diff --git a/weight_dir/DpoWeight/DPOP_Fix_ND3V1/checkpoint-1300/README.md b/weight_dir/DpoWeight/DPOP_Fix_ND3V1/checkpoint-1300/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..fca4e142c186eed8cc62de799f497379c37cf395
--- /dev/null
+++ b/weight_dir/DpoWeight/DPOP_Fix_ND3V1/checkpoint-1300/README.md
@@ -0,0 +1,202 @@
+---
+library_name: peft
+base_model: ./output_dir/fix_codeLlama
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.11.1
\ No newline at end of file
diff --git a/weight_dir/DpoWeight/DPOP_Fix_ND3V1/checkpoint-1300/adapter_config.json b/weight_dir/DpoWeight/DPOP_Fix_ND3V1/checkpoint-1300/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..575a448f63fb53fbd72f81c498d8a392a21028f0
--- /dev/null
+++ b/weight_dir/DpoWeight/DPOP_Fix_ND3V1/checkpoint-1300/adapter_config.json
@@ -0,0 +1,34 @@
+{
+  "alpha_pattern": {},
+  "auto_mapping": null,
+  "base_model_name_or_path": "./output_dir/fix_codeLlama",
+  "bias": "none",
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 16,
+  "lora_dropout": 0.05,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "r": 8,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "fc_out",
+    "q_proj",
+    "k_proj",
+    "fc_in",
+    "wte",
+    "v_proj",
+    "out_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "use_dora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/weight_dir/DpoWeight/DPOP_Fix_ND3V1/checkpoint-1300/adapter_model.safetensors b/weight_dir/DpoWeight/DPOP_Fix_ND3V1/checkpoint-1300/adapter_model.safetensors
new file mode 100644
index 0000000000000000000000000000000000000000..c2d3a426dd48435a4235416bda666af1cc3fc1ae
--- /dev/null
+++ b/weight_dir/DpoWeight/DPOP_Fix_ND3V1/checkpoint-1300/adapter_model.safetensors
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b3bc6527e61727868deb335a388d96da4e34312a12e2813a5de86145429eae49
+size 25191360
diff --git a/weight_dir/DpoWeight/DPOP_Fix_ND3V1/checkpoint-1300/added_tokens.json b/weight_dir/DpoWeight/DPOP_Fix_ND3V1/checkpoint-1300/added_tokens.json
new file mode 100644
index 0000000000000000000000000000000000000000..c09203d8a52151e0d3b5f3c8e6daedc5b60832b5
--- /dev/null
+++ b/weight_dir/DpoWeight/DPOP_Fix_ND3V1/checkpoint-1300/added_tokens.json
@@ -0,0 +1,3 @@
+{
+  "<PAD>": 32016
+}
diff --git a/weight_dir/DpoWeight/DPOP_Fix_ND3V1/checkpoint-1300/optimizer.pt b/weight_dir/DpoWeight/DPOP_Fix_ND3V1/checkpoint-1300/optimizer.pt
new file mode 100644
index 0000000000000000000000000000000000000000..3050b773e6fb2b6a3c6a32256e096ddaee9fdae4
--- /dev/null
+++ b/weight_dir/DpoWeight/DPOP_Fix_ND3V1/checkpoint-1300/optimizer.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:10ccd2e39e796e699e4a18d72274aa953b12ab9ff05892840247e7f21689b079
+size 50445242
diff --git a/weight_dir/DpoWeight/DPOP_Fix_ND3V1/checkpoint-1300/rng_state_0.pth b/weight_dir/DpoWeight/DPOP_Fix_ND3V1/checkpoint-1300/rng_state_0.pth
new file mode 100644
index 0000000000000000000000000000000000000000..821db9d32dc925919011ebfc6eaa9775b9c28453
--- /dev/null
+++ b/weight_dir/DpoWeight/DPOP_Fix_ND3V1/checkpoint-1300/rng_state_0.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d70f4816d9856643f7986b709d87aa477584114329562b62b70a8ee6bedbc11b
+size 14960
diff --git a/weight_dir/DpoWeight/DPOP_Fix_ND3V1/checkpoint-1300/rng_state_1.pth b/weight_dir/DpoWeight/DPOP_Fix_ND3V1/checkpoint-1300/rng_state_1.pth
new file mode 100644
index 0000000000000000000000000000000000000000..49e1e337a584b0935e42549c91b360c37b575900
--- /dev/null
