Upload folder using huggingface_hub

added_tokens.json

@@ -0,0 +1,25 @@
+{
+  "</tool_call>": 151658,
+  "<pad>": 151665,
+  "<tool_call>": 151657,
+  "<|box_end|>": 151649,
+  "<|box_start|>": 151648,
+  "<|endoftext|>": 151643,
+  "<|file_sep|>": 151664,
+  "<|fim_middle|>": 151660,
+  "<|fim_pad|>": 151662,
+  "<|fim_prefix|>": 151659,
+  "<|fim_suffix|>": 151661,
+  "<|im_end|>": 151645,
+  "<|im_start|>": 151644,
+  "<|image_pad|>": 151655,
+  "<|object_ref_end|>": 151647,
+  "<|object_ref_start|>": 151646,
+  "<|quad_end|>": 151651,
+  "<|quad_start|>": 151650,
+  "<|repo_name|>": 151663,
+  "<|video_pad|>": 151656,
+  "<|vision_end|>": 151653,
+  "<|vision_pad|>": 151654,
+  "<|vision_start|>": 151652
+}

config.json

@@ -0,0 +1,32 @@
+{
+  "_name_or_path": "/apdcephfs/share_302520718/share_info/peisongwang/saved_models/saved_models/Qwen2.5-Math-7B-YARN_sft_qwen2.5_math_7B_base_0127_new_veri_3k_raw_20250204_022818/checkpoint-412",
+  "architectures": [
+    "Qwen2ForCausalLM"
+  ],
+  "attention_dropout": 0.0,
+  "bos_token_id": 128245,
+  "eos_token": "<|im_end|>",
+  "eos_token_id": 151645,
+  "hidden_act": "silu",
+  "hidden_size": 3584,
+  "initializer_range": 0.02,
+  "intermediate_size": 18944,
+  "max_position_embeddings": 4096,
+  "max_window_layers": 28,
+  "model_type": "qwen2",
+  "num_attention_heads": 28,
+  "num_hidden_layers": 28,
+  "num_key_value_heads": 4,
+  "pad_token_id": 151665,
+  "rms_norm_eps": 1e-06,
+  "rope_scaling": null,
+  "rope_theta": 10000,
+  "sliding_window": null,
+  "tie_word_embeddings": false,
+  "torch_dtype": "bfloat16",
+  "transformers_version": "4.46.3",
+  "use_cache": false,
+  "use_mrope": false,
+  "use_sliding_window": false,
+  "vocab_size": 151666
+}

generation_config.json

@@ -0,0 +1,6 @@
+{
+  "bos_token_id": 151643,
+  "eos_token_id": 151645,
+  "max_new_tokens": 2048,
+  "transformers_version": "4.46.3"
+}

latest

@@ -0,0 +1 @@
+global_step376

model-00001-of-00004.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:eb5bf902e7ef52acecc9902d538b1d71cc3210309045e3c9cdc9ad6a0745de9e
+size 4874807912

model-00002-of-00004.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0bce25f9a19a4c9a45de39b4e2062105544a710e39b8bba7af189373d0992c20
+size 4932751008

model-00003-of-00004.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0a746af0e1339b14159c2102e509ed9e57958436ba18aad359c9af7fda3c375c
+size 4330865200

model-00004-of-00004.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:27e3f624fa417c5184f69dff839d6b05f7f757729ca48ee06697c85fafcd6d77
+size 1087142016

model.safetensors.index.json

@@ -0,0 +1,346 @@
+{
+  "metadata": {
+    "total_size": 15225527296
+  },
+  "weight_map": {
+    "lm_head.weight": "model-00004-of-00004.safetensors",
+    "model.embed_tokens.weight": "model-00001-of-00004.safetensors",
+    "model.layers.0.input_layernorm.weight": "model-00001-of-00004.safetensors",
+    "model.layers.0.mlp.down_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.0.mlp.gate_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.0.mlp.up_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.0.post_attention_layernorm.weight": "model-00001-of-00004.safetensors",
+    "model.layers.0.self_attn.k_proj.bias": "model-00001-of-00004.safetensors",
+    "model.layers.0.self_attn.k_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.0.self_attn.o_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.0.self_attn.q_proj.bias": "model-00001-of-00004.safetensors",
+    "model.layers.0.self_attn.q_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.0.self_attn.v_proj.bias": "model-00001-of-00004.safetensors",
+    "model.layers.0.self_attn.v_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.1.input_layernorm.weight": "model-00001-of-00004.safetensors",
+    "model.layers.1.mlp.down_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.1.mlp.gate_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.1.mlp.up_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.1.post_attention_layernorm.weight": "model-00001-of-00004.safetensors",
+    "model.layers.1.self_attn.k_proj.bias": "model-00001-of-00004.safetensors",
+    "model.layers.1.self_attn.k_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.1.self_attn.o_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.1.self_attn.q_proj.bias": "model-00001-of-00004.safetensors",
+    "model.layers.1.self_attn.q_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.1.self_attn.v_proj.bias": "model-00001-of-00004.safetensors",
+    "model.layers.1.self_attn.v_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.10.input_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.10.mlp.down_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.10.mlp.gate_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.10.mlp.up_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.10.post_attention_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.10.self_attn.k_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.10.self_attn.k_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.10.self_attn.o_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.10.self_attn.q_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.10.self_attn.q_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.10.self_attn.v_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.10.self_attn.v_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.11.input_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.11.mlp.down_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.11.mlp.gate_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.11.mlp.up_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.11.post_attention_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.11.self_attn.k_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.11.self_attn.k_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.11.self_attn.o_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.11.self_attn.q_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.11.self_attn.q_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.11.self_attn.v_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.11.self_attn.v_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.12.input_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.12.mlp.down_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.12.mlp.gate_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.12.mlp.up_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.12.post_attention_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.12.self_attn.k_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.12.self_attn.k_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.12.self_attn.o_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.12.self_attn.q_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.12.self_attn.q_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.12.self_attn.v_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.12.self_attn.v_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.13.input_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.13.mlp.down_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.13.mlp.gate_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.13.mlp.up_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.13.post_attention_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.13.self_attn.k_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.13.self_attn.k_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.13.self_attn.o_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.13.self_attn.q_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.13.self_attn.q_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.13.self_attn.v_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.13.self_attn.v_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.14.input_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.14.mlp.down_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.14.mlp.gate_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.14.mlp.up_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.14.post_attention_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.14.self_attn.k_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.14.self_attn.k_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.14.self_attn.o_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.14.self_attn.q_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.14.self_attn.q_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.14.self_attn.v_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.14.self_attn.v_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.15.input_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.15.mlp.down_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.15.mlp.gate_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.15.mlp.up_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.15.post_attention_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.15.self_attn.k_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.15.self_attn.k_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.15.self_attn.o_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.15.self_attn.q_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.15.self_attn.q_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.15.self_attn.v_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.15.self_attn.v_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.16.input_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.16.mlp.down_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.16.mlp.gate_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.16.mlp.up_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.16.post_attention_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.16.self_attn.k_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.16.self_attn.k_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.16.self_attn.o_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.16.self_attn.q_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.16.self_attn.q_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.16.self_attn.v_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.16.self_attn.v_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.17.input_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.17.mlp.down_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.17.mlp.gate_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.17.mlp.up_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.17.post_attention_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.17.self_attn.k_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.17.self_attn.k_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.17.self_attn.o_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.17.self_attn.q_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.17.self_attn.q_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.17.self_attn.v_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.17.self_attn.v_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.18.input_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.18.mlp.down_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.18.mlp.gate_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.18.mlp.up_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.18.post_attention_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.18.self_attn.k_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.18.self_attn.k_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.18.self_attn.o_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.18.self_attn.q_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.18.self_attn.q_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.18.self_attn.v_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.18.self_attn.v_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.19.input_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.19.mlp.down_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.19.mlp.gate_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.19.mlp.up_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.19.post_attention_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.19.self_attn.k_proj.bias": "model-00003-of-00004.safetensors",
+    "model.layers.19.self_attn.k_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.19.self_attn.o_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.19.self_attn.q_proj.bias": "model-00003-of-00004.safetensors",
+    "model.layers.19.self_attn.q_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.19.self_attn.v_proj.bias": "model-00003-of-00004.safetensors",
+    "model.layers.19.self_attn.v_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.2.input_layernorm.weight": "model-00001-of-00004.safetensors",
+    "model.layers.2.mlp.down_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.2.mlp.gate_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.2.mlp.up_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.2.post_attention_layernorm.weight": "model-00001-of-00004.safetensors",
+    "model.layers.2.self_attn.k_proj.bias": "model-00001-of-00004.safetensors",
+    "model.layers.2.self_attn.k_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.2.self_attn.o_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.2.self_attn.q_proj.bias": "model-00001-of-00004.safetensors",
+    "model.layers.2.self_attn.q_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.2.self_attn.v_proj.bias": "model-00001-of-00004.safetensors",
+    "model.layers.2.self_attn.v_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.20.input_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.20.mlp.down_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.20.mlp.gate_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.20.mlp.up_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.20.post_attention_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.20.self_attn.k_proj.bias": "model-00003-of-00004.safetensors",
+    "model.layers.20.self_attn.k_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.20.self_attn.o_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.20.self_attn.q_proj.bias": "model-00003-of-00004.safetensors",
+    "model.layers.20.self_attn.q_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.20.self_attn.v_proj.bias": "model-00003-of-00004.safetensors",
+    "model.layers.20.self_attn.v_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.21.input_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.21.mlp.down_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.21.mlp.gate_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.21.mlp.up_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.21.post_attention_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.21.self_attn.k_proj.bias": "model-00003-of-00004.safetensors",
+    "model.layers.21.self_attn.k_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.21.self_attn.o_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.21.self_attn.q_proj.bias": "model-00003-of-00004.safetensors",
+    "model.layers.21.self_attn.q_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.21.self_attn.v_proj.bias": "model-00003-of-00004.safetensors",
+    "model.layers.21.self_attn.v_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.22.input_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.22.mlp.down_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.22.mlp.gate_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.22.mlp.up_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.22.post_attention_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.22.self_attn.k_proj.bias": "model-00003-of-00004.safetensors",
+    "model.layers.22.self_attn.k_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.22.self_attn.o_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.22.self_attn.q_proj.bias": "model-00003-of-00004.safetensors",
+    "model.layers.22.self_attn.q_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.22.self_attn.v_proj.bias": "model-00003-of-00004.safetensors",
+    "model.layers.22.self_attn.v_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.23.input_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.23.mlp.down_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.23.mlp.gate_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.23.mlp.up_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.23.post_attention_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.23.self_attn.k_proj.bias": "model-00003-of-00004.safetensors",
+    "model.layers.23.self_attn.k_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.23.self_attn.o_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.23.self_attn.q_proj.bias": "model-00003-of-00004.safetensors",
+    "model.layers.23.self_attn.q_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.23.self_attn.v_proj.bias": "model-00003-of-00004.safetensors",
+    "model.layers.23.self_attn.v_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.24.input_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.24.mlp.down_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.24.mlp.gate_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.24.mlp.up_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.24.post_attention_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.24.self_attn.k_proj.bias": "model-00003-of-00004.safetensors",
+    "model.layers.24.self_attn.k_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.24.self_attn.o_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.24.self_attn.q_proj.bias": "model-00003-of-00004.safetensors",
+    "model.layers.24.self_attn.q_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.24.self_attn.v_proj.bias": "model-00003-of-00004.safetensors",
+    "model.layers.24.self_attn.v_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.25.input_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.25.mlp.down_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.25.mlp.gate_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.25.mlp.up_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.25.post_attention_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.25.self_attn.k_proj.bias": "model-00003-of-00004.safetensors",
+    "model.layers.25.self_attn.k_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.25.self_attn.o_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.25.self_attn.q_proj.bias": "model-00003-of-00004.safetensors",
+    "model.layers.25.self_attn.q_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.25.self_attn.v_proj.bias": "model-00003-of-00004.safetensors",
+    "model.layers.25.self_attn.v_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.26.input_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.26.mlp.down_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.26.mlp.gate_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.26.mlp.up_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.26.post_attention_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.26.self_attn.k_proj.bias": "model-00003-of-00004.safetensors",
+    "model.layers.26.self_attn.k_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.26.self_attn.o_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.26.self_attn.q_proj.bias": "model-00003-of-00004.safetensors",
+    "model.layers.26.self_attn.q_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.26.self_attn.v_proj.bias": "model-00003-of-00004.safetensors",
+    "model.layers.26.self_attn.v_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.27.input_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.27.mlp.down_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.27.mlp.gate_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.27.mlp.up_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.27.post_attention_layernorm.weight": "model-00003-of-00004.safetensors",
+    "model.layers.27.self_attn.k_proj.bias": "model-00003-of-00004.safetensors",
+    "model.layers.27.self_attn.k_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.27.self_attn.o_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.27.self_attn.q_proj.bias": "model-00003-of-00004.safetensors",
+    "model.layers.27.self_attn.q_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.27.self_attn.v_proj.bias": "model-00003-of-00004.safetensors",
+    "model.layers.27.self_attn.v_proj.weight": "model-00003-of-00004.safetensors",
+    "model.layers.3.input_layernorm.weight": "model-00001-of-00004.safetensors",
+    "model.layers.3.mlp.down_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.3.mlp.gate_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.3.mlp.up_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.3.post_attention_layernorm.weight": "model-00001-of-00004.safetensors",
+    "model.layers.3.self_attn.k_proj.bias": "model-00001-of-00004.safetensors",
+    "model.layers.3.self_attn.k_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.3.self_attn.o_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.3.self_attn.q_proj.bias": "model-00001-of-00004.safetensors",
+    "model.layers.3.self_attn.q_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.3.self_attn.v_proj.bias": "model-00001-of-00004.safetensors",
+    "model.layers.3.self_attn.v_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.4.input_layernorm.weight": "model-00001-of-00004.safetensors",
+    "model.layers.4.mlp.down_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.4.mlp.gate_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.4.mlp.up_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.4.post_attention_layernorm.weight": "model-00001-of-00004.safetensors",
+    "model.layers.4.self_attn.k_proj.bias": "model-00001-of-00004.safetensors",
+    "model.layers.4.self_attn.k_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.4.self_attn.o_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.4.self_attn.q_proj.bias": "model-00001-of-00004.safetensors",
+    "model.layers.4.self_attn.q_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.4.self_attn.v_proj.bias": "model-00001-of-00004.safetensors",
+    "model.layers.4.self_attn.v_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.5.input_layernorm.weight": "model-00001-of-00004.safetensors",
+    "model.layers.5.mlp.down_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.5.mlp.gate_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.5.mlp.up_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.5.post_attention_layernorm.weight": "model-00001-of-00004.safetensors",
+    "model.layers.5.self_attn.k_proj.bias": "model-00001-of-00004.safetensors",
+    "model.layers.5.self_attn.k_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.5.self_attn.o_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.5.self_attn.q_proj.bias": "model-00001-of-00004.safetensors",
+    "model.layers.5.self_attn.q_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.5.self_attn.v_proj.bias": "model-00001-of-00004.safetensors",
+    "model.layers.5.self_attn.v_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.6.input_layernorm.weight": "model-00001-of-00004.safetensors",
+    "model.layers.6.mlp.down_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.6.mlp.gate_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.6.mlp.up_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.6.post_attention_layernorm.weight": "model-00001-of-00004.safetensors",
+    "model.layers.6.self_attn.k_proj.bias": "model-00001-of-00004.safetensors",
+    "model.layers.6.self_attn.k_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.6.self_attn.o_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.6.self_attn.q_proj.bias": "model-00001-of-00004.safetensors",
+    "model.layers.6.self_attn.q_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.6.self_attn.v_proj.bias": "model-00001-of-00004.safetensors",
+    "model.layers.6.self_attn.v_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.7.input_layernorm.weight": "model-00001-of-00004.safetensors",
+    "model.layers.7.mlp.down_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.7.mlp.gate_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.7.mlp.up_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.7.post_attention_layernorm.weight": "model-00001-of-00004.safetensors",
+    "model.layers.7.self_attn.k_proj.bias": "model-00001-of-00004.safetensors",
+    "model.layers.7.self_attn.k_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.7.self_attn.o_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.7.self_attn.q_proj.bias": "model-00001-of-00004.safetensors",
+    "model.layers.7.self_attn.q_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.7.self_attn.v_proj.bias": "model-00001-of-00004.safetensors",
+    "model.layers.7.self_attn.v_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.8.input_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.8.mlp.down_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.8.mlp.gate_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.8.mlp.up_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.8.post_attention_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.8.self_attn.k_proj.bias": "model-00001-of-00004.safetensors",
+    "model.layers.8.self_attn.k_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.8.self_attn.o_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.8.self_attn.q_proj.bias": "model-00001-of-00004.safetensors",
+    "model.layers.8.self_attn.q_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.8.self_attn.v_proj.bias": "model-00001-of-00004.safetensors",
+    "model.layers.8.self_attn.v_proj.weight": "model-00001-of-00004.safetensors",
+    "model.layers.9.input_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.9.mlp.down_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.9.mlp.gate_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.9.mlp.up_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.9.post_attention_layernorm.weight": "model-00002-of-00004.safetensors",
+    "model.layers.9.self_attn.k_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.9.self_attn.k_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.9.self_attn.o_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.9.self_attn.q_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.9.self_attn.q_proj.weight": "model-00002-of-00004.safetensors",
+    "model.layers.9.self_attn.v_proj.bias": "model-00002-of-00004.safetensors",
+    "model.layers.9.self_attn.v_proj.weight": "model-00002-of-00004.safetensors",
+    "model.norm.weight": "model-00003-of-00004.safetensors"
+  }
+}

rng_state_0.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e3865ab44b225a84cdef8a56a07f193e923f6c2cb1e09aee98ed64ef85c639ed
+size 15984

rng_state_1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6257e72c2e7cda4379fbeec9ec66509aaed74cf14e61417761b6e01c6e329859
+size 15984

rng_state_10.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c6e3c3abc5314ec4d345235a009987052d95e8a0fd9e411c3ed3678e67bc0875
+size 15997

rng_state_11.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:90ac37496fb33ff91ed29a1e54bc016750c0942f18e177fb6365ab2ca5f6802e
+size 15997

rng_state_12.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4675ea563799d3b7cf0d0c5be07812551eb2deea0f9dee7debe06b28772d8913
+size 15997

rng_state_13.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cf5c589702ffa8f4c3a5f039e41b819c8c496e72ac96764e636c38ec89fc9894
+size 15997

rng_state_14.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ac9adeb39bf94e3ee34687c97d1db16c159423f96c71e0369b502b2a8e3a1992
+size 15997

rng_state_15.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5d4342945efd1ee15d50acaa53084a340312edc686594af6e63020ce696c5112
+size 15997

rng_state_16.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4e9a8ab9c0c7ef22fdeced8d96f6eeb77011809d5c9c7905413eaaa89544ca60
+size 15997

rng_state_17.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:73179f17caa252e75516078a8b6206803b67a6bcf71331dcfaf6c915f3312e66
+size 15997

rng_state_18.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:36718ff596188e02272f8995dceeffba8aea2b15f87c183d2bb35ff73494e616
+size 15997

rng_state_19.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2c5e7a5c4fa1b4f55a2b169e6f17111789defb7f14000937c0abd007011ed8bf
+size 15997

rng_state_2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:89e174547ec147003fc11cd9f5546ff753874b55807a9c0458cde18da168f8d7
+size 15984

