Upload folder using huggingface_hub

.gitattributes

@@ -33,3 +33,4 @@ 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.json filter=lfs diff=lfs merge=lfs -text

config.json

@@ -0,0 +1,36 @@
+{
+  "architectures": [
+    "ArceeForCausalLM"
+  ],
+  "attention_bias": false,
+  "attention_dropout": 0.0,
+  "bos_token_id": 128000,
+  "eos_token_id": 128001,
+  "head_dim": 128,
+  "hidden_act": "relu2",
+  "hidden_size": 2560,
+  "initializer_range": 0.02,
+  "intermediate_size": 18432,
+  "max_position_embeddings": 65536,
+  "mlp_bias": false,
+  "model_type": "arcee",
+  "num_attention_heads": 20,
+  "num_hidden_layers": 36,
+  "num_key_value_heads": 4,
+  "rms_norm_eps": 1e-05,
+  "rope_scaling": {
+    "beta_fast": 32.0,
+    "beta_slow": 1.0,
+    "factor": 20.0,
+    "mscale": 1.0,
+    "original_max_position_embeddings": 4096,
+    "rope_type": "yarn",
+    "type": "yarn"
+  },
+  "rope_theta": 10000.0,
+  "tie_word_embeddings": false,
+  "torch_dtype": "bfloat16",
+  "transformers_version": "4.55.0",
+  "use_cache": false,
+  "vocab_size": 128004
+}

generation_config.json

@@ -0,0 +1,8 @@
+{
+  "_from_model_config": true,
+  "bos_token_id": 128000,
+  "do_sample": true,
+  "eos_token_id": 128001,
+  "transformers_version": "4.55.0",
+  "use_cache": false
+}

latest

@@ -0,0 +1 @@
+global_step238

model-00001-of-00002.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:657e3aacd08d18144f553bab26bcbffb71920cd668ac2e29e047fd811b956a50
+size 4965240752

model-00002-of-00002.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b0f31983fddea56d5ab63f637d88eb28b5bf59b936e5bf3617937b5f559b7e70
+size 4273162352

model.safetensors.index.json

@@ -0,0 +1,299 @@
+{
+  "metadata": {
+    "total_parameters": 186880,
+    "total_size": 9238369280
+  },
+  "weight_map": {
+    "lm_head.weight": "model-00002-of-00002.safetensors",
+    "model.embed_tokens.weight": "model-00001-of-00002.safetensors",
+    "model.layers.0.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.0.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.0.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.0.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.0.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.0.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.0.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.0.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.1.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.1.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.1.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.1.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.1.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.1.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.1.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.1.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.10.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.10.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.10.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.10.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.10.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.10.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.10.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.10.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.11.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.11.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.11.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.11.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.11.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.11.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.11.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.11.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.12.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.12.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.12.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.12.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.12.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.12.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.12.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.12.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.13.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.13.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.13.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.13.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.13.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.13.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.13.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.13.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.14.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.14.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.14.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.14.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.14.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.14.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.14.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.14.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.15.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.16.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.16.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.16.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.16.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.16.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.16.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.16.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.16.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.17.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.17.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.17.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.17.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.17.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.17.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.17.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.17.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.18.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.18.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.18.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.18.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.18.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.18.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.18.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.18.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.19.input_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.19.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.19.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.19.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.19.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.19.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.19.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.19.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.2.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.2.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.2.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.2.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.2.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.2.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.2.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.2.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.20.input_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.20.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.20.mlp.up_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.20.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.20.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.20.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.20.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.20.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.21.input_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.21.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.21.mlp.up_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.21.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.21.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.21.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.21.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.21.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.22.input_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.22.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.22.mlp.up_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.22.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.22.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.22.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.22.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.22.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.23.input_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.23.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.23.mlp.up_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.23.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.23.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.23.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.23.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.23.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.24.input_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.24.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.24.mlp.up_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.24.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.24.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.24.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.24.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.24.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.25.input_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.25.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.25.mlp.up_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.25.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.25.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.25.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.25.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.25.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.26.input_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.26.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.26.mlp.up_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.26.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.26.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.26.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.26.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.26.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.27.input_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.27.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.27.mlp.up_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.27.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.27.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.27.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.27.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.27.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.28.input_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.28.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.28.mlp.up_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.28.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.28.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.28.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.28.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.28.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.29.input_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.29.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.29.mlp.up_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.29.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.29.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.29.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.29.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.29.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.3.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.3.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.3.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.3.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.3.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.3.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.3.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.3.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.30.input_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.30.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.30.mlp.up_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.30.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.30.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.30.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.30.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.30.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.31.input_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.31.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.31.mlp.up_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.31.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.31.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.31.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.31.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.31.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.32.input_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.32.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.32.mlp.up_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.32.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.32.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.32.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.32.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.32.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.33.input_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.33.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.33.mlp.up_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.33.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.33.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.33.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.33.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.33.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.34.input_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.34.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.34.mlp.up_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.34.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.34.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.34.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.34.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.34.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.35.input_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.35.mlp.down_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.35.mlp.up_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.35.post_attention_layernorm.weight": "model-00002-of-00002.safetensors",
+    "model.layers.35.self_attn.k_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.35.self_attn.o_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.35.self_attn.q_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.35.self_attn.v_proj.weight": "model-00002-of-00002.safetensors",
+    "model.layers.4.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.4.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.4.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.4.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.4.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.4.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.4.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.4.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.5.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.5.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.5.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.5.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.5.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.5.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.5.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.5.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.6.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.6.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.6.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.6.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.6.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.6.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.6.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.6.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.7.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.7.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.7.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.7.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.7.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.7.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.7.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.7.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.8.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.8.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.8.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.8.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.8.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.8.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.8.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.8.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.9.input_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.9.mlp.down_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.9.mlp.up_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.9.post_attention_layernorm.weight": "model-00001-of-00002.safetensors",
+    "model.layers.9.self_attn.k_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.9.self_attn.o_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.9.self_attn.q_proj.weight": "model-00001-of-00002.safetensors",
+    "model.layers.9.self_attn.v_proj.weight": "model-00001-of-00002.safetensors",
+    "model.norm.weight": "model-00002-of-00002.safetensors"
+  }
+}

rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9892a41432d4e48929c5376b8225d5373bb9afe9f14665fbcb3934dc45cb330a
+size 15429

rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b6f3949e2b25ad05905da16bc2d58aeadc9a9f1db0b0e97a04ceb63467acff1d
+size 15429

rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5c3812c5e7f13c9d32f97d83cd625428061b1c9486e869caffd453d83ac07ceb
+size 15429

rng_state_3.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6eda756b33768639c1e1932ae00c5d43269de5705c09162bc09acdf71d3533c0
+size 15429

scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2167771333f3651d3a9e822987a7a5fc9e8592122c6e569de0f5445a8d9b3690
+size 1465

special_tokens_map.json

@@ -0,0 +1,23 @@
+{
+  "bos_token": {
+    "content": "<|begin_of_text|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "eos_token": {
+    "content": "<|end_of_text|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": {
+    "content": "<|end_of_text|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}

tokenizer.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2f7e8b61d1cb49d927a5f5aaeecdac9cb173cccb3ec1897cb5d7bb08bda54907
+size 17157913

tokenizer_config.json

@@ -0,0 +1,47 @@
+{
+  "added_tokens_decoder": {
+    "128000": {
+      "content": "<|begin_of_text|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "128001": {
+      "content": "<|end_of_text|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "128002": {
+      "content": "<|im_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "128003": {
+      "content": "<|im_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    }
+  },
+  "bos_token": "<|begin_of_text|>",
+  "clean_up_tokenization_spaces": true,
+  "eos_token": "<|end_of_text|>",
+  "extra_special_tokens": {},
+  "model_input_names": [
+    "input_ids",
+    "attention_mask"
+  ],
+  "model_max_length": 65536,
+  "pad_token": "<|end_of_text|>",
+  "tokenizer_class": "PreTrainedTokenizerFast"
+}

