e-06,
+      "loss": 0.4616,
+      "step": 101
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 3.4485981308411216e-06,
+      "loss": 0.4635,
+      "step": 102
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.420560747663551e-06,
+      "loss": 0.4468,
+      "step": 103
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 3.3925233644859815e-06,
+      "loss": 0.4683,
+      "step": 104
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 3.364485981308411e-06,
+      "loss": 0.4622,
+      "step": 105
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 3.3364485981308414e-06,
+      "loss": 0.4897,
+      "step": 106
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 3.308411214953271e-06,
+      "loss": 0.4764,
+      "step": 107
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 3.280373831775701e-06,
+      "loss": 0.4696,
+      "step": 108
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 3.2523364485981305e-06,
+      "loss": 0.4531,
+      "step": 109
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 3.224299065420561e-06,
+      "loss": 0.485,
+      "step": 110
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 3.1962616822429904e-06,
+      "loss": 0.449,
+      "step": 111
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 3.1682242990654208e-06,
+      "loss": 0.4156,
+      "step": 112
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 3.1401869158878503e-06,
+      "loss": 0.4264,
+      "step": 113
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 3.1121495327102807e-06,
+      "loss": 0.4018,
+      "step": 114
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 3.08411214953271e-06,
+      "loss": 0.3894,
+      "step": 115
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 3.05607476635514e-06,
+      "loss": 0.3905,
+      "step": 116
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 3.0280373831775697e-06,
+      "loss": 0.4005,
+      "step": 117
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 3e-06,
+      "loss": 0.4057,
+      "step": 118
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 2.97196261682243e-06,
+      "loss": 0.4114,
+      "step": 119
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 2.94392523364486e-06,
+      "loss": 0.3942,
+      "step": 120
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 2.91588785046729e-06,
+      "loss": 0.3643,
+      "step": 121
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 2.8878504672897195e-06,
+      "loss": 0.3757,
+      "step": 122
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 2.8598130841121494e-06,
+      "loss": 0.3906,
+      "step": 123
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 2.8317757009345794e-06,
+      "loss": 0.4167,
+      "step": 124
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 2.8037383177570094e-06,
+      "loss": 0.3954,
+      "step": 125
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.7757009345794393e-06,
+      "loss": 0.4028,
+      "step": 126
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.7476635514018693e-06,
+      "loss": 0.3882,
+      "step": 127
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.7196261682242992e-06,
+      "loss": 0.4066,
+      "step": 128
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.691588785046729e-06,
+      "loss": 0.3983,
+      "step": 129
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.6635514018691587e-06,
+      "loss": 0.4104,
+      "step": 130
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.6355140186915887e-06,
+      "loss": 0.3803,
+      "step": 131
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.6074766355140186e-06,
+      "loss": 0.3948,
+      "step": 132
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.5794392523364486e-06,
+      "loss": 0.3896,
+      "step": 133
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.5514018691588785e-06,
+      "loss": 0.3839,
+      "step": 134
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.5233644859813085e-06,
+      "loss": 0.4102,
+      "step": 135
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.4953271028037385e-06,
+      "loss": 0.3917,
+      "step": 136
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.4672897196261684e-06,
+      "loss": 0.4054,
+      "step": 137
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.439252336448598e-06,
+      "loss": 0.3944,
+      "step": 138
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.411214953271028e-06,
+      "loss": 0.3979,
+      "step": 139
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.383177570093458e-06,
+      "loss": 0.3756,
+      "step": 140
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.355140186915888e-06,
+      "loss": 0.3889,
+      "step": 141
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.3271028037383178e-06,
+      "loss": 0.3918,
+      "step": 142
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.2990654205607477e-06,
+      "loss": 0.3882,
+      "step": 143
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.2710280373831777e-06,
+      "loss": 0.391,
+      "step": 144
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.2429906542056077e-06,
+      "loss": 0.4049,
+      "step": 145
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.214953271028037e-06,
+      "loss": 0.3966,
+      "step": 146
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.186915887850467e-06,
+      "loss": 0.399,
+      "step": 147
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.158878504672897e-06,
+      "loss": 0.3811,
+      "step": 148
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.130841121495327e-06,
+      "loss": 0.3883,
+      "step": 149
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.102803738317757e-06,
+      "loss": 0.3651,
+      "step": 150
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.074766355140187e-06,
+      "loss": 0.387,
+      "step": 151
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 2.046728971962617e-06,
+      "loss": 0.3863,
+      "step": 152
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 2.0186915887850465e-06,
+      "loss": 0.389,
+      "step": 153
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.9906542056074764e-06,
+      "loss": 0.3871,
+      "step": 154
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.9626168224299064e-06,
+      "loss": 0.3777,
+      "step": 155
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 1.9345794392523363e-06,
+      "loss": 0.3974,
+      "step": 156
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 1.9065420560747663e-06,
+      "loss": 0.409,
+      "step": 157
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.8785046728971962e-06,
+      "loss": 0.3907,
+      "step": 158
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.850467289719626e-06,
+      "loss": 0.3903,
+      "step": 159
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.822429906542056e-06,
+      "loss": 0.3808,
+      "step": 160
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 1.794392523364486e-06,
+      "loss": 0.3995,
+      "step": 161
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 1.7663551401869159e-06,
+      "loss": 0.4188,
+      "step": 162
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 1.7383177570093456e-06,
+      "loss": 0.3734,
+      "step": 163
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 1.7102803738317756e-06,
+      "loss": 0.3829,
+      "step": 164
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.6822429906542055e-06,
+      "loss": 0.4018,
+      "step": 165
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.6542056074766355e-06,
+      "loss": 0.3687,
+      "step": 166
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.6261682242990652e-06,
+      "loss": 0.394,
+      "step": 167
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 1.5981308411214952e-06,
+      "loss": 0.3903,
+      "step": 168
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 1.5700934579439251e-06,
+      "loss": 0.3535,
+      "step": 169
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 1.542056074766355e-06,
+      "loss": 0.3492,
+      "step": 170
+    }
+  ],
+  "logging_steps": 1,
+  "max_steps": 224,
+  "num_train_epochs": 4,
+  "save_steps": 10,
+  "total_flos": 1.4259778418324275e+18,
+  "trial_name": null,
+  "trial_params": null
+}

checkpoint-170/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3f772f6ccbcbbefdc206f1ef8cd2122f95829c41904e09540b951c4e4a2b54d6
+size 5947

checkpoint-170/zero_to_fp32.py

@@ -0,0 +1,587 @@
+#!/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):
+    """
+    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)
+    elif zero_stage == 3:
+        return _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states)
+
+
+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 _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()
+            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):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    _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):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    _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):
+    """
+    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``
+
+    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)
+
+
+def convert_zero_checkpoint_to_fp32_state_dict(checkpoint_dir, output_file, tag=None):
+    """
+    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``
+    """
+
+    state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag)
+    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("-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)