rng_state_20.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d02c16ad1318c21ccf2b684a4f6d306d23138930cdc0f290242642b645f21b9b
+size 15997

rng_state_21.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:32841e4bcddeb3309edafcd924cbdfc39e7afd26b2eb54cbad011dda24a3e500
+size 15997

rng_state_22.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:67c9707e2ff2e45e4ad0bca86707ad0d9497994a7e40379e32b5f419fb449612
+size 15997

rng_state_23.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c9b57b8c5a1e6b8a0c0f95c94654f9b3da52127b8e3b0d2b5c3f457542efcdf3
+size 15997

rng_state_24.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d81987ce1755885e27a974603f8619ca9c18794e797b0dc956b234767fe11e11
+size 15997

rng_state_25.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f311571a69ccbd02dee979b30720fc4cbab94a0ede0eb1e2c8f3f102d883b2f2
+size 15997

rng_state_26.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:36c624004cac8de9ccde6ba5bdf44d4f463fa51e1a1a5b551c1be1e98f794ed1
+size 15997

rng_state_27.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e8ba76f40b6dee5fd3a8c47b446ddc0ba7dacd1b3c0cfb2b9b14238996ed72ee
+size 15997

rng_state_28.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fe38217854d0ff7533528edb83ada94af37505f6283ec014fbf15686282c572a
+size 15997

rng_state_29.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:1a84b7c72cf78f062a1e6d2865d2ec8762116ba019fe01a64550fe2c0b82d454
+size 15997

rng_state_3.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cdf6256678c0cc9ae5152a3d9dd7fac6df68703d551c36d4ac5a0645432703ce
+size 15984

rng_state_30.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:14b5dce8c5363374b240aeef15e82aa93ff4999df856880efdeaa7e59eb37826
+size 15997

rng_state_31.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dd52f47665dfd973fff54b99621bcfa9bc1d7465668170a8f0a295ff5e0448da
+size 15997

rng_state_4.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4fafa67829e9acb85aa859262bfed7166ee8f88b9d1cad25d099204e92a8f04e
+size 15984

rng_state_5.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:379301d5a6aa21be095f95c1aae3170bbc391c9f0931dc9aafa58f86aa4c8e47
+size 15984

rng_state_6.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a767192b6b63edd00ca9259b2e3341b065c130772f4a7da1c3807926745c4385
+size 15984

rng_state_7.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e492e5c7ef3e5e752569c07471c8c376be9b958b90a2bfdbcb286e81a23f74e2
+size 15984

rng_state_8.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0a268e7c1b7baefa6b6c9149d281e2490398214668d41a7d15a44c0c310dddc9
+size 15984

rng_state_9.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4a6f8caa1b30da6eabb8b58d63a4348509fde2a2ed17853623afbbf3f8a36ecd
+size 15984

special_tokens_map.json

@@ -0,0 +1,39 @@
+{
+  "additional_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "bos_token": {
+    "content": "<s>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "eos_token": {
+    "content": "<|im_end|>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": "<pad>",
+  "unk_token": {
+    "content": "<unk>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  }
+}

tokenizer_config.json

@@ -0,0 +1,233 @@
+{
+  "add_bos_token": false,
+  "add_prefix_space": false,
+  "added_tokens_decoder": {
+    "128244": {
+      "content": "<unk>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "128245": {
+      "content": "<s>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151643": {
+      "content": "<|endoftext|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151644": {
+      "content": "<|im_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151645": {
+      "content": "<|im_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151646": {
+      "content": "<|object_ref_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151647": {
+      "content": "<|object_ref_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151648": {
+      "content": "<|box_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151649": {
+      "content": "<|box_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151650": {
+      "content": "<|quad_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151651": {
+      "content": "<|quad_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151652": {
+      "content": "<|vision_start|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151653": {
+      "content": "<|vision_end|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151654": {
+      "content": "<|vision_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151655": {
+      "content": "<|image_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151656": {
+      "content": "<|video_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "151657": {
+      "content": "<tool_call>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151658": {
+      "content": "</tool_call>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151659": {
+      "content": "<|fim_prefix|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151660": {
+      "content": "<|fim_middle|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151661": {
+      "content": "<|fim_suffix|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151662": {
+      "content": "<|fim_pad|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151663": {
+      "content": "<|repo_name|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151664": {
+      "content": "<|file_sep|>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": false
+    },
+    "151665": {
+      "content": "<pad>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    }
+  },
+  "additional_special_tokens": [
+    "<|im_start|>",
+    "<|im_end|>",
+    "<|object_ref_start|>",
+    "<|object_ref_end|>",
+    "<|box_start|>",
+    "<|box_end|>",
+    "<|quad_start|>",
+    "<|quad_end|>",
+    "<|vision_start|>",
+    "<|vision_end|>",
+    "<|vision_pad|>",
+    "<|image_pad|>",
+    "<|video_pad|>"
+  ],
+  "bos_token": "<s>",
+  "chat_template": "{%- if tools %}\n    {{- '<|im_start|>system\\n' }}\n    {%- if messages[0]['role'] == 'system' %}\n        {{- messages[0]['content'] }}\n    {%- else %}\n        {{- 'Please reason step by step, and put your final answer within \\\\boxed{}.' }}\n    {%- endif %}\n    {{- \"\\n\\n# Tools\\n\\nYou may call one or more functions to assist with the user query.\\n\\nYou are provided with function signatures within <tools></tools> XML tags:\\n<tools>\" }}\n    {%- for tool in tools %}\n        {{- \"\\n\" }}\n        {{- tool | tojson }}\n    {%- endfor %}\n    {{- \"\\n</tools>\\n\\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\\n<tool_call>\\n{\\\"name\\\": <function-name>, \\\"arguments\\\": <args-json-object>}\\n</tool_call><|im_end|>\\n\" }}\n{%- else %}\n    {%- if messages[0]['role'] == 'system' %}\n        {{- '<|im_start|>system\\n' + messages[0]['content'] + '<|im_end|>\\n' }}\n    {%- else %}\n        {{- '<|im_start|>system\\nPlease reason step by step, and put your final answer within \\\\boxed{}.<|im_end|>\\n' }}\n    {%- endif %}\n{%- endif %}\n{%- for message in messages %}\n    {%- if (message.role == \"user\") or (message.role == \"system\" and not loop.first) or (message.role == \"assistant\" and not message.tool_calls) %}\n        {{- '<|im_start|>' + message.role + '\\n' + message.content + '<|im_end|>' + '\\n' }}\n    {%- elif message.role == \"assistant\" %}\n        {{- '<|im_start|>' + message.role }}\n        {%- if message.content %}\n            {{- '\\n' + message.content }}\n        {%- endif %}\n        {%- for tool_call in message.tool_calls %}\n            {%- if tool_call.function is defined %}\n                {%- set tool_call = tool_call.function %}\n            {%- endif %}\n            {{- '\\n<tool_call>\\n{\"name\": \"' }}\n            {{- tool_call.name }}\n            {{- '\", \"arguments\": ' }}\n            {{- tool_call.arguments | tojson }}\n            {{- '}\\n</tool_call>' }}\n        {%- endfor %}\n        {{- '<|im_end|>\\n' }}\n    {%- elif message.role == \"tool\" %}\n        {%- if (loop.index0 == 0) or (messages[loop.index0 - 1].role != \"tool\") %}\n            {{- '<|im_start|>user' }}\n        {%- endif %}\n        {{- '\\n<tool_response>\\n' }}\n        {{- message.content }}\n        {{- '\\n</tool_response>' }}\n        {%- if loop.last or (messages[loop.index0 + 1].role != \"tool\") %}\n            {{- '<|im_end|>\\n' }}\n        {%- endif %}\n    {%- endif %}\n{%- endfor %}\n{%- if add_generation_prompt %}\n    {{- '<|im_start|>assistant\\n' }}\n{%- endif %}\n",
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|im_end|>",
+  "errors": "replace",
+  "model_max_length": 8000,
+  "pad_token": "<pad>",
+  "padding_side": "left",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "truncation_side": "left",
+  "unk_token": "<unk>"
+}