trainer_state.json

@@ -0,0 +1,2414 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.5002627430373096,
+  "eval_steps": 500,
+  "global_step": 238,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0021019442984760903,
+      "grad_norm": 0.7624253401313755,
+      "learning_rate": 0.0,
+      "loss": 2.2955,
+      "memory/device_mem_reserved(gib)": 68.22,
+      "memory/max_mem_active(gib)": 63.52,
+      "memory/max_mem_allocated(gib)": 62.82,
+      "step": 1
+    },
+    {
+      "epoch": 0.004203888596952181,
+      "grad_norm": 0.7189116302541813,
+      "learning_rate": 2e-08,
+      "loss": 2.2824,
+      "memory/device_mem_reserved(gib)": 68.31,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 2
+    },
+    {
+      "epoch": 0.006305832895428271,
+      "grad_norm": 0.8431595274838072,
+      "learning_rate": 4e-08,
+      "loss": 2.3101,
+      "memory/device_mem_reserved(gib)": 68.94,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 3
+    },
+    {
+      "epoch": 0.008407777193904361,
+      "grad_norm": 0.8637289443313003,
+      "learning_rate": 6e-08,
+      "loss": 2.3514,
+      "memory/device_mem_reserved(gib)": 68.94,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 4
+    },
+    {
+      "epoch": 0.010509721492380452,
+      "grad_norm": 0.8128827491990301,
+      "learning_rate": 8e-08,
+      "loss": 2.3621,
+      "memory/device_mem_reserved(gib)": 68.94,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 5
+    },
+    {
+      "epoch": 0.012611665790856543,
+      "grad_norm": 0.9504830158009488,
+      "learning_rate": 1e-07,
+      "loss": 2.4108,
+      "memory/device_mem_reserved(gib)": 68.94,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 6
+    },
+    {
+      "epoch": 0.014713610089332634,
+      "grad_norm": 0.9140479063802851,
+      "learning_rate": 1.2e-07,
+      "loss": 2.2224,
+      "memory/device_mem_reserved(gib)": 68.94,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 7
+    },
+    {
+      "epoch": 0.016815554387808723,
+      "grad_norm": 0.8632210617655338,
+      "learning_rate": 1.4e-07,
+      "loss": 2.3589,
+      "memory/device_mem_reserved(gib)": 68.94,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 8
+    },
+    {
+      "epoch": 0.018917498686284815,
+      "grad_norm": 0.8747745167339828,
+      "learning_rate": 1.6e-07,
+      "loss": 2.2411,
+      "memory/device_mem_reserved(gib)": 68.94,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 9
+    },
+    {
+      "epoch": 0.021019442984760904,
+      "grad_norm": 0.7739481898974889,
+      "learning_rate": 1.8e-07,
+      "loss": 2.2522,
+      "memory/device_mem_reserved(gib)": 68.94,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 10
+    },
+    {
+      "epoch": 0.023121387283236993,
+      "grad_norm": 0.8301921811025426,
+      "learning_rate": 2e-07,
+      "loss": 2.3565,
+      "memory/device_mem_reserved(gib)": 68.94,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 11
+    },
+    {
+      "epoch": 0.025223331581713086,
+      "grad_norm": 0.7677035533090953,
+      "learning_rate": 2.1999999999999998e-07,
+      "loss": 2.2208,
+      "memory/device_mem_reserved(gib)": 68.94,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 12
+    },
+    {
+      "epoch": 0.027325275880189175,
+      "grad_norm": 0.7834629656153209,
+      "learning_rate": 2.4e-07,
+      "loss": 2.2526,
+      "memory/device_mem_reserved(gib)": 68.94,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 13
+    },
+    {
+      "epoch": 0.029427220178665267,
+      "grad_norm": 0.776588932490268,
+      "learning_rate": 2.6e-07,
+      "loss": 2.2727,
+      "memory/device_mem_reserved(gib)": 68.94,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 14
+    },
+    {
+      "epoch": 0.03152916447714135,
+      "grad_norm": 0.7753598356301531,
+      "learning_rate": 2.8e-07,
+      "loss": 2.3564,
+      "memory/device_mem_reserved(gib)": 68.94,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 15
+    },
+    {
+      "epoch": 0.033631108775617445,
+      "grad_norm": 0.7165697716264268,
+      "learning_rate": 3e-07,
+      "loss": 2.3331,
+      "memory/device_mem_reserved(gib)": 68.94,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 16
+    },
+    {
+      "epoch": 0.03573305307409354,
+      "grad_norm": 0.729205845829164,
+      "learning_rate": 3.2e-07,
+      "loss": 2.322,
+      "memory/device_mem_reserved(gib)": 68.94,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 17
+    },
+    {
+      "epoch": 0.03783499737256963,
+      "grad_norm": 0.8495080537327478,
+      "learning_rate": 3.4000000000000003e-07,
+      "loss": 2.4952,
+      "memory/device_mem_reserved(gib)": 68.94,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 18
+    },
+    {
+      "epoch": 0.039936941671045716,
+      "grad_norm": 0.7578372584471679,
+      "learning_rate": 3.6e-07,
+      "loss": 2.3132,
+      "memory/device_mem_reserved(gib)": 68.94,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 19
+    },
+    {
+      "epoch": 0.04203888596952181,
+      "grad_norm": 0.6681608647353537,
+      "learning_rate": 3.7999999999999996e-07,
+      "loss": 2.3086,
+      "memory/device_mem_reserved(gib)": 68.94,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 20
+    },
+    {
+      "epoch": 0.0441408302679979,
+      "grad_norm": 0.7042221303721394,
+      "learning_rate": 4e-07,
+      "loss": 2.3644,
+      "memory/device_mem_reserved(gib)": 68.94,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 21
+    },
+    {
+      "epoch": 0.046242774566473986,
+      "grad_norm": 0.725951911870576,
+      "learning_rate": 4.1999999999999995e-07,
+      "loss": 2.3208,
+      "memory/device_mem_reserved(gib)": 68.94,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 22
+    },
+    {
+      "epoch": 0.04834471886495008,
+      "grad_norm": 0.6347256826610295,
+      "learning_rate": 4.3999999999999997e-07,
+      "loss": 2.328,
+      "memory/device_mem_reserved(gib)": 68.94,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 23
+    },
+    {
+      "epoch": 0.05044666316342617,
+      "grad_norm": 0.5682080423503054,
+      "learning_rate": 4.6e-07,
+      "loss": 2.2008,
+      "memory/device_mem_reserved(gib)": 68.94,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 24
+    },
+    {
+      "epoch": 0.05254860746190226,
+      "grad_norm": 0.5787647024012217,
+      "learning_rate": 4.8e-07,
+      "loss": 2.2841,
+      "memory/device_mem_reserved(gib)": 68.94,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 25
+    },
+    {
+      "epoch": 0.05465055176037835,
+      "grad_norm": 0.5807941293103913,
+      "learning_rate": 5e-07,
+      "loss": 2.2661,
+      "memory/device_mem_reserved(gib)": 68.94,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 26
+    },
+    {
+      "epoch": 0.05675249605885444,
+      "grad_norm": 0.5238787661221586,
+      "learning_rate": 5.2e-07,
+      "loss": 2.292,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 27
+    },
+    {
+      "epoch": 0.058854440357330534,
+      "grad_norm": 0.6607378470156829,
+      "learning_rate": 5.4e-07,
+      "loss": 2.346,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 28
+    },
+    {
+      "epoch": 0.06095638465580662,
+      "grad_norm": 0.5949442201958344,
+      "learning_rate": 5.6e-07,
+      "loss": 2.2856,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 29
+    },
+    {
+      "epoch": 0.0630583289542827,
+      "grad_norm": 0.6213907595973902,
+      "learning_rate": 5.8e-07,
+      "loss": 2.3527,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 30
+    },
+    {
+      "epoch": 0.0651602732527588,
+      "grad_norm": 0.6574213245120029,
+      "learning_rate": 6e-07,
+      "loss": 2.2896,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 31
+    },
+    {
+      "epoch": 0.06726221755123489,
+      "grad_norm": 0.7904069125236015,
+      "learning_rate": 6.2e-07,
+      "loss": 2.4192,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 32
+    },
+    {
+      "epoch": 0.06936416184971098,
+      "grad_norm": 0.6912774106481298,
+      "learning_rate": 6.4e-07,
+      "loss": 2.3085,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 33
+    },
+    {
+      "epoch": 0.07146610614818708,
+      "grad_norm": 0.6819796440725628,
+      "learning_rate": 6.6e-07,
+      "loss": 2.2756,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 34
+    },
+    {
+      "epoch": 0.07356805044666316,
+      "grad_norm": 0.7580978517321655,
+      "learning_rate": 6.800000000000001e-07,
+      "loss": 2.3645,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 35
+    },
+    {
+      "epoch": 0.07566999474513926,
+      "grad_norm": 0.6791446776516942,
+      "learning_rate": 7e-07,
+      "loss": 2.2628,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 36
+    },
+    {
+      "epoch": 0.07777193904361535,
+      "grad_norm": 0.569840280711906,
+      "learning_rate": 7.2e-07,
+      "loss": 2.2602,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 37
+    },
+    {
+      "epoch": 0.07987388334209143,
+      "grad_norm": 0.5498888556096215,
+      "learning_rate": 7.4e-07,
+      "loss": 2.3167,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 38
+    },
+    {
+      "epoch": 0.08197582764056753,
+      "grad_norm": 0.5268765090754575,
+      "learning_rate": 7.599999999999999e-07,
+      "loss": 2.2378,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 39
+    },
+    {
+      "epoch": 0.08407777193904362,
+      "grad_norm": 0.4848125502462646,
+      "learning_rate": 7.799999999999999e-07,
+      "loss": 2.2882,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 40
+    },
+    {
+      "epoch": 0.0861797162375197,
+      "grad_norm": 0.5814992292096023,
+      "learning_rate": 8e-07,
+      "loss": 2.3471,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 41
+    },
+    {
+      "epoch": 0.0882816605359958,
+      "grad_norm": 0.6166392360245904,
+      "learning_rate": 8.199999999999999e-07,
+      "loss": 2.441,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 42
+    },
+    {
+      "epoch": 0.09038360483447189,
+      "grad_norm": 0.6377322312855411,
+      "learning_rate": 8.399999999999999e-07,
+      "loss": 2.3912,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 43
+    },
+    {
+      "epoch": 0.09248554913294797,
+      "grad_norm": 0.5055719418643514,
+      "learning_rate": 8.599999999999999e-07,
+      "loss": 2.2561,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 44
+    },
+    {
+      "epoch": 0.09458749343142407,
+      "grad_norm": 0.49178646668795084,
+      "learning_rate": 8.799999999999999e-07,
+      "loss": 2.2599,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 45
+    },
+    {
+      "epoch": 0.09668943772990016,
+      "grad_norm": 0.47537370207387974,
+      "learning_rate": 9e-07,
+      "loss": 2.3064,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 46