checkpoint-220/config.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:44e972929ad46908648b92c2feb180fb201dab6bdc232a5bc5776a74d47e0e0a
+size 620

checkpoint-220/generation_config.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:afcf2a1da583dbeb66bb367a759713901e23ce19ed83ed8f5a21b0e877addd5b
+size 116

checkpoint-220/latest

@@ -0,0 +1 @@
+global_step220

checkpoint-220/pytorch_model-00001-of-00002.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1171e5f39f71c8486edb47cb2832409a5c3e362a605b9018951a59ee44cce2dc
+size 9943077196

checkpoint-220/pytorch_model-00002-of-00002.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:49524e486aa7327a0fbc5339a8d44df4a39bb31a9723f5367f8a53448d80c3f0
+size 4540584799

checkpoint-220/pytorch_model.bin.index.json

@@ -0,0 +1,298 @@
+{
+  "metadata": {
+    "total_size": 14483562496
+  },
+  "weight_map": {
+    "lm_head.weight": "pytorch_model-00002-of-00002.bin",
+    "model.embed_tokens.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.0.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.0.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.0.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.0.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.0.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.0.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.0.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.0.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.0.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.1.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.1.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.1.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.1.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.1.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.1.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.1.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.1.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.1.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.10.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.10.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.10.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.10.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.10.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.10.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.10.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.10.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.10.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.11.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.11.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.11.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.11.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.11.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.11.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.11.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.11.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.11.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.12.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.12.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.12.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.12.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.12.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.12.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.12.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.12.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.12.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.13.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.13.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.13.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.13.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.13.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.13.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.13.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.13.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.13.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.14.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.14.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.14.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.14.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.14.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.14.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.14.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.14.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.14.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.15.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.15.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.15.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.15.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.15.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.15.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.15.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.15.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.15.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.16.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.16.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.16.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.16.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.16.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.16.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.16.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.16.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.16.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.17.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.17.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.17.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.17.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.17.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.17.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.17.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.17.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.17.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.18.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.18.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.18.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.18.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.18.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.18.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.18.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.18.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.18.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.19.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.19.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.19.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.19.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.19.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.19.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.19.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.19.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.19.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.2.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.2.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.2.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.2.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.2.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.2.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.2.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.2.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.2.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.20.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.20.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.20.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.20.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.20.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.20.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.20.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.20.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.20.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.21.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.21.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.21.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.21.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.21.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.21.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.21.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.21.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.21.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.22.input_layernorm.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.22.mlp.down_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.22.mlp.gate_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.22.mlp.up_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.22.post_attention_layernorm.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.22.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.22.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.22.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.22.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.23.input_layernorm.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.23.mlp.down_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.23.mlp.gate_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.23.mlp.up_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.23.post_attention_layernorm.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.23.self_attn.k_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.23.self_attn.o_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.23.self_attn.q_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.23.self_attn.v_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.24.input_layernorm.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.24.mlp.down_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.24.mlp.gate_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.24.mlp.up_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.24.post_attention_layernorm.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.24.self_attn.k_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.24.self_attn.o_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.24.self_attn.q_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.24.self_attn.v_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.25.input_layernorm.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.25.mlp.down_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.25.mlp.gate_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.25.mlp.up_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.25.post_attention_layernorm.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.25.self_attn.k_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.25.self_attn.o_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.25.self_attn.q_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.25.self_attn.v_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.26.input_layernorm.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.26.mlp.down_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.26.mlp.gate_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.26.mlp.up_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.26.post_attention_layernorm.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.26.self_attn.k_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.26.self_attn.o_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.26.self_attn.q_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.26.self_attn.v_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.27.input_layernorm.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.27.mlp.down_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.27.mlp.gate_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.27.mlp.up_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.27.post_attention_layernorm.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.27.self_attn.k_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.27.self_attn.o_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.27.self_attn.q_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.27.self_attn.v_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.28.input_layernorm.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.28.mlp.down_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.28.mlp.gate_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.28.mlp.up_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.28.post_attention_layernorm.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.28.self_attn.k_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.28.self_attn.o_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.28.self_attn.q_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.28.self_attn.v_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.29.input_layernorm.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.29.mlp.down_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.29.mlp.gate_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.29.mlp.up_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.29.post_attention_layernorm.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.29.self_attn.k_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.29.self_attn.o_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.29.self_attn.q_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.29.self_attn.v_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.3.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.3.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.3.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.3.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.3.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.3.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.3.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.3.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.3.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.30.input_layernorm.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.30.mlp.down_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.30.mlp.gate_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.30.mlp.up_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.30.post_attention_layernorm.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.30.self_attn.k_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.30.self_attn.o_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.30.self_attn.q_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.30.self_attn.v_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.31.input_layernorm.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.31.mlp.down_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.31.mlp.gate_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.31.mlp.up_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.31.post_attention_layernorm.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.31.self_attn.k_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.31.self_attn.o_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.31.self_attn.q_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.31.self_attn.v_proj.weight": "pytorch_model-00002-of-00002.bin",
+    "model.layers.4.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.4.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.4.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.4.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.4.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.4.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.4.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.4.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.4.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.5.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.5.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.5.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.5.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.5.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.5.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.5.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.5.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.5.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.6.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.6.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.6.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.6.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.6.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.6.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.6.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.6.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.6.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.7.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.7.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.7.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.7.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.7.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.7.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.7.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.7.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.7.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.8.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.8.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.8.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.8.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.8.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.8.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.8.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.8.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.8.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.9.input_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.9.mlp.down_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.9.mlp.gate_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.9.mlp.up_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.9.post_attention_layernorm.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.9.self_attn.k_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.9.self_attn.o_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.9.self_attn.q_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.layers.9.self_attn.v_proj.weight": "pytorch_model-00001-of-00002.bin",
+    "model.norm.weight": "pytorch_model-00002-of-00002.bin"
+  }
+}

checkpoint-220/rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:db779cf9587a6d610fb5a77992eaff5b7023434a619f30e062d99dea1ff2918c
+size 17655

checkpoint-220/rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6e811b297222dd684a3ed489577251e81e4ee09e43d649c972861259cbc5b070
+size 17655

checkpoint-220/rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ae7f943d9c364bdc8d9a868e3c9a8a1d71d15aeaa438fb1461daa8ad5833d4ee
+size 17655

checkpoint-220/rng_state_3.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d72265887dfe6344a4ca37a0c48b3dfa9066a8410d2cb192dc651b3ad539316e
+size 17655