trainer_state.json

@@ -0,0 +1,3176 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.27376564277588167,
+  "eval_steps": 500,
+  "global_step": 376,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.0029124004550625713,
+      "importance_ratio": 0.9999997019767761,
+      "kl_div_avg": 0.000832047313451767,
+      "learning_rate": 0.0,
+      "loss_func": "stage2",
+      "step": 1
+    },
+    {
+      "epoch": 0.0029124004550625713,
+      "importance_ratio": 1.0000362396240234,
+      "kl_div_avg": 0.0005252331611700356,
+      "learning_rate": 2.153382790366965e-07,
+      "loss_func": "stage2",
+      "step": 2
+    },
+    {
+      "epoch": 0.0029124004550625713,
+      "importance_ratio": 1.0001579523086548,
+      "kl_div_avg": 0.00018331335741095245,
+      "learning_rate": 3.4130309724299266e-07,
+      "loss_func": "stage2",
+      "step": 3
+    },
+    {
+      "epoch": 0.0029124004550625713,
+      "importance_ratio": 0.9999959468841553,
+      "kl_div_avg": 0.005200453568249941,
+      "learning_rate": 4.30676558073393e-07,
+      "loss_func": "stage2",
+      "step": 4
+    },
+    {
+      "epoch": 0.0058248009101251426,
+      "importance_ratio": 1.0,
+      "kl_div_avg": 0.000816567218862474,
+      "learning_rate": 5e-07,
+      "loss_func": "stage2",
+      "step": 5
+    },
+    {
+      "epoch": 0.0058248009101251426,
+      "importance_ratio": 1.0000829696655273,
+      "kl_div_avg": 0.0006388744222931564,
+      "learning_rate": 5e-07,
+      "loss_func": "stage2",
+      "step": 6
+    },
+    {
+      "epoch": 0.0058248009101251426,
+      "importance_ratio": 0.9999892711639404,
+      "kl_div_avg": 0.0024963156320154667,
+      "learning_rate": 4.999271455631648e-07,
+      "loss_func": "stage2",
+      "step": 7
+    },
+    {
+      "epoch": 0.0058248009101251426,
+      "importance_ratio": 1.0001212358474731,
+      "kl_div_avg": 0.0005678388406522572,
+      "learning_rate": 4.998542911263296e-07,
+      "loss_func": "stage2",
+      "step": 8
+    },
+    {
+      "epoch": 0.008737201365187713,
+      "importance_ratio": 1.0,
+      "kl_div_avg": 0.0009334093192592263,
+      "learning_rate": 4.997814366894943e-07,
+      "loss_func": "stage2",
+      "step": 9,
+      "total_loss": -0.65625
+    },
+    {
+      "epoch": 0.008737201365187713,
+      "importance_ratio": 1.0001897811889648,
+      "kl_div_avg": 0.001136686885729432,
+      "learning_rate": 4.997085822526592e-07,
+      "loss_func": "stage2",
+      "step": 10,
+      "total_loss": -0.66015625
+    },
+    {
+      "epoch": 0.008737201365187713,
+      "importance_ratio": 0.9997982382774353,
+      "kl_div_avg": 0.001998334191739559,
+      "learning_rate": 4.99635727815824e-07,
+      "loss_func": "stage2",
+      "step": 11,
+      "total_loss": 0.578125
+    },
+    {
+      "epoch": 0.008737201365187713,
+      "importance_ratio": 1.0004870891571045,
+      "kl_div_avg": 0.0075031365267932415,
+      "learning_rate": 4.995628733789887e-07,
+      "loss_func": "stage2",
+      "step": 12,
+      "total_loss": 0.609375
+    },
+    {
+      "epoch": 0.011649601820250285,
+      "importance_ratio": 1.0,
+      "kl_div_avg": 0.005973694380372763,
+      "learning_rate": 4.994900189421535e-07,
+      "loss_func": "stage2",
+      "step": 13,
+      "total_loss": 0.490234375
+    },
+    {
+      "epoch": 0.011649601820250285,
+      "importance_ratio": 1.0000145435333252,
+      "kl_div_avg": 0.0021390169858932495,
+      "learning_rate": 4.994171645053183e-07,
+      "loss_func": "stage2",
+      "step": 14
+    },
+    {
+      "epoch": 0.011649601820250285,
+      "importance_ratio": 1.0002269744873047,
+      "kl_div_avg": 0.001650022342801094,
+      "learning_rate": 4.993443100684832e-07,
+      "loss_func": "stage2",
+      "step": 15
+    },
+    {
+      "epoch": 0.011649601820250285,
+      "importance_ratio": 0.9983251690864563,
+      "kl_div_avg": 0.009172441437840462,
+      "learning_rate": 4.99271455631648e-07,
+      "loss_func": "stage2",
+      "step": 16,
+      "total_loss": -1.1796875
+    },
+    {
+      "epoch": 0.014562002275312855,
+      "importance_ratio": 1.0,
+      "kl_div_avg": 0.00243795826099813,
+      "learning_rate": 4.991986011948127e-07,
+      "loss_func": "stage2",
+      "step": 17
+    },
+    {
+      "epoch": 0.014562002275312855,
+      "importance_ratio": 1.000211477279663,
+      "kl_div_avg": 0.0024358637165278196,
+      "learning_rate": 4.991257467579775e-07,
+      "loss_func": "stage2",
+      "step": 18
+    },
+    {
+      "epoch": 0.014562002275312855,
+      "importance_ratio": 1.000443935394287,
+      "kl_div_avg": 0.0023228239733725786,
+      "learning_rate": 4.990528923211423e-07,
+      "loss_func": "stage2",
+      "step": 19
+    },
+    {
+      "epoch": 0.014562002275312855,
+      "importance_ratio": 1.0005327463150024,
+      "kl_div_avg": 0.00688572833314538,
+      "learning_rate": 4.989800378843072e-07,
+      "loss_func": "stage2",
+      "step": 20
+    },
+    {
+      "epoch": 0.017474402730375427,
+      "importance_ratio": 1.0000183582305908,
+      "kl_div_avg": 0.007346789818257093,
+      "learning_rate": 4.98907183447472e-07,
+      "loss_func": "stage2",
+      "step": 21
+    },
+    {
+      "epoch": 0.017474402730375427,
+      "importance_ratio": 1.0002293586730957,
+      "kl_div_avg": 0.00521897291764617,
+      "learning_rate": 4.988343290106367e-07,
+      "loss_func": "stage2",
+      "step": 22,
+      "total_loss": 0.53515625
+    },
+    {
+      "epoch": 0.017474402730375427,
+      "importance_ratio": 1.000298261642456,
+      "kl_div_avg": 0.0083134351298213,
+      "learning_rate": 4.987614745738015e-07,
+      "loss_func": "stage2",
+      "step": 23,
+      "total_loss": 0.185546875
+    },
+    {
+      "epoch": 0.017474402730375427,
+      "importance_ratio": 1.0002543926239014,
+      "kl_div_avg": 0.007159438915550709,
+      "learning_rate": 4.986886201369663e-07,
+      "loss_func": "stage2",
+      "step": 24,
+      "total_loss": -0.91015625
+    },
+    {
+      "epoch": 0.020386803185437997,
+      "importance_ratio": 1.0,
+      "kl_div_avg": 0.0036798412911593914,
+      "learning_rate": 4.986157657001312e-07,
+      "loss_func": "stage2",
+      "step": 25,
+      "total_loss": -1.953125
+    },
+    {
+      "epoch": 0.020386803185437997,
+      "importance_ratio": 1.0004552602767944,
+      "kl_div_avg": 0.004781949333846569,
+      "learning_rate": 4.985429112632959e-07,
+      "loss_func": "stage2",
+      "step": 26,
+      "total_loss": 0.50390625
+    },
+    {
+      "epoch": 0.020386803185437997,
+      "importance_ratio": 1.0013015270233154,
+      "kl_div_avg": 0.006170668639242649,
+      "learning_rate": 4.984700568264607e-07,
+      "loss_func": "stage2",
+      "step": 27
+    },
+    {
+      "epoch": 0.020386803185437997,
+      "importance_ratio": 1.0008548498153687,
+      "kl_div_avg": 0.0056500802747905254,
+      "learning_rate": 4.983972023896255e-07,
+      "loss_func": "stage2",
+      "step": 28,
+      "total_loss": 0.59375
+    },
+    {
+      "epoch": 0.02329920364050057,
+      "importance_ratio": 1.0,
+      "kl_div_avg": 0.012528179213404655,
+      "learning_rate": 4.983243479527903e-07,
+      "loss_func": "stage2",
+      "step": 29,
+      "total_loss": -0.64453125
+    },
+    {
+      "epoch": 0.02329920364050057,
+      "importance_ratio": 1.0004550218582153,
+      "kl_div_avg": 0.012631883844733238,
+      "learning_rate": 4.98251493515955e-07,
+      "loss_func": "stage2",
+      "step": 30,
+      "total_loss": 1.921875
+    },
+    {
+      "epoch": 0.02329920364050057,
+      "importance_ratio": 1.0013270378112793,
+      "kl_div_avg": 0.014947270974516869,
+      "learning_rate": 4.981786390791199e-07,
+      "loss_func": "stage2",
+      "step": 31,
+      "total_loss": -0.6875
+    },
+    {
+      "epoch": 0.02329920364050057,
+      "importance_ratio": 1.0007331371307373,
+      "kl_div_avg": 0.014372318051755428,
+      "learning_rate": 4.981057846422847e-07,
+      "loss_func": "stage2",
+      "step": 32,
+      "total_loss": -0.71484375
+    },
+    {
+      "epoch": 0.02621160409556314,
+      "importance_ratio": 0.9999833106994629,
+      "kl_div_avg": 0.013135651126503944,
+      "learning_rate": 4.980329302054495e-07,
+      "loss_func": "stage2",
+      "step": 33,
+      "total_loss": 0.49609375
+    },
+    {
+      "epoch": 0.02621160409556314,
+      "importance_ratio": 1.0011787414550781,
+      "kl_div_avg": 0.012072667479515076,
+      "learning_rate": 4.979600757686143e-07,
+      "loss_func": "stage2",
+      "step": 34,
+      "total_loss": 0.83984375
+    },
+    {
+      "epoch": 0.02621160409556314,
+      "importance_ratio": 1.0003936290740967,
+      "kl_div_avg": 0.014411704614758492,
+      "learning_rate": 4.97887221331779e-07,
+      "loss_func": "stage2",
+      "step": 35,
+      "total_loss": 0.50390625
+    },
+    {
+      "epoch": 0.02621160409556314,
+      "importance_ratio": 1.0030162334442139,
+      "kl_div_avg": 0.014976001344621181,
+      "learning_rate": 4.978143668949439e-07,
+      "loss_func": "stage2",
+      "step": 36,
+      "total_loss": -1.890625
+    },
+    {
+      "epoch": 0.02912400455062571,
+      "importance_ratio": 0.9999978542327881,
+      "kl_div_avg": 0.01573282852768898,
+      "learning_rate": 4.977415124581087e-07,
+      "loss_func": "stage2",
+      "step": 37
+    },
+    {
+      "epoch": 0.02912400455062571,
+      "importance_ratio": 0.9999271035194397,
+      "kl_div_avg": 0.01654386892914772,
+      "learning_rate": 4.976686580212735e-07,
+      "loss_func": "stage2",
+      "step": 38,
+      "total_loss": 0.359375
+    },
+    {
+      "epoch": 0.02912400455062571,
+      "importance_ratio": 1.0013998746871948,
+      "kl_div_avg": 0.01607615500688553,
+      "learning_rate": 4.975958035844383e-07,
+      "loss_func": "stage2",
+      "step": 39,
+      "total_loss": -0.5625
+    },
+    {
+      "epoch": 0.02912400455062571,
+      "importance_ratio": 1.0015034675598145,
+      "kl_div_avg": 0.028678957372903824,
+      "learning_rate": 4.97522949147603e-07,
+      "loss_func": "stage2",
+      "step": 40,
+      "total_loss": -1.4140625
+    },
+    {
+      "epoch": 0.03203640500568828,
+      "importance_ratio": 0.9999984502792358,
+      "kl_div_avg": 0.020423095673322678,
+      "learning_rate": 4.974500947107679e-07,
+      "loss_func": "stage2",
+      "step": 41
+    },
+    {
+      "epoch": 0.03203640500568828,
+      "importance_ratio": 1.000805139541626,
+      "kl_div_avg": 0.026921488344669342,
+      "learning_rate": 4.973772402739327e-07,
+      "loss_func": "stage2",
+      "step": 42
+    },
+    {
+      "epoch": 0.03203640500568828,
+      "importance_ratio": 1.0019316673278809,
+      "kl_div_avg": 0.024319637566804886,
+      "learning_rate": 4.973043858370975e-07,
+      "loss_func": "stage2",
+      "step": 43
+    },
+    {
+      "epoch": 0.03203640500568828,
+      "importance_ratio": 1.002344012260437,
+      "kl_div_avg": 0.022771870717406273,
+      "learning_rate": 4.972315314002623e-07,
+      "loss_func": "stage2",
+      "step": 44
+    },
+    {
+      "epoch": 0.034948805460750854,
+      "importance_ratio": 1.0,
+      "kl_div_avg": 0.016621489077806473,
+      "learning_rate": 4.97158676963427e-07,
+      "loss_func": "stage2",
+      "step": 45
+    },
+    {
+      "epoch": 0.034948805460750854,
+      "importance_ratio": 1.0002620220184326,
+      "kl_div_avg": 0.016239261254668236,
+      "learning_rate": 4.970858225265919e-07,
+      "loss_func": "stage2",
+      "step": 46
+    },
+    {
+      "epoch": 0.034948805460750854,
+      "importance_ratio": 1.0005873441696167,
+      "kl_div_avg": 0.013311684131622314,
+      "learning_rate": 4.970129680897566e-07,
+      "loss_func": "stage2",
+      "step": 47
+    },
+    {
+      "epoch": 0.034948805460750854,
+      "importance_ratio": 1.0010583400726318,
+      "kl_div_avg": 0.014977280981838703,
+      "learning_rate": 4.969401136529215e-07,
+      "loss_func": "stage2",
+      "step": 48
+    },
+    {
+      "epoch": 0.03786120591581343,
+      "importance_ratio": 0.9999948740005493,
+      "kl_div_avg": 0.026462934911251068,
+      "learning_rate": 4.968672592160862e-07,
+      "loss_func": "stage2",
+      "step": 49
+    },
+    {
+      "epoch": 0.03786120591581343,
+      "importance_ratio": 1.0000110864639282,
+      "kl_div_avg": 0.02019820362329483,
+      "learning_rate": 4.96794404779251e-07,
+      "loss_func": "stage2",
+      "step": 50,
+      "total_loss": 0.27734375
+    },
+    {
+      "epoch": 0.03786120591581343,
+      "importance_ratio": 1.000244379043579,
+      "kl_div_avg": 0.017759989947080612,
+      "learning_rate": 4.967215503424158e-07,
+      "loss_func": "stage2",
+      "step": 51,
+      "total_loss": -0.447265625
+    },
+    {
+      "epoch": 0.03786120591581343,
+      "importance_ratio": 1.0000954866409302,
+      "kl_div_avg": 0.0189061239361763,
+      "learning_rate": 4.966486959055806e-07,
+      "loss_func": "stage2",
+      "step": 52,
+      "total_loss": -1.28125
+    },
+    {
+      "epoch": 0.040773606370875994,
+      "importance_ratio": 1.0,
+      "kl_div_avg": 0.04083386808633804,
+      "learning_rate": 4.965758414687455e-07,
+      "loss_func": "stage2",
+      "step": 53
+    },
+    {
+      "epoch": 0.040773606370875994,
+      "importance_ratio": 0.9995175004005432,
+      "kl_div_avg": 0.03316584229469299,
+      "learning_rate": 4.965029870319102e-07,
+      "loss_func": "stage2",
+      "step": 54,
+      "total_loss": 1.5
+    },
+    {
+      "epoch": 0.040773606370875994,
+      "importance_ratio": 0.999799370765686,
+      "kl_div_avg": 0.02658572420477867,
+      "learning_rate": 4.96430132595075e-07,
+      "loss_func": "stage2",
+      "step": 55,
+      "total_loss": 1.15625
+    },
+    {
+      "epoch": 0.040773606370875994,
+      "importance_ratio": 0.9997727870941162,
+      "kl_div_avg": 0.02100951597094536,
+      "learning_rate": 4.963572781582398e-07,
+      "loss_func": "stage2",
+      "step": 56,
+      "total_loss": -1.109375
+    },
+    {
+      "epoch": 0.04368600682593857,
+      "importance_ratio": 1.0,
+      "kl_div_avg": 0.04962018132209778,
+      "learning_rate": 4.962844237214046e-07,
+      "loss_func": "stage2",
+      "step": 57
+    },
+    {
+      "epoch": 0.04368600682593857,
+      "importance_ratio": 0.9993993043899536,
+      "kl_div_avg": 0.027939818799495697,
+      "learning_rate": 4.962115692845694e-07,
+      "loss_func": "stage2",
+      "step": 58,
+      "total_loss": -2.609375
+    },
+    {
+      "epoch": 0.04368600682593857,
+      "importance_ratio": 0.9994364976882935,
+      "kl_div_avg": 0.03141516447067261,
+      "learning_rate": 4.961387148477342e-07,
+      "loss_func": "stage2",
+      "step": 59,
+      "total_loss": 0.8984375
+    },
+    {
+      "epoch": 0.04368600682593857,
+      "importance_ratio": 0.9994009137153625,
+      "kl_div_avg": 0.022545481100678444,
+      "learning_rate": 4.96065860410899e-07,
+      "loss_func": "stage2",
+      "step": 60,
+      "total_loss": 0.734375
+    },
+    {
+      "epoch": 0.04659840728100114,
+      "importance_ratio": 1.0,
+      "kl_div_avg": 0.03172942250967026,
+      "learning_rate": 4.959930059740638e-07,
+      "loss_func": "stage2",
+      "step": 61
+    },
+    {
+      "epoch": 0.04659840728100114,
+      "importance_ratio": 0.9994012117385864,
+      "kl_div_avg": 0.020564470440149307,
+      "learning_rate": 4.959201515372286e-07,
+      "loss_func": "stage2",
+      "step": 62
+    },
+    {
+      "epoch": 0.04659840728100114,
+      "importance_ratio": 0.9986348152160645,
+      "kl_div_avg": 0.02474614419043064,
+      "learning_rate": 4.958472971003934e-07,
+      "loss_func": "stage2",
+      "step": 63
+    },
+    {
+      "epoch": 0.04659840728100114,
+      "importance_ratio": 0.9994561672210693,
+      "kl_div_avg": 0.02151934802532196,
+      "learning_rate": 4.957744426635582e-07,
+      "loss_func": "stage2",
+      "step": 64
+    },
+    {
+      "epoch": 0.04951080773606371,
+      "importance_ratio": 1.0,
+      "kl_div_avg": 0.03217197209596634,
+      "learning_rate": 4.95701588226723e-07,
+      "loss_func": "stage2",
+      "step": 65
+    },
+    {
+      "epoch": 0.04951080773606371,
+      "importance_ratio": 1.0001919269561768,
+      "kl_div_avg": 0.020508602261543274,
+      "learning_rate": 4.956287337898878e-07,
+      "loss_func": "stage2",
+      "step": 66
+    },
+    {
+      "epoch": 0.04951080773606371,
+      "importance_ratio": 1.0006890296936035,
+      "kl_div_avg": 0.02224303036928177,
+      "learning_rate": 4.955558793530525e-07,
+      "loss_func": "stage2",
+      "step": 67
+    },
+    {
+      "epoch": 0.04951080773606371,
+      "importance_ratio": 1.0008221864700317,
+      "kl_div_avg": 0.020413096994161606,
+      "learning_rate": 4.954830249162173e-07,
+      "loss_func": "stage2",
+      "step": 68
+    },
+    {
+      "epoch": 0.05242320819112628,
+      "importance_ratio": 0.9999940395355225,
+      "kl_div_avg": 0.022445013746619225,
+      "learning_rate": 4.954101704793822e-07,
+      "loss_func": "stage2",
+      "step": 69
+    },
+    {
+      "epoch": 0.05242320819112628,
+      "importance_ratio": 0.999756932258606,
+      "kl_div_avg": 0.014056399464607239,
+      "learning_rate": 4.95337316042547e-07,
+      "loss_func": "stage2",
+      "step": 70
+    },
+    {
+      "epoch": 0.05242320819112628,
+      "importance_ratio": 0.9997715950012207,
+      "kl_div_avg": 0.012785263359546661,
+      "learning_rate": 4.952644616057118e-07,
+      "loss_func": "stage2",
+      "step": 71
+    },
+    {
+      "epoch": 0.05242320819112628,
+      "importance_ratio": 0.9999353885650635,
+      "kl_div_avg": 0.024628497660160065,
+      "learning_rate": 4.951916071688765e-07,
+      "loss_func": "stage2",
+      "step": 72
+    },
+    {
+      "epoch": 0.055335608646188854,
+      "importance_ratio": 1.000006914138794,
+      "kl_div_avg": 0.02687786892056465,
+      "learning_rate": 4.951187527320413e-07,
+      "loss_func": "stage2",
+      "step": 73,
+      "total_loss": 1.2734375
+    },
+    {
+      "epoch": 0.055335608646188854,
+      "importance_ratio": 0.9989913702011108,
+      "kl_div_avg": 0.03032485395669937,
+      "learning_rate": 4.950458982952062e-07,
+      "loss_func": "stage2",
+      "step": 74,