+++ b/weight_dir/DpoWeight/DPOP_Fix_ND3V1/checkpoint-1300/rng_state_1.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:861efdf0dceef25cf08029e4518f52f53768678173fcd28906388233c3fb3b02
+size 14960
diff --git a/weight_dir/DpoWeight/DPOP_Fix_ND3V1/checkpoint-1300/rng_state_2.pth b/weight_dir/DpoWeight/DPOP_Fix_ND3V1/checkpoint-1300/rng_state_2.pth
new file mode 100644
index 0000000000000000000000000000000000000000..19c08616c8adbf113b6c07188c76e959b541c970
--- /dev/null
+++ b/weight_dir/DpoWeight/DPOP_Fix_ND3V1/checkpoint-1300/rng_state_2.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ebdb27a64a2feccfa8a360d15c28e460fbe78b54e9d7e847e02a63fc8fa3eb33
+size 14960
diff --git a/weight_dir/DpoWeight/DPOP_Fix_ND3V1/checkpoint-1300/rng_state_3.pth b/weight_dir/DpoWeight/DPOP_Fix_ND3V1/checkpoint-1300/rng_state_3.pth
new file mode 100644
index 0000000000000000000000000000000000000000..25edf7d4a6204cc520c57c2eda32dddccd4be297
--- /dev/null
+++ b/weight_dir/DpoWeight/DPOP_Fix_ND3V1/checkpoint-1300/rng_state_3.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ca8dd3637e6e034724517d6f63740144aba4bf2af43d797bbf16033e48567a69
+size 14960
diff --git a/weight_dir/DpoWeight/DPOP_Fix_ND3V1/checkpoint-1300/scheduler.pt b/weight_dir/DpoWeight/DPOP_Fix_ND3V1/checkpoint-1300/scheduler.pt
new file mode 100644
index 0000000000000000000000000000000000000000..265180afd994aa0d0089f5d8f496a23cf7feb4ed
--- /dev/null
+++ b/weight_dir/DpoWeight/DPOP_Fix_ND3V1/checkpoint-1300/scheduler.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d0355f6d2acfbddd46776507259ab052a22b6ee6d2b38dce536d622af0a82d05
+size 1064
diff --git a/weight_dir/DpoWeight/DPOP_Fix_ND3V1/checkpoint-1300/special_tokens_map.json b/weight_dir/DpoWeight/DPOP_Fix_ND3V1/checkpoint-1300/special_tokens_map.json
new file mode 100644
index 0000000000000000000000000000000000000000..2ece23e993f7e1c7063cb51148b6fa5c6c224775
--- /dev/null
+++ b/weight_dir/DpoWeight/DPOP_Fix_ND3V1/checkpoint-1300/special_tokens_map.json
@@ -0,0 +1,30 @@
+{
+  "bos_token": {
+    "content": "<s>",
+    "lstrip": false,
+    "normalized": true,
+    "rstrip": false,
+    "single_word": false
+  },
+  "eos_token": {
+    "content": "</s>",
+    "lstrip": false,
+    "normalized": true,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": {
+    "content": "<PAD>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "unk_token": {
+    "content": "<unk>",
+    "lstrip": false,
+    "normalized": true,
+    "rstrip": false,
+    "single_word": false
+  }
+}
diff --git a/weight_dir/DpoWeight/DPOP_Fix_ND3V1/checkpoint-1300/tokenizer.model b/weight_dir/DpoWeight/DPOP_Fix_ND3V1/checkpoint-1300/tokenizer.model
new file mode 100644
index 0000000000000000000000000000000000000000..f6722e8b170230ebdd4c0f5f2ce03f219be536d4
--- /dev/null
+++ b/weight_dir/DpoWeight/DPOP_Fix_ND3V1/checkpoint-1300/tokenizer.model
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:45ccb9c8b6b561889acea59191d66986d314e7cbd6a78abc6e49b139ca91c1e6
+size 500058
diff --git a/weight_dir/DpoWeight/DPOP_Fix_ND3V1/checkpoint-1300/tokenizer_config.json b/weight_dir/DpoWeight/DPOP_Fix_ND3V1/checkpoint-1300/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..dc4d9a55d56b17600905bfc69a0e45e2fae7a1c4
--- /dev/null
+++ b/weight_dir/DpoWeight/DPOP_Fix_ND3V1/checkpoint-1300/tokenizer_config.json
@@ -0,0 +1,50 @@
+{
+  "add_bos_token": true,
+  "add_eos_token": false,
+  "add_prefix_space": true,
+  "added_tokens_decoder": {
+    "0": {
+      "content": "<unk>",
+      "lstrip": false,