checkpoint-220/trainer_state.json

@@ -0,0 +1,1339 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 3.928571428571429,
+  "eval_steps": 250,
+  "global_step": 220,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.02,
+      "learning_rate": 0.0,
+      "loss": 1.3708,
+      "step": 1
+    },
+    {
+      "epoch": 0.04,
+      "learning_rate": 6.000000000000001e-07,
+      "loss": 1.0095,
+      "step": 2
+    },
+    {
+      "epoch": 0.05,
+      "learning_rate": 1.2000000000000002e-06,
+      "loss": 1.033,
+      "step": 3
+    },
+    {
+      "epoch": 0.07,
+      "learning_rate": 1.8e-06,
+      "loss": 0.919,
+      "step": 4
+    },
+    {
+      "epoch": 0.09,
+      "learning_rate": 2.4000000000000003e-06,
+      "loss": 0.8426,
+      "step": 5
+    },
+    {
+      "epoch": 0.11,
+      "learning_rate": 3e-06,
+      "loss": 0.7499,
+      "step": 6
+    },
+    {
+      "epoch": 0.12,
+      "learning_rate": 3.6e-06,
+      "loss": 0.7977,
+      "step": 7
+    },
+    {
+      "epoch": 0.14,
+      "learning_rate": 4.2e-06,
+      "loss": 0.7178,
+      "step": 8
+    },
+    {
+      "epoch": 0.16,
+      "learning_rate": 4.800000000000001e-06,
+      "loss": 0.721,
+      "step": 9
+    },
+    {
+      "epoch": 0.18,
+      "learning_rate": 5.4e-06,
+      "loss": 0.686,
+      "step": 10
+    },
+    {
+      "epoch": 0.2,
+      "learning_rate": 6e-06,
+      "loss": 0.7123,
+      "step": 11
+    },
+    {
+      "epoch": 0.21,
+      "learning_rate": 5.9719626168224305e-06,
+      "loss": 0.6778,
+      "step": 12
+    },
+    {
+      "epoch": 0.23,
+      "learning_rate": 5.94392523364486e-06,
+      "loss": 0.6724,
+      "step": 13
+    },
+    {
+      "epoch": 0.25,
+      "learning_rate": 5.9158878504672904e-06,
+      "loss": 0.6661,
+      "step": 14
+    },
+    {
+      "epoch": 0.27,
+      "learning_rate": 5.88785046728972e-06,
+      "loss": 0.6647,
+      "step": 15
+    },
+    {
+      "epoch": 0.29,
+      "learning_rate": 5.85981308411215e-06,
+      "loss": 0.6411,
+      "step": 16
+    },
+    {
+      "epoch": 0.3,
+      "learning_rate": 5.83177570093458e-06,
+      "loss": 0.6361,
+      "step": 17
+    },
+    {
+      "epoch": 0.32,
+      "learning_rate": 5.8037383177570094e-06,
+      "loss": 0.6205,
+      "step": 18
+    },
+    {
+      "epoch": 0.34,
+      "learning_rate": 5.775700934579439e-06,
+      "loss": 0.6062,
+      "step": 19
+    },
+    {
+      "epoch": 0.36,
+      "learning_rate": 5.747663551401869e-06,
+      "loss": 0.6578,
+      "step": 20
+    },
+    {
+      "epoch": 0.38,
+      "learning_rate": 5.719626168224299e-06,
+      "loss": 0.6056,
+      "step": 21
+    },
+    {
+      "epoch": 0.39,
+      "learning_rate": 5.691588785046729e-06,
+      "loss": 0.6033,
+      "step": 22
+    },
+    {
+      "epoch": 0.41,
+      "learning_rate": 5.663551401869159e-06,
+      "loss": 0.618,
+      "step": 23
+    },
+    {
+      "epoch": 0.43,
+      "learning_rate": 5.635514018691589e-06,
+      "loss": 0.6132,
+      "step": 24
+    },
+    {
+      "epoch": 0.45,
+      "learning_rate": 5.607476635514019e-06,
+      "loss": 0.6046,
+      "step": 25
+    },
+    {
+      "epoch": 0.46,
+      "learning_rate": 5.579439252336449e-06,
+      "loss": 0.6043,
+      "step": 26
+    },
+    {
+      "epoch": 0.48,
+      "learning_rate": 5.551401869158879e-06,
+      "loss": 0.552,
+      "step": 27
+    },
+    {
+      "epoch": 0.5,
+      "learning_rate": 5.523364485981309e-06,
+      "loss": 0.5772,
+      "step": 28
+    },
+    {
+      "epoch": 0.52,
+      "learning_rate": 5.4953271028037385e-06,
+      "loss": 0.5818,
+      "step": 29
+    },
+    {
+      "epoch": 0.54,
+      "learning_rate": 5.467289719626169e-06,
+      "loss": 0.5933,
+      "step": 30
+    },
+    {
+      "epoch": 0.55,
+      "learning_rate": 5.4392523364485985e-06,
+      "loss": 0.5727,
+      "step": 31
+    },
+    {
+      "epoch": 0.57,
+      "learning_rate": 5.411214953271029e-06,
+      "loss": 0.5911,
+      "step": 32
+    },
+    {
+      "epoch": 0.59,
+      "learning_rate": 5.383177570093458e-06,
+      "loss": 0.5568,
+      "step": 33
+    },
+    {
+      "epoch": 0.61,
+      "learning_rate": 5.355140186915888e-06,
+      "loss": 0.5238,
+      "step": 34
+    },
+    {
+      "epoch": 0.62,
+      "learning_rate": 5.3271028037383174e-06,
+      "loss": 0.5666,
+      "step": 35
+    },
+    {
+      "epoch": 0.64,
+      "learning_rate": 5.299065420560748e-06,
+      "loss": 0.5509,
+      "step": 36
+    },
+    {
+      "epoch": 0.66,
+      "learning_rate": 5.271028037383177e-06,
+      "loss": 0.5887,
+      "step": 37
+    },
+    {
+      "epoch": 0.68,
+      "learning_rate": 5.242990654205608e-06,
+      "loss": 0.5499,
+      "step": 38
+    },
+    {
+      "epoch": 0.7,
+      "learning_rate": 5.214953271028037e-06,
+      "loss": 0.5745,
+      "step": 39
+    },
+    {
+      "epoch": 0.71,
+      "learning_rate": 5.186915887850468e-06,
+      "loss": 0.5301,
+      "step": 40
+    },
+    {
+      "epoch": 0.73,
+      "learning_rate": 5.158878504672897e-06,
+      "loss": 0.556,
+      "step": 41
+    },
+    {
+      "epoch": 0.75,
+      "learning_rate": 5.1308411214953276e-06,
+      "loss": 0.5634,
+      "step": 42
+    },
+    {
+      "epoch": 0.77,
+      "learning_rate": 5.102803738317757e-06,
+      "loss": 0.5457,
+      "step": 43
+    },
+    {
+      "epoch": 0.79,
+      "learning_rate": 5.0747663551401875e-06,
+      "loss": 0.5901,
+      "step": 44
+    },