+      "total_loss": -0.8046875
+    },
+    {
+      "epoch": 0.055335608646188854,
+      "importance_ratio": 1.0003002882003784,
+      "kl_div_avg": 0.02986188791692257,
+      "learning_rate": 4.94973043858371e-07,
+      "loss_func": "stage2",
+      "step": 75,
+      "total_loss": -0.2177734375
+    },
+    {
+      "epoch": 0.055335608646188854,
+      "importance_ratio": 1.0016403198242188,
+      "kl_div_avg": 0.024808024987578392,
+      "learning_rate": 4.949001894215358e-07,
+      "loss_func": "stage2",
+      "step": 76
+    },
+    {
+      "epoch": 0.05824800910125142,
+      "importance_ratio": 1.0,
+      "kl_div_avg": 0.022816969081759453,
+      "learning_rate": 4.948273349847005e-07,
+      "loss_func": "stage2",
+      "step": 77
+    },
+    {
+      "epoch": 0.05824800910125142,
+      "importance_ratio": 1.0000745058059692,
+      "kl_div_avg": 0.022778337821364403,
+      "learning_rate": 4.947544805478653e-07,
+      "loss_func": "stage2",
+      "step": 78
+    },
+    {
+      "epoch": 0.05824800910125142,
+      "importance_ratio": 1.00030517578125,
+      "kl_div_avg": 0.024456456303596497,
+      "learning_rate": 4.946816261110302e-07,
+      "loss_func": "stage2",
+      "step": 79
+    },
+    {
+      "epoch": 0.05824800910125142,
+      "importance_ratio": 0.9999829530715942,
+      "kl_div_avg": 0.010129483416676521,
+      "learning_rate": 4.94608771674195e-07,
+      "loss_func": "stage2",
+      "step": 80
+    },
+    {
+      "epoch": 0.061160409556313994,
+      "importance_ratio": 1.0,
+      "kl_div_avg": 0.01612446829676628,
+      "learning_rate": 4.945359172373597e-07,
+      "loss_func": "stage2",
+      "step": 81,
+      "total_loss": 0.92578125
+    },
+    {
+      "epoch": 0.061160409556313994,
+      "importance_ratio": 1.0004006624221802,
+      "kl_div_avg": 0.03181140497326851,
+      "learning_rate": 4.944630628005245e-07,
+      "loss_func": "stage2",
+      "step": 82,
+      "total_loss": -0.39453125
+    },
+    {
+      "epoch": 0.061160409556313994,
+      "importance_ratio": 1.0004996061325073,
+      "kl_div_avg": 0.03759714588522911,
+      "learning_rate": 4.943902083636893e-07,
+      "loss_func": "stage2",
+      "step": 83
+    },
+    {
+      "epoch": 0.061160409556313994,
+      "importance_ratio": 1.000779390335083,
+      "kl_div_avg": 0.03193162754178047,
+      "learning_rate": 4.943173539268542e-07,
+      "loss_func": "stage2",
+      "step": 84,
+      "total_loss": -0.98828125
+    },
+    {
+      "epoch": 0.06407281001137656,
+      "importance_ratio": 1.0,
+      "kl_div_avg": 0.03282681852579117,
+      "learning_rate": 4.942444994900189e-07,
+      "loss_func": "stage2",
+      "step": 85
+    },
+    {
+      "epoch": 0.06407281001137656,
+      "importance_ratio": 1.000492811203003,
+      "kl_div_avg": 0.030602332204580307,
+      "learning_rate": 4.941716450531837e-07,
+      "loss_func": "stage2",
+      "step": 86
+    },
+    {
+      "epoch": 0.06407281001137656,
+      "importance_ratio": 0.9995746612548828,
+      "kl_div_avg": 0.020411640405654907,
+      "learning_rate": 4.940987906163485e-07,
+      "loss_func": "stage2",
+      "step": 87
+    },
+    {
+      "epoch": 0.06407281001137656,
+      "importance_ratio": 1.0005005598068237,
+      "kl_div_avg": 0.03345700725913048,
+      "learning_rate": 4.940259361795133e-07,
+      "loss_func": "stage2",
+      "step": 88
+    },
+    {
+      "epoch": 0.06698521046643914,
+      "importance_ratio": 1.0,
+      "kl_div_avg": 0.03304775059223175,
+      "learning_rate": 4.939530817426781e-07,
+      "loss_func": "stage2",
+      "step": 89
+    },
+    {
+      "epoch": 0.06698521046643914,
+      "importance_ratio": 1.0001821517944336,
+      "kl_div_avg": 0.023516833782196045,
+      "learning_rate": 4.938802273058428e-07,
+      "loss_func": "stage2",
+      "step": 90
+    },
+    {
+      "epoch": 0.06698521046643914,
+      "importance_ratio": 0.9999701976776123,
+      "kl_div_avg": 0.019087690860033035,
+      "learning_rate": 4.938073728690077e-07,
+      "loss_func": "stage2",
+      "step": 91
+    },
+    {
+      "epoch": 0.06698521046643914,
+      "importance_ratio": 1.0003143548965454,
+      "kl_div_avg": 0.030453892424702644,
+      "learning_rate": 4.937345184321725e-07,
+      "loss_func": "stage2",
+      "step": 92
+    },
+    {
+      "epoch": 0.06989761092150171,
+      "importance_ratio": 0.9999972581863403,
+      "kl_div_avg": 0.017626766115427017,
+      "learning_rate": 4.936616639953373e-07,
+      "loss_func": "stage2",
+      "step": 93,
+      "total_loss": -0.400390625
+    },
+    {
+      "epoch": 0.06989761092150171,
+      "importance_ratio": 1.0001485347747803,
+      "kl_div_avg": 0.02021140605211258,
+      "learning_rate": 4.935888095585021e-07,
+      "loss_func": "stage2",
+      "step": 94,
+      "total_loss": -0.330078125
+    },
+    {
+      "epoch": 0.06989761092150171,
+      "importance_ratio": 1.0001258850097656,
+      "kl_div_avg": 0.016257211565971375,
+      "learning_rate": 4.935159551216668e-07,
+      "loss_func": "stage2",
+      "step": 95,
+      "total_loss": -0.056396484375
+    },
+    {
+      "epoch": 0.06989761092150171,
+      "importance_ratio": 1.000044584274292,
+      "kl_div_avg": 0.013729836791753769,
+      "learning_rate": 4.934431006848317e-07,
+      "loss_func": "stage2",
+      "step": 96,
+      "total_loss": -2.09375
+    },
+    {
+      "epoch": 0.07281001137656427,
+      "importance_ratio": 1.0,
+      "kl_div_avg": 0.016564857214689255,
+      "learning_rate": 4.933702462479965e-07,
+      "loss_func": "stage2",
+      "step": 97,
+      "total_loss": -0.55859375
+    },
+    {
+      "epoch": 0.07281001137656427,
+      "importance_ratio": 1.0003982782363892,
+      "kl_div_avg": 0.020946823060512543,
+      "learning_rate": 4.932973918111613e-07,
+      "loss_func": "stage2",
+      "step": 98,
+      "total_loss": -0.330078125
+    },
+    {
+      "epoch": 0.07281001137656427,
+      "importance_ratio": 1.0006924867630005,
+      "kl_div_avg": 0.02827777899801731,
+      "learning_rate": 4.932245373743261e-07,
+      "loss_func": "stage2",
+      "step": 99,
+      "total_loss": 1.0078125
+    },
+    {
+      "epoch": 0.07281001137656427,
+      "importance_ratio": 1.0015332698822021,
+      "kl_div_avg": 0.022303098812699318,
+      "learning_rate": 4.931516829374908e-07,
+      "loss_func": "stage2",
+      "step": 100,
+      "total_loss": -0.447265625
+    },
+    {
+      "epoch": 0.07572241183162685,
+      "importance_ratio": 1.0,
+      "kl_div_avg": 0.025843780487775803,
+      "learning_rate": 4.930788285006557e-07,
+      "loss_func": "stage2",
+      "step": 101
+    },
+    {
+      "epoch": 0.07572241183162685,
+      "importance_ratio": 1.0007214546203613,
+      "kl_div_avg": 0.02605029195547104,
+      "learning_rate": 4.930059740638205e-07,
+      "loss_func": "stage2",
+      "step": 102
+    },
+    {
+      "epoch": 0.07572241183162685,
+      "importance_ratio": 1.0010168552398682,
+      "kl_div_avg": 0.014528175815939903,
+      "learning_rate": 4.929331196269853e-07,
+      "loss_func": "stage2",
+      "step": 103
+    },
+    {
+      "epoch": 0.07572241183162685,
+      "importance_ratio": 1.0012537240982056,
+      "kl_div_avg": 0.017674319446086884,
+      "learning_rate": 4.9286026519015e-07,
+      "loss_func": "stage2",
+      "step": 104
+    },
+    {
+      "epoch": 0.07863481228668942,
+      "importance_ratio": 1.0,
+      "kl_div_avg": 0.0374433770775795,
+      "learning_rate": 4.927874107533148e-07,
+      "loss_func": "stage2",
+      "step": 105
+    },
+    {
+      "epoch": 0.07863481228668942,
+      "importance_ratio": 1.0003745555877686,
+      "kl_div_avg": 0.03026525117456913,
+      "learning_rate": 4.927145563164796e-07,
+      "loss_func": "stage2",
+      "step": 106
+    },
+    {
+      "epoch": 0.07863481228668942,
+      "importance_ratio": 1.0008736848831177,
+      "kl_div_avg": 0.032609131187200546,
+      "learning_rate": 4.926417018796445e-07,
+      "loss_func": "stage2",
+      "step": 107
+    },
+    {
+      "epoch": 0.07863481228668942,
+      "importance_ratio": 1.0013196468353271,
+      "kl_div_avg": 0.030896253883838654,
+      "learning_rate": 4.925688474428093e-07,
+      "loss_func": "stage2",
+      "step": 108
+    },
+    {
+      "epoch": 0.08154721274175199,
+      "importance_ratio": 1.0,
+      "kl_div_avg": 0.014346327632665634,
+      "learning_rate": 4.92495993005974e-07,
+      "loss_func": "stage2",
+      "step": 109
+    },
+    {
+      "epoch": 0.08154721274175199,
+      "importance_ratio": 1.0002002716064453,
+      "kl_div_avg": 0.05834633857011795,
+      "learning_rate": 4.924231385691388e-07,
+      "loss_func": "stage2",
+      "step": 110
+    },
+    {
+      "epoch": 0.08154721274175199,
+      "importance_ratio": 1.0007352828979492,
+      "kl_div_avg": 0.035077303647994995,
+      "learning_rate": 4.923502841323036e-07,
+      "loss_func": "stage2",
+      "step": 111
+    },
+    {
+      "epoch": 0.08154721274175199,
+      "importance_ratio": 1.000758409500122,
+      "kl_div_avg": 0.03604833036661148,
+      "learning_rate": 4.922774296954685e-07,
+      "loss_func": "stage2",
+      "step": 112
+    },
+    {
+      "epoch": 0.08445961319681457,
+      "importance_ratio": 1.0,
+      "kl_div_avg": 0.052136775106191635,
+      "learning_rate": 4.922045752586333e-07,
+      "loss_func": "stage2",
+      "step": 113
+    },
+    {
+      "epoch": 0.08445961319681457,
+      "importance_ratio": 0.9997671842575073,
+      "kl_div_avg": 0.03220074623823166,
+      "learning_rate": 4.92131720821798e-07,
+      "loss_func": "stage2",
+      "step": 114
+    },
+    {
+      "epoch": 0.08445961319681457,
+      "importance_ratio": 1.0004247426986694,
+      "kl_div_avg": 0.08191032707691193,
+      "learning_rate": 4.920588663849628e-07,
+      "loss_func": "stage2",
+      "step": 115
+    },
+    {
+      "epoch": 0.08445961319681457,
+      "importance_ratio": 1.0003342628479004,
+      "kl_div_avg": 0.048977646976709366,
+      "learning_rate": 4.919860119481276e-07,
+      "loss_func": "stage2",
+      "step": 116
+    },
+    {
+      "epoch": 0.08737201365187713,
+      "importance_ratio": 1.0,
+      "kl_div_avg": 0.03430851176381111,
+      "learning_rate": 4.919131575112925e-07,
+      "loss_func": "stage2",
+      "step": 117
+    },
+    {
+      "epoch": 0.08737201365187713,
+      "importance_ratio": 1.0001081228256226,
+      "kl_div_avg": 0.02755298838019371,
+      "learning_rate": 4.918403030744572e-07,
+      "loss_func": "stage2",
+      "step": 118
+    },
+    {
+      "epoch": 0.08737201365187713,
+      "importance_ratio": 1.001303791999817,
+      "kl_div_avg": 0.021847058087587357,
+      "learning_rate": 4.91767448637622e-07,
+      "loss_func": "stage2",
+      "step": 119
+    },
+    {
+      "epoch": 0.08737201365187713,
+      "importance_ratio": 1.002398133277893,
+      "kl_div_avg": 0.027827974408864975,
+      "learning_rate": 4.916945942007868e-07,
+      "loss_func": "stage2",
+      "step": 120
+    },
+    {
+      "epoch": 0.0902844141069397,
+      "importance_ratio": 1.0,
+      "kl_div_avg": 0.030544161796569824,
+      "learning_rate": 4.916217397639516e-07,
+      "loss_func": "stage2",
+      "step": 121
+    },
+    {
+      "epoch": 0.0902844141069397,
+      "importance_ratio": 0.9998140335083008,
+      "kl_div_avg": 0.022375673055648804,
+      "learning_rate": 4.915488853271165e-07,
+      "loss_func": "stage2",
+      "step": 122,
+      "total_loss": -0.2255859375
+    },
+    {
+      "epoch": 0.0902844141069397,
+      "importance_ratio": 0.999756932258606,
+      "kl_div_avg": 0.024012772366404533,
+      "learning_rate": 4.914760308902812e-07,
+      "loss_func": "stage2",
+      "step": 123,
+      "total_loss": 1.9375
+    },
+    {
+      "epoch": 0.0902844141069397,
+      "importance_ratio": 0.9983590245246887,
+      "kl_div_avg": 0.030688390135765076,
+      "learning_rate": 4.91403176453446e-07,
+      "loss_func": "stage2",
+      "step": 124,
+      "total_loss": -0.3125
+    },
+    {
+      "epoch": 0.09319681456200228,
+      "importance_ratio": 1.0000343322753906,
+      "kl_div_avg": 0.0227835550904274,
+      "learning_rate": 4.913303220166108e-07,
+      "loss_func": "stage2",
+      "step": 125,
+      "total_loss": 1.390625
+    },
+    {
+      "epoch": 0.09319681456200228,
+      "importance_ratio": 0.998842179775238,
+      "kl_div_avg": 0.039185911417007446,
+      "learning_rate": 4.912574675797756e-07,
+      "loss_func": "stage2",
+      "step": 126,
+      "total_loss": -1.1328125
+    },
+    {
+      "epoch": 0.09319681456200228,
+      "importance_ratio": 0.9984805583953857,
+      "kl_div_avg": 0.03698894754052162,
+      "learning_rate": 4.911846131429404e-07,
+      "loss_func": "stage2",
+      "step": 127
+    },
+    {
+      "epoch": 0.09319681456200228,
+      "importance_ratio": 0.9969534277915955,
+      "kl_div_avg": 0.03554018214344978,
+      "learning_rate": 4.911117587061051e-07,
+      "loss_func": "stage2",
+      "step": 128,
+      "total_loss": -1.3359375
+    },
+    {
+      "epoch": 0.09610921501706485,
+      "importance_ratio": 1.0,
+      "kl_div_avg": 0.040138669312000275,
+      "learning_rate": 4.9103890426927e-07,
+      "loss_func": "stage2",
+      "step": 129
+    },
+    {
+      "epoch": 0.09610921501706485,
+      "importance_ratio": 0.9992395639419556,
+      "kl_div_avg": 0.022615844383835793,
+      "learning_rate": 4.909660498324348e-07,
+      "loss_func": "stage2",
+      "step": 130
+    },
+    {
+      "epoch": 0.09610921501706485,
+      "importance_ratio": 0.9979618787765503,
+      "kl_div_avg": 0.03186524659395218,
+      "learning_rate": 4.908931953955996e-07,
+      "loss_func": "stage2",
+      "step": 131
+    },
+    {
+      "epoch": 0.09610921501706485,
+      "importance_ratio": 0.9972538948059082,
+      "kl_div_avg": 0.02623588964343071,
+      "learning_rate": 4.908203409587643e-07,
+      "loss_func": "stage2",
+      "step": 132
+    },
+    {
+      "epoch": 0.09902161547212741,
+      "importance_ratio": 1.0,
+      "kl_div_avg": 0.041892606765031815,
+      "learning_rate": 4.907474865219291e-07,
+      "loss_func": "stage2",
+      "step": 133
+    },
+    {
+      "epoch": 0.09902161547212741,
+      "importance_ratio": 0.9995802640914917,
+      "kl_div_avg": 0.030310548841953278,
+      "learning_rate": 4.90674632085094e-07,
+      "loss_func": "stage2",
+      "step": 134
+    },
+    {
+      "epoch": 0.09902161547212741,
+      "importance_ratio": 0.9992105960845947,
+      "kl_div_avg": 0.06457407772541046,
+      "learning_rate": 4.906017776482588e-07,
+      "loss_func": "stage2",
+      "step": 135
+    },
+    {
+      "epoch": 0.09902161547212741,
+      "importance_ratio": 0.9980916380882263,
+      "kl_div_avg": 0.03993566334247589,
+      "learning_rate": 4.905289232114235e-07,
+      "loss_func": "stage2",
+      "step": 136
+    },
+    {
+      "epoch": 0.10193401592719,
+      "importance_ratio": 1.0,
+      "kl_div_avg": 0.021149709820747375,
+      "learning_rate": 4.904560687745883e-07,
+      "loss_func": "stage2",
+      "step": 137
+    },
+    {
+      "epoch": 0.10193401592719,
+      "importance_ratio": 0.9998972415924072,
+      "kl_div_avg": 0.04991449788212776,
+      "learning_rate": 4.903832143377531e-07,
+      "loss_func": "stage2",
+      "step": 138
+    },
+    {
+      "epoch": 0.10193401592719,
+      "importance_ratio": 0.9993065595626831,
+      "kl_div_avg": 0.03638123348355293,
+      "learning_rate": 4.90310359900918e-07,
+      "loss_func": "stage2",
+      "step": 139
+    },
+    {
+      "epoch": 0.10193401592719,
+      "importance_ratio": 0.9985758066177368,
+      "kl_div_avg": 0.04586644098162651,
+      "learning_rate": 4.902375054640828e-07,
+      "loss_func": "stage2",
+      "step": 140
+    },
+    {
+      "epoch": 0.10484641638225256,
+      "importance_ratio": 0.9999945759773254,
+      "kl_div_avg": 0.025496020913124084,
+      "learning_rate": 4.901646510272475e-07,
+      "loss_func": "stage2",
+      "step": 141,
+      "total_loss": -1.21875
+    },
+    {
+      "epoch": 0.10484641638225256,
+      "importance_ratio": 0.9999227523803711,
+      "kl_div_avg": 0.025891993194818497,
+      "learning_rate": 4.900917965904123e-07,
+      "loss_func": "stage2",
+      "step": 142,
+      "total_loss": 1.2890625
+    },
+    {
+      "epoch": 0.10484641638225256,
+      "importance_ratio": 0.9999605417251587,
+      "kl_div_avg": 0.02385072410106659,
+      "learning_rate": 4.900189421535771e-07,
+      "loss_func": "stage2",
+      "step": 143
+    },
+    {
+      "epoch": 0.10484641638225256,
+      "importance_ratio": 1.0007202625274658,
+      "kl_div_avg": 0.027996843680739403,
+      "learning_rate": 4.89946087716742e-07,
+      "loss_func": "stage2",
+      "step": 144,
+      "total_loss": -1.4765625
+    },
+    {
+      "epoch": 0.10775881683731513,
+      "importance_ratio": 1.0,
+      "kl_div_avg": 0.01974659040570259,
+      "learning_rate": 4.898732332799068e-07,
+      "loss_func": "stage2",
+      "step": 145
+    },
+    {
+      "epoch": 0.10775881683731513,
+      "importance_ratio": 1.0009050369262695,
+      "kl_div_avg": 0.0290310550481081,
+      "learning_rate": 4.898003788430715e-07,
+      "loss_func": "stage2",
+      "step": 146
+    },
+    {
+      "epoch": 0.10775881683731513,
+      "importance_ratio": 1.0006165504455566,
+      "kl_div_avg": 0.025887252762913704,
+      "learning_rate": 4.897275244062363e-07,
+      "loss_func": "stage2",
+      "step": 147
+    },
+    {
+      "epoch": 0.10775881683731513,
+      "importance_ratio": 1.0016820430755615,
+      "kl_div_avg": 0.021247221156954765,
+      "learning_rate": 4.896546699694011e-07,
+      "loss_func": "stage2",
+      "step": 148
+    },
+    {
+      "epoch": 0.11067121729237771,
+      "importance_ratio": 1.0,
+      "kl_div_avg": 0.040258921682834625,
+      "learning_rate": 4.895818155325659e-07,
+      "loss_func": "stage2",
+      "step": 149,
+      "total_loss": 0.013671875
+    },
+    {
+      "epoch": 0.11067121729237771,
+      "importance_ratio": 1.0006983280181885,
+      "kl_div_avg": 0.025034813210368156,
+      "learning_rate": 4.895089610957307e-07,
+      "loss_func": "stage2",
+      "step": 150,
+      "total_loss": -0.08203125