+      "normalized": true,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "1": {
+      "content": "<s>",
+      "lstrip": false,
+      "normalized": true,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "2": {
+      "content": "</s>",
+      "lstrip": false,
+      "normalized": true,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "32016": {
+      "content": "<PAD>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    }
+  },
+  "bos_token": "<s>",
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "</s>",
+  "legacy": true,
+  "model_max_length": 1000000000000000019884624838656,
+  "pad_token": "<PAD>",
+  "sp_model_kwargs": {},
+  "spaces_between_special_tokens": false,
+  "tokenizer_class": "LlamaTokenizer",
+  "unk_token": "<unk>",
+  "use_default_system_prompt": false
+}
diff --git a/weight_dir/DpoWeight/DPOP_Fix_ND3V1/checkpoint-1300/trainer_state.json b/weight_dir/DpoWeight/DPOP_Fix_ND3V1/checkpoint-1300/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..a5114f6b35fbdb36d91275c19358e6dedccb3585
--- /dev/null
+++ b/weight_dir/DpoWeight/DPOP_Fix_ND3V1/checkpoint-1300/trainer_state.json
@@ -0,0 +1,2031 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 16.883116883116884,
+  "eval_steps": 400,
+  "global_step": 1300,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.12987012987012986,
+      "grad_norm": 106.52960205078125,
+      "learning_rate": 5.000000000000001e-07,
+      "logits/chosen": -2.8220086097717285,
+      "logits/rejected": -2.813664674758911,
+      "logps/chosen": -6.851963043212891,
+      "logps/rejected": -22.329076766967773,
+      "loss": 1.7802,
+      "rewards/accuracies": 0.42500001192092896,
+      "rewards/chosen": 0.0019315744284540415,
+      "rewards/margins": -0.00016089165001176298,
+      "rewards/rejected": 0.0020924662239849567,
+      "step": 10
+    },
+    {
+      "epoch": 0.2597402597402597,
+      "grad_norm": 249.00958251953125,
+      "learning_rate": 1.0000000000000002e-06,
+      "logits/chosen": -2.8003315925598145,
+      "logits/rejected": -2.791141986846924,
+      "logps/chosen": -7.967398166656494,
+      "logps/rejected": -21.916549682617188,
+      "loss": 3.02,
+      "rewards/accuracies": 0.44999998807907104,
+      "rewards/chosen": -0.0019492346327751875,
+      "rewards/margins": -0.0002492312341928482,
+      "rewards/rejected": -0.0017000030493363738,
+      "step": 20
+    },
+    {
+      "epoch": 0.38961038961038963,
+      "grad_norm": 231.99557495117188,
+      "learning_rate": 1.5e-06,
+      "logits/chosen": -2.801818609237671,
+      "logits/rejected": -2.795116424560547,
+      "logps/chosen": -8.219635009765625,
+      "logps/rejected": -22.34463882446289,
+      "loss": 2.6654,
+      "rewards/accuracies": 0.4000000059604645,
+      "rewards/chosen": 0.000893857330083847,
+      "rewards/margins": -0.003195409430190921,
+      "rewards/rejected": 0.004089267458766699,
+      "step": 30
+    },
+    {
+      "epoch": 0.5194805194805194,
+      "grad_norm": 94.2333755493164,
+      "learning_rate": 2.0000000000000003e-06,
+      "logits/chosen": -2.8080127239227295,
+      "logits/rejected": -2.804051637649536,
+      "logps/chosen": -9.42918586730957,
+      "logps/rejected": -22.583698272705078,
+      "loss": 2.3196,
+      "rewards/accuracies": 0.574999988079071,
+      "rewards/chosen": 0.0048586102202534676,
+      "rewards/margins": 0.002601353684440255,
+      "rewards/rejected": 0.0022572563029825687,
+      "step": 40
+    },
+    {
+      "epoch": 0.6493506493506493,