+    {
+      "epoch": 0.8,
+      "learning_rate": 5.046728971962617e-06,
+      "loss": 0.5568,
+      "step": 45
+    },
+    {
+      "epoch": 0.82,
+      "learning_rate": 5.018691588785047e-06,
+      "loss": 0.5588,
+      "step": 46
+    },
+    {
+      "epoch": 0.84,
+      "learning_rate": 4.990654205607477e-06,
+      "loss": 0.5605,
+      "step": 47
+    },
+    {
+      "epoch": 0.86,
+      "learning_rate": 4.962616822429907e-06,
+      "loss": 0.5529,
+      "step": 48
+    },
+    {
+      "epoch": 0.88,
+      "learning_rate": 4.934579439252337e-06,
+      "loss": 0.5553,
+      "step": 49
+    },
+    {
+      "epoch": 0.89,
+      "learning_rate": 4.906542056074766e-06,
+      "loss": 0.5185,
+      "step": 50
+    },
+    {
+      "epoch": 0.91,
+      "learning_rate": 4.878504672897196e-06,
+      "loss": 0.5436,
+      "step": 51
+    },
+    {
+      "epoch": 0.93,
+      "learning_rate": 4.850467289719626e-06,
+      "loss": 0.529,
+      "step": 52
+    },
+    {
+      "epoch": 0.95,
+      "learning_rate": 4.822429906542056e-06,
+      "loss": 0.5315,
+      "step": 53
+    },
+    {
+      "epoch": 0.96,
+      "learning_rate": 4.794392523364486e-06,
+      "loss": 0.5261,
+      "step": 54
+    },
+    {
+      "epoch": 0.98,
+      "learning_rate": 4.766355140186916e-06,
+      "loss": 0.5634,
+      "step": 55
+    },
+    {
+      "epoch": 1.0,
+      "learning_rate": 4.738317757009346e-06,
+      "loss": 0.5225,
+      "step": 56
+    },
+    {
+      "epoch": 1.02,
+      "learning_rate": 4.710280373831776e-06,
+      "loss": 0.4673,
+      "step": 57
+    },
+    {
+      "epoch": 1.04,
+      "learning_rate": 4.682242990654206e-06,
+      "loss": 0.5062,
+      "step": 58
+    },
+    {
+      "epoch": 1.05,
+      "learning_rate": 4.6542056074766356e-06,
+      "loss": 0.4767,
+      "step": 59
+    },
+    {
+      "epoch": 1.07,
+      "learning_rate": 4.626168224299066e-06,
+      "loss": 0.4665,
+      "step": 60
+    },
+    {
+      "epoch": 1.09,
+      "learning_rate": 4.5981308411214955e-06,
+      "loss": 0.4807,
+      "step": 61
+    },
+    {
+      "epoch": 1.11,
+      "learning_rate": 4.570093457943926e-06,
+      "loss": 0.4797,
+      "step": 62
+    },
+    {
+      "epoch": 1.12,
+      "learning_rate": 4.542056074766355e-06,
+      "loss": 0.445,
+      "step": 63
+    },
+    {
+      "epoch": 1.14,
+      "learning_rate": 4.514018691588785e-06,
+      "loss": 0.4966,
+      "step": 64
+    },
+    {
+      "epoch": 1.16,
+      "learning_rate": 4.485981308411215e-06,
+      "loss": 0.4892,
+      "step": 65
+    },
+    {
+      "epoch": 1.18,
+      "learning_rate": 4.457943925233645e-06,
+      "loss": 0.5169,
+      "step": 66
+    },
+    {
+      "epoch": 1.2,
+      "learning_rate": 4.429906542056074e-06,
+      "loss": 0.4722,
+      "step": 67
+    },
+    {
+      "epoch": 1.21,
+      "learning_rate": 4.401869158878505e-06,
+      "loss": 0.4431,
+      "step": 68
+    },
+    {
+      "epoch": 1.23,
+      "learning_rate": 4.373831775700934e-06,
+      "loss": 0.4756,
+      "step": 69
+    },
+    {
+      "epoch": 1.25,
+      "learning_rate": 4.345794392523365e-06,
+      "loss": 0.4779,
+      "step": 70
+    },
+    {
+      "epoch": 1.27,
+      "learning_rate": 4.317757009345794e-06,
+      "loss": 0.4775,
+      "step": 71
+    },
+    {
+      "epoch": 1.29,
+      "learning_rate": 4.289719626168225e-06,
+      "loss": 0.4802,
+      "step": 72
+    },
+    {
+      "epoch": 1.3,
+      "learning_rate": 4.261682242990654e-06,
+      "loss": 0.4698,
+      "step": 73
+    },
+    {
+      "epoch": 1.32,
+      "learning_rate": 4.2336448598130845e-06,
+      "loss": 0.4514,
+      "step": 74
+    },
+    {
+      "epoch": 1.34,
+      "learning_rate": 4.205607476635514e-06,
+      "loss": 0.4423,
+      "step": 75
+    },
+    {
+      "epoch": 1.36,
+      "learning_rate": 4.177570093457944e-06,
+      "loss": 0.4484,
+      "step": 76
+    },
+    {
+      "epoch": 1.38,
+      "learning_rate": 4.149532710280374e-06,
+      "loss": 0.4898,
+      "step": 77
+    },
+    {
+      "epoch": 1.39,
+      "learning_rate": 4.121495327102804e-06,
+      "loss": 0.4627,
+      "step": 78
+    },
+    {
+      "epoch": 1.41,
+      "learning_rate": 4.093457943925234e-06,
+      "loss": 0.4528,
+      "step": 79
+    },
+    {
+      "epoch": 1.43,
+      "learning_rate": 4.065420560747663e-06,
+      "loss": 0.4737,
+      "step": 80
+    },
+    {
+      "epoch": 1.45,
+      "learning_rate": 4.037383177570093e-06,
+      "loss": 0.4742,
+      "step": 81
+    },
+    {
+      "epoch": 1.46,
+      "learning_rate": 4.009345794392523e-06,
+      "loss": 0.4621,
+      "step": 82
+    },
+    {
+      "epoch": 1.48,
+      "learning_rate": 3.981308411214953e-06,
+      "loss": 0.4683,
+      "step": 83
+    },
+    {
+      "epoch": 1.5,
+      "learning_rate": 3.953271028037383e-06,
+      "loss": 0.4616,
+      "step": 84
+    },
+    {
+      "epoch": 1.52,
+      "learning_rate": 3.925233644859813e-06,
+      "loss": 0.4376,
+      "step": 85
+    },
+    {
+      "epoch": 1.54,
+      "learning_rate": 3.897196261682243e-06,
+      "loss": 0.4539,
+      "step": 86
+    },
+    {
+      "epoch": 1.55,
+      "learning_rate": 3.869158878504673e-06,
+      "loss": 0.4366,
+      "step": 87
+    },
+    {
+      "epoch": 1.57,
+      "learning_rate": 3.841121495327103e-06,
+      "loss": 0.434,
+      "step": 88
+    },
+    {
+      "epoch": 1.59,
+      "learning_rate": 3.8130841121495326e-06,