+    },
+    {
+      "epoch": 0.11067121729237771,
+      "importance_ratio": 1.0025944709777832,
+      "kl_div_avg": 0.034798409789800644,
+      "learning_rate": 4.894361066588955e-07,
+      "loss_func": "stage2",
+      "step": 151,
+      "total_loss": -0.484375
+    },
+    {
+      "epoch": 0.11067121729237771,
+      "importance_ratio": 1.0026025772094727,
+      "kl_div_avg": 0.02111036330461502,
+      "learning_rate": 4.893632522220603e-07,
+      "loss_func": "stage2",
+      "step": 152,
+      "total_loss": -0.142578125
+    },
+    {
+      "epoch": 0.11358361774744027,
+      "importance_ratio": 1.0,
+      "kl_div_avg": 0.03777769207954407,
+      "learning_rate": 4.892903977852251e-07,
+      "loss_func": "stage2",
+      "step": 153
+    },
+    {
+      "epoch": 0.11358361774744027,
+      "importance_ratio": 1.0004066228866577,
+      "kl_div_avg": 0.027016688138246536,
+      "learning_rate": 4.892175433483899e-07,
+      "loss_func": "stage2",
+      "step": 154
+    },
+    {
+      "epoch": 0.11358361774744027,
+      "importance_ratio": 1.000730037689209,
+      "kl_div_avg": 0.026569224894046783,
+      "learning_rate": 4.891446889115547e-07,
+      "loss_func": "stage2",
+      "step": 155
+    },
+    {
+      "epoch": 0.11358361774744027,
+      "importance_ratio": 1.0018224716186523,
+      "kl_div_avg": 0.038676515221595764,
+      "learning_rate": 4.890718344747195e-07,
+      "loss_func": "stage2",
+      "step": 156
+    },
+    {
+      "epoch": 0.11649601820250284,
+      "importance_ratio": 1.0,
+      "kl_div_avg": 0.052732765674591064,
+      "learning_rate": 4.889989800378843e-07,
+      "loss_func": "stage2",
+      "step": 157,
+      "total_loss": -1.03125
+    },
+    {
+      "epoch": 0.11649601820250284,
+      "importance_ratio": 0.9996334314346313,
+      "kl_div_avg": 0.05555739998817444,
+      "learning_rate": 4.889261256010491e-07,
+      "loss_func": "stage2",
+      "step": 158,
+      "total_loss": -1.0625
+    },
+    {
+      "epoch": 0.11649601820250284,
+      "importance_ratio": 0.9995920658111572,
+      "kl_div_avg": 0.03014589659869671,
+      "learning_rate": 4.888532711642138e-07,
+      "loss_func": "stage2",
+      "step": 159,
+      "total_loss": -0.80859375
+    },
+    {
+      "epoch": 0.11649601820250284,
+      "importance_ratio": 0.9997560977935791,
+      "kl_div_avg": 0.06232505291700363,
+      "learning_rate": 4.887804167273787e-07,
+      "loss_func": "stage2",
+      "step": 160,
+      "total_loss": -1.125
+    },
+    {
+      "epoch": 0.11940841865756542,
+      "importance_ratio": 1.0,
+      "kl_div_avg": 0.05529934912919998,
+      "learning_rate": 4.887075622905435e-07,
+      "loss_func": "stage2",
+      "step": 161
+    },
+    {
+      "epoch": 0.11940841865756542,
+      "importance_ratio": 1.0000139474868774,
+      "kl_div_avg": 0.022508345544338226,
+      "learning_rate": 4.886347078537083e-07,
+      "loss_func": "stage2",
+      "step": 162
+    },
+    {
+      "epoch": 0.11940841865756542,
+      "importance_ratio": 1.0001630783081055,
+      "kl_div_avg": 0.015631631016731262,
+      "learning_rate": 4.885618534168731e-07,
+      "loss_func": "stage2",
+      "step": 163
+    },
+    {
+      "epoch": 0.11940841865756542,
+      "importance_ratio": 1.0005483627319336,
+      "kl_div_avg": 0.06399040669202805,
+      "learning_rate": 4.884889989800378e-07,
+      "loss_func": "stage2",
+      "step": 164
+    },
+    {
+      "epoch": 0.12232081911262799,
+      "importance_ratio": 1.0,
+      "kl_div_avg": 0.03488156944513321,
+      "learning_rate": 4.884161445432027e-07,
+      "loss_func": "stage2",
+      "step": 165
+    },
+    {
+      "epoch": 0.12232081911262799,
+      "importance_ratio": 1.000325322151184,
+      "kl_div_avg": 0.026072338223457336,
+      "learning_rate": 4.883432901063674e-07,
+      "loss_func": "stage2",
+      "step": 166
+    },
+    {
+      "epoch": 0.12232081911262799,
+      "importance_ratio": 1.0011181831359863,
+      "kl_div_avg": 0.031591013073921204,
+      "learning_rate": 4.882704356695323e-07,
+      "loss_func": "stage2",
+      "step": 167
+    },
+    {
+      "epoch": 0.12232081911262799,
+      "importance_ratio": 1.0013492107391357,
+      "kl_div_avg": 0.03676939755678177,
+      "learning_rate": 4.881975812326971e-07,
+      "loss_func": "stage2",
+      "step": 168
+    },
+    {
+      "epoch": 0.12523321956769057,
+      "importance_ratio": 1.0,
+      "kl_div_avg": 0.030846180394291878,
+      "learning_rate": 4.881247267958618e-07,
+      "loss_func": "stage2",
+      "step": 169
+    },
+    {
+      "epoch": 0.12523321956769057,
+      "importance_ratio": 0.9997780323028564,
+      "kl_div_avg": 0.030299291014671326,
+      "learning_rate": 4.880518723590266e-07,
+      "loss_func": "stage2",
+      "step": 170,
+      "total_loss": 1.0
+    },
+    {
+      "epoch": 0.12523321956769057,
+      "importance_ratio": 0.9996495246887207,
+      "kl_div_avg": 0.03033524751663208,
+      "learning_rate": 4.879790179221914e-07,
+      "loss_func": "stage2",
+      "step": 171,
+      "total_loss": 1.1953125
+    },
+    {
+      "epoch": 0.12523321956769057,
+      "importance_ratio": 0.9985495805740356,
+      "kl_div_avg": 0.024776607751846313,
+      "learning_rate": 4.879061634853563e-07,
+      "loss_func": "stage2",
+      "step": 172,
+      "total_loss": -1.5703125
+    },
+    {
+      "epoch": 0.12814562002275312,
+      "importance_ratio": 1.0,
+      "kl_div_avg": 0.03331543505191803,
+      "learning_rate": 4.87833309048521e-07,
+      "loss_func": "stage2",
+      "step": 173
+    },
+    {
+      "epoch": 0.12814562002275312,
+      "importance_ratio": 0.9997560381889343,
+      "kl_div_avg": 0.027614161372184753,
+      "learning_rate": 4.877604546116858e-07,
+      "loss_func": "stage2",
+      "step": 174
+    },
+    {
+      "epoch": 0.12814562002275312,
+      "importance_ratio": 0.9999554753303528,
+      "kl_div_avg": 0.0482001006603241,
+      "learning_rate": 4.876876001748506e-07,
+      "loss_func": "stage2",
+      "step": 175
+    },
+    {
+      "epoch": 0.12814562002275312,
+      "importance_ratio": 1.0001144409179688,
+      "kl_div_avg": 0.03714786469936371,
+      "learning_rate": 4.876147457380154e-07,
+      "loss_func": "stage2",
+      "step": 176
+    },
+    {
+      "epoch": 0.1310580204778157,
+      "importance_ratio": 1.0,
+      "kl_div_avg": 0.05323777347803116,
+      "learning_rate": 4.875418913011803e-07,
+      "loss_func": "stage2",
+      "step": 177,
+      "total_loss": -0.48828125
+    },
+    {
+      "epoch": 0.1310580204778157,
+      "importance_ratio": 1.0003318786621094,
+      "kl_div_avg": 0.03846876323223114,
+      "learning_rate": 4.87469036864345e-07,
+      "loss_func": "stage2",
+      "step": 178,
+      "total_loss": 0.640625
+    },
+    {
+      "epoch": 0.1310580204778157,
+      "importance_ratio": 1.0011920928955078,
+      "kl_div_avg": 0.041027769446372986,
+      "learning_rate": 4.873961824275098e-07,
+      "loss_func": "stage2",
+      "step": 179,
+      "total_loss": -0.55078125
+    },
+    {
+      "epoch": 0.1310580204778157,
+      "importance_ratio": 1.0019876956939697,
+      "kl_div_avg": 0.045351505279541016,
+      "learning_rate": 4.873233279906746e-07,
+      "loss_func": "stage2",
+      "step": 180
+    },
+    {
+      "epoch": 0.13397042093287828,
+      "importance_ratio": 1.0,
+      "kl_div_avg": 0.02115241438150406,
+      "learning_rate": 4.872504735538394e-07,
+      "loss_func": "stage2",
+      "step": 181,
+      "total_loss": 0.8046875
+    },
+    {
+      "epoch": 0.13397042093287828,
+      "importance_ratio": 1.0003323554992676,
+      "kl_div_avg": 0.026031356304883957,
+      "learning_rate": 4.871776191170043e-07,
+      "loss_func": "stage2",
+      "step": 182
+    },
+    {
+      "epoch": 0.13397042093287828,
+      "importance_ratio": 1.0005035400390625,
+      "kl_div_avg": 0.014925890602171421,
+      "learning_rate": 4.87104764680169e-07,
+      "loss_func": "stage2",
+      "step": 183,
+      "total_loss": 0.8125
+    },
+    {
+      "epoch": 0.13397042093287828,
+      "importance_ratio": 1.0016844272613525,
+      "kl_div_avg": 0.02387421950697899,
+      "learning_rate": 4.870319102433338e-07,
+      "loss_func": "stage2",
+      "step": 184,
+      "total_loss": -1.6484375
+    },
+    {
+      "epoch": 0.13688282138794083,
+      "importance_ratio": 1.0,
+      "kl_div_avg": 0.04454325884580612,
+      "learning_rate": 4.869590558064986e-07,
+      "loss_func": "stage2",
+      "step": 185
+    },
+    {
+      "epoch": 0.13688282138794083,
+      "importance_ratio": 1.0004349946975708,
+      "kl_div_avg": 0.037973642349243164,
+      "learning_rate": 4.868862013696634e-07,
+      "loss_func": "stage2",
+      "step": 186
+    },
+    {
+      "epoch": 0.13688282138794083,
+      "importance_ratio": 1.0010215044021606,
+      "kl_div_avg": 0.043117135763168335,
+      "learning_rate": 4.868133469328281e-07,
+      "loss_func": "stage2",
+      "step": 187
+    },
+    {
+      "epoch": 0.13688282138794083,
+      "importance_ratio": 1.0012775659561157,
+      "kl_div_avg": 0.033980466425418854,
+      "learning_rate": 4.86740492495993e-07,
+      "loss_func": "stage2",
+      "step": 188
+    },
+    {
+      "epoch": 0.13979522184300341,
+      "importance_ratio": 1.0,
+      "kl_div_avg": 0.03627289831638336,
+      "learning_rate": 4.866676380591578e-07,
+      "loss_func": "stage2",
+      "step": 189,
+      "total_loss": 1.046875
+    },
+    {
+      "epoch": 0.13979522184300341,
+      "importance_ratio": 1.0005991458892822,
+      "kl_div_avg": 0.0504533126950264,
+      "learning_rate": 4.865947836223226e-07,
+      "loss_func": "stage2",
+      "step": 190,
+      "total_loss": 1.265625
+    },
+    {
+      "epoch": 0.13979522184300341,
+      "importance_ratio": 1.0003464221954346,
+      "kl_div_avg": 0.04870688170194626,
+      "learning_rate": 4.865219291854874e-07,
+      "loss_func": "stage2",
+      "step": 191,
+      "total_loss": -1.1875
+    },
+    {
+      "epoch": 0.13979522184300341,
+      "importance_ratio": 1.0007530450820923,
+      "kl_div_avg": 0.038728028535842896,
+      "learning_rate": 4.864490747486521e-07,
+      "loss_func": "stage2",
+      "step": 192,
+      "total_loss": 1.2109375
+    },
+    {
+      "epoch": 0.142707622298066,
+      "importance_ratio": 1.0000061988830566,
+      "kl_div_avg": 0.028430147096514702,
+      "learning_rate": 4.86376220311817e-07,
+      "loss_func": "stage2",
+      "step": 193
+    },
+    {
+      "epoch": 0.142707622298066,
+      "importance_ratio": 1.0001733303070068,
+      "kl_div_avg": 0.01401587575674057,
+      "learning_rate": 4.863033658749818e-07,
+      "loss_func": "stage2",
+      "step": 194
+    },
+    {
+      "epoch": 0.142707622298066,
+      "importance_ratio": 1.0002660751342773,
+      "kl_div_avg": 0.024514583870768547,
+      "learning_rate": 4.862305114381466e-07,
+      "loss_func": "stage2",
+      "step": 195
+    },
+    {
+      "epoch": 0.142707622298066,
+      "importance_ratio": 1.0008165836334229,
+      "kl_div_avg": 0.034122683107852936,
+      "learning_rate": 4.861576570013113e-07,
+      "loss_func": "stage2",
+      "step": 196
+    },
+    {
+      "epoch": 0.14562002275312855,
+      "importance_ratio": 1.0,
+      "kl_div_avg": 0.04862019419670105,
+      "learning_rate": 4.860848025644761e-07,
+      "loss_func": "stage2",
+      "step": 197
+    },
+    {
+      "epoch": 0.14562002275312855,
+      "importance_ratio": 0.9995243549346924,
+      "kl_div_avg": 0.02397969551384449,
+      "learning_rate": 4.86011948127641e-07,
+      "loss_func": "stage2",
+      "step": 198,
+      "total_loss": -0.4609375
+    },
+    {
+      "epoch": 0.14562002275312855,
+      "importance_ratio": 0.9991512894630432,
+      "kl_div_avg": 0.027180787175893784,
+      "learning_rate": 4.859390936908058e-07,
+      "loss_func": "stage2",
+      "step": 199,
+      "total_loss": -0.5390625
+    },
+    {
+      "epoch": 0.14562002275312855,
+      "importance_ratio": 0.999755859375,
+      "kl_div_avg": 0.02702735923230648,
+      "learning_rate": 4.858662392539706e-07,
+      "loss_func": "stage2",
+      "step": 200,
+      "total_loss": 0.90625
+    },
+    {
+      "epoch": 0.14853242320819113,
+      "importance_ratio": 1.0,
+      "kl_div_avg": 0.06592931598424911,
+      "learning_rate": 4.857933848171353e-07,
+      "loss_func": "stage2",
+      "step": 201
+    },
+    {
+      "epoch": 0.14853242320819113,
+      "importance_ratio": 0.9992879629135132,
+      "kl_div_avg": 0.02779082953929901,
+      "learning_rate": 4.857205303803001e-07,
+      "loss_func": "stage2",
+      "step": 202
+    },
+    {
+      "epoch": 0.14853242320819113,
+      "importance_ratio": 0.999265193939209,
+      "kl_div_avg": 0.013107150793075562,
+      "learning_rate": 4.85647675943465e-07,
+      "loss_func": "stage2",
+      "step": 203
+    },
+    {
+      "epoch": 0.14853242320819113,
+      "importance_ratio": 0.9988900423049927,
+      "kl_div_avg": 0.04416199028491974,
+      "learning_rate": 4.855748215066297e-07,
+      "loss_func": "stage2",
+      "step": 204
+    },
+    {
+      "epoch": 0.1514448236632537,
+      "importance_ratio": 1.0,
+      "kl_div_avg": 0.015596572309732437,
+      "learning_rate": 4.855019670697946e-07,
+      "loss_func": "stage2",
+      "step": 205,
+      "total_loss": -0.3125
+    },
+    {
+      "epoch": 0.1514448236632537,
+      "importance_ratio": 0.9995862245559692,
+      "kl_div_avg": 0.026362475007772446,
+      "learning_rate": 4.854291126329593e-07,
+      "loss_func": "stage2",
+      "step": 206,
+      "total_loss": -0.294921875
+    },
+    {
+      "epoch": 0.1514448236632537,
+      "importance_ratio": 0.9992890357971191,
+      "kl_div_avg": 0.025417888537049294,
+      "learning_rate": 4.853562581961241e-07,
+      "loss_func": "stage2",
+      "step": 207,
+      "total_loss": 0.9609375
+    },
+    {
+      "epoch": 0.1514448236632537,
+      "importance_ratio": 0.9989533424377441,
+      "kl_div_avg": 0.022100646048784256,
+      "learning_rate": 4.85283403759289e-07,
+      "loss_func": "stage2",
+      "step": 208
+    },
+    {
+      "epoch": 0.15435722411831626,
+      "importance_ratio": 1.0,
+      "kl_div_avg": 0.035768553614616394,
+      "learning_rate": 4.852105493224537e-07,
+      "loss_func": "stage2",
+      "step": 209
+    },
+    {
+      "epoch": 0.15435722411831626,
+      "importance_ratio": 0.9997737407684326,
+      "kl_div_avg": 0.016606420278549194,
+      "learning_rate": 4.851376948856185e-07,
+      "loss_func": "stage2",
+      "step": 210
+    },
+    {
+      "epoch": 0.15435722411831626,
+      "importance_ratio": 0.9995913505554199,
+      "kl_div_avg": 0.012671421281993389,
+      "learning_rate": 4.850648404487833e-07,
+      "loss_func": "stage2",
+      "step": 211
+    },
+    {
+      "epoch": 0.15435722411831626,
+      "importance_ratio": 0.9999702572822571,
+      "kl_div_avg": 0.03230913355946541,
+      "learning_rate": 4.849919860119481e-07,
+      "loss_func": "stage2",
+      "step": 212
+    },
+    {
+      "epoch": 0.15726962457337884,
+      "importance_ratio": 0.9999392628669739,
+      "kl_div_avg": 0.04738989472389221,
+      "learning_rate": 4.849191315751129e-07,
+      "loss_func": "stage2",
+      "step": 213,
+      "total_loss": -0.443359375
+    },
+    {
+      "epoch": 0.15726962457337884,
+      "importance_ratio": 0.9999153017997742,
+      "kl_div_avg": 0.027265094220638275,
+      "learning_rate": 4.848462771382777e-07,
+      "loss_func": "stage2",
+      "step": 214,
+      "total_loss": -0.474609375
+    },
+    {
+      "epoch": 0.15726962457337884,
+      "importance_ratio": 0.999463677406311,
+      "kl_div_avg": 0.042790867388248444,
+      "learning_rate": 4.847734227014425e-07,
+      "loss_func": "stage2",
+      "step": 215
+    },
+    {
+      "epoch": 0.15726962457337884,
+      "importance_ratio": 0.9999215006828308,
+      "kl_div_avg": 0.013197468593716621,
+      "learning_rate": 4.847005682646073e-07,
+      "loss_func": "stage2",
+      "step": 216,
+      "total_loss": 0.7109375
+    },
+    {
+      "epoch": 0.16018202502844142,
+      "importance_ratio": 1.0,
+      "kl_div_avg": 0.019805949181318283,
+      "learning_rate": 4.846277138277721e-07,
+      "loss_func": "stage2",
+      "step": 217
+    },
+    {
+      "epoch": 0.16018202502844142,
+      "importance_ratio": 1.0000206232070923,
+      "kl_div_avg": 0.02629798650741577,
+      "learning_rate": 4.845548593909369e-07,
+      "loss_func": "stage2",
+      "step": 218,
+      "total_loss": -0.4765625
+    },
+    {
+      "epoch": 0.16018202502844142,
+      "importance_ratio": 1.0013692378997803,
+      "kl_div_avg": 0.04414498805999756,
+      "learning_rate": 4.844820049541016e-07,
+      "loss_func": "stage2",
+      "step": 219,
+      "total_loss": 0.890625
+    },
+    {
+      "epoch": 0.16018202502844142,
+      "importance_ratio": 1.0012259483337402,
+      "kl_div_avg": 0.03558802232146263,
+      "learning_rate": 4.844091505172665e-07,
+      "loss_func": "stage2",
+      "step": 220,
+      "total_loss": -0.302734375
+    },
+    {
+      "epoch": 0.16309442548350397,
+      "importance_ratio": 1.0,
+      "kl_div_avg": 0.022240914404392242,
+      "learning_rate": 4.843362960804313e-07,
+      "loss_func": "stage2",
+      "step": 221
+    },
+    {
+      "epoch": 0.16309442548350397,
+      "importance_ratio": 1.00046706199646,
+      "kl_div_avg": 0.06476189196109772,
+      "learning_rate": 4.842634416435961e-07,
+      "loss_func": "stage2",
+      "step": 222
+    },
+    {
+      "epoch": 0.16309442548350397,
+      "importance_ratio": 0.9995772838592529,
+      "kl_div_avg": 0.033390507102012634,
+      "learning_rate": 4.841905872067609e-07,
+      "loss_func": "stage2",
+      "step": 223
+    },
+    {
+      "epoch": 0.16309442548350397,
+      "importance_ratio": 1.0012643337249756,
+      "kl_div_avg": 0.0590643510222435,
+      "learning_rate": 4.841177327699256e-07,
+      "loss_func": "stage2",
+      "step": 224
+    },
+    {
+      "epoch": 0.16600682593856655,
+      "importance_ratio": 1.0,
+      "kl_div_avg": 0.034556955099105835,
+      "learning_rate": 4.840448783330905e-07,
+      "loss_func": "stage2",
+      "step": 225
+    },
+    {
+      "epoch": 0.16600682593856655,