+    },
+    {
+      "epoch": 0.09879138202837624,
+      "grad_norm": 0.5089053853006482,
+      "learning_rate": 9.2e-07,
+      "loss": 2.391,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 47
+    },
+    {
+      "epoch": 0.10089332632685234,
+      "grad_norm": 0.4728302009023318,
+      "learning_rate": 9.399999999999999e-07,
+      "loss": 2.3139,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 48
+    },
+    {
+      "epoch": 0.10299527062532843,
+      "grad_norm": 0.4974785018291372,
+      "learning_rate": 9.6e-07,
+      "loss": 2.3599,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 49
+    },
+    {
+      "epoch": 0.10509721492380451,
+      "grad_norm": 0.5140106787374947,
+      "learning_rate": 9.8e-07,
+      "loss": 2.4427,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 50
+    },
+    {
+      "epoch": 0.10719915922228061,
+      "grad_norm": 0.5361457578321233,
+      "learning_rate": 1e-06,
+      "loss": 2.3295,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 51
+    },
+    {
+      "epoch": 0.1093011035207567,
+      "grad_norm": 0.49844160829734835,
+      "learning_rate": 9.999863397100894e-07,
+      "loss": 2.2672,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 52
+    },
+    {
+      "epoch": 0.11140304781923278,
+      "grad_norm": 0.5385578770440957,
+      "learning_rate": 9.999453595867715e-07,
+      "loss": 2.3261,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 53
+    },
+    {
+      "epoch": 0.11350499211770888,
+      "grad_norm": 0.497092836247932,
+      "learning_rate": 9.998770618692484e-07,
+      "loss": 2.3326,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 54
+    },
+    {
+      "epoch": 0.11560693641618497,
+      "grad_norm": 0.539870187568986,
+      "learning_rate": 9.997814502893856e-07,
+      "loss": 2.2381,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 55
+    },
+    {
+      "epoch": 0.11770888071466107,
+      "grad_norm": 0.5073884711048833,
+      "learning_rate": 9.996585300715115e-07,
+      "loss": 2.3122,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 56
+    },
+    {
+      "epoch": 0.11981082501313715,
+      "grad_norm": 0.5162826315178152,
+      "learning_rate": 9.99508307932129e-07,
+      "loss": 2.2719,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 57
+    },
+    {
+      "epoch": 0.12191276931161324,
+      "grad_norm": 0.5135640558488429,
+      "learning_rate": 9.9933079207955e-07,
+      "loss": 2.4354,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 58
+    },
+    {
+      "epoch": 0.12401471361008934,
+      "grad_norm": 0.48495138081284994,
+      "learning_rate": 9.991259922134465e-07,
+      "loss": 2.2913,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 59
+    },
+    {
+      "epoch": 0.1261166579085654,
+      "grad_norm": 0.5031121760605395,
+      "learning_rate": 9.98893919524321e-07,
+      "loss": 2.293,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 60
+    },
+    {
+      "epoch": 0.1282186022070415,
+      "grad_norm": 0.4053272758920918,
+      "learning_rate": 9.98634586692894e-07,
+      "loss": 2.2873,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 61
+    },
+    {
+      "epoch": 0.1303205465055176,
+      "grad_norm": 0.4532646932853173,
+      "learning_rate": 9.983480078894123e-07,
+      "loss": 2.3065,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 62
+    },
+    {
+      "epoch": 0.13242249080399368,
+      "grad_norm": 0.4496821436560576,
+      "learning_rate": 9.98034198772874e-07,
+      "loss": 2.2886,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 63
+    },
+    {
+      "epoch": 0.13452443510246978,
+      "grad_norm": 0.48430661978532813,
+      "learning_rate": 9.976931764901733e-07,
+      "loss": 2.3404,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 64
+    },
+    {
+      "epoch": 0.13662637940094588,
+      "grad_norm": 0.5163168950805126,
+      "learning_rate": 9.97324959675163e-07,
+      "loss": 2.286,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 65
+    },
+    {
+      "epoch": 0.13872832369942195,
+      "grad_norm": 0.4385342628062273,
+      "learning_rate": 9.969295684476368e-07,
+      "loss": 2.2923,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 66
+    },
+    {
+      "epoch": 0.14083026799789805,
+      "grad_norm": 0.4476245967273303,
+      "learning_rate": 9.9650702441223e-07,
+      "loss": 2.2454,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 67
+    },
+    {
+      "epoch": 0.14293221229637415,
+      "grad_norm": 0.4493507785126621,
+      "learning_rate": 9.960573506572389e-07,
+      "loss": 2.3361,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 68
+    },
+    {
+      "epoch": 0.14503415659485025,
+      "grad_norm": 0.4676473798188462,
+      "learning_rate": 9.955805717533585e-07,
+      "loss": 2.3795,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 69
+    },
+    {
+      "epoch": 0.14713610089332632,
+      "grad_norm": 0.5003504816633514,
+      "learning_rate": 9.950767137523416e-07,
+      "loss": 2.3638,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 70
+    },
+    {
+      "epoch": 0.14923804519180242,
+      "grad_norm": 0.41298653135277646,
+      "learning_rate": 9.94545804185573e-07,
+      "loss": 2.2986,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 71
+    },
+    {
+      "epoch": 0.15133998949027852,
+      "grad_norm": 0.48549576119983434,
+      "learning_rate": 9.939878720625673e-07,
+      "loss": 2.3772,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 72
+    },
+    {
+      "epoch": 0.1534419337887546,
+      "grad_norm": 0.37069853589006974,
+      "learning_rate": 9.93402947869383e-07,
+      "loss": 2.2609,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 73
+    },
+    {
+      "epoch": 0.1555438780872307,
+      "grad_norm": 0.3822824223589903,
+      "learning_rate": 9.927910635669561e-07,
+      "loss": 2.3263,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 74
+    },
+    {
+      "epoch": 0.1576458223857068,
+      "grad_norm": 0.4645424064190486,
+      "learning_rate": 9.921522525893547e-07,
+      "loss": 2.421,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 75
+    },
+    {
+      "epoch": 0.15974776668418286,
+      "grad_norm": 0.40728550126377283,
+      "learning_rate": 9.91486549841951e-07,
+      "loss": 2.3488,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 76
+    },
+    {
+      "epoch": 0.16184971098265896,
+      "grad_norm": 0.39534534329560483,
+      "learning_rate": 9.907939916995152e-07,
+      "loss": 2.2277,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 77
+    },
+    {
+      "epoch": 0.16395165528113506,
+      "grad_norm": 0.3994213467776548,
+      "learning_rate": 9.900746160042272e-07,
+      "loss": 2.3751,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 78
+    },
+    {
+      "epoch": 0.16605359957961113,
+      "grad_norm": 0.3952978443639354,
+      "learning_rate": 9.893284620636098e-07,
+      "loss": 2.3407,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 79
+    },
+    {
+      "epoch": 0.16815554387808723,
+      "grad_norm": 0.3847266788854899,
+      "learning_rate": 9.88555570648379e-07,
+      "loss": 2.2882,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 80
+    },
+    {
+      "epoch": 0.17025748817656333,
+      "grad_norm": 0.3942404109616697,
+      "learning_rate": 9.877559839902183e-07,
+      "loss": 2.3809,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 81
+    },
+    {
+      "epoch": 0.1723594324750394,
+      "grad_norm": 0.3726144315608755,
+      "learning_rate": 9.869297457794698e-07,
+      "loss": 2.2965,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 82
+    },
+    {
+      "epoch": 0.1744613767735155,
+      "grad_norm": 0.4044669149844896,
+      "learning_rate": 9.860769011627474e-07,
+      "loss": 2.3778,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 83
+    },
+    {
+      "epoch": 0.1765633210719916,
+      "grad_norm": 0.44263984303122605,
+      "learning_rate": 9.851974967404702e-07,
+      "loss": 2.3655,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 84
+    },
+    {
+      "epoch": 0.17866526537046767,
+      "grad_norm": 0.3800348736088796,
+      "learning_rate": 9.842915805643156e-07,
+      "loss": 2.2951,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 85
+    },
+    {
+      "epoch": 0.18076720966894377,
+      "grad_norm": 0.38644114608168073,
+      "learning_rate": 9.833592021345937e-07,
+      "loss": 2.3567,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 86
+    },
+    {
+      "epoch": 0.18286915396741987,
+      "grad_norm": 0.5360672745714498,
+      "learning_rate": 9.824004123975434e-07,
+      "loss": 2.3769,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 87
+    },
+    {
+      "epoch": 0.18497109826589594,
+      "grad_norm": 0.3826183850679395,
+      "learning_rate": 9.814152637425477e-07,
+      "loss": 2.2676,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 88
+    },
+    {
+      "epoch": 0.18707304256437204,
+      "grad_norm": 0.3874657198676833,
+      "learning_rate": 9.804038099992716e-07,
+      "loss": 2.2044,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 89
+    },
+    {
+      "epoch": 0.18917498686284814,
+      "grad_norm": 0.42284650951618596,
+      "learning_rate": 9.793661064347204e-07,
+      "loss": 2.2791,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 90
+    },
+    {
+      "epoch": 0.19127693116132422,
+      "grad_norm": 0.4012146632153047,
+      "learning_rate": 9.783022097502203e-07,
+      "loss": 2.2554,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 91
+    },
+    {
+      "epoch": 0.19337887545980031,
+      "grad_norm": 0.37104574503246424,
+      "learning_rate": 9.772121780783201e-07,
+      "loss": 2.2696,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 92
+    },
+    {
+      "epoch": 0.19548081975827641,
+      "grad_norm": 0.4115506199685101,
+      "learning_rate": 9.76096070979614e-07,
+      "loss": 2.292,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 93
+    },
+    {
+      "epoch": 0.19758276405675249,
+      "grad_norm": 0.4949212386577297,
+      "learning_rate": 9.749539494394885e-07,
+      "loss": 2.3154,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 94
+    },
+    {
+      "epoch": 0.19968470835522859,
+      "grad_norm": 0.5305093079330326,
+      "learning_rate": 9.737858758647889e-07,
+      "loss": 2.3967,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 95
+    },
+    {
+      "epoch": 0.20178665265370468,
+      "grad_norm": 0.39802845026570083,
+      "learning_rate": 9.725919140804098e-07,
+      "loss": 2.3833,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 96
+    },
+    {
+      "epoch": 0.20388859695218076,
+      "grad_norm": 0.392517153138478,
+      "learning_rate": 9.713721293258078e-07,