+      "loss": 0.4848,
+      "step": 89
+    },
+    {
+      "epoch": 1.61,
+      "learning_rate": 3.785046728971963e-06,
+      "loss": 0.4676,
+      "step": 90
+    },
+    {
+      "epoch": 1.62,
+      "learning_rate": 3.7570093457943925e-06,
+      "loss": 0.4743,
+      "step": 91
+    },
+    {
+      "epoch": 1.64,
+      "learning_rate": 3.7289719626168225e-06,
+      "loss": 0.4555,
+      "step": 92
+    },
+    {
+      "epoch": 1.66,
+      "learning_rate": 3.700934579439252e-06,
+      "loss": 0.4616,
+      "step": 93
+    },
+    {
+      "epoch": 1.68,
+      "learning_rate": 3.6728971962616824e-06,
+      "loss": 0.4527,
+      "step": 94
+    },
+    {
+      "epoch": 1.7,
+      "learning_rate": 3.644859813084112e-06,
+      "loss": 0.4281,
+      "step": 95
+    },
+    {
+      "epoch": 1.71,
+      "learning_rate": 3.6168224299065423e-06,
+      "loss": 0.4759,
+      "step": 96
+    },
+    {
+      "epoch": 1.73,
+      "learning_rate": 3.588785046728972e-06,
+      "loss": 0.4906,
+      "step": 97
+    },
+    {
+      "epoch": 1.75,
+      "learning_rate": 3.560747663551402e-06,
+      "loss": 0.485,
+      "step": 98
+    },
+    {
+      "epoch": 1.77,
+      "learning_rate": 3.5327102803738317e-06,
+      "loss": 0.4704,
+      "step": 99
+    },
+    {
+      "epoch": 1.79,
+      "learning_rate": 3.5046728971962617e-06,
+      "loss": 0.4604,
+      "step": 100
+    },
+    {
+      "epoch": 1.8,
+      "learning_rate": 3.4766355140186912e-06,
+      "loss": 0.4616,
+      "step": 101
+    },
+    {
+      "epoch": 1.82,
+      "learning_rate": 3.4485981308411216e-06,
+      "loss": 0.4635,
+      "step": 102
+    },
+    {
+      "epoch": 1.84,
+      "learning_rate": 3.420560747663551e-06,
+      "loss": 0.4468,
+      "step": 103
+    },
+    {
+      "epoch": 1.86,
+      "learning_rate": 3.3925233644859815e-06,
+      "loss": 0.4683,
+      "step": 104
+    },
+    {
+      "epoch": 1.88,
+      "learning_rate": 3.364485981308411e-06,
+      "loss": 0.4622,
+      "step": 105
+    },
+    {
+      "epoch": 1.89,
+      "learning_rate": 3.3364485981308414e-06,
+      "loss": 0.4897,
+      "step": 106
+    },
+    {
+      "epoch": 1.91,
+      "learning_rate": 3.308411214953271e-06,
+      "loss": 0.4764,
+      "step": 107
+    },
+    {
+      "epoch": 1.93,
+      "learning_rate": 3.280373831775701e-06,
+      "loss": 0.4696,
+      "step": 108
+    },
+    {
+      "epoch": 1.95,
+      "learning_rate": 3.2523364485981305e-06,
+      "loss": 0.4531,
+      "step": 109
+    },
+    {
+      "epoch": 1.96,
+      "learning_rate": 3.224299065420561e-06,
+      "loss": 0.485,
+      "step": 110
+    },
+    {
+      "epoch": 1.98,
+      "learning_rate": 3.1962616822429904e-06,
+      "loss": 0.449,
+      "step": 111
+    },
+    {
+      "epoch": 2.0,
+      "learning_rate": 3.1682242990654208e-06,
+      "loss": 0.4156,
+      "step": 112
+    },
+    {
+      "epoch": 2.02,
+      "learning_rate": 3.1401869158878503e-06,
+      "loss": 0.4264,
+      "step": 113
+    },
+    {
+      "epoch": 2.04,
+      "learning_rate": 3.1121495327102807e-06,
+      "loss": 0.4018,
+      "step": 114
+    },
+    {
+      "epoch": 2.05,
+      "learning_rate": 3.08411214953271e-06,
+      "loss": 0.3894,
+      "step": 115
+    },
+    {
+      "epoch": 2.07,
+      "learning_rate": 3.05607476635514e-06,
+      "loss": 0.3905,
+      "step": 116
+    },
+    {
+      "epoch": 2.09,
+      "learning_rate": 3.0280373831775697e-06,
+      "loss": 0.4005,
+      "step": 117
+    },
+    {
+      "epoch": 2.11,
+      "learning_rate": 3e-06,
+      "loss": 0.4057,
+      "step": 118
+    },
+    {
+      "epoch": 2.12,
+      "learning_rate": 2.97196261682243e-06,
+      "loss": 0.4114,
+      "step": 119
+    },
+    {
+      "epoch": 2.14,
+      "learning_rate": 2.94392523364486e-06,
+      "loss": 0.3942,
+      "step": 120
+    },
+    {
+      "epoch": 2.16,
+      "learning_rate": 2.91588785046729e-06,
+      "loss": 0.3643,
+      "step": 121
+    },
+    {
+      "epoch": 2.18,
+      "learning_rate": 2.8878504672897195e-06,
+      "loss": 0.3757,
+      "step": 122
+    },
+    {
+      "epoch": 2.2,
+      "learning_rate": 2.8598130841121494e-06,
+      "loss": 0.3906,
+      "step": 123
+    },
+    {
+      "epoch": 2.21,
+      "learning_rate": 2.8317757009345794e-06,
+      "loss": 0.4167,
+      "step": 124
+    },
+    {
+      "epoch": 2.23,
+      "learning_rate": 2.8037383177570094e-06,
+      "loss": 0.3954,
+      "step": 125
+    },
+    {
+      "epoch": 2.25,
+      "learning_rate": 2.7757009345794393e-06,
+      "loss": 0.4028,
+      "step": 126
+    },
+    {
+      "epoch": 2.27,
+      "learning_rate": 2.7476635514018693e-06,
+      "loss": 0.3882,
+      "step": 127
+    },
+    {
+      "epoch": 2.29,
+      "learning_rate": 2.7196261682242992e-06,
+      "loss": 0.4066,
+      "step": 128
+    },
+    {
+      "epoch": 2.3,
+      "learning_rate": 2.691588785046729e-06,
+      "loss": 0.3983,
+      "step": 129
+    },
+    {
+      "epoch": 2.32,
+      "learning_rate": 2.6635514018691587e-06,
+      "loss": 0.4104,
+      "step": 130
+    },
+    {
+      "epoch": 2.34,
+      "learning_rate": 2.6355140186915887e-06,
+      "loss": 0.3803,
+      "step": 131
+    },
+    {
+      "epoch": 2.36,
+      "learning_rate": 2.6074766355140186e-06,
+      "loss": 0.3948,
+      "step": 132
+    },
+    {
+      "epoch": 2.38,
+      "learning_rate": 2.5794392523364486e-06,
+      "loss": 0.3896,
+      "step": 133
+    },
+    {