+      "importance_ratio": 1.0007654428482056,
+      "kl_div_avg": 0.031194303184747696,
+      "learning_rate": 4.839720238962553e-07,
+      "loss_func": "stage2",
+      "step": 226
+    },
+    {
+      "epoch": 0.16600682593856655,
+      "importance_ratio": 1.0008078813552856,
+      "kl_div_avg": 0.03249838203191757,
+      "learning_rate": 4.838991694594201e-07,
+      "loss_func": "stage2",
+      "step": 227
+    },
+    {
+      "epoch": 0.16600682593856655,
+      "importance_ratio": 1.0013104677200317,
+      "kl_div_avg": 0.02999018132686615,
+      "learning_rate": 4.838263150225848e-07,
+      "loss_func": "stage2",
+      "step": 228
+    },
+    {
+      "epoch": 0.16891922639362913,
+      "importance_ratio": 0.9999874830245972,
+      "kl_div_avg": 0.055338919162750244,
+      "learning_rate": 4.837534605857496e-07,
+      "loss_func": "stage2",
+      "step": 229,
+      "total_loss": 0.62109375
+    },
+    {
+      "epoch": 0.16891922639362913,
+      "importance_ratio": 1.001804232597351,
+      "kl_div_avg": 0.0552295483648777,
+      "learning_rate": 4.836806061489144e-07,
+      "loss_func": "stage2",
+      "step": 230
+    },
+    {
+      "epoch": 0.16891922639362913,
+      "importance_ratio": 1.0008022785186768,
+      "kl_div_avg": 0.03091678023338318,
+      "learning_rate": 4.836077517120793e-07,
+      "loss_func": "stage2",
+      "step": 231,
+      "total_loss": 0.69140625
+    },
+    {
+      "epoch": 0.16891922639362913,
+      "importance_ratio": 1.0017752647399902,
+      "kl_div_avg": 0.02759787067770958,
+      "learning_rate": 4.835348972752441e-07,
+      "loss_func": "stage2",
+      "step": 232,
+      "total_loss": -0.75390625
+    },
+    {
+      "epoch": 0.1718316268486917,
+      "importance_ratio": 1.0,
+      "kl_div_avg": 0.04960617050528526,
+      "learning_rate": 4.834620428384088e-07,
+      "loss_func": "stage2",
+      "step": 233
+    },
+    {
+      "epoch": 0.1718316268486917,
+      "importance_ratio": 0.9998636245727539,
+      "kl_div_avg": 0.04251248389482498,
+      "learning_rate": 4.833891884015736e-07,
+      "loss_func": "stage2",
+      "step": 234
+    },
+    {
+      "epoch": 0.1718316268486917,
+      "importance_ratio": 0.9993151426315308,
+      "kl_div_avg": 0.047904349863529205,
+      "learning_rate": 4.833163339647384e-07,
+      "loss_func": "stage2",
+      "step": 235
+    },
+    {
+      "epoch": 0.1718316268486917,
+      "importance_ratio": 0.9979664087295532,
+      "kl_div_avg": 0.04585729539394379,
+      "learning_rate": 4.832434795279033e-07,
+      "loss_func": "stage2",
+      "step": 236
+    },
+    {
+      "epoch": 0.17474402730375427,
+      "importance_ratio": 0.9999979734420776,
+      "kl_div_avg": 0.032474417239427567,
+      "learning_rate": 4.831706250910681e-07,
+      "loss_func": "stage2",
+      "step": 237,
+      "total_loss": 0.66796875
+    },
+    {
+      "epoch": 0.17474402730375427,
+      "importance_ratio": 0.9992362856864929,
+      "kl_div_avg": 0.04708264023065567,
+      "learning_rate": 4.830977706542328e-07,
+      "loss_func": "stage2",
+      "step": 238,
+      "total_loss": 1.0078125
+    },
+    {
+      "epoch": 0.17474402730375427,
+      "importance_ratio": 0.9996246695518494,
+      "kl_div_avg": 0.03764179348945618,
+      "learning_rate": 4.830249162173976e-07,
+      "loss_func": "stage2",
+      "step": 239,
+      "total_loss": 0.9375
+    },
+    {
+      "epoch": 0.17474402730375427,
+      "importance_ratio": 0.9991611838340759,
+      "kl_div_avg": 0.022797293961048126,
+      "learning_rate": 4.829520617805624e-07,
+      "loss_func": "stage2",
+      "step": 240,
+      "total_loss": -0.8515625
+    },
+    {
+      "epoch": 0.17765642775881685,
+      "importance_ratio": 1.0,
+      "kl_div_avg": 0.030025824904441833,
+      "learning_rate": 4.828792073437273e-07,
+      "loss_func": "stage2",
+      "step": 241
+    },
+    {
+      "epoch": 0.17765642775881685,
+      "importance_ratio": 0.9989254474639893,
+      "kl_div_avg": 0.04463012516498566,
+      "learning_rate": 4.82806352906892e-07,
+      "loss_func": "stage2",
+      "step": 242
+    },
+    {
+      "epoch": 0.17765642775881685,
+      "importance_ratio": 0.9987425804138184,
+      "kl_div_avg": 0.048046525567770004,
+      "learning_rate": 4.827334984700568e-07,
+      "loss_func": "stage2",
+      "step": 243
+    },
+    {
+      "epoch": 0.17765642775881685,
+      "importance_ratio": 0.9986125230789185,
+      "kl_div_avg": 0.02986510843038559,
+      "learning_rate": 4.826606440332216e-07,
+      "loss_func": "stage2",
+      "step": 244
+    },
+    {
+      "epoch": 0.1805688282138794,
+      "importance_ratio": 1.0,
+      "kl_div_avg": 0.06527923047542572,
+      "learning_rate": 4.825877895963864e-07,
+      "loss_func": "stage2",
+      "step": 245
+    },
+    {
+      "epoch": 0.1805688282138794,
+      "importance_ratio": 0.9997293949127197,
+      "kl_div_avg": 0.03313666582107544,
+      "learning_rate": 4.825149351595513e-07,
+      "loss_func": "stage2",
+      "step": 246
+    },
+    {
+      "epoch": 0.1805688282138794,
+      "importance_ratio": 0.9996707439422607,
+      "kl_div_avg": 0.020555861294269562,
+      "learning_rate": 4.824420807227159e-07,
+      "loss_func": "stage2",
+      "step": 247
+    },
+    {
+      "epoch": 0.1805688282138794,
+      "importance_ratio": 1.0002386569976807,
+      "kl_div_avg": 0.03867045417428017,
+      "learning_rate": 4.823692262858808e-07,
+      "loss_func": "stage2",
+      "step": 248
+    },
+    {
+      "epoch": 0.18348122866894198,
+      "importance_ratio": 1.0,
+      "kl_div_avg": 0.05499115586280823,
+      "learning_rate": 4.822963718490456e-07,
+      "loss_func": "stage2",
+      "step": 249
+    },
+    {
+      "epoch": 0.18348122866894198,
+      "importance_ratio": 0.9998311996459961,
+      "kl_div_avg": 0.049019381403923035,
+      "learning_rate": 4.822235174122104e-07,
+      "loss_func": "stage2",
+      "step": 250
+    },
+    {
+      "epoch": 0.18348122866894198,
+      "importance_ratio": 1.0000450611114502,
+      "kl_div_avg": 0.020925089716911316,
+      "learning_rate": 4.821506629753751e-07,
+      "loss_func": "stage2",
+      "step": 251
+    },
+    {
+      "epoch": 0.18348122866894198,
+      "importance_ratio": 1.0005130767822266,
+      "kl_div_avg": 0.05002495273947716,
+      "learning_rate": 4.820778085385399e-07,
+      "loss_func": "stage2",
+      "step": 252
+    },
+    {
+      "epoch": 0.18639362912400456,
+      "importance_ratio": 1.0,
+      "kl_div_avg": 0.03229670226573944,
+      "learning_rate": 4.820049541017048e-07,
+      "loss_func": "stage2",
+      "step": 253
+    },
+    {
+      "epoch": 0.18639362912400456,
+      "importance_ratio": 1.0004810094833374,
+      "kl_div_avg": 0.05420541763305664,
+      "learning_rate": 4.819320996648696e-07,
+      "loss_func": "stage2",
+      "step": 254
+    },
+    {
+      "epoch": 0.18639362912400456,
+      "importance_ratio": 1.0008883476257324,
+      "kl_div_avg": 0.027021722868084908,
+      "learning_rate": 4.818592452280344e-07,
+      "loss_func": "stage2",
+      "step": 255
+    },
+    {
+      "epoch": 0.18639362912400456,
+      "importance_ratio": 1.0022274255752563,
+      "kl_div_avg": 0.03898981586098671,
+      "learning_rate": 4.817863907911991e-07,
+      "loss_func": "stage2",
+      "step": 256
+    },
+    {
+      "epoch": 0.18930602957906711,
+      "importance_ratio": 1.0,
+      "kl_div_avg": 0.07298167794942856,
+      "learning_rate": 4.817135363543639e-07,
+      "loss_func": "stage2",
+      "step": 257
+    },
+    {
+      "epoch": 0.18930602957906711,
+      "importance_ratio": 1.0023469924926758,
+      "kl_div_avg": 0.055954575538635254,
+      "learning_rate": 4.816406819175288e-07,
+      "loss_func": "stage2",
+      "step": 258
+    },
+    {
+      "epoch": 0.18930602957906711,
+      "importance_ratio": 1.0012736320495605,
+      "kl_div_avg": 0.04164516180753708,
+      "learning_rate": 4.815678274806936e-07,
+      "loss_func": "stage2",
+      "step": 259
+    },
+    {
+      "epoch": 0.18930602957906711,
+      "importance_ratio": 1.00413978099823,
+      "kl_div_avg": 0.07184624671936035,
+      "learning_rate": 4.814949730438584e-07,
+      "loss_func": "stage2",
+      "step": 260
+    },
+    {
+      "epoch": 0.1922184300341297,
+      "importance_ratio": 1.0,
+      "kl_div_avg": 0.027168840169906616,
+      "learning_rate": 4.814221186070231e-07,
+      "loss_func": "stage2",
+      "step": 261,
+      "total_loss": -0.287109375
+    },
+    {
+      "epoch": 0.1922184300341297,
+      "importance_ratio": 1.0015285015106201,
+      "kl_div_avg": 0.038711898028850555,
+      "learning_rate": 4.813492641701879e-07,
+      "loss_func": "stage2",
+      "step": 262,
+      "total_loss": 0.77734375
+    },
+    {
+      "epoch": 0.1922184300341297,
+      "importance_ratio": 1.0005645751953125,
+      "kl_div_avg": 0.021953891962766647,
+      "learning_rate": 4.812764097333528e-07,
+      "loss_func": "stage2",
+      "step": 263,
+      "total_loss": -0.376953125
+    },
+    {
+      "epoch": 0.1922184300341297,
+      "importance_ratio": 1.0069310665130615,
+      "kl_div_avg": 0.04371698200702667,
+      "learning_rate": 4.812035552965176e-07,
+      "loss_func": "stage2",
+      "step": 264,
+      "total_loss": 1.2890625
+    },
+    {
+      "epoch": 0.19513083048919228,
+      "importance_ratio": 1.0,
+      "kl_div_avg": 0.023378241807222366,
+      "learning_rate": 4.811307008596823e-07,
+      "loss_func": "stage2",
+      "step": 265,
+      "total_loss": -0.6875
+    },
+    {
+      "epoch": 0.19513083048919228,
+      "importance_ratio": 1.0003318786621094,
+      "kl_div_avg": 0.030648723244667053,
+      "learning_rate": 4.810578464228471e-07,
+      "loss_func": "stage2",
+      "step": 266,
+      "total_loss": 0.6328125
+    },
+    {
+      "epoch": 0.19513083048919228,
+      "importance_ratio": 0.9998455047607422,
+      "kl_div_avg": 0.04823487997055054,
+      "learning_rate": 4.809849919860119e-07,
+      "loss_func": "stage2",
+      "step": 267,
+      "total_loss": 1.1796875
+    },
+    {
+      "epoch": 0.19513083048919228,
+      "importance_ratio": 0.999752402305603,
+      "kl_div_avg": 0.024743907153606415,
+      "learning_rate": 4.809121375491767e-07,
+      "loss_func": "stage2",
+      "step": 268,
+      "total_loss": 0.32421875
+    },
+    {
+      "epoch": 0.19804323094425483,
+      "importance_ratio": 1.0,
+      "kl_div_avg": 0.017089594155550003,
+      "learning_rate": 4.808392831123416e-07,
+      "loss_func": "stage2",
+      "step": 269,
+      "total_loss": 0.9375
+    },
+    {
+      "epoch": 0.19804323094425483,
+      "importance_ratio": 0.9995547533035278,
+      "kl_div_avg": 0.019818518310785294,
+      "learning_rate": 4.807664286755063e-07,
+      "loss_func": "stage2",
+      "step": 270
+    },
+    {
+      "epoch": 0.19804323094425483,
+      "importance_ratio": 0.9994271993637085,
+      "kl_div_avg": 0.020417192950844765,
+      "learning_rate": 4.806935742386711e-07,
+      "loss_func": "stage2",
+      "step": 271,
+      "total_loss": -0.46875
+    },
+    {
+      "epoch": 0.19804323094425483,
+      "importance_ratio": 0.9990317821502686,
+      "kl_div_avg": 0.020819377154111862,
+      "learning_rate": 4.806207198018359e-07,
+      "loss_func": "stage2",
+      "step": 272,
+      "total_loss": -0.546875
+    },
+    {
+      "epoch": 0.2009556313993174,
+      "importance_ratio": 1.0,
+      "kl_div_avg": 0.04535255581140518,
+      "learning_rate": 4.805478653650007e-07,
+      "loss_func": "stage2",
+      "step": 273
+    },
+    {
+      "epoch": 0.2009556313993174,
+      "importance_ratio": 0.9979629516601562,
+      "kl_div_avg": 0.05327494442462921,
+      "learning_rate": 4.804750109281656e-07,
+      "loss_func": "stage2",
+      "step": 274
+    },
+    {
+      "epoch": 0.2009556313993174,
+      "importance_ratio": 0.9990555047988892,
+      "kl_div_avg": 0.041617073118686676,
+      "learning_rate": 4.804021564913303e-07,
+      "loss_func": "stage2",
+      "step": 275
+    },
+    {
+      "epoch": 0.2009556313993174,
+      "importance_ratio": 0.9964962005615234,
+      "kl_div_avg": 0.043679993599653244,
+      "learning_rate": 4.803293020544951e-07,
+      "loss_func": "stage2",
+      "step": 276
+    },
+    {
+      "epoch": 0.20386803185438,
+      "importance_ratio": 1.0,
+      "kl_div_avg": 0.038318827748298645,
+      "learning_rate": 4.802564476176599e-07,
+      "loss_func": "stage2",
+      "step": 277
+    },
+    {
+      "epoch": 0.20386803185438,
+      "importance_ratio": 0.9991254806518555,
+      "kl_div_avg": 0.03668826445937157,
+      "learning_rate": 4.801835931808247e-07,
+      "loss_func": "stage2",
+      "step": 278
+    },
+    {
+      "epoch": 0.20386803185438,
+      "importance_ratio": 0.9992523193359375,
+      "kl_div_avg": 0.031174367293715477,
+      "learning_rate": 4.801107387439895e-07,
+      "loss_func": "stage2",
+      "step": 279
+    },
+    {
+      "epoch": 0.20386803185438,
+      "importance_ratio": 0.9991823434829712,
+      "kl_div_avg": 0.022506345063447952,
+      "learning_rate": 4.800378843071543e-07,
+      "loss_func": "stage2",
+      "step": 280
+    },
+    {
+      "epoch": 0.20678043230944254,
+      "importance_ratio": 1.0000108480453491,
+      "kl_div_avg": 0.07250766456127167,
+      "learning_rate": 4.799650298703191e-07,
+      "loss_func": "stage2",
+      "step": 281
+    },
+    {
+      "epoch": 0.20678043230944254,
+      "importance_ratio": 0.9996242523193359,
+      "kl_div_avg": 0.04273686558008194,
+      "learning_rate": 4.798921754334839e-07,
+      "loss_func": "stage2",
+      "step": 282
+    },
+    {
+      "epoch": 0.20678043230944254,
+      "importance_ratio": 1.0001927614212036,
+      "kl_div_avg": 0.024231867864727974,
+      "learning_rate": 4.798193209966487e-07,
+      "loss_func": "stage2",
+      "step": 283
+    },
+    {
+      "epoch": 0.20678043230944254,
+      "importance_ratio": 0.9994685649871826,
+      "kl_div_avg": 0.05078791826963425,
+      "learning_rate": 4.797464665598135e-07,
+      "loss_func": "stage2",
+      "step": 284
+    },
+    {
+      "epoch": 0.20969283276450512,
+      "importance_ratio": 0.9999988079071045,
+      "kl_div_avg": 0.05624654144048691,
+      "learning_rate": 4.796736121229782e-07,
+      "loss_func": "stage2",
+      "step": 285
+    },
+    {
+      "epoch": 0.20969283276450512,
+      "importance_ratio": 1.000368595123291,
+      "kl_div_avg": 0.06209776550531387,
+      "learning_rate": 4.796007576861431e-07,
+      "loss_func": "stage2",
+      "step": 286
+    },
+    {
+      "epoch": 0.20969283276450512,
+      "importance_ratio": 1.0004286766052246,
+      "kl_div_avg": 0.05481367185711861,
+      "learning_rate": 4.795279032493079e-07,
+      "loss_func": "stage2",
+      "step": 287
+    },
+    {
+      "epoch": 0.20969283276450512,
+      "importance_ratio": 1.0034512281417847,
+      "kl_div_avg": 0.053617849946022034,
+      "learning_rate": 4.794550488124726e-07,
+      "loss_func": "stage2",
+      "step": 288
+    },
+    {
+      "epoch": 0.2126052332195677,
+      "importance_ratio": 1.0,
+      "kl_div_avg": 0.029997631907463074,
+      "learning_rate": 4.793821943756374e-07,
+      "loss_func": "stage2",
+      "step": 289
+    },
+    {
+      "epoch": 0.2126052332195677,
+      "importance_ratio": 1.0002427101135254,
+      "kl_div_avg": 0.0545087493956089,
+      "learning_rate": 4.793093399388022e-07,
+      "loss_func": "stage2",
+      "step": 290
+    },
+    {
+      "epoch": 0.2126052332195677,
+      "importance_ratio": 0.9994891881942749,
+      "kl_div_avg": 0.04038421809673309,
+      "learning_rate": 4.792364855019671e-07,
+      "loss_func": "stage2",
+      "step": 291
+    },
+    {
+      "epoch": 0.2126052332195677,
+      "importance_ratio": 0.9999635815620422,
+      "kl_div_avg": 0.03762605041265488,
+      "learning_rate": 4.791636310651319e-07,
+      "loss_func": "stage2",
+      "step": 292
+    },
+    {
+      "epoch": 0.21551763367463025,
+      "importance_ratio": 1.0,
+      "kl_div_avg": 0.027985263615846634,
+      "learning_rate": 4.790907766282966e-07,
+      "loss_func": "stage2",
+      "step": 293
+    },
+    {
+      "epoch": 0.21551763367463025,
+      "importance_ratio": 1.0003306865692139,
+      "kl_div_avg": 0.040808938443660736,
+      "learning_rate": 4.790179221914614e-07,
+      "loss_func": "stage2",
+      "step": 294
+    },
+    {
+      "epoch": 0.21551763367463025,
+      "importance_ratio": 1.00089693069458,
+      "kl_div_avg": 0.0335797518491745,
+      "learning_rate": 4.789450677546262e-07,
+      "loss_func": "stage2",
+      "step": 295
+    },
+    {
+      "epoch": 0.21551763367463025,
+      "importance_ratio": 1.0001624822616577,
+      "kl_div_avg": 0.015699325129389763,
+      "learning_rate": 4.788722133177911e-07,
+      "loss_func": "stage2",
+      "step": 296
+    },
+    {
+      "epoch": 0.21843003412969283,
+      "importance_ratio": 1.0,
+      "kl_div_avg": 0.05593610554933548,
+      "learning_rate": 4.787993588809558e-07,
+      "loss_func": "stage2",
+      "step": 297,
+      "total_loss": -0.08349609375
+    },
+    {
+      "epoch": 0.21843003412969283,
+      "importance_ratio": 1.0008994340896606,
+      "kl_div_avg": 0.04691431298851967,
+      "learning_rate": 4.787265044441206e-07,
+      "loss_func": "stage2",
+      "step": 298,
+      "total_loss": -0.56640625
+    },
+    {
+      "epoch": 0.21843003412969283,
+      "importance_ratio": 1.0015584230422974,
+      "kl_div_avg": 0.04217691347002983,
+      "learning_rate": 4.786536500072854e-07,
+      "loss_func": "stage2",
+      "step": 299,
+      "total_loss": -0.408203125
+    },
+    {
+      "epoch": 0.21843003412969283,
+      "importance_ratio": 1.0016425848007202,
+      "kl_div_avg": 0.05044947564601898,
+      "learning_rate": 4.785807955704502e-07,
+      "loss_func": "stage2",
+      "step": 300,
+      "total_loss": 1.8984375
+    },
+    {
+      "epoch": 0.22134243458475542,
+      "importance_ratio": 1.0,
+      "kl_div_avg": 0.13301140069961548,
+      "learning_rate": 4.785079411336151e-07,
+      "loss_func": "stage2",
+      "step": 301
+    },
+    {
+      "epoch": 0.22134243458475542,
+      "importance_ratio": 0.9997519254684448,