+      "grad_norm": 213.9645538330078,
+      "learning_rate": 2.5e-06,
+      "logits/chosen": -2.805826187133789,
+      "logits/rejected": -2.7987563610076904,
+      "logps/chosen": -7.2015862464904785,
+      "logps/rejected": -23.5020694732666,
+      "loss": 2.6731,
+      "rewards/accuracies": 0.5,
+      "rewards/chosen": 0.0019127244595438242,
+      "rewards/margins": 0.002104248385876417,
+      "rewards/rejected": -0.000191524246474728,
+      "step": 50
+    },
+    {
+      "epoch": 0.7792207792207793,
+      "grad_norm": 110.76453399658203,
+      "learning_rate": 3e-06,
+      "logits/chosen": -2.801133394241333,
+      "logits/rejected": -2.7940406799316406,
+      "logps/chosen": -9.063034057617188,
+      "logps/rejected": -20.352069854736328,
+      "loss": 2.5189,
+      "rewards/accuracies": 0.44999998807907104,
+      "rewards/chosen": -0.003256978467106819,
+      "rewards/margins": -0.0020958157256245613,
+      "rewards/rejected": -0.0011611627414822578,
+      "step": 60
+    },
+    {
+      "epoch": 0.9090909090909091,
+      "grad_norm": 174.06761169433594,
+      "learning_rate": 3.5e-06,
+      "logits/chosen": -2.815706968307495,
+      "logits/rejected": -2.8055267333984375,
+      "logps/chosen": -7.804632663726807,
+      "logps/rejected": -22.440189361572266,
+      "loss": 1.7947,
+      "rewards/accuracies": 0.3499999940395355,
+      "rewards/chosen": 0.002698649885132909,
+      "rewards/margins": -0.0054735904559493065,
+      "rewards/rejected": 0.008172241039574146,
+      "step": 70
+    },
+    {
+      "epoch": 1.0389610389610389,
+      "grad_norm": 87.83943939208984,
+      "learning_rate": 4.000000000000001e-06,
+      "logits/chosen": -2.8100368976593018,
+      "logits/rejected": -2.798546075820923,
+      "logps/chosen": -7.676709175109863,
+      "logps/rejected": -21.805500030517578,
+      "loss": 1.678,
+      "rewards/accuracies": 0.4000000059604645,
+      "rewards/chosen": 0.00427329121157527,
+      "rewards/margins": -0.006891848053783178,
+      "rewards/rejected": 0.011165140196681023,
+      "step": 80
+    },
+    {
+      "epoch": 1.1688311688311688,
+      "grad_norm": 116.58035278320312,
+      "learning_rate": 4.5e-06,
+      "logits/chosen": -2.8018991947174072,
+      "logits/rejected": -2.7894458770751953,
+      "logps/chosen": -8.77466869354248,
+      "logps/rejected": -22.66843032836914,
+      "loss": 1.872,
+      "rewards/accuracies": 0.30000001192092896,
+      "rewards/chosen": 0.0032441741786897182,
+      "rewards/margins": -0.009967333637177944,
+      "rewards/rejected": 0.01321150828152895,
+      "step": 90
+    },
+    {
+      "epoch": 1.2987012987012987,
+      "grad_norm": 94.74404907226562,
+      "learning_rate": 5e-06,
+      "logits/chosen": -2.8042826652526855,
+      "logits/rejected": -2.7987048625946045,
+      "logps/chosen": -7.8375115394592285,
+      "logps/rejected": -21.124469757080078,
+      "loss": 1.7389,
+      "rewards/accuracies": 0.25,
+      "rewards/chosen": 0.00530009251087904,
+      "rewards/margins": -0.010122200474143028,
+      "rewards/rejected": 0.015422293916344643,
+      "step": 100
+    },
+    {
+      "epoch": 1.4285714285714286,
+      "grad_norm": 36.4909782409668,
+      "learning_rate": 4.997252228714279e-06,
+      "logits/chosen": -2.8379225730895996,
+      "logits/rejected": -2.831869602203369,
+      "logps/chosen": -8.803790092468262,
+      "logps/rejected": -20.375246047973633,
+      "loss": 1.3703,