+      "epoch": 2.39,
+      "learning_rate": 2.5514018691588785e-06,
+      "loss": 0.3839,
+      "step": 134
+    },
+    {
+      "epoch": 2.41,
+      "learning_rate": 2.5233644859813085e-06,
+      "loss": 0.4102,
+      "step": 135
+    },
+    {
+      "epoch": 2.43,
+      "learning_rate": 2.4953271028037385e-06,
+      "loss": 0.3917,
+      "step": 136
+    },
+    {
+      "epoch": 2.45,
+      "learning_rate": 2.4672897196261684e-06,
+      "loss": 0.4054,
+      "step": 137
+    },
+    {
+      "epoch": 2.46,
+      "learning_rate": 2.439252336448598e-06,
+      "loss": 0.3944,
+      "step": 138
+    },
+    {
+      "epoch": 2.48,
+      "learning_rate": 2.411214953271028e-06,
+      "loss": 0.3979,
+      "step": 139
+    },
+    {
+      "epoch": 2.5,
+      "learning_rate": 2.383177570093458e-06,
+      "loss": 0.3756,
+      "step": 140
+    },
+    {
+      "epoch": 2.52,
+      "learning_rate": 2.355140186915888e-06,
+      "loss": 0.3889,
+      "step": 141
+    },
+    {
+      "epoch": 2.54,
+      "learning_rate": 2.3271028037383178e-06,
+      "loss": 0.3918,
+      "step": 142
+    },
+    {
+      "epoch": 2.55,
+      "learning_rate": 2.2990654205607477e-06,
+      "loss": 0.3882,
+      "step": 143
+    },
+    {
+      "epoch": 2.57,
+      "learning_rate": 2.2710280373831777e-06,
+      "loss": 0.391,
+      "step": 144
+    },
+    {
+      "epoch": 2.59,
+      "learning_rate": 2.2429906542056077e-06,
+      "loss": 0.4049,
+      "step": 145
+    },
+    {
+      "epoch": 2.61,
+      "learning_rate": 2.214953271028037e-06,
+      "loss": 0.3966,
+      "step": 146
+    },
+    {
+      "epoch": 2.62,
+      "learning_rate": 2.186915887850467e-06,
+      "loss": 0.399,
+      "step": 147
+    },
+    {
+      "epoch": 2.64,
+      "learning_rate": 2.158878504672897e-06,
+      "loss": 0.3811,
+      "step": 148
+    },
+    {
+      "epoch": 2.66,
+      "learning_rate": 2.130841121495327e-06,
+      "loss": 0.3883,
+      "step": 149
+    },
+    {
+      "epoch": 2.68,
+      "learning_rate": 2.102803738317757e-06,
+      "loss": 0.3651,
+      "step": 150
+    },
+    {
+      "epoch": 2.7,
+      "learning_rate": 2.074766355140187e-06,
+      "loss": 0.387,
+      "step": 151
+    },
+    {
+      "epoch": 2.71,
+      "learning_rate": 2.046728971962617e-06,
+      "loss": 0.3863,
+      "step": 152
+    },
+    {
+      "epoch": 2.73,
+      "learning_rate": 2.0186915887850465e-06,
+      "loss": 0.389,
+      "step": 153
+    },
+    {
+      "epoch": 2.75,
+      "learning_rate": 1.9906542056074764e-06,
+      "loss": 0.3871,
+      "step": 154
+    },
+    {
+      "epoch": 2.77,
+      "learning_rate": 1.9626168224299064e-06,
+      "loss": 0.3777,
+      "step": 155
+    },
+    {
+      "epoch": 2.79,
+      "learning_rate": 1.9345794392523363e-06,
+      "loss": 0.3974,
+      "step": 156
+    },
+    {
+      "epoch": 2.8,
+      "learning_rate": 1.9065420560747663e-06,
+      "loss": 0.409,
+      "step": 157
+    },
+    {
+      "epoch": 2.82,
+      "learning_rate": 1.8785046728971962e-06,
+      "loss": 0.3907,
+      "step": 158
+    },
+    {
+      "epoch": 2.84,
+      "learning_rate": 1.850467289719626e-06,
+      "loss": 0.3903,
+      "step": 159
+    },
+    {
+      "epoch": 2.86,
+      "learning_rate": 1.822429906542056e-06,
+      "loss": 0.3808,
+      "step": 160
+    },
+    {
+      "epoch": 2.88,
+      "learning_rate": 1.794392523364486e-06,
+      "loss": 0.3995,
+      "step": 161
+    },
+    {
+      "epoch": 2.89,
+      "learning_rate": 1.7663551401869159e-06,
+      "loss": 0.4188,
+      "step": 162
+    },
+    {
+      "epoch": 2.91,
+      "learning_rate": 1.7383177570093456e-06,
+      "loss": 0.3734,
+      "step": 163
+    },
+    {
+      "epoch": 2.93,
+      "learning_rate": 1.7102803738317756e-06,
+      "loss": 0.3829,
+      "step": 164
+    },
+    {
+      "epoch": 2.95,
+      "learning_rate": 1.6822429906542055e-06,
+      "loss": 0.4018,
+      "step": 165
+    },
+    {
+      "epoch": 2.96,
+      "learning_rate": 1.6542056074766355e-06,
+      "loss": 0.3687,
+      "step": 166
+    },
+    {
+      "epoch": 2.98,
+      "learning_rate": 1.6261682242990652e-06,
+      "loss": 0.394,
+      "step": 167
+    },
+    {
+      "epoch": 3.0,
+      "learning_rate": 1.5981308411214952e-06,
+      "loss": 0.3903,
+      "step": 168
+    },
+    {
+      "epoch": 3.02,
+      "learning_rate": 1.5700934579439251e-06,
+      "loss": 0.3535,
+      "step": 169
+    },
+    {
+      "epoch": 3.04,
+      "learning_rate": 1.542056074766355e-06,
+      "loss": 0.3492,
+      "step": 170
+    },
+    {
+      "epoch": 3.05,
+      "learning_rate": 1.5140186915887848e-06,
+      "loss": 0.3518,
+      "step": 171
+    },
+    {
+      "epoch": 3.07,
+      "learning_rate": 1.485981308411215e-06,
+      "loss": 0.3239,
+      "step": 172
+    },
+    {
+      "epoch": 3.09,
+      "learning_rate": 1.457943925233645e-06,
+      "loss": 0.3549,
+      "step": 173
+    },
+    {
+      "epoch": 3.11,
+      "learning_rate": 1.4299065420560747e-06,
+      "loss": 0.3258,
+      "step": 174
+    },
+    {
+      "epoch": 3.12,
+      "learning_rate": 1.4018691588785047e-06,
+      "loss": 0.336,
+      "step": 175
+    },
+    {
+      "epoch": 3.14,
+      "learning_rate": 1.3738317757009346e-06,
+      "loss": 0.327,
+      "step": 176
+    },
+    {
+      "epoch": 3.16,
+      "learning_rate": 1.3457943925233646e-06,
+      "loss": 0.329,
+      "step": 177
+    },
+    {
+      "epoch": 3.18,