+      "kl_div_avg": 0.0724569633603096,
+      "learning_rate": 4.784350866967798e-07,
+      "loss_func": "stage2",
+      "step": 302,
+      "total_loss": 0.8984375
+    },
+    {
+      "epoch": 0.22134243458475542,
+      "importance_ratio": 0.9995059967041016,
+      "kl_div_avg": 0.07791668176651001,
+      "learning_rate": 4.783622322599446e-07,
+      "loss_func": "stage2",
+      "step": 303,
+      "total_loss": 0.90234375
+    },
+    {
+      "epoch": 0.22134243458475542,
+      "importance_ratio": 0.9976062774658203,
+      "kl_div_avg": 0.03134729713201523,
+      "learning_rate": 4.782893778231094e-07,
+      "loss_func": "stage2",
+      "step": 304,
+      "total_loss": -0.3984375
+    },
+    {
+      "epoch": 0.22425483503981797,
+      "importance_ratio": 1.0,
+      "kl_div_avg": 0.023282647132873535,
+      "learning_rate": 4.782165233862742e-07,
+      "loss_func": "stage2",
+      "step": 305
+    },
+    {
+      "epoch": 0.22425483503981797,
+      "importance_ratio": 0.9994572401046753,
+      "kl_div_avg": 0.039033204317092896,
+      "learning_rate": 4.78143668949439e-07,
+      "loss_func": "stage2",
+      "step": 306
+    },
+    {
+      "epoch": 0.22425483503981797,
+      "importance_ratio": 0.9989634156227112,
+      "kl_div_avg": 0.04119700938463211,
+      "learning_rate": 4.780708145126038e-07,
+      "loss_func": "stage2",
+      "step": 307
+    },
+    {
+      "epoch": 0.22425483503981797,
+      "importance_ratio": 0.9977949261665344,
+      "kl_div_avg": 0.06166260689496994,
+      "learning_rate": 4.779979600757686e-07,
+      "loss_func": "stage2",
+      "step": 308
+    },
+    {
+      "epoch": 0.22716723549488055,
+      "importance_ratio": 1.0,
+      "kl_div_avg": 0.023296231403946877,
+      "learning_rate": 4.779251056389334e-07,
+      "loss_func": "stage2",
+      "step": 309,
+      "total_loss": -0.75
+    },
+    {
+      "epoch": 0.22716723549488055,
+      "importance_ratio": 0.9992837905883789,
+      "kl_div_avg": 0.03709561377763748,
+      "learning_rate": 4.778522512020982e-07,
+      "loss_func": "stage2",
+      "step": 310,
+      "total_loss": -1.375
+    },
+    {
+      "epoch": 0.22716723549488055,
+      "importance_ratio": 0.9993754625320435,
+      "kl_div_avg": 0.0337868295609951,
+      "learning_rate": 4.777793967652629e-07,
+      "loss_func": "stage2",
+      "step": 311
+    },
+    {
+      "epoch": 0.22716723549488055,
+      "importance_ratio": 0.9995378851890564,
+      "kl_div_avg": 0.03989873826503754,
+      "learning_rate": 4.777065423284278e-07,
+      "loss_func": "stage2",
+      "step": 312,
+      "total_loss": 2.1875
+    },
+    {
+      "epoch": 0.23007963594994313,
+      "importance_ratio": 1.0,
+      "kl_div_avg": 0.039063937962055206,
+      "learning_rate": 4.776336878915926e-07,
+      "loss_func": "stage2",
+      "step": 313
+    },
+    {
+      "epoch": 0.23007963594994313,
+      "importance_ratio": 1.0004396438598633,
+      "kl_div_avg": 0.05700904130935669,
+      "learning_rate": 4.775608334547574e-07,
+      "loss_func": "stage2",
+      "step": 314
+    },
+    {
+      "epoch": 0.23007963594994313,
+      "importance_ratio": 1.0002377033233643,
+      "kl_div_avg": 0.03964870795607567,
+      "learning_rate": 4.774879790179222e-07,
+      "loss_func": "stage2",
+      "step": 315
+    },
+    {
+      "epoch": 0.23007963594994313,
+      "importance_ratio": 1.0020824670791626,
+      "kl_div_avg": 0.0543329194188118,
+      "learning_rate": 4.774151245810869e-07,
+      "loss_func": "stage2",
+      "step": 316
+    },
+    {
+      "epoch": 0.23299203640500568,
+      "importance_ratio": 1.0,
+      "kl_div_avg": 0.06052602827548981,
+      "learning_rate": 4.773422701442518e-07,
+      "loss_func": "stage2",
+      "step": 317
+    },
+    {
+      "epoch": 0.23299203640500568,
+      "importance_ratio": 0.9997171759605408,
+      "kl_div_avg": 0.11223512887954712,
+      "learning_rate": 4.772694157074166e-07,
+      "loss_func": "stage2",
+      "step": 318
+    },
+    {
+      "epoch": 0.23299203640500568,
+      "importance_ratio": 1.00046706199646,
+      "kl_div_avg": 0.04581636190414429,
+      "learning_rate": 4.771965612705814e-07,
+      "loss_func": "stage2",
+      "step": 319
+    },
+    {
+      "epoch": 0.23299203640500568,
+      "importance_ratio": 1.002081036567688,
+      "kl_div_avg": 0.038961946964263916,
+      "learning_rate": 4.771237068337461e-07,
+      "loss_func": "stage2",
+      "step": 320
+    },
+    {
+      "epoch": 0.23590443686006826,
+      "importance_ratio": 0.9999942183494568,
+      "kl_div_avg": 0.023084010928869247,
+      "learning_rate": 4.770508523969109e-07,
+      "loss_func": "stage2",
+      "step": 321,
+      "total_loss": -1.359375
+    },
+    {
+      "epoch": 0.23590443686006826,
+      "importance_ratio": 1.000396490097046,
+      "kl_div_avg": 0.024995721876621246,
+      "learning_rate": 4.769779979600758e-07,
+      "loss_func": "stage2",
+      "step": 322,
+      "total_loss": -1.09375
+    },
+    {
+      "epoch": 0.23590443686006826,
+      "importance_ratio": 1.0004132986068726,
+      "kl_div_avg": 0.03009365126490593,
+      "learning_rate": 4.769051435232405e-07,
+      "loss_func": "stage2",
+      "step": 323
+    },
+    {
+      "epoch": 0.23590443686006826,
+      "importance_ratio": 1.0022618770599365,
+      "kl_div_avg": 0.040596380829811096,
+      "learning_rate": 4.768322890864053e-07,
+      "loss_func": "stage2",
+      "step": 324,
+      "total_loss": 1.09375
+    },
+    {
+      "epoch": 0.23881683731513084,
+      "importance_ratio": 1.0,
+      "kl_div_avg": 0.04582751914858818,
+      "learning_rate": 4.7675943464957014e-07,
+      "loss_func": "stage2",
+      "step": 325,
+      "total_loss": -0.8515625
+    },
+    {
+      "epoch": 0.23881683731513084,
+      "importance_ratio": 0.9998639225959778,
+      "kl_div_avg": 0.024669548496603966,
+      "learning_rate": 4.7668658021273497e-07,
+      "loss_func": "stage2",
+      "step": 326,
+      "total_loss": -0.58203125
+    },
+    {
+      "epoch": 0.23881683731513084,
+      "importance_ratio": 0.9992770552635193,
+      "kl_div_avg": 0.04793988913297653,
+      "learning_rate": 4.7661372577589975e-07,
+      "loss_func": "stage2",
+      "step": 327,
+      "total_loss": 1.7890625
+    },
+    {
+      "epoch": 0.23881683731513084,
+      "importance_ratio": 0.9995605945587158,
+      "kl_div_avg": 0.043279629200696945,
+      "learning_rate": 4.765408713390645e-07,
+      "loss_func": "stage2",
+      "step": 328,
+      "total_loss": -0.462890625
+    },
+    {
+      "epoch": 0.2417292377701934,
+      "importance_ratio": 1.0,
+      "kl_div_avg": 0.0608009472489357,
+      "learning_rate": 4.764680169022293e-07,
+      "loss_func": "stage2",
+      "step": 329
+    },
+    {
+      "epoch": 0.2417292377701934,
+      "importance_ratio": 1.000767707824707,
+      "kl_div_avg": 0.06964662671089172,
+      "learning_rate": 4.7639516246539413e-07,
+      "loss_func": "stage2",
+      "step": 330
+    },
+    {
+      "epoch": 0.2417292377701934,
+      "importance_ratio": 1.0013978481292725,
+      "kl_div_avg": 0.07270665466785431,
+      "learning_rate": 4.7632230802855896e-07,
+      "loss_func": "stage2",
+      "step": 331
+    },
+    {
+      "epoch": 0.2417292377701934,
+      "importance_ratio": 1.000546932220459,
+      "kl_div_avg": 0.06220786273479462,
+      "learning_rate": 4.762494535917237e-07,
+      "loss_func": "stage2",
+      "step": 332
+    },
+    {
+      "epoch": 0.24464163822525598,
+      "importance_ratio": 1.0,
+      "kl_div_avg": 0.050163887441158295,
+      "learning_rate": 4.761765991548885e-07,
+      "loss_func": "stage2",
+      "step": 333,
+      "total_loss": 0.84375
+    },
+    {
+      "epoch": 0.24464163822525598,
+      "importance_ratio": 0.9995476007461548,
+      "kl_div_avg": 0.028794296085834503,
+      "learning_rate": 4.761037447180533e-07,
+      "loss_func": "stage2",
+      "step": 334,
+      "total_loss": 0.984375
+    },
+    {
+      "epoch": 0.24464163822525598,
+      "importance_ratio": 0.9990682601928711,
+      "kl_div_avg": 0.016825854778289795,
+      "learning_rate": 4.760308902812181e-07,
+      "loss_func": "stage2",
+      "step": 335,
+      "total_loss": -0.5859375
+    },
+    {
+      "epoch": 0.24464163822525598,
+      "importance_ratio": 0.9990837574005127,
+      "kl_div_avg": 0.0576476976275444,
+      "learning_rate": 4.759580358443829e-07,
+      "loss_func": "stage2",
+      "step": 336
+    },
+    {
+      "epoch": 0.24755403868031856,
+      "importance_ratio": 1.0,
+      "kl_div_avg": 0.03124171867966652,
+      "learning_rate": 4.758851814075477e-07,
+      "loss_func": "stage2",
+      "step": 337,
+      "total_loss": -0.5625
+    },
+    {
+      "epoch": 0.24755403868031856,
+      "importance_ratio": 0.9997870326042175,
+      "kl_div_avg": 0.030996575951576233,
+      "learning_rate": 4.758123269707125e-07,
+      "loss_func": "stage2",
+      "step": 338,
+      "total_loss": -0.7578125
+    },
+    {
+      "epoch": 0.24755403868031856,
+      "importance_ratio": 0.9987279176712036,
+      "kl_div_avg": 0.04082472622394562,
+      "learning_rate": 4.757394725338773e-07,
+      "loss_func": "stage2",
+      "step": 339,
+      "total_loss": -0.5078125
+    },
+    {
+      "epoch": 0.24755403868031856,
+      "importance_ratio": 0.9984229803085327,
+      "kl_div_avg": 0.04259004443883896,
+      "learning_rate": 4.756666180970421e-07,
+      "loss_func": "stage2",
+      "step": 340,
+      "total_loss": 1.921875
+    },
+    {
+      "epoch": 0.25046643913538114,
+      "importance_ratio": 1.000009536743164,
+      "kl_div_avg": 0.0397428497672081,
+      "learning_rate": 4.755937636602069e-07,
+      "loss_func": "stage2",
+      "step": 341,
+      "total_loss": -0.78125
+    },
+    {
+      "epoch": 0.25046643913538114,
+      "importance_ratio": 1.0004072189331055,
+      "kl_div_avg": 0.0358537882566452,
+      "learning_rate": 4.7552090922337167e-07,
+      "loss_func": "stage2",
+      "step": 342,
+      "total_loss": -0.265625
+    },
+    {
+      "epoch": 0.25046643913538114,
+      "importance_ratio": 0.9997845888137817,
+      "kl_div_avg": 0.02491338737308979,
+      "learning_rate": 4.754480547865365e-07,
+      "loss_func": "stage2",
+      "step": 343
+    },
+    {
+      "epoch": 0.25046643913538114,
+      "importance_ratio": 1.0003361701965332,
+      "kl_div_avg": 0.01933516189455986,
+      "learning_rate": 4.753752003497013e-07,
+      "loss_func": "stage2",
+      "step": 344,
+      "total_loss": 2.171875
+    },
+    {
+      "epoch": 0.25337883959044366,
+      "importance_ratio": 1.0,
+      "kl_div_avg": 0.055748093873262405,
+      "learning_rate": 4.7530234591286605e-07,
+      "loss_func": "stage2",
+      "step": 345,
+      "total_loss": 0.15625
+    },
+    {
+      "epoch": 0.25337883959044366,
+      "importance_ratio": 1.001150131225586,
+      "kl_div_avg": 0.06542669236660004,
+      "learning_rate": 4.752294914760309e-07,
+      "loss_func": "stage2",
+      "step": 346,
+      "total_loss": -0.341796875
+    },
+    {
+      "epoch": 0.25337883959044366,
+      "importance_ratio": 1.0002367496490479,
+      "kl_div_avg": 0.02955247461795807,
+      "learning_rate": 4.7515663703919566e-07,
+      "loss_func": "stage2",
+      "step": 347
+    },
+    {
+      "epoch": 0.25337883959044366,
+      "importance_ratio": 1.0002845525741577,
+      "kl_div_avg": 0.028858419507741928,
+      "learning_rate": 4.750837826023605e-07,
+      "loss_func": "stage2",
+      "step": 348
+    },
+    {
+      "epoch": 0.25629124004550624,
+      "importance_ratio": 1.0,
+      "kl_div_avg": 0.050181884318590164,
+      "learning_rate": 4.750109281655252e-07,
+      "loss_func": "stage2",
+      "step": 349
+    },
+    {
+      "epoch": 0.25629124004550624,
+      "importance_ratio": 1.0005871057510376,
+      "kl_div_avg": 0.04680076614022255,
+      "learning_rate": 4.7493807372869004e-07,
+      "loss_func": "stage2",
+      "step": 350
+    },
+    {
+      "epoch": 0.25629124004550624,
+      "importance_ratio": 1.0019506216049194,
+      "kl_div_avg": 0.06889910995960236,
+      "learning_rate": 4.748652192918549e-07,
+      "loss_func": "stage2",
+      "step": 351
+    },
+    {
+      "epoch": 0.25629124004550624,
+      "importance_ratio": 1.0027525424957275,
+      "kl_div_avg": 0.05746535211801529,
+      "learning_rate": 4.7479236485501965e-07,
+      "loss_func": "stage2",
+      "step": 352
+    },
+    {
+      "epoch": 0.2592036405005688,
+      "importance_ratio": 1.0,
+      "kl_div_avg": 0.042875513434410095,
+      "learning_rate": 4.7471951041818443e-07,
+      "loss_func": "stage2",
+      "step": 353
+    },
+    {
+      "epoch": 0.2592036405005688,
+      "importance_ratio": 1.0002708435058594,
+      "kl_div_avg": 0.028494440019130707,
+      "learning_rate": 4.746466559813492e-07,
+      "loss_func": "stage2",
+      "step": 354
+    },
+    {
+      "epoch": 0.2592036405005688,
+      "importance_ratio": 1.0010781288146973,
+      "kl_div_avg": 0.036078453063964844,
+      "learning_rate": 4.7457380154451404e-07,
+      "loss_func": "stage2",
+      "step": 355
+    },
+    {
+      "epoch": 0.2592036405005688,
+      "importance_ratio": 1.0010154247283936,
+      "kl_div_avg": 0.0418880432844162,
+      "learning_rate": 4.7450094710767887e-07,
+      "loss_func": "stage2",
+      "step": 356
+    },
+    {
+      "epoch": 0.2621160409556314,
+      "importance_ratio": 1.0,
+      "kl_div_avg": 0.034274179488420486,
+      "learning_rate": 4.7442809267084364e-07,
+      "loss_func": "stage2",
+      "step": 357
+    },
+    {
+      "epoch": 0.2621160409556314,
+      "importance_ratio": 1.0006074905395508,
+      "kl_div_avg": 0.06455402821302414,
+      "learning_rate": 4.743552382340084e-07,
+      "loss_func": "stage2",
+      "step": 358
+    },
+    {
+      "epoch": 0.2621160409556314,
+      "importance_ratio": 1.000532627105713,
+      "kl_div_avg": 0.06488654017448425,
+      "learning_rate": 4.7428238379717325e-07,
+      "loss_func": "stage2",
+      "step": 359
+    },
+    {
+      "epoch": 0.2621160409556314,
+      "importance_ratio": 1.0015405416488647,
+      "kl_div_avg": 0.05848488584160805,
+      "learning_rate": 4.7420952936033803e-07,
+      "loss_func": "stage2",
+      "step": 360
+    },
+    {
+      "epoch": 0.265028441410694,
+      "importance_ratio": 1.0,
+      "kl_div_avg": 0.056837569922208786,
+      "learning_rate": 4.7413667492350286e-07,
+      "loss_func": "stage2",
+      "step": 361
+    },
+    {
+      "epoch": 0.265028441410694,
+      "importance_ratio": 0.9998044967651367,
+      "kl_div_avg": 0.06221582740545273,
+      "learning_rate": 4.740638204866676e-07,
+      "loss_func": "stage2",
+      "step": 362
+    },
+    {
+      "epoch": 0.265028441410694,
+      "importance_ratio": 0.9996775388717651,
+      "kl_div_avg": 0.060226716101169586,
+      "learning_rate": 4.739909660498324e-07,
+      "loss_func": "stage2",
+      "step": 363
+    },
+    {
+      "epoch": 0.265028441410694,
+      "importance_ratio": 0.9990496039390564,
+      "kl_div_avg": 0.07773536443710327,
+      "learning_rate": 4.7391811161299724e-07,
+      "loss_func": "stage2",
+      "step": 364
+    },
+    {
+      "epoch": 0.26794084186575656,
+      "importance_ratio": 1.0,
+      "kl_div_avg": 0.0389438197016716,
+      "learning_rate": 4.73845257176162e-07,
+      "loss_func": "stage2",
+      "step": 365
+    },
+    {
+      "epoch": 0.26794084186575656,
+      "importance_ratio": 0.999747633934021,
+      "kl_div_avg": 0.059490106999874115,
+      "learning_rate": 4.737724027393268e-07,
+      "loss_func": "stage2",
+      "step": 366
+    },
+    {
+      "epoch": 0.26794084186575656,
+      "importance_ratio": 0.9999392032623291,
+      "kl_div_avg": 0.04581678286194801,
+      "learning_rate": 4.736995483024916e-07,
+      "loss_func": "stage2",
+      "step": 367
+    },
+    {
+      "epoch": 0.26794084186575656,
+      "importance_ratio": 0.9997276067733765,
+      "kl_div_avg": 0.04072655737400055,
+      "learning_rate": 4.736266938656564e-07,
+      "loss_func": "stage2",
+      "step": 368
+    },
+    {
+      "epoch": 0.2708532423208191,
+      "importance_ratio": 1.0,
+      "kl_div_avg": 0.04005524516105652,
+      "learning_rate": 4.7355383942882123e-07,
+      "loss_func": "stage2",
+      "step": 369
+    },
+    {
+      "epoch": 0.2708532423208191,
+      "importance_ratio": 0.9993170499801636,
+      "kl_div_avg": 0.06568346917629242,
+      "learning_rate": 4.7348098499198596e-07,
+      "loss_func": "stage2",
+      "step": 370
+    },
+    {
+      "epoch": 0.2708532423208191,
+      "importance_ratio": 0.9980641603469849,
+      "kl_div_avg": 0.05638229474425316,
+      "learning_rate": 4.734081305551508e-07,
+      "loss_func": "stage2",
+      "step": 371
+    },
+    {
+      "epoch": 0.2708532423208191,
+      "importance_ratio": 0.9988054037094116,
+      "kl_div_avg": 0.01818370260298252,
+      "learning_rate": 4.7333527611831557e-07,
+      "loss_func": "stage2",
+      "step": 372
+    },
+    {
+      "epoch": 0.27376564277588167,
+      "importance_ratio": 1.0,
+      "kl_div_avg": 0.04747728630900383,
+      "learning_rate": 4.732624216814804e-07,
+      "loss_func": "stage2",
+      "step": 373
+    },
+    {
+      "epoch": 0.27376564277588167,
+      "importance_ratio": 0.9997256398200989,
+      "kl_div_avg": 0.05248579382896423,
+      "learning_rate": 4.731895672446452e-07,
+      "loss_func": "stage2",
+      "step": 374
+    },
+    {
+      "epoch": 0.27376564277588167,
+      "importance_ratio": 0.9994120001792908,
+      "kl_div_avg": 0.04088594764471054,
+      "learning_rate": 4.7311671280780995e-07,
+      "loss_func": "stage2",
+      "step": 375
+    },
+    {
+      "epoch": 0.27376564277588167,
+      "importance_ratio": 0.9979708194732666,
+      "kl_div_avg": 0.0623336061835289,
+      "learning_rate": 4.730438583709748e-07,
+      "loss_func": "stage2",
+      "step": 376
+    }
+  ],
+  "logging_steps": 1.0,
+  "max_steps": 6868,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 20.0,
+  "save_steps": 50,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": false,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 0,
+  "train_batch_size": 1,
+  "trial_name": null,
+  "trial_params": null
+}