+      "rewards/accuracies": 0.42500001192092896,
+      "rewards/chosen": 0.01235434040427208,
+      "rewards/margins": -0.006987647153437138,
+      "rewards/rejected": 0.019341984763741493,
+      "step": 110
+    },
+    {
+      "epoch": 1.5584415584415585,
+      "grad_norm": 39.18604278564453,
+      "learning_rate": 4.989014955054746e-06,
+      "logits/chosen": -2.817542552947998,
+      "logits/rejected": -2.8139848709106445,
+      "logps/chosen": -7.656688690185547,
+      "logps/rejected": -20.828929901123047,
+      "loss": 1.1774,
+      "rewards/accuracies": 0.30000001192092896,
+      "rewards/chosen": 0.01700626313686371,
+      "rewards/margins": -0.011658851057291031,
+      "rewards/rejected": 0.02866511419415474,
+      "step": 120
+    },
+    {
+      "epoch": 1.6883116883116882,
+      "grad_norm": 65.63705444335938,
+      "learning_rate": 4.975306286336628e-06,
+      "logits/chosen": -2.8285748958587646,
+      "logits/rejected": -2.823660373687744,
+      "logps/chosen": -7.743475914001465,
+      "logps/rejected": -20.759389877319336,
+      "loss": 1.3562,
+      "rewards/accuracies": 0.2750000059604645,
+      "rewards/chosen": 0.020265722647309303,
+      "rewards/margins": -0.01016728300601244,
+      "rewards/rejected": 0.03043300285935402,
+      "step": 130
+    },
+    {
+      "epoch": 1.8181818181818183,
+      "grad_norm": 170.55665588378906,
+      "learning_rate": 4.95615635718894e-06,
+      "logits/chosen": -2.821213722229004,
+      "logits/rejected": -2.8140625953674316,
+      "logps/chosen": -8.116856575012207,
+      "logps/rejected": -19.581222534179688,
+      "loss": 1.7842,
+      "rewards/accuracies": 0.2750000059604645,
+      "rewards/chosen": 0.017684193328022957,
+      "rewards/margins": -0.010469591245055199,
+      "rewards/rejected": 0.028153782710433006,
+      "step": 140
+    },
+    {
+      "epoch": 1.948051948051948,
+      "grad_norm": 71.12332916259766,
+      "learning_rate": 4.931607263312033e-06,
+      "logits/chosen": -2.823256015777588,
+      "logits/rejected": -2.812701463699341,
+      "logps/chosen": -6.964502811431885,
+      "logps/rejected": -21.734210968017578,
+      "loss": 1.5393,
+      "rewards/accuracies": 0.22499999403953552,
+      "rewards/chosen": 0.010581249371170998,
+      "rewards/margins": -0.01839107647538185,
+      "rewards/rejected": 0.0289723239839077,
+      "step": 150
+    },
+    {
+      "epoch": 2.0779220779220777,
+      "grad_norm": 173.57005310058594,
+      "learning_rate": 4.901712968942101e-06,
+      "logits/chosen": -2.808175563812256,
+      "logits/rejected": -2.798382520675659,
+      "logps/chosen": -7.128758907318115,
+      "logps/rejected": -22.715810775756836,
+      "loss": 1.4333,
+      "rewards/accuracies": 0.30000001192092896,
+      "rewards/chosen": 0.012118353508412838,
+      "rewards/margins": -0.01597173698246479,
+      "rewards/rejected": 0.028090089559555054,
+      "step": 160
+    },
+    {
+      "epoch": 2.207792207792208,
+      "grad_norm": 119.67935180664062,
+      "learning_rate": 4.866539188226086e-06,
+      "logits/chosen": -2.788814067840576,
+      "logits/rejected": -2.781687021255493,
+      "logps/chosen": -9.231825828552246,
+      "logps/rejected": -23.992216110229492,
+      "loss": 1.2406,
+      "rewards/accuracies": 0.25,
+      "rewards/chosen": 0.01427288819104433,
+      "rewards/margins": -0.011877929791808128,
+      "rewards/rejected": 0.026150818914175034,
+      "step": 170
+    },
+    {