+      "learning_rate": 1.3177570093457943e-06,
+      "loss": 0.3492,
+      "step": 178
+    },
+    {
+      "epoch": 3.2,
+      "learning_rate": 1.2897196261682243e-06,
+      "loss": 0.357,
+      "step": 179
+    },
+    {
+      "epoch": 3.21,
+      "learning_rate": 1.2616822429906543e-06,
+      "loss": 0.3419,
+      "step": 180
+    },
+    {
+      "epoch": 3.23,
+      "learning_rate": 1.2336448598130842e-06,
+      "loss": 0.318,
+      "step": 181
+    },
+    {
+      "epoch": 3.25,
+      "learning_rate": 1.205607476635514e-06,
+      "loss": 0.3176,
+      "step": 182
+    },
+    {
+      "epoch": 3.27,
+      "learning_rate": 1.177570093457944e-06,
+      "loss": 0.3391,
+      "step": 183
+    },
+    {
+      "epoch": 3.29,
+      "learning_rate": 1.1495327102803739e-06,
+      "loss": 0.364,
+      "step": 184
+    },
+    {
+      "epoch": 3.3,
+      "learning_rate": 1.1214953271028038e-06,
+      "loss": 0.3409,
+      "step": 185
+    },
+    {
+      "epoch": 3.32,
+      "learning_rate": 1.0934579439252336e-06,
+      "loss": 0.3336,
+      "step": 186
+    },
+    {
+      "epoch": 3.34,
+      "learning_rate": 1.0654205607476635e-06,
+      "loss": 0.3552,
+      "step": 187
+    },
+    {
+      "epoch": 3.36,
+      "learning_rate": 1.0373831775700935e-06,
+      "loss": 0.3473,
+      "step": 188
+    },
+    {
+      "epoch": 3.38,
+      "learning_rate": 1.0093457943925232e-06,
+      "loss": 0.3384,
+      "step": 189
+    },
+    {
+      "epoch": 3.39,
+      "learning_rate": 9.813084112149532e-07,
+      "loss": 0.3273,
+      "step": 190
+    },
+    {
+      "epoch": 3.41,
+      "learning_rate": 9.532710280373831e-07,
+      "loss": 0.3488,
+      "step": 191
+    },
+    {
+      "epoch": 3.43,
+      "learning_rate": 9.25233644859813e-07,
+      "loss": 0.3374,
+      "step": 192
+    },
+    {
+      "epoch": 3.45,
+      "learning_rate": 8.97196261682243e-07,
+      "loss": 0.3247,
+      "step": 193
+    },
+    {
+      "epoch": 3.46,
+      "learning_rate": 8.691588785046728e-07,
+      "loss": 0.3444,
+      "step": 194
+    },
+    {
+      "epoch": 3.48,
+      "learning_rate": 8.411214953271028e-07,
+      "loss": 0.3518,
+      "step": 195
+    },
+    {
+      "epoch": 3.5,
+      "learning_rate": 8.130841121495326e-07,
+      "loss": 0.3563,
+      "step": 196
+    },
+    {
+      "epoch": 3.52,
+      "learning_rate": 7.850467289719626e-07,
+      "loss": 0.3516,
+      "step": 197
+    },
+    {
+      "epoch": 3.54,
+      "learning_rate": 7.570093457943924e-07,
+      "loss": 0.339,
+      "step": 198
+    },
+    {
+      "epoch": 3.55,
+      "learning_rate": 7.289719626168225e-07,
+      "loss": 0.3305,
+      "step": 199
+    },
+    {
+      "epoch": 3.57,
+      "learning_rate": 7.009345794392523e-07,
+      "loss": 0.3325,
+      "step": 200
+    },
+    {
+      "epoch": 3.59,
+      "learning_rate": 6.728971962616823e-07,
+      "loss": 0.3621,
+      "step": 201
+    },
+    {
+      "epoch": 3.61,
+      "learning_rate": 6.448598130841121e-07,
+      "loss": 0.3207,
+      "step": 202
+    },
+    {
+      "epoch": 3.62,
+      "learning_rate": 6.168224299065421e-07,
+      "loss": 0.3458,
+      "step": 203
+    },
+    {
+      "epoch": 3.64,
+      "learning_rate": 5.88785046728972e-07,
+      "loss": 0.331,
+      "step": 204
+    },
+    {
+      "epoch": 3.66,
+      "learning_rate": 5.607476635514019e-07,
+      "loss": 0.3307,
+      "step": 205
+    },
+    {
+      "epoch": 3.68,
+      "learning_rate": 5.327102803738318e-07,
+      "loss": 0.3531,
+      "step": 206
+    },
+    {
+      "epoch": 3.7,
+      "learning_rate": 5.046728971962616e-07,
+      "loss": 0.3312,
+      "step": 207
+    },
+    {
+      "epoch": 3.71,
+      "learning_rate": 4.7663551401869157e-07,
+      "loss": 0.3485,
+      "step": 208
+    },
+    {
+      "epoch": 3.73,
+      "learning_rate": 4.485981308411215e-07,
+      "loss": 0.3285,
+      "step": 209
+    },
+    {
+      "epoch": 3.75,
+      "learning_rate": 4.205607476635514e-07,
+      "loss": 0.3267,
+      "step": 210
+    },
+    {
+      "epoch": 3.77,
+      "learning_rate": 3.925233644859813e-07,
+      "loss": 0.3211,
+      "step": 211
+    },
+    {
+      "epoch": 3.79,
+      "learning_rate": 3.6448598130841124e-07,
+      "loss": 0.3286,
+      "step": 212
+    },
+    {
+      "epoch": 3.8,
+      "learning_rate": 3.3644859813084115e-07,
+      "loss": 0.3511,
+      "step": 213
+    },
+    {
+      "epoch": 3.82,
+      "learning_rate": 3.0841121495327105e-07,
+      "loss": 0.3375,
+      "step": 214
+    },
+    {
+      "epoch": 3.84,
+      "learning_rate": 2.8037383177570096e-07,
+      "loss": 0.3544,
+      "step": 215
+    },
+    {
+      "epoch": 3.86,
+      "learning_rate": 2.523364485981308e-07,
+      "loss": 0.3212,
+      "step": 216
+    },
+    {
+      "epoch": 3.88,
+      "learning_rate": 2.2429906542056074e-07,
+      "loss": 0.3364,
+      "step": 217
+    },
+    {
+      "epoch": 3.89,
+      "learning_rate": 1.9626168224299064e-07,
+      "loss": 0.3521,
+      "step": 218
+    },
+    {
+      "epoch": 3.91,
+      "learning_rate": 1.6822429906542057e-07,
+      "loss": 0.3453,
+      "step": 219
+    },
+    {
+      "epoch": 3.93,
+      "learning_rate": 1.4018691588785048e-07,
+      "loss": 0.3297,
+      "step": 220
+    }
+  ],
+  "logging_steps": 1,
+  "max_steps": 224,
+  "num_train_epochs": 4,
+  "save_steps": 10,
+  "total_flos": 1.8453830894302003e+18,
+  "trial_name": null,
+  "trial_params": null
+}