+      "loss": 2.3458,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 97
+    },
+    {
+      "epoch": 0.20599054125065686,
+      "grad_norm": 0.38285765355194634,
+      "learning_rate": 9.70126588251436e-07,
+      "loss": 2.2321,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 98
+    },
+    {
+      "epoch": 0.20809248554913296,
+      "grad_norm": 0.42890083185292094,
+      "learning_rate": 9.688553589151037e-07,
+      "loss": 2.2823,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 99
+    },
+    {
+      "epoch": 0.21019442984760903,
+      "grad_norm": 0.3788992789108253,
+      "learning_rate": 9.675585107782555e-07,
+      "loss": 2.2955,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 100
+    },
+    {
+      "epoch": 0.21229637414608513,
+      "grad_norm": 0.40025954957804155,
+      "learning_rate": 9.66236114702178e-07,
+      "loss": 2.3454,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 101
+    },
+    {
+      "epoch": 0.21439831844456123,
+      "grad_norm": 0.4040329751371346,
+      "learning_rate": 9.648882429441256e-07,
+      "loss": 2.3362,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 102
+    },
+    {
+      "epoch": 0.2165002627430373,
+      "grad_norm": 0.35667806143435715,
+      "learning_rate": 9.635149691533747e-07,
+      "loss": 2.3089,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 103
+    },
+    {
+      "epoch": 0.2186022070415134,
+      "grad_norm": 0.42503183804867145,
+      "learning_rate": 9.621163683671978e-07,
+      "loss": 2.3024,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 104
+    },
+    {
+      "epoch": 0.2207041513399895,
+      "grad_norm": 0.3833710476470682,
+      "learning_rate": 9.606925170067636e-07,
+      "loss": 2.2944,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 105
+    },
+    {
+      "epoch": 0.22280609563846557,
+      "grad_norm": 0.38645757625412946,
+      "learning_rate": 9.592434928729615e-07,
+      "loss": 2.2595,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 106
+    },
+    {
+      "epoch": 0.22490803993694167,
+      "grad_norm": 0.37424692884672933,
+      "learning_rate": 9.577693751421505e-07,
+      "loss": 2.3025,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 107
+    },
+    {
+      "epoch": 0.22700998423541777,
+      "grad_norm": 0.3776781851494623,
+      "learning_rate": 9.562702443618331e-07,
+      "loss": 2.2724,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 108
+    },
+    {
+      "epoch": 0.22911192853389384,
+      "grad_norm": 0.392564325121222,
+      "learning_rate": 9.547461824462533e-07,
+      "loss": 2.3737,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 109
+    },
+    {
+      "epoch": 0.23121387283236994,
+      "grad_norm": 0.3480699195596026,
+      "learning_rate": 9.531972726719215e-07,
+      "loss": 2.2591,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 110
+    },
+    {
+      "epoch": 0.23331581713084604,
+      "grad_norm": 0.3563697131151561,
+      "learning_rate": 9.516235996730644e-07,
+      "loss": 2.3639,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 111
+    },
+    {
+      "epoch": 0.23541776142932214,
+      "grad_norm": 0.4943435915920374,
+      "learning_rate": 9.500252494369991e-07,
+      "loss": 2.3605,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 112
+    },
+    {
+      "epoch": 0.2375197057277982,
+      "grad_norm": 0.3975018845059572,
+      "learning_rate": 9.484023092994364e-07,
+      "loss": 2.4139,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 113
+    },
+    {
+      "epoch": 0.2396216500262743,
+      "grad_norm": 0.37314820478206834,
+      "learning_rate": 9.467548679397071e-07,
+      "loss": 2.293,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 114
+    },
+    {
+      "epoch": 0.2417235943247504,
+      "grad_norm": 0.39240855452269136,
+      "learning_rate": 9.450830153759176e-07,
+      "loss": 2.3568,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 115
+    },
+    {
+      "epoch": 0.24382553862322648,
+      "grad_norm": 0.36202032847414545,
+      "learning_rate": 9.433868429600309e-07,
+      "loss": 2.36,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 116
+    },
+    {
+      "epoch": 0.24592748292170258,
+      "grad_norm": 0.3852500669591038,
+      "learning_rate": 9.416664433728748e-07,
+      "loss": 2.335,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 117
+    },
+    {
+      "epoch": 0.24802942722017868,
+      "grad_norm": 0.35255828976101133,
+      "learning_rate": 9.399219106190775e-07,
+      "loss": 2.3367,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 118
+    },
+    {
+      "epoch": 0.2501313715186548,
+      "grad_norm": 0.4145689168519548,
+      "learning_rate": 9.381533400219317e-07,
+      "loss": 2.3807,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 119
+    },
+    {
+      "epoch": 0.2522333158171308,
+      "grad_norm": 0.3638037446306906,
+      "learning_rate": 9.363608282181861e-07,
+      "loss": 2.2441,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 120
+    },
+    {
+      "epoch": 0.2543352601156069,
+      "grad_norm": 0.3892269635122991,
+      "learning_rate": 9.345444731527641e-07,
+      "loss": 2.3285,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 121
+    },
+    {
+      "epoch": 0.256437204414083,
+      "grad_norm": 0.3848382071231666,
+      "learning_rate": 9.327043740734128e-07,
+      "loss": 2.2713,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 122
+    },
+    {
+      "epoch": 0.2585391487125591,
+      "grad_norm": 0.3602411460013298,
+      "learning_rate": 9.308406315252798e-07,
+      "loss": 2.32,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 123
+    },
+    {
+      "epoch": 0.2606410930110352,
+      "grad_norm": 0.36833348877975325,
+      "learning_rate": 9.289533473454192e-07,
+      "loss": 2.1967,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 124
+    },
+    {
+      "epoch": 0.2627430373095113,
+      "grad_norm": 0.3585772049526573,
+      "learning_rate": 9.270426246572272e-07,
+      "loss": 2.3642,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 125
+    },
+    {
+      "epoch": 0.26484498160798736,
+      "grad_norm": 0.34020805834208123,
+      "learning_rate": 9.251085678648071e-07,
+      "loss": 2.237,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 126
+    },
+    {
+      "epoch": 0.26694692590646346,
+      "grad_norm": 0.38311234004852174,
+      "learning_rate": 9.23151282647265e-07,
+      "loss": 2.2439,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 127
+    },
+    {
+      "epoch": 0.26904887020493956,
+      "grad_norm": 0.40490379874064303,
+      "learning_rate": 9.211708759529346e-07,
+      "loss": 2.3447,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 128
+    },
+    {
+      "epoch": 0.27115081450341566,
+      "grad_norm": 0.38814226346705333,
+      "learning_rate": 9.191674559935347e-07,
+      "loss": 2.2642,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 129
+    },
+    {
+      "epoch": 0.27325275880189176,
+      "grad_norm": 0.392535018684069,
+      "learning_rate": 9.171411322382551e-07,
+      "loss": 2.4222,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 130
+    },
+    {
+      "epoch": 0.27535470310036786,
+      "grad_norm": 0.36293069595975763,
+      "learning_rate": 9.150920154077753e-07,
+      "loss": 2.2375,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 131
+    },
+    {
+      "epoch": 0.2774566473988439,
+      "grad_norm": 0.3827224228744126,
+      "learning_rate": 9.130202174682153e-07,
+      "loss": 2.3121,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 132
+    },
+    {
+      "epoch": 0.27955859169732,
+      "grad_norm": 0.39154739093650776,
+      "learning_rate": 9.109258516250171e-07,
+      "loss": 2.3246,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 133
+    },
+    {
+      "epoch": 0.2816605359957961,
+      "grad_norm": 0.35430283896633147,
+      "learning_rate": 9.08809032316759e-07,
+      "loss": 2.2922,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 134
+    },
+    {
+      "epoch": 0.2837624802942722,
+      "grad_norm": 0.39840449294712393,
+      "learning_rate": 9.066698752089028e-07,
+      "loss": 2.34,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 135
+    },
+    {
+      "epoch": 0.2858644245927483,
+      "grad_norm": 0.3657527770786503,
+      "learning_rate": 9.045084971874737e-07,
+      "loss": 2.3127,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 136
+    },
+    {
+      "epoch": 0.2879663688912244,
+      "grad_norm": 0.40390275500061623,
+      "learning_rate": 9.02325016352673e-07,
+      "loss": 2.2761,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 137
+    },
+    {
+      "epoch": 0.2900683131897005,
+      "grad_norm": 0.3486049947257035,
+      "learning_rate": 9.001195520124255e-07,
+      "loss": 2.2909,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 138
+    },
+    {
+      "epoch": 0.29217025748817654,
+      "grad_norm": 0.369271223650673,
+      "learning_rate": 8.978922246758606e-07,
+      "loss": 2.3146,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 139
+    },
+    {
+      "epoch": 0.29427220178665264,
+      "grad_norm": 0.34559908408986584,
+      "learning_rate": 8.956431560467266e-07,
+      "loss": 2.3861,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 140
+    },
+    {
+      "epoch": 0.29637414608512874,
+      "grad_norm": 0.40663130251420515,
+      "learning_rate": 8.933724690167416e-07,
+      "loss": 2.3351,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 141
+    },
+    {
+      "epoch": 0.29847609038360484,
+      "grad_norm": 0.36345254242299446,
+      "learning_rate": 8.910802876588781e-07,
+      "loss": 2.2782,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 142
+    },
+    {
+      "epoch": 0.30057803468208094,
+      "grad_norm": 0.38393881395986873,
+      "learning_rate": 8.887667372205838e-07,
+      "loss": 2.2808,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 143
+    },
+    {
+      "epoch": 0.30267997898055704,
+      "grad_norm": 0.35972360098945216,
+      "learning_rate": 8.864319441169372e-07,
+      "loss": 2.2753,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 144
+    },
+    {
+      "epoch": 0.3047819232790331,
+      "grad_norm": 0.4197014359486705,
+      "learning_rate": 8.840760359237411e-07,
+      "loss": 2.3163,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 145
+    },
+    {
+      "epoch": 0.3068838675775092,
+      "grad_norm": 0.3698464136493578,
+      "learning_rate": 8.816991413705514e-07,
+      "loss": 2.3585,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 146
+    },
+    {
+      "epoch": 0.3089858118759853,
+      "grad_norm": 0.38628726944167563,
+      "learning_rate": 8.793013903336427e-07,
+      "loss": 2.2954,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 147
+    },
+    {