+      "epoch": 2.3376623376623376,
+      "grad_norm": 78.3140869140625,
+      "learning_rate": 4.826163240767717e-06,
+      "logits/chosen": -2.819803476333618,
+      "logits/rejected": -2.812873363494873,
+      "logps/chosen": -8.85770320892334,
+      "logps/rejected": -21.352455139160156,
+      "loss": 1.6327,
+      "rewards/accuracies": 0.25,
+      "rewards/chosen": 0.01637933775782585,
+      "rewards/margins": -0.017695654183626175,
+      "rewards/rejected": 0.034074995666742325,
+      "step": 180
+    },
+    {
+      "epoch": 2.4675324675324677,
+      "grad_norm": 34.724788665771484,
+      "learning_rate": 4.780673881662242e-06,
+      "logits/chosen": -2.802525043487549,
+      "logits/rejected": -2.7943992614746094,
+      "logps/chosen": -9.141613006591797,
+      "logps/rejected": -21.16362762451172,
+      "loss": 1.3604,
+      "rewards/accuracies": 0.30000001192092896,
+      "rewards/chosen": 0.019131088629364967,
+      "rewards/margins": -0.012987020425498486,
+      "rewards/rejected": 0.03211811184883118,
+      "step": 190
+    },
+    {
+      "epoch": 2.5974025974025974,
+      "grad_norm": 155.53683471679688,
+      "learning_rate": 4.730171106393466e-06,
+      "logits/chosen": -2.8240277767181396,
+      "logits/rejected": -2.8137762546539307,
+      "logps/chosen": -6.279611587524414,
+      "logps/rejected": -22.48813247680664,
+      "loss": 1.3818,
+      "rewards/accuracies": 0.3499999940395355,
+      "rewards/chosen": 0.018246710300445557,
+      "rewards/margins": -0.011857626028358936,
+      "rewards/rejected": 0.03010433353483677,
+      "step": 200
+    },
+    {
+      "epoch": 2.7272727272727275,
+      "grad_norm": 64.57522583007812,
+      "learning_rate": 4.674765931021976e-06,
+      "logits/chosen": -2.8272862434387207,
+      "logits/rejected": -2.8214235305786133,
+      "logps/chosen": -8.027566909790039,
+      "logps/rejected": -20.517822265625,
+      "loss": 1.2211,
+      "rewards/accuracies": 0.25,
+      "rewards/chosen": 0.01772388443350792,
+      "rewards/margins": -0.015222163870930672,
+      "rewards/rejected": 0.03294604271650314,
+      "step": 210
+    },
+    {
+      "epoch": 2.857142857142857,
+      "grad_norm": 138.96739196777344,
+      "learning_rate": 4.614580148147744e-06,
+      "logits/chosen": -2.8115546703338623,
+      "logits/rejected": -2.804287910461426,
+      "logps/chosen": -6.3580474853515625,
+      "logps/rejected": -20.281545639038086,
+      "loss": 1.3669,
+      "rewards/accuracies": 0.32499998807907104,
+      "rewards/chosen": 0.0185546912252903,
+      "rewards/margins": -0.011832709424197674,
+      "rewards/rejected": 0.03038739785552025,
+      "step": 220
+    },
+    {
+      "epoch": 2.987012987012987,
+      "grad_norm": 55.348880767822266,
+      "learning_rate": 4.5497460591835615e-06,
+      "logits/chosen": -2.8048064708709717,
+      "logits/rejected": -2.799748420715332,
+      "logps/chosen": -8.203588485717773,
+      "logps/rejected": -20.907344818115234,
+      "loss": 1.1073,
+      "rewards/accuracies": 0.44999998807907104,
+      "rewards/chosen": 0.02281924895942211,
+      "rewards/margins": -0.005632441025227308,
+      "rewards/rejected": 0.02845168672502041,
+      "step": 230
+    },
+    {
+      "epoch": 3.116883116883117,
+      "grad_norm": 114.07877349853516,
+      "learning_rate": 4.480406183527823e-06,
+      "logits/chosen": -2.820413589477539,
+      "logits/rejected": -2.810023307800293,
+      "logps/chosen": -7.28661584854126,