training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bf5202dc42116948ce5e1e9705b424475d1ad4c2c99c84501c25f6f13751af8b
+size 11640

zero_to_fp32.py

@@ -0,0 +1,587 @@
+#!/usr/bin/env python
+
+# Copyright (c) Microsoft Corporation.
+# SPDX-License-Identifier: Apache-2.0
+
+# DeepSpeed Team
+
+# This script extracts fp32 consolidated weights from a zero 1, 2 and 3 DeepSpeed checkpoints. It gets
+# copied into the top level checkpoint dir, so the user can easily do the conversion at any point in
+# the future. Once extracted, the weights don't require DeepSpeed and can be used in any
+# application.
+#
+# example: python zero_to_fp32.py . pytorch_model.bin
+
+import argparse
+import torch
+import glob
+import math
+import os
+import re
+from collections import OrderedDict
+from dataclasses import dataclass
+
+# while this script doesn't use deepspeed to recover data, since the checkpoints are pickled with
+# DeepSpeed data structures it has to be available in the current python environment.
+from deepspeed.utils import logger
+from deepspeed.checkpoint.constants import (DS_VERSION, OPTIMIZER_STATE_DICT, SINGLE_PARTITION_OF_FP32_GROUPS,
+                                            FP32_FLAT_GROUPS, ZERO_STAGE, PARTITION_COUNT, PARAM_SHAPES, BUFFER_NAMES,
+                                            FROZEN_PARAM_SHAPES, FROZEN_PARAM_FRAGMENTS)
+
+
+@dataclass
+class zero_model_state:
+    buffers: dict()
+    param_shapes: dict()
+    shared_params: list
+    ds_version: int
+    frozen_param_shapes: dict()
+    frozen_param_fragments: dict()
+
+
+debug = 0
+
+# load to cpu
+device = torch.device('cpu')
+
+
+def atoi(text):
+    return int(text) if text.isdigit() else text
+
+
+def natural_keys(text):
+    '''
+    alist.sort(key=natural_keys) sorts in human order
+    http://nedbatchelder.com/blog/200712/human_sorting.html
+    (See Toothy's implementation in the comments)
+    '''
+    return [atoi(c) for c in re.split(r'(\d+)', text)]
+
+
+def get_model_state_file(checkpoint_dir, zero_stage):
+    if not os.path.isdir(checkpoint_dir):
+        raise FileNotFoundError(f"Directory '{checkpoint_dir}' doesn't exist")
+
+    # there should be only one file
+    if zero_stage <= 2:
+        file = os.path.join(checkpoint_dir, "mp_rank_00_model_states.pt")
+    elif zero_stage == 3:
+        file = os.path.join(checkpoint_dir, "zero_pp_rank_0_mp_rank_00_model_states.pt")
+
+    if not os.path.exists(file):
+        raise FileNotFoundError(f"can't find model states file at '{file}'")
+
+    return file
+
+
+def get_checkpoint_files(checkpoint_dir, glob_pattern):
+    # XXX: need to test that this simple glob rule works for multi-node setup too
+    ckpt_files = sorted(glob.glob(os.path.join(checkpoint_dir, glob_pattern)), key=natural_keys)
+
+    if len(ckpt_files) == 0:
+        raise FileNotFoundError(f"can't find {glob_pattern} files in directory '{checkpoint_dir}'")
+
+    return ckpt_files
+
+
+def get_optim_files(checkpoint_dir):
+    return get_checkpoint_files(checkpoint_dir, "*_optim_states.pt")
+
+
+def get_model_state_files(checkpoint_dir):
+    return get_checkpoint_files(checkpoint_dir, "*_model_states.pt")
+
+
+def parse_model_states(files):
+    zero_model_states = []
+    for file in files:
+        state_dict = torch.load(file, map_location=device)
+
+        if BUFFER_NAMES not in state_dict:
+            raise ValueError(f"{file} is not a model state checkpoint")
+        buffer_names = state_dict[BUFFER_NAMES]
+        if debug:
+            print("Found buffers:", buffer_names)
+
+        # recover just the buffers while restoring them to fp32 if they were saved in fp16
+        buffers = {k: v.float() for k, v in state_dict["module"].items() if k in buffer_names}
+        param_shapes = state_dict[PARAM_SHAPES]
+
+        # collect parameters that are included in param_shapes
+        param_names = []
+        for s in param_shapes:
+            for name in s.keys():
+                param_names.append(name)
+
+        # update with frozen parameters
+        frozen_param_shapes = state_dict.get(FROZEN_PARAM_SHAPES, None)
+        if frozen_param_shapes is not None:
+            if debug:
+                print(f"Found frozen_param_shapes: {frozen_param_shapes}")
+            param_names += list(frozen_param_shapes.keys())
+
+        # handle shared params
+        shared_params = [[k, v] for k, v in state_dict["shared_params"].items()]
+
+        ds_version = state_dict.get(DS_VERSION, None)
+
+        frozen_param_fragments = state_dict.get(FROZEN_PARAM_FRAGMENTS, None)
+
+        z_model_state = zero_model_state(buffers=buffers,
+                                         param_shapes=param_shapes,
+                                         shared_params=shared_params,
+                                         ds_version=ds_version,
+                                         frozen_param_shapes=frozen_param_shapes,
+                                         frozen_param_fragments=frozen_param_fragments)
+        zero_model_states.append(z_model_state)
+
+    return zero_model_states
+
+
+def parse_optim_states(files, ds_checkpoint_dir):
+
+    total_files = len(files)
+    state_dicts = []
+    for f in files:
+        state_dict = torch.load(f, map_location=device)
+        # immediately discard the potentially huge 2 optimizer states as we only care for fp32 master weights
+        # and also handle the case where it was already removed by another helper script
+        state_dict["optimizer_state_dict"].pop("optimizer_state_dict", None)
+        state_dicts.append(state_dict)
+
+    if not ZERO_STAGE in state_dicts[0][OPTIMIZER_STATE_DICT]:
+        raise ValueError(f"{files[0]} is not a zero checkpoint")
+    zero_stage = state_dicts[0][OPTIMIZER_STATE_DICT][ZERO_STAGE]
+    world_size = state_dicts[0][OPTIMIZER_STATE_DICT][PARTITION_COUNT]
+
+    # For ZeRO-2 each param group can have different partition_count as data parallelism for expert
+    # parameters can be different from data parallelism for non-expert parameters. So we can just
+    # use the max of the partition_count to get the dp world_size.
+
+    if type(world_size) is list:
+        world_size = max(world_size)
+
+    if world_size != total_files:
+        raise ValueError(
+            f"Expected {world_size} of '*_optim_states.pt' under '{ds_checkpoint_dir}' but found {total_files} files. "
+            "Possibly due to an overwrite of an old checkpoint, or a checkpoint didn't get saved by one or more processes."
+        )
+
+    # the groups are named differently in each stage
+    if zero_stage <= 2:
+        fp32_groups_key = SINGLE_PARTITION_OF_FP32_GROUPS
+    elif zero_stage == 3:
+        fp32_groups_key = FP32_FLAT_GROUPS
+    else:
+        raise ValueError(f"unknown zero stage {zero_stage}")
+
+    if zero_stage <= 2:
+        fp32_flat_groups = [state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key] for i in range(len(state_dicts))]
+    elif zero_stage == 3:
+        # if there is more than one param group, there will be multiple flattened tensors - one
+        # flattened tensor per group - for simplicity merge them into a single tensor
+        #
+        # XXX: could make the script more memory efficient for when there are multiple groups - it
+        # will require matching the sub-lists of param_shapes for each param group flattened tensor
+
+        fp32_flat_groups = [
+            torch.cat(state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key], 0) for i in range(len(state_dicts))
+        ]
+
+    return zero_stage, world_size, fp32_flat_groups
+
+
+def _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir):
+    """
+    Returns fp32 state_dict reconstructed from ds checkpoint
+
+    Args:
+        - ``ds_checkpoint_dir``: path to the deepspeed checkpoint folder (where the optimizer files are)
+
+    """
+    print(f"Processing zero checkpoint '{ds_checkpoint_dir}'")
+
+    optim_files = get_optim_files(ds_checkpoint_dir)
+    zero_stage, world_size, fp32_flat_groups = parse_optim_states(optim_files, ds_checkpoint_dir)
+    print(f"Detected checkpoint of type zero stage {zero_stage}, world_size: {world_size}")
+
+    model_files = get_model_state_files(ds_checkpoint_dir)
+
+    zero_model_states = parse_model_states(model_files)
+    print(f'Parsing checkpoint created by deepspeed=={zero_model_states[0].ds_version}')
+
+    if zero_stage <= 2:
+        return _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states)
+    elif zero_stage == 3:
+        return _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states)
+
+
+def _zero2_merge_frozen_params(state_dict, zero_model_states):
+    if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
+        return
+
+    frozen_param_shapes = zero_model_states[0].frozen_param_shapes
+    frozen_param_fragments = zero_model_states[0].frozen_param_fragments
+
+    if debug:
+        num_elem = sum(s.numel() for s in frozen_param_shapes.values())
+        print(f'rank 0: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
+
+        wanted_params = len(frozen_param_shapes)
+        wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
+        avail_numel = sum([p.numel() for p in frozen_param_fragments.values()])
+        print(f'Frozen params: Have {avail_numel} numels to process.')
+        print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
+
+    total_params = 0
+    total_numel = 0
+    for name, shape in frozen_param_shapes.items():
+        total_params += 1
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+
+        state_dict[name] = frozen_param_fragments[name]
+
+        if debug:
+            print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
+
+    print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
+    param_shapes = zero_model_states[0].param_shapes
+
+    # Reconstruction protocol:
+    #
+    # XXX: document this
+
+    if debug:
+        for i in range(world_size):
+            for j in range(len(fp32_flat_groups[0])):
+                print(f"{FP32_FLAT_GROUPS}[{i}][{j}].shape={fp32_flat_groups[i][j].shape}")
+
+    # XXX: memory usage doubles here (zero2)
+    num_param_groups = len(fp32_flat_groups[0])
+    merged_single_partition_of_fp32_groups = []
+    for i in range(num_param_groups):
+        merged_partitions = [sd[i] for sd in fp32_flat_groups]
+        full_single_fp32_vector = torch.cat(merged_partitions, 0)
+        merged_single_partition_of_fp32_groups.append(full_single_fp32_vector)
+    avail_numel = sum(
+        [full_single_fp32_vector.numel() for full_single_fp32_vector in merged_single_partition_of_fp32_groups])
+
+    if debug:
+        wanted_params = sum([len(shapes) for shapes in param_shapes])
+        wanted_numel = sum([sum(shape.numel() for shape in shapes.values()) for shapes in param_shapes])
+        # not asserting if there is a mismatch due to possible padding
+        print(f"Have {avail_numel} numels to process.")
+        print(f"Need {wanted_numel} numels in {wanted_params} params.")
+
+    # params
+    # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
+    # out-of-core computing solution
+    total_numel = 0
+    total_params = 0
+    for shapes, full_single_fp32_vector in zip(param_shapes, merged_single_partition_of_fp32_groups):
+        offset = 0
+        avail_numel = full_single_fp32_vector.numel()
+        for name, shape in shapes.items():
+
+            unpartitioned_numel = shape.numel()
+            total_numel += unpartitioned_numel
+            total_params += 1
+
+            if debug:
+                print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
+            state_dict[name] = full_single_fp32_vector.narrow(0, offset, unpartitioned_numel).view(shape)
+            offset += unpartitioned_numel
+
+        # Z2 started to align to 2*world_size to improve nccl performance. Therefore both offset and
+        # avail_numel can differ by anywhere between 0..2*world_size. Due to two unrelated complex
+        # paddings performed in the code it's almost impossible to predict the exact numbers w/o the
+        # live optimizer object, so we are checking that the numbers are within the right range
+        align_to = 2 * world_size
+
+        def zero2_align(x):
+            return align_to * math.ceil(x / align_to)
+
+        if debug:
+            print(f"original offset={offset}, avail_numel={avail_numel}")
+
+        offset = zero2_align(offset)
+        avail_numel = zero2_align(avail_numel)
+
+        if debug:
+            print(f"aligned  offset={offset}, avail_numel={avail_numel}")
+
+        # Sanity check
+        if offset != avail_numel:
+            raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
+
+    print(f"Reconstructed fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    _zero2_merge_frozen_params(state_dict, zero_model_states)
+
+    _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
+
+    # recover shared parameters
+    for pair in zero_model_states[0].shared_params:
+        if pair[1] in state_dict:
+            state_dict[pair[0]] = state_dict[pair[1]]
+
+    return state_dict
+
+
+def zero3_partitioned_param_info(unpartitioned_numel, world_size):
+    remainder = unpartitioned_numel % world_size
+    padding_numel = (world_size - remainder) if remainder else 0
+    partitioned_numel = math.ceil(unpartitioned_numel / world_size)
+    return partitioned_numel, padding_numel
+
+
+def _zero3_merge_frozen_params(state_dict, world_size, zero_model_states):
+    if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
+        return
+
+    if debug:
+        for i in range(world_size):
+            num_elem = sum(s.numel() for s in zero_model_states[i].frozen_param_fragments.values())
+            print(f'rank {i}: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
+
+        frozen_param_shapes = zero_model_states[0].frozen_param_shapes
+        wanted_params = len(frozen_param_shapes)
+        wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
+        avail_numel = sum([p.numel() for p in zero_model_states[0].frozen_param_fragments.values()]) * world_size
+        print(f'Frozen params: Have {avail_numel} numels to process.')
+        print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
+
+    total_params = 0
+    total_numel = 0
+    for name, shape in zero_model_states[0].frozen_param_shapes.items():
+        total_params += 1
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+
+        param_frags = tuple(model_state.frozen_param_fragments[name] for model_state in zero_model_states)
+        state_dict[name] = torch.cat(param_frags, 0).narrow(0, 0, unpartitioned_numel).view(shape)
+
+        partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
+
+        if debug:
+            print(
+                f"Frozen params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
+            )
+
+    print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
+    param_shapes = zero_model_states[0].param_shapes
+    avail_numel = fp32_flat_groups[0].numel() * world_size
+    # Reconstruction protocol: For zero3 we need to zip the partitions together at boundary of each
+    # param, re-consolidating each param, while dealing with padding if any
+
+    # merge list of dicts, preserving order
+    param_shapes = {k: v for d in param_shapes for k, v in d.items()}
+
+    if debug:
+        for i in range(world_size):
+            print(f"{FP32_FLAT_GROUPS}[{i}].shape={fp32_flat_groups[i].shape}")
+
+        wanted_params = len(param_shapes)
+        wanted_numel = sum(shape.numel() for shape in param_shapes.values())
+        # not asserting if there is a mismatch due to possible padding
+        avail_numel = fp32_flat_groups[0].numel() * world_size
+        print(f"Trainable params: Have {avail_numel} numels to process.")
+        print(f"Trainable params: Need {wanted_numel} numels in {wanted_params} params.")
+
+    # params
+    # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
+    # out-of-core computing solution
+    offset = 0
+    total_numel = 0
+    total_params = 0
+    for name, shape in param_shapes.items():
+
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+        total_params += 1
+
+        partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
+
+        if debug:
+            print(
+                f"Trainable params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
+            )
+
+        # XXX: memory usage doubles here
+        state_dict[name] = torch.cat(
+            tuple(fp32_flat_groups[i].narrow(0, offset, partitioned_numel) for i in range(world_size)),
+            0).narrow(0, 0, unpartitioned_numel).view(shape)
+        offset += partitioned_numel
+
+    offset *= world_size
+
+    # Sanity check
+    if offset != avail_numel:
+        raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
+
+    print(f"Reconstructed Trainable fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    _zero3_merge_frozen_params(state_dict, world_size, zero_model_states)
+
+    _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
+
+    # recover shared parameters
+    for pair in zero_model_states[0].shared_params:
+        if pair[1] in state_dict:
+            state_dict[pair[0]] = state_dict[pair[1]]
+
+    return state_dict
+
+
+def get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag=None):
+    """
+    Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated state_dict that can be loaded with
+    ``load_state_dict()`` and used for training without DeepSpeed or shared with others, for example
+    via a model hub.
+
+    Args:
+        - ``checkpoint_dir``: path to the desired checkpoint folder
+        - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in 'latest' file. e.g., ``global_step14``
+
+    Returns:
+        - pytorch ``state_dict``
+
+    Note: this approach may not work if your application doesn't have sufficient free CPU memory and
+    you may need to use the offline approach using the ``zero_to_fp32.py`` script that is saved with
+    the checkpoint.
+
+    A typical usage might be ::
+
+        from deepspeed.utils.zero_to_fp32 import get_fp32_state_dict_from_zero_checkpoint
+        # do the training and checkpoint saving
+        state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir) # already on cpu
+        model = model.cpu() # move to cpu
+        model.load_state_dict(state_dict)
+        # submit to model hub or save the model to share with others
+
+    In this example the ``model`` will no longer be usable in the deepspeed context of the same
+    application. i.e. you will need to re-initialize the deepspeed engine, since
+    ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
+
+    If you want it all done for you, use ``load_state_dict_from_zero_checkpoint`` instead.
+
+    """
+    if tag is None:
+        latest_path = os.path.join(checkpoint_dir, 'latest')
+        if os.path.isfile(latest_path):
+            with open(latest_path, 'r') as fd:
+                tag = fd.read().strip()
+        else:
+            raise ValueError(f"Unable to find 'latest' file at {latest_path}")
+
+    ds_checkpoint_dir = os.path.join(checkpoint_dir, tag)
+
+    if not os.path.isdir(ds_checkpoint_dir):
+        raise FileNotFoundError(f"Directory '{ds_checkpoint_dir}' doesn't exist")
+
+    return _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir)
+
+
+def convert_zero_checkpoint_to_fp32_state_dict(checkpoint_dir, output_file, tag=None):
+    """
+    Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict`` file that can be
+    loaded with ``torch.load(file)`` + ``load_state_dict()`` and used for training without DeepSpeed.
+
+    Args:
+        - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
+        - ``output_file``: path to the pytorch fp32 state_dict output file (e.g. path/pytorch_model.bin)
+        - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
+    """
+
+    state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag)
+    print(f"Saving fp32 state dict to {output_file}")
+    torch.save(state_dict, output_file)
+
+
+def load_state_dict_from_zero_checkpoint(model, checkpoint_dir, tag=None):
+    """
+    1. Put the provided model to cpu
+    2. Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict``
+    3. Load it into the provided model
+
+    Args:
+        - ``model``: the model object to update
+        - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
+        - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
+
+    Returns:
+        - ``model`: modified model
+
+    Make sure you have plenty of CPU memory available before you call this function. If you don't
+    have enough use the ``zero_to_fp32.py`` utility to do the conversion. You will find it
+    conveniently placed for you in the checkpoint folder.
+
+    A typical usage might be ::
+
+        from deepspeed.utils.zero_to_fp32 import load_state_dict_from_zero_checkpoint
+        model = load_state_dict_from_zero_checkpoint(trainer.model, checkpoint_dir)
+        # submit to model hub or save the model to share with others
+
+    Note, that once this was run, the ``model`` will no longer be usable in the deepspeed context
+    of the same application. i.e. you will need to re-initialize the deepspeed engine, since
+    ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
+
+    """
+    logger.info(f"Extracting fp32 weights")
+    state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag)
+
+    logger.info(f"Overwriting model with fp32 weights")
+    model = model.cpu()
+    model.load_state_dict(state_dict, strict=False)
+
+    return model
+
+
+if __name__ == "__main__":
+
+    parser = argparse.ArgumentParser()
+    parser.add_argument("checkpoint_dir",
+                        type=str,
+                        help="path to the desired checkpoint folder, e.g., path/checkpoint-12")
+    parser.add_argument(
+        "output_file",
+        type=str,
+        help="path to the pytorch fp32 state_dict output file (e.g. path/checkpoint-12/pytorch_model.bin)")
+    parser.add_argument("-t",
+                        "--tag",
+                        type=str,
+                        default=None,
+                        help="checkpoint tag used as a unique identifier for checkpoint. e.g., global_step1")
+    parser.add_argument("-d", "--debug", action='store_true', help="enable debug")
+    args = parser.parse_args()
+
+    debug = args.debug
+
+    convert_zero_checkpoint_to_fp32_state_dict(args.checkpoint_dir, args.output_file, tag=args.tag)