+      "epoch": 0.3110877561744614,
+      "grad_norm": 0.33899721461114324,
+      "learning_rate": 8.768829138289122e-07,
+      "loss": 2.2799,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 148
+    },
+    {
+      "epoch": 0.3131897004729375,
+      "grad_norm": 0.39286568836433555,
+      "learning_rate": 8.744438440047206e-07,
+      "loss": 2.3867,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 149
+    },
+    {
+      "epoch": 0.3152916447714136,
+      "grad_norm": 0.36680644419068636,
+      "learning_rate": 8.719843141346717e-07,
+      "loss": 2.2539,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 150
+    },
+    {
+      "epoch": 0.3173935890698896,
+      "grad_norm": 0.4226555891529418,
+      "learning_rate": 8.695044586103295e-07,
+      "loss": 2.4062,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 151
+    },
+    {
+      "epoch": 0.3194955333683657,
+      "grad_norm": 0.343763121119237,
+      "learning_rate": 8.67004412933876e-07,
+      "loss": 2.2993,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 152
+    },
+    {
+      "epoch": 0.3215974776668418,
+      "grad_norm": 0.34716852552812194,
+      "learning_rate": 8.644843137107057e-07,
+      "loss": 2.3404,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 153
+    },
+    {
+      "epoch": 0.3236994219653179,
+      "grad_norm": 0.3968883598563259,
+      "learning_rate": 8.619442986419629e-07,
+      "loss": 2.3012,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 154
+    },
+    {
+      "epoch": 0.325801366263794,
+      "grad_norm": 0.33889705699000894,
+      "learning_rate": 8.593845065170163e-07,
+      "loss": 2.2621,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 155
+    },
+    {
+      "epoch": 0.3279033105622701,
+      "grad_norm": 0.351512969072057,
+      "learning_rate": 8.568050772058761e-07,
+      "loss": 2.357,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 156
+    },
+    {
+      "epoch": 0.33000525486074617,
+      "grad_norm": 0.3668822383961036,
+      "learning_rate": 8.542061516515511e-07,
+      "loss": 2.3499,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 157
+    },
+    {
+      "epoch": 0.33210719915922227,
+      "grad_norm": 0.3813104081247767,
+      "learning_rate": 8.515878718623473e-07,
+      "loss": 2.3762,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 158
+    },
+    {
+      "epoch": 0.33420914345769837,
+      "grad_norm": 0.3555623840160132,
+      "learning_rate": 8.489503809041087e-07,
+      "loss": 2.2511,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 159
+    },
+    {
+      "epoch": 0.33631108775617446,
+      "grad_norm": 0.3426684159571787,
+      "learning_rate": 8.462938228923999e-07,
+      "loss": 2.3354,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 160
+    },
+    {
+      "epoch": 0.33841303205465056,
+      "grad_norm": 0.3491214060448838,
+      "learning_rate": 8.436183429846313e-07,
+      "loss": 2.2395,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 161
+    },
+    {
+      "epoch": 0.34051497635312666,
+      "grad_norm": 0.4563165572220967,
+      "learning_rate": 8.409240873721276e-07,
+      "loss": 2.3872,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 162
+    },
+    {
+      "epoch": 0.3426169206516027,
+      "grad_norm": 0.33319192852547314,
+      "learning_rate": 8.382112032721398e-07,
+      "loss": 2.3122,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 163
+    },
+    {
+      "epoch": 0.3447188649500788,
+      "grad_norm": 0.38388679911113793,
+      "learning_rate": 8.354798389198012e-07,
+      "loss": 2.3693,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 164
+    },
+    {
+      "epoch": 0.3468208092485549,
+      "grad_norm": 0.39313380831692907,
+      "learning_rate": 8.327301435600272e-07,
+      "loss": 2.3085,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 165
+    },
+    {
+      "epoch": 0.348922753547031,
+      "grad_norm": 0.41915800484281546,
+      "learning_rate": 8.299622674393614e-07,
+      "loss": 2.3851,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 166
+    },
+    {
+      "epoch": 0.3510246978455071,
+      "grad_norm": 0.35317676640002343,
+      "learning_rate": 8.271763617977641e-07,
+      "loss": 2.2271,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 167
+    },
+    {
+      "epoch": 0.3531266421439832,
+      "grad_norm": 0.3528294909167197,
+      "learning_rate": 8.243725788603508e-07,
+      "loss": 2.3087,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 168
+    },
+    {
+      "epoch": 0.35522858644245925,
+      "grad_norm": 0.38320812537843413,
+      "learning_rate": 8.215510718290723e-07,
+      "loss": 2.2441,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 169
+    },
+    {
+      "epoch": 0.35733053074093535,
+      "grad_norm": 0.403856066195845,
+      "learning_rate": 8.187119948743449e-07,
+      "loss": 2.3326,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 170
+    },
+    {
+      "epoch": 0.35943247503941145,
+      "grad_norm": 0.35785694946938973,
+      "learning_rate": 8.158555031266254e-07,
+      "loss": 2.332,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 171
+    },
+    {
+      "epoch": 0.36153441933788755,
+      "grad_norm": 0.34400894235353774,
+      "learning_rate": 8.129817526679357e-07,
+      "loss": 2.2897,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 172
+    },
+    {
+      "epoch": 0.36363636363636365,
+      "grad_norm": 0.4126959309133071,
+      "learning_rate": 8.100909005233334e-07,
+      "loss": 2.3507,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 173
+    },
+    {
+      "epoch": 0.36573830793483975,
+      "grad_norm": 0.42717818377517935,
+      "learning_rate": 8.071831046523318e-07,
+      "loss": 2.3917,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 174
+    },
+    {
+      "epoch": 0.3678402522333158,
+      "grad_norm": 0.3579933408679328,
+      "learning_rate": 8.042585239402697e-07,
+      "loss": 2.2518,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 175
+    },
+    {
+      "epoch": 0.3699421965317919,
+      "grad_norm": 0.39551576662619,
+      "learning_rate": 8.013173181896282e-07,
+      "loss": 2.4125,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 176
+    },
+    {
+      "epoch": 0.372044140830268,
+      "grad_norm": 0.365049869078283,
+      "learning_rate": 7.983596481113005e-07,
+      "loss": 2.2727,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 177
+    },
+    {
+      "epoch": 0.3741460851287441,
+      "grad_norm": 0.4011873410931577,
+      "learning_rate": 7.953856753158094e-07,
+      "loss": 2.3436,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 178
+    },
+    {
+      "epoch": 0.3762480294272202,
+      "grad_norm": 0.3842765318105432,
+      "learning_rate": 7.923955623044775e-07,
+      "loss": 2.3529,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 179
+    },
+    {
+      "epoch": 0.3783499737256963,
+      "grad_norm": 0.3554244239833299,
+      "learning_rate": 7.893894724605468e-07,
+      "loss": 2.2397,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 180
+    },
+    {
+      "epoch": 0.3804519180241724,
+      "grad_norm": 0.4463726239175773,
+      "learning_rate": 7.863675700402526e-07,
+      "loss": 2.3635,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 181
+    },
+    {
+      "epoch": 0.38255386232264843,
+      "grad_norm": 0.36826384823166514,
+      "learning_rate": 7.833300201638474e-07,
+      "loss": 2.3262,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 182
+    },
+    {
+      "epoch": 0.38465580662112453,
+      "grad_norm": 0.40131324496051124,
+      "learning_rate": 7.802769888065789e-07,
+      "loss": 2.3718,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 183
+    },
+    {
+      "epoch": 0.38675775091960063,
+      "grad_norm": 0.3849897857523413,
+      "learning_rate": 7.772086427896211e-07,
+      "loss": 2.2332,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 184
+    },
+    {
+      "epoch": 0.38885969521807673,
+      "grad_norm": 0.36493755016771345,
+      "learning_rate": 7.741251497709583e-07,
+      "loss": 2.3377,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 185
+    },
+    {
+      "epoch": 0.39096163951655283,
+      "grad_norm": 0.35102308079227,
+      "learning_rate": 7.710266782362247e-07,
+      "loss": 2.3105,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 186
+    },
+    {
+      "epoch": 0.3930635838150289,
+      "grad_norm": 0.38998005813653297,
+      "learning_rate": 7.679133974894982e-07,
+      "loss": 2.3349,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 187
+    },
+    {
+      "epoch": 0.39516552811350497,
+      "grad_norm": 0.379125931962091,
+      "learning_rate": 7.647854776440495e-07,
+      "loss": 2.2724,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 188
+    },
+    {
+      "epoch": 0.39726747241198107,
+      "grad_norm": 0.3947787843638888,
+      "learning_rate": 7.616430896130455e-07,
+      "loss": 2.337,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 189
+    },
+    {
+      "epoch": 0.39936941671045717,
+      "grad_norm": 0.37487637035067606,
+      "learning_rate": 7.584864051002126e-07,
+      "loss": 2.3746,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 190
+    },
+    {
+      "epoch": 0.40147136100893327,
+      "grad_norm": 0.366432821290813,
+      "learning_rate": 7.553155965904534e-07,
+      "loss": 2.3042,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 191
+    },
+    {
+      "epoch": 0.40357330530740937,
+      "grad_norm": 0.34566325498775646,
+      "learning_rate": 7.521308373404217e-07,
+      "loss": 2.2799,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 192
+    },
+    {
+      "epoch": 0.40567524960588547,
+      "grad_norm": 0.360571812024321,
+      "learning_rate": 7.489323013690561e-07,
+      "loss": 2.1848,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 193
+    },
+    {
+      "epoch": 0.4077771939043615,
+      "grad_norm": 0.38102049467871574,
+      "learning_rate": 7.457201634480712e-07,
+      "loss": 2.3506,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 194
+    },
+    {
+      "epoch": 0.4098791382028376,
+      "grad_norm": 0.4157458990557322,
+      "learning_rate": 7.424945990924079e-07,
+      "loss": 2.2602,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 195
+    },
+    {
+      "epoch": 0.4119810825013137,
+      "grad_norm": 0.3815988927632132,
+      "learning_rate": 7.392557845506432e-07,
+      "loss": 2.39,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 196
+    },
+    {
+      "epoch": 0.4140830267997898,
+      "grad_norm": 0.3575553199145919,
+      "learning_rate": 7.360038967953597e-07,
+      "loss": 2.3257,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 197
+    },
+    {
+      "epoch": 0.4161849710982659,
+      "grad_norm": 0.37179609481335857,