checkpoint-220/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3f772f6ccbcbbefdc206f1ef8cd2122f95829c41904e09540b951c4e4a2b54d6
+size 5947

checkpoint-220/zero_to_fp32.py

@@ -0,0 +1,587 @@
+#!/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):
+    """
+    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)
+    elif zero_stage == 3:
+        return _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states)
+
+
+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 _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()
+            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):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    _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):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    _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):
+    """
+    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``
+
+    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)
+
+
+def convert_zero_checkpoint_to_fp32_state_dict(checkpoint_dir, output_file, tag=None):
+    """
+    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``
+    """
+
+    state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag)
+    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("-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)

config.json

@@ -1,25 +1,3 @@
-{
-  "_name_or_path": "mistralai/Mistral-7B-v0.1",
-  "architectures": [
-    "MistralForCausalLM"
-  ],
-  "bos_token_id": 1,
-  "eos_token_id": 2,
-  "hidden_act": "silu",
-  "hidden_size": 4096,
-  "initializer_range": 0.02,
-  "intermediate_size": 14336,
-  "max_position_embeddings": 32768,
-  "model_type": "mistral",
-  "num_attention_heads": 32,
-  "num_hidden_layers": 32,
-  "num_key_value_heads": 8,
-  "rms_norm_eps": 1e-05,
-  "rope_theta": 10000.0,
-  "sliding_window": 4096,
-  "tie_word_embeddings": false,
-  "torch_dtype": "bfloat16",
-  "transformers_version": "4.34.0.dev0",
-  "use_cache": false,
-  "vocab_size": 32006
-}
+version https://git-lfs.github.com/spec/v1
+oid sha256:44e972929ad46908648b92c2feb180fb201dab6bdc232a5bc5776a74d47e0e0a
+size 620

generation_config.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:afcf2a1da583dbeb66bb367a759713901e23ce19ed83ed8f5a21b0e877addd5b
+size 116

pytorch_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7a5617e2395028d62117f2b331666efc9f56999932c1857c0449287e9c9de3a2
+size 14483602725

special_tokens_map.json

@@ -1,6 +1,3 @@
-{
-  "bos_token": "<s>",
-  "eos_token": "<|im_end|>",
-  "pad_token": "</s>",
-  "unk_token": "<unk>"
-}
+version https://git-lfs.github.com/spec/v1
+oid sha256:b454cf92b8686901e71f611ed4e6b93e70a2692dd3a5bb50b1c408efb9a47a03
+size 101

tokenizer_config.json

@@ -1,93 +1,3 @@
-{
-  "add_bos_token": true,
-  "add_eos_token": false,
-  "added_tokens_decoder": {
-    "0": {
-      "content": "<unk>",
-      "lstrip": true,
-      "normalized": false,
-      "rstrip": true,
-      "single_word": false,
-      "special": true
-    },
-    "1": {
-      "content": "<s>",
-      "lstrip": true,
-      "normalized": false,
-      "rstrip": true,
-      "single_word": false,
-      "special": true
-    },
-    "2": {
-      "content": "</s>",
-      "lstrip": false,
-      "normalized": false,
-      "rstrip": false,
-      "single_word": false,
-      "special": true
-    },
-    "32000": {
-      "content": "<|im_end|>",
-      "lstrip": true,
-      "normalized": true,
-      "rstrip": true,
-      "single_word": false,
-      "special": false
-    },
-    "32001": {
-      "content": "<|im_start|>",
-      "lstrip": true,
-      "normalized": true,
-      "rstrip": true,
-      "single_word": false,
-      "special": false
-    },
-    "32002": {
-      "content": "<thinking>",
-      "lstrip": true,
-      "normalized": true,
-      "rstrip": true,
-      "single_word": false,
-      "special": false
-    },
-    "32003": {
-      "content": "</thinking>",
-      "lstrip": true,
-      "normalized": true,
-      "rstrip": true,
-      "single_word": false,
-      "special": false
-    },
-    "32004": {
-      "content": "<talking>",
-      "lstrip": true,
-      "normalized": true,
-      "rstrip": true,
-      "single_word": false,
-      "special": false
-    },
-    "32005": {
-      "content": "</talking>",
-      "lstrip": true,
-      "normalized": true,
-      "rstrip": true,
-      "single_word": false,
-      "special": false
-    }
-  },
-  "additional_special_tokens": [],
-  "bos_token": "<s>",
-  "clean_up_tokenization_spaces": false,
-  "eos_token": "</s>",
-  "legacy": true,
-  "model_max_length": 1000000000000000019884624838656,
-  "pad_token": null,
-  "sp_model_kwargs": {},
-  "spaces_between_special_tokens": false,
-  "tokenizer_class": "LlamaTokenizer",
-  "tokenizer_file": "/root/.cache/huggingface/hub/models--mistralai--Mistral-7B-v0.1/snapshots/78814a934d8ce59ee93599378f3e929ff137da06/tokenizer.json",
-  "trust_remote_code": false,
-  "unk_token": "<unk>",
-  "use_default_system_prompt": true,
-  "use_fast": true
-}
+version https://git-lfs.github.com/spec/v1
+oid sha256:8a9c27902a9367ab1f5bf6b21f20b0617070ab1a546d2f16fdafacd816cee4e1
+size 2211