+      "logps/rejected": -22.23759651184082,
+      "loss": 1.2872,
+      "rewards/accuracies": 0.5,
+      "rewards/chosen": 0.02074371837079525,
+      "rewards/margins": -0.00726896058768034,
+      "rewards/rejected": 0.028012678027153015,
+      "step": 240
+    },
+    {
+      "epoch": 3.2467532467532467,
+      "grad_norm": 114.84168243408203,
+      "learning_rate": 4.406712945275955e-06,
+      "logits/chosen": -2.8098714351654053,
+      "logits/rejected": -2.8057053089141846,
+      "logps/chosen": -8.777189254760742,
+      "logps/rejected": -23.82990074157715,
+      "loss": 1.2292,
+      "rewards/accuracies": 0.32499998807907104,
+      "rewards/chosen": 0.020751068368554115,
+      "rewards/margins": -0.011817799881100655,
+      "rewards/rejected": 0.03256886824965477,
+      "step": 250
+    },
+    {
+      "epoch": 3.3766233766233764,
+      "grad_norm": 125.23298645019531,
+      "learning_rate": 4.328828338159173e-06,
+      "logits/chosen": -2.8179001808166504,
+      "logits/rejected": -2.809523820877075,
+      "logps/chosen": -7.79488468170166,
+      "logps/rejected": -21.827838897705078,
+      "loss": 1.0056,
+      "rewards/accuracies": 0.44999998807907104,
+      "rewards/chosen": 0.018612362444400787,
+      "rewards/margins": -0.005726366303861141,
+      "rewards/rejected": 0.024338727816939354,
+      "step": 260
+    },
+    {
+      "epoch": 3.5064935064935066,
+      "grad_norm": 118.24935150146484,
+      "learning_rate": 4.246923569447105e-06,
+      "logits/chosen": -2.827606201171875,
+      "logits/rejected": -2.8264949321746826,
+      "logps/chosen": -9.784859657287598,
+      "logps/rejected": -19.85146141052246,
+      "loss": 1.2282,
+      "rewards/accuracies": 0.22499999403953552,
+      "rewards/chosen": 0.01882827654480934,
+      "rewards/margins": -0.013741184957325459,
+      "rewards/rejected": 0.032569460570812225,
+      "step": 270
+    },
+    {
+      "epoch": 3.6363636363636362,
+      "grad_norm": 231.2564697265625,
+      "learning_rate": 4.161178683597055e-06,
+      "logits/chosen": -2.8154025077819824,
+      "logits/rejected": -2.808724880218506,
+      "logps/chosen": -10.802755355834961,
+      "logps/rejected": -20.200098037719727,
+      "loss": 1.2421,
+      "rewards/accuracies": 0.42500001192092896,
+      "rewards/chosen": 0.016843315213918686,
+      "rewards/margins": -0.009146241471171379,
+      "rewards/rejected": 0.025989552959799767,
+      "step": 280
+    },
+    {
+      "epoch": 3.7662337662337664,
+      "grad_norm": 178.36927795410156,
+      "learning_rate": 4.071782166477213e-06,
+      "logits/chosen": -2.827821969985962,
+      "logits/rejected": -2.817737102508545,
+      "logps/chosen": -7.973962306976318,
+      "logps/rejected": -20.839879989624023,
+      "loss": 0.9562,
+      "rewards/accuracies": 0.550000011920929,
+      "rewards/chosen": 0.02151758037507534,
+      "rewards/margins": 0.0009062625467777252,
+      "rewards/rejected": 0.020611315965652466,
+      "step": 290
+    },
+    {
+      "epoch": 3.896103896103896,
+      "grad_norm": 177.05746459960938,
+      "learning_rate": 3.978930531033807e-06,
+      "logits/chosen": -2.81160306930542,
+      "logits/rejected": -2.807624340057373,
+      "logps/chosen": -7.694178581237793,
+      "logps/rejected": -20.271114349365234,
+      "loss": 1.2155,
+      "rewards/accuracies": 0.3499999940395355,
+      "rewards/chosen": 0.02011878788471222,
+      "rewards/margins": -0.011032785288989544,