+      "learning_rate": 7.327391135134749e-07,
+      "loss": 2.3281,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 198
+    },
+    {
+      "epoch": 0.418286915396742,
+      "grad_norm": 0.35686209920084183,
+      "learning_rate": 7.294616130965336e-07,
+      "loss": 2.2884,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 199
+    },
+    {
+      "epoch": 0.42038885969521805,
+      "grad_norm": 0.3966932403444605,
+      "learning_rate": 7.261715746309593e-07,
+      "loss": 2.3668,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 200
+    },
+    {
+      "epoch": 0.42249080399369415,
+      "grad_norm": 0.37119910362955255,
+      "learning_rate": 7.228691778882692e-07,
+      "loss": 2.216,
+      "memory/device_mem_reserved(gib)": 69.0,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 201
+    },
+    {
+      "epoch": 0.42459274829217025,
+      "grad_norm": 0.34835664991688975,
+      "learning_rate": 7.195546033152506e-07,
+      "loss": 2.3013,
+      "memory/device_mem_reserved(gib)": 69.04,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 202
+    },
+    {
+      "epoch": 0.42669469259064635,
+      "grad_norm": 0.36756486717782244,
+      "learning_rate": 7.162280320241019e-07,
+      "loss": 2.2983,
+      "memory/device_mem_reserved(gib)": 69.04,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 203
+    },
+    {
+      "epoch": 0.42879663688912245,
+      "grad_norm": 0.3580849549174155,
+      "learning_rate": 7.128896457825363e-07,
+      "loss": 2.2168,
+      "memory/device_mem_reserved(gib)": 69.04,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 204
+    },
+    {
+      "epoch": 0.43089858118759855,
+      "grad_norm": 0.38919198730377413,
+      "learning_rate": 7.095396270038492e-07,
+      "loss": 2.3673,
+      "memory/device_mem_reserved(gib)": 69.04,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 205
+    },
+    {
+      "epoch": 0.4330005254860746,
+      "grad_norm": 0.45321125836545045,
+      "learning_rate": 7.061781587369518e-07,
+      "loss": 2.2495,
+      "memory/device_mem_reserved(gib)": 69.04,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 206
+    },
+    {
+      "epoch": 0.4351024697845507,
+      "grad_norm": 0.3390158068246942,
+      "learning_rate": 7.028054246563678e-07,
+      "loss": 2.2959,
+      "memory/device_mem_reserved(gib)": 69.04,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 207
+    },
+    {
+      "epoch": 0.4372044140830268,
+      "grad_norm": 0.3932391160329032,
+      "learning_rate": 6.99421609052199e-07,
+      "loss": 2.3348,
+      "memory/device_mem_reserved(gib)": 69.04,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 208
+    },
+    {
+      "epoch": 0.4393063583815029,
+      "grad_norm": 0.35196191595880966,
+      "learning_rate": 6.960268968200538e-07,
+      "loss": 2.3416,
+      "memory/device_mem_reserved(gib)": 69.04,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 209
+    },
+    {
+      "epoch": 0.441408302679979,
+      "grad_norm": 0.3970691259787115,
+      "learning_rate": 6.92621473450945e-07,
+      "loss": 2.3328,
+      "memory/device_mem_reserved(gib)": 69.04,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 210
+    },
+    {
+      "epoch": 0.4435102469784551,
+      "grad_norm": 0.3362178241906251,
+      "learning_rate": 6.892055250211551e-07,
+      "loss": 2.2666,
+      "memory/device_mem_reserved(gib)": 69.04,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 211
+    },
+    {
+      "epoch": 0.44561219127693114,
+      "grad_norm": 0.3370093871143424,
+      "learning_rate": 6.857792381820672e-07,
+      "loss": 2.3654,
+      "memory/device_mem_reserved(gib)": 69.04,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 212
+    },
+    {
+      "epoch": 0.44771413557540723,
+      "grad_norm": 0.3439137104265468,
+      "learning_rate": 6.823428001499676e-07,
+      "loss": 2.3236,
+      "memory/device_mem_reserved(gib)": 69.1,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 213
+    },
+    {
+      "epoch": 0.44981607987388333,
+      "grad_norm": 0.363363512278423,
+      "learning_rate": 6.788963986958152e-07,
+      "loss": 2.3153,
+      "memory/device_mem_reserved(gib)": 69.1,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 214
+    },
+    {
+      "epoch": 0.45191802417235943,
+      "grad_norm": 0.3550856155819428,
+      "learning_rate": 6.754402221349825e-07,
+      "loss": 2.3337,
+      "memory/device_mem_reserved(gib)": 69.1,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 215
+    },
+    {
+      "epoch": 0.45401996847083553,
+      "grad_norm": 0.43364470288014306,
+      "learning_rate": 6.71974459316964e-07,
+      "loss": 2.3817,
+      "memory/device_mem_reserved(gib)": 69.1,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 216
+    },
+    {
+      "epoch": 0.45612191276931163,
+      "grad_norm": 0.3594186404992842,
+      "learning_rate": 6.684992996150598e-07,
+      "loss": 2.282,
+      "memory/device_mem_reserved(gib)": 69.1,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 217
+    },
+    {
+      "epoch": 0.4582238570677877,
+      "grad_norm": 0.348193721582919,
+      "learning_rate": 6.650149329160257e-07,
+      "loss": 2.3266,
+      "memory/device_mem_reserved(gib)": 69.1,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 218
+    },
+    {
+      "epoch": 0.4603258013662638,
+      "grad_norm": 0.36563818617010935,
+      "learning_rate": 6.615215496096986e-07,
+      "loss": 2.2706,
+      "memory/device_mem_reserved(gib)": 69.1,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 219
+    },
+    {
+      "epoch": 0.4624277456647399,
+      "grad_norm": 0.3766707141167757,
+      "learning_rate": 6.580193405785938e-07,
+      "loss": 2.2786,
+      "memory/device_mem_reserved(gib)": 69.1,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 220
+    },
+    {
+      "epoch": 0.464529689963216,
+      "grad_norm": 0.37040693778721345,
+      "learning_rate": 6.545084971874736e-07,
+      "loss": 2.3041,
+      "memory/device_mem_reserved(gib)": 69.1,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 221
+    },
+    {
+      "epoch": 0.4666316342616921,
+      "grad_norm": 0.3814375380964394,
+      "learning_rate": 6.509892112728928e-07,
+      "loss": 2.2896,
+      "memory/device_mem_reserved(gib)": 69.1,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 222
+    },
+    {
+      "epoch": 0.4687335785601682,
+      "grad_norm": 0.38809585401355357,
+      "learning_rate": 6.474616751327142e-07,
+      "loss": 2.407,
+      "memory/device_mem_reserved(gib)": 69.1,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 223
+    },
+    {
+      "epoch": 0.4708355228586443,
+      "grad_norm": 0.3450197035654617,
+      "learning_rate": 6.439260815156038e-07,
+      "loss": 2.3212,
+      "memory/device_mem_reserved(gib)": 69.1,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 224
+    },
+    {
+      "epoch": 0.4729374671571203,
+      "grad_norm": 0.3638524400528564,
+      "learning_rate": 6.403826236104965e-07,
+      "loss": 2.3958,
+      "memory/device_mem_reserved(gib)": 69.1,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 225
+    },
+    {
+      "epoch": 0.4750394114555964,
+      "grad_norm": 11.416588328761524,
+      "learning_rate": 6.368314950360415e-07,
+      "loss": 2.4091,
+      "memory/device_mem_reserved(gib)": 69.1,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 226
+    },
+    {
+      "epoch": 0.4771413557540725,
+      "grad_norm": 0.38009234188930396,
+      "learning_rate": 6.33272889830022e-07,
+      "loss": 2.3481,
+      "memory/device_mem_reserved(gib)": 69.1,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 227
+    },
+    {
+      "epoch": 0.4792433000525486,
+      "grad_norm": 0.3874884323158228,
+      "learning_rate": 6.297070024387534e-07,
+      "loss": 2.2936,
+      "memory/device_mem_reserved(gib)": 69.1,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 228
+    },
+    {
+      "epoch": 0.4813452443510247,
+      "grad_norm": 0.3875050817077963,
+      "learning_rate": 6.261340277064578e-07,
+      "loss": 2.2781,
+      "memory/device_mem_reserved(gib)": 69.1,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 229
+    },
+    {
+      "epoch": 0.4834471886495008,
+      "grad_norm": 0.35862524615310853,
+      "learning_rate": 6.225541608646179e-07,
+      "loss": 2.317,
+      "memory/device_mem_reserved(gib)": 69.1,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 230
+    },
+    {
+      "epoch": 0.48554913294797686,
+      "grad_norm": 0.3684856860526338,
+      "learning_rate": 6.189675975213093e-07,
+      "loss": 2.2496,
+      "memory/device_mem_reserved(gib)": 69.1,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 231
+    },
+    {
+      "epoch": 0.48765107724645296,
+      "grad_norm": 0.3592072791824982,
+      "learning_rate": 6.153745336505124e-07,
+      "loss": 2.3916,
+      "memory/device_mem_reserved(gib)": 69.1,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 232
+    },
+    {
+      "epoch": 0.48975302154492906,
+      "grad_norm": 0.3492976591005929,
+      "learning_rate": 6.117751655814037e-07,
+      "loss": 2.3432,
+      "memory/device_mem_reserved(gib)": 69.1,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 233
+    },
+    {
+      "epoch": 0.49185496584340516,
+      "grad_norm": 0.39657164130018213,
+      "learning_rate": 6.081696899876281e-07,
+      "loss": 2.2399,
+      "memory/device_mem_reserved(gib)": 69.1,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 234
+    },
+    {
+      "epoch": 0.49395691014188126,
+      "grad_norm": 0.33918396528061745,
+      "learning_rate": 6.045583038765537e-07,
+      "loss": 2.2886,
+      "memory/device_mem_reserved(gib)": 69.1,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 235
+    },
+    {
+      "epoch": 0.49605885444035736,
+      "grad_norm": 0.3812617838396709,
+      "learning_rate": 6.009412045785051e-07,
+      "loss": 2.3345,
+      "memory/device_mem_reserved(gib)": 69.1,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 236
+    },
+    {
+      "epoch": 0.4981607987388334,
+      "grad_norm": 0.35316898092715904,
+      "learning_rate": 5.973185897359827e-07,
+      "loss": 2.3495,
+      "memory/device_mem_reserved(gib)": 69.1,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 237
+    },
+    {
+      "epoch": 0.5002627430373096,
+      "grad_norm": 0.35687689358418945,
+      "learning_rate": 5.936906572928624e-07,
+      "loss": 2.3206,
+      "memory/device_mem_reserved(gib)": 69.1,
+      "memory/max_mem_active(gib)": 63.57,
+      "memory/max_mem_allocated(gib)": 62.86,
+      "step": 238
+    }
+  ],
+  "logging_steps": 1,
+  "max_steps": 475,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 1,
+  "save_steps": 238,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 7.365676632322867e+17,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}

training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:841cfbb6b5cef1df538288e0b332863b7380582685cc768f5302c09eddd9f463
+size 9169

zero_to_fp32.py

@@ -0,0 +1,760 @@
+#!/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 . output_dir/
+#   or
+#   python zero_to_fp32.py . output_dir/ --safe_serialization
+
+import argparse
+import torch
+import glob
+import math
+import os
+import re
+import gc
+import json
+import numpy as np
+from tqdm import tqdm
+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, weights_only=False)
+
+        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 tqdm(files, desc='Loading checkpoint shards'):
+        state_dict = torch.load(f, map_location=device, mmap=True, weights_only=False)
+        # 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}")
+
+    fp32_flat_groups = [state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key] for i in range(len(state_dicts))]
+    return zero_stage, world_size, fp32_flat_groups
+
+
+def _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir, exclude_frozen_parameters):
+    """
+    Returns fp32 state_dict reconstructed from ds checkpoint
+
+    Args:
+        - ``ds_checkpoint_dir``: path to the deepspeed checkpoint folder (where the optimizer files are)
+
+    """
+    print(f"Processing zero checkpoint '{ds_checkpoint_dir}'")
+
+    optim_files = get_optim_files(ds_checkpoint_dir)
+    zero_stage, world_size, fp32_flat_groups = parse_optim_states(optim_files, ds_checkpoint_dir)
+    print(f"Detected checkpoint of type zero stage {zero_stage}, world_size: {world_size}")
+
+    model_files = get_model_state_files(ds_checkpoint_dir)
+
+    zero_model_states = parse_model_states(model_files)
+    print(f'Parsing checkpoint created by deepspeed=={zero_model_states[0].ds_version}')
+
+    if zero_stage <= 2:
+        return _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                                          exclude_frozen_parameters)
+    elif zero_stage == 3:
+        return _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                                          exclude_frozen_parameters)
+
+
+def _zero2_merge_frozen_params(state_dict, zero_model_states):
+    if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
+        return
+
+    frozen_param_shapes = zero_model_states[0].frozen_param_shapes
+    frozen_param_fragments = zero_model_states[0].frozen_param_fragments
+
+    if debug:
+        num_elem = sum(s.numel() for s in frozen_param_shapes.values())
+        print(f'rank 0: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
+
+        wanted_params = len(frozen_param_shapes)
+        wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
+        avail_numel = sum([p.numel() for p in frozen_param_fragments.values()])
+        print(f'Frozen params: Have {avail_numel} numels to process.')
+        print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
+
+    total_params = 0
+    total_numel = 0
+    for name, shape in frozen_param_shapes.items():
+        total_params += 1
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+
+        state_dict[name] = frozen_param_fragments[name]
+
+        if debug:
+            print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
+
+    print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _has_callable(obj, fn):
+    attr = getattr(obj, fn, None)
+    return callable(attr)
+
+
+def _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
+    param_shapes = zero_model_states[0].param_shapes
+
+    # Reconstruction protocol:
+    #
+    # XXX: document this
+
+    if debug:
+        for i in range(world_size):
+            for j in range(len(fp32_flat_groups[0])):
+                print(f"{FP32_FLAT_GROUPS}[{i}][{j}].shape={fp32_flat_groups[i][j].shape}")
+
+    # XXX: memory usage doubles here (zero2)
+    num_param_groups = len(fp32_flat_groups[0])
+    merged_single_partition_of_fp32_groups = []
+    for i in range(num_param_groups):
+        merged_partitions = [sd[i] for sd in fp32_flat_groups]
+        full_single_fp32_vector = torch.cat(merged_partitions, 0)
+        merged_single_partition_of_fp32_groups.append(full_single_fp32_vector)
+    avail_numel = sum(
+        [full_single_fp32_vector.numel() for full_single_fp32_vector in merged_single_partition_of_fp32_groups])
+
+    if debug:
+        wanted_params = sum([len(shapes) for shapes in param_shapes])
+        wanted_numel = sum([sum(shape.numel() for shape in shapes.values()) for shapes in param_shapes])
+        # not asserting if there is a mismatch due to possible padding
+        print(f"Have {avail_numel} numels to process.")
+        print(f"Need {wanted_numel} numels in {wanted_params} params.")
+
+    # params
+    # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
+    # out-of-core computing solution
+    total_numel = 0
+    total_params = 0
+    for shapes, full_single_fp32_vector in zip(param_shapes, merged_single_partition_of_fp32_groups):
+        offset = 0
+        avail_numel = full_single_fp32_vector.numel()
+        for name, shape in shapes.items():
+
+            unpartitioned_numel = shape.numel() if _has_callable(shape, 'numel') else math.prod(shape)
+            total_numel += unpartitioned_numel
+            total_params += 1
+
+            if debug:
+                print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
+            state_dict[name] = full_single_fp32_vector.narrow(0, offset, unpartitioned_numel).view(shape)
+            offset += unpartitioned_numel
+
+        # Z2 started to align to 2*world_size to improve nccl performance. Therefore both offset and
+        # avail_numel can differ by anywhere between 0..2*world_size. Due to two unrelated complex
+        # paddings performed in the code it's almost impossible to predict the exact numbers w/o the
+        # live optimizer object, so we are checking that the numbers are within the right range
+        align_to = 2 * world_size
+
+        def zero2_align(x):
+            return align_to * math.ceil(x / align_to)
+
+        if debug:
+            print(f"original offset={offset}, avail_numel={avail_numel}")
+
+        offset = zero2_align(offset)
+        avail_numel = zero2_align(avail_numel)
+
+        if debug:
+            print(f"aligned  offset={offset}, avail_numel={avail_numel}")
+
+        # Sanity check
+        if offset != avail_numel:
+            raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
+
+    print(f"Reconstructed fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                               exclude_frozen_parameters):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    if not exclude_frozen_parameters:
+        _zero2_merge_frozen_params(state_dict, zero_model_states)
+
+    _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
+
+    # recover shared parameters
+    for pair in zero_model_states[0].shared_params:
+        if pair[1] in state_dict:
+            state_dict[pair[0]] = state_dict[pair[1]]
+
+    return state_dict
+
+
+def zero3_partitioned_param_info(unpartitioned_numel, world_size):
+    remainder = unpartitioned_numel % world_size
+    padding_numel = (world_size - remainder) if remainder else 0
+    partitioned_numel = math.ceil(unpartitioned_numel / world_size)
+    return partitioned_numel, padding_numel
+
+
+def _zero3_merge_frozen_params(state_dict, world_size, zero_model_states):
+    if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
+        return
+
+    if debug:
+        for i in range(world_size):
+            num_elem = sum(s.numel() for s in zero_model_states[i].frozen_param_fragments.values())
+            print(f'rank {i}: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
+
+        frozen_param_shapes = zero_model_states[0].frozen_param_shapes
+        wanted_params = len(frozen_param_shapes)
+        wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
+        avail_numel = sum([p.numel() for p in zero_model_states[0].frozen_param_fragments.values()]) * world_size
+        print(f'Frozen params: Have {avail_numel} numels to process.')
+        print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
+
+    total_params = 0
+    total_numel = 0
+    for name, shape in zero_model_states[0].frozen_param_shapes.items():
+        total_params += 1
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+
+        param_frags = tuple(model_state.frozen_param_fragments[name] for model_state in zero_model_states)
+        state_dict[name] = torch.cat(param_frags, 0).narrow(0, 0, unpartitioned_numel).view(shape)
+
+        partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
+
+        if debug:
+            print(
+                f"Frozen params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
+            )
+
+    print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
+
+
+class GatheredTensor:
+    """
+    A pseudo tensor that collects partitioned weights.
+    It is more memory efficient when there are multiple groups.
+    """
+
+    def __init__(self, flat_groups, flat_groups_offset, offset, partitioned_numel, shape):
+        self.flat_groups = flat_groups
+        self.flat_groups_offset = flat_groups_offset
+        self.offset = offset
+        self.partitioned_numel = partitioned_numel
+        self.shape = shape
+        self.dtype = self.flat_groups[0][0].dtype
+
+    def contiguous(self):
+        """
+        Merge partitioned weights from flat_groups into a single tensor.
+        """
+        end_idx = self.offset + self.partitioned_numel
+        world_size = len(self.flat_groups)
+        pad_flat_param_chunks = []
+
+        for rank_i in range(world_size):
+            # for each rank, we need to collect weights from related group/groups
+            flat_groups_at_rank_i = self.flat_groups[rank_i]
+            start_group_id = None
+            end_group_id = None
+            for group_id in range(len(self.flat_groups_offset)):
+                if self.flat_groups_offset[group_id] <= self.offset < self.flat_groups_offset[group_id + 1]:
+                    start_group_id = group_id
+                if self.flat_groups_offset[group_id] < end_idx <= self.flat_groups_offset[group_id + 1]:
+                    end_group_id = group_id
+                    break
+            # collect weights from related group/groups
+            for group_id in range(start_group_id, end_group_id + 1):
+                flat_tensor = flat_groups_at_rank_i[group_id]
+                start_offset = self.offset - self.flat_groups_offset[group_id]
+                end_offset = min(end_idx, self.flat_groups_offset[group_id + 1]) - self.flat_groups_offset[group_id]
+                pad_flat_param_chunks.append(flat_tensor[start_offset:end_offset])
+
+        # collect weights from all ranks
+        pad_flat_param = torch.cat(pad_flat_param_chunks, dim=0)
+        param = pad_flat_param[:self.shape.numel()].view(self.shape).contiguous()
+        return param
+
+
+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 = sum([flat_group.numel() for flat_group in fp32_flat_groups[0]]) * 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
+    flat_groups_offset = [0] + list(np.cumsum([flat_tensor.numel() for flat_tensor in fp32_flat_groups[0]]))
+    for name, shape in tqdm(param_shapes.items(), desc='Gathering sharded weights'):
+        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}"
+            )
+
+        # memory efficient tensor
+        tensor = GatheredTensor(fp32_flat_groups, flat_groups_offset, offset, partitioned_numel, shape)
+        state_dict[name] = tensor
+        offset += partitioned_numel
+
+    offset *= world_size
+
+    # Sanity check
+    if offset != avail_numel:
+        raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
+
+    print(f"Reconstructed Trainable fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                               exclude_frozen_parameters):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    if not exclude_frozen_parameters:
+        _zero3_merge_frozen_params(state_dict, world_size, zero_model_states)
+
+    _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
+
+    # recover shared parameters
+    for pair in zero_model_states[0].shared_params:
+        if pair[1] in state_dict:
+            state_dict[pair[0]] = state_dict[pair[1]]
+
+    return state_dict
+
+
+def to_torch_tensor(state_dict, return_empty_tensor=False):
+    """
+    Convert state_dict of GatheredTensor to torch tensor
+    """
+    torch_state_dict = {}
+    converted_tensors = {}
+    for name, tensor in state_dict.items():
+        tensor_id = id(tensor)
+        if tensor_id in converted_tensors:  # shared tensors
+            shared_tensor = torch_state_dict[converted_tensors[tensor_id]]
+            torch_state_dict[name] = shared_tensor
+        else:
+            converted_tensors[tensor_id] = name
+            if return_empty_tensor:
+                torch_state_dict[name] = torch.empty(tensor.shape, dtype=tensor.dtype)
+            else:
+                torch_state_dict[name] = tensor.contiguous()
+    return torch_state_dict
+
+
+def get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir,
+                                             tag=None,
+                                             exclude_frozen_parameters=False,
+                                             lazy_mode=False):
+    """
+    Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated state_dict that can be loaded with
+    ``load_state_dict()`` and used for training without DeepSpeed or shared with others, for example
+    via a model hub.
+
+    Args:
+        - ``checkpoint_dir``: path to the desired checkpoint folder
+        - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in 'latest' file. e.g., ``global_step14``
+        - ``exclude_frozen_parameters``: exclude frozen parameters
+        - ``lazy_mode``: get state_dict in lazy mode. It returns a dict of pesduo tensor instead of torch tensor, which is more memory efficient.
+          Convert the pesduo tensor to torch tensor by ``.contiguous()``
+
+    Returns:
+        - pytorch ``state_dict``
+
+    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.
+
+    Note: the above usage may not work if your application doesn't have sufficient free CPU memory.
+    You may need to use the offline approach using the ``zero_to_fp32.py`` script that is saved with
+    the checkpoint. Or you can load state_dict in lazy mode ::
+
+        from deepspeed.utils.zero_to_fp32 import get_fp32_state_dict_from_zero_checkpoint
+        state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, lazy_mode=True) # not on cpu
+        for name, lazy_tensor in state_dict.item():
+            tensor = lazy_tensor.contiguous()  # to cpu
+            print(name, tensor)
+            # del tensor to release memory if it no longer in use
+    """
+    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")
+
+    state_dict = _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir, exclude_frozen_parameters)
+    if lazy_mode:
+        return state_dict
+    else:
+        return to_torch_tensor(state_dict)
+
+
+def convert_zero_checkpoint_to_fp32_state_dict(checkpoint_dir,
+                                               output_dir,
+                                               max_shard_size="5GB",
+                                               safe_serialization=False,
+                                               tag=None,
+                                               exclude_frozen_parameters=False):
+    """
+    Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict`` file that can be
+    loaded with ``torch.load(file)`` + ``load_state_dict()`` and used for training without DeepSpeed.
+
+    Args:
+        - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
+        - ``output_dir``: directory to the pytorch fp32 state_dict output files
+        - ``max_shard_size``: the maximum size for a checkpoint before being sharded, default value is 5GB
+        - ``safe_serialization``:  whether to save the model using `safetensors` or the traditional PyTorch way (that uses `pickle`).
+        - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
+        - ``exclude_frozen_parameters``: exclude frozen parameters
+    """
+
+    # Dependency pre-check
+    if safe_serialization:
+        try:
+            from safetensors.torch import save_file
+        except ImportError:
+            print('If you want to use `safe_serialization`, please `pip install safetensors`')
+            raise
+    if max_shard_size is not None:
+        try:
+            from huggingface_hub import split_torch_state_dict_into_shards
+        except ImportError:
+            print('If you want to use `max_shard_size`, please `pip install huggingface_hub`')
+            raise
+
+    # Convert zero checkpoint to state_dict
+    state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir,
+                                                          tag,
+                                                          exclude_frozen_parameters,
+                                                          lazy_mode=True)
+
+    # Shard the model if it is too big.
+    weights_name = "model.safetensors" if safe_serialization else "pytorch_model.bin"
+    if max_shard_size is not None:
+        filename_pattern = weights_name.replace(".bin", "{suffix}.bin").replace(".safetensors", "{suffix}.safetensors")
+        # an memory-efficient approach for sharding
+        empty_state_dict = to_torch_tensor(state_dict, return_empty_tensor=True)
+        state_dict_split = split_torch_state_dict_into_shards(empty_state_dict,
+                                                              filename_pattern=filename_pattern,
+                                                              max_shard_size=max_shard_size)
+    else:
+        from collections import namedtuple
+        StateDictSplit = namedtuple("StateDictSplit", ["is_sharded", "filename_to_tensors"])
+        state_dict_split = StateDictSplit(is_sharded=False,
+                                          filename_to_tensors={weights_name: list(state_dict.keys())})
+
+    # Save the model by shard
+    os.makedirs(output_dir, exist_ok=True)
+    filename_to_tensors = state_dict_split.filename_to_tensors.items()
+    for shard_file, tensors in tqdm(filename_to_tensors, desc="Saving checkpoint shards"):
+        shard_state_dict = {tensor_name: state_dict[tensor_name] for tensor_name in tensors}
+        shard_state_dict = to_torch_tensor(shard_state_dict)
+        output_path = os.path.join(output_dir, shard_file)
+        if safe_serialization:
+            save_file(shard_state_dict, output_path, metadata={"format": "pt"})
+        else:
+            torch.save(shard_state_dict, output_path)
+        # release the memory of current shard
+        for tensor_name in list(shard_state_dict.keys()):
+            del state_dict[tensor_name]
+            del shard_state_dict[tensor_name]
+        del shard_state_dict
+        gc.collect()
+
+    # Save index if sharded
+    if state_dict_split.is_sharded:
+        index = {
+            "metadata": state_dict_split.metadata,
+            "weight_map": state_dict_split.tensor_to_filename,
+        }
+        save_index_file = "model.safetensors.index.json" if safe_serialization else "pytorch_model.bin.index.json"
+        save_index_file = os.path.join(output_dir, save_index_file)
+        with open(save_index_file, "w", encoding="utf-8") as f:
+            content = json.dumps(index, indent=2, sort_keys=True) + "\n"
+            f.write(content)
+
+
+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_dir",
+                        type=str,
+                        help="directory to the pytorch fp32 state_dict output files"
+                        "(e.g. path/checkpoint-12-output/)")
+    parser.add_argument(
+        "--max_shard_size",
+        type=str,
+        default="5GB",
+        help="The maximum size for a checkpoint before being sharded. Checkpoints shard will then be each of size"
+        "lower than this size. If expressed as a string, needs to be digits followed by a unit (like `5MB`"
+        "We default it to 5GB in order for models to be able to run easily on free-tier google colab instances"
+        "without CPU OOM issues.")
+    parser.add_argument(
+        "--safe_serialization",
+        default=False,
+        action='store_true',
+        help="Whether to save the model using `safetensors` or the traditional PyTorch way (that uses `pickle`).")
+    parser.add_argument("-t",
+                        "--tag",
+                        type=str,
+                        default=None,
+                        help="checkpoint tag used as a unique identifier for checkpoint. e.g., global_step1")
+    parser.add_argument("--exclude_frozen_parameters", action='store_true', help="exclude frozen parameters")
+    parser.add_argument("-d", "--debug", action='store_true', help="enable debug")
+    args = parser.parse_args()
+
+    debug = args.debug
+
+    convert_zero_checkpoint_to_fp32_state_dict(args.checkpoint_dir,
+                                               args.output_dir,
+                                               max_shard_size=args.max_shard_size,
+                                               safe_serialization=args.safe_serialization,
+                                               tag=args.tag,
+                                               exclude_frozen_parameters=args.exclude_frozen_parameters)