+      "rewards/rejected": 0.03115157224237919,
+      "step": 300
+    },
+    {
+      "epoch": 4.025974025974026,
+      "grad_norm": 85.1644287109375,
+      "learning_rate": 3.882827885312999e-06,
+      "logits/chosen": -2.826159954071045,
+      "logits/rejected": -2.8190231323242188,
+      "logps/chosen": -7.839417934417725,
+      "logps/rejected": -20.949127197265625,
+      "loss": 1.1714,
+      "rewards/accuracies": 0.5249999761581421,
+      "rewards/chosen": 0.018428083509206772,
+      "rewards/margins": -0.0033446471206843853,
+      "rewards/rejected": 0.021772734820842743,
+      "step": 310
+    },
+    {
+      "epoch": 4.1558441558441555,
+      "grad_norm": 96.2293472290039,
+      "learning_rate": 3.783685483787105e-06,
+      "logits/chosen": -2.7987990379333496,
+      "logits/rejected": -2.7919256687164307,
+      "logps/chosen": -8.851567268371582,
+      "logps/rejected": -21.075613021850586,
+      "loss": 1.0528,
+      "rewards/accuracies": 0.550000011920929,
+      "rewards/chosen": 0.027806680649518967,
+      "rewards/margins": 0.007172185927629471,
+      "rewards/rejected": 0.020634492859244347,
+      "step": 320
+    },
+    {
+      "epoch": 4.285714285714286,
+      "grad_norm": 64.6633071899414,
+      "learning_rate": 3.6817212629714135e-06,
+      "logits/chosen": -2.8131086826324463,
+      "logits/rejected": -2.8076319694519043,
+      "logps/chosen": -8.776361465454102,
+      "logps/rejected": -21.563058853149414,
+      "loss": 0.9623,
+      "rewards/accuracies": 0.32499998807907104,
+      "rewards/chosen": 0.02237451635301113,
+      "rewards/margins": -0.005439778324216604,
+      "rewards/rejected": 0.027814293280243874,
+      "step": 330
+    },
+    {
+      "epoch": 4.415584415584416,
+      "grad_norm": 61.17387771606445,
+      "learning_rate": 3.5771593623524263e-06,
+      "logits/chosen": -2.8211700916290283,
+      "logits/rejected": -2.8142738342285156,
+      "logps/chosen": -8.312505722045898,
+      "logps/rejected": -22.2635440826416,
+      "loss": 0.8663,
+      "rewards/accuracies": 0.5249999761581421,
+      "rewards/chosen": 0.022781318053603172,
+      "rewards/margins": -0.0001775051496224478,
+      "rewards/rejected": 0.022958822548389435,
+      "step": 340
+    },
+    {
+      "epoch": 4.545454545454545,
+      "grad_norm": 244.74371337890625,
+      "learning_rate": 3.4702296316806243e-06,
+      "logits/chosen": -2.8051793575286865,
+      "logits/rejected": -2.8000447750091553,
+      "logps/chosen": -8.497454643249512,
+      "logps/rejected": -21.671079635620117,
+      "loss": 0.9413,
+      "rewards/accuracies": 0.5249999761581421,
+      "rewards/chosen": 0.018070612102746964,
+      "rewards/margins": -0.0006186591344885528,
+      "rewards/rejected": 0.018689271062612534,
+      "step": 350
+    },
+    {
+      "epoch": 4.675324675324675,
+      "grad_norm": 180.42735290527344,
+      "learning_rate": 3.3611671257108323e-06,
+      "logits/chosen": -2.8005387783050537,
+      "logits/rejected": -2.7906429767608643,
+      "logps/chosen": -8.441190719604492,
+      "logps/rejected": -22.472454071044922,
+      "loss": 1.0874,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": 0.02211115136742592,
+      "rewards/margins": 0.009135196916759014,
+      "rewards/rejected": 0.012975958175957203,
+      "step": 360
+    },
+    {
+      "epoch": 4.805194805194805,
+      "grad_norm": 37.385963439941406,
+      "learning_rate": 3.2502115875008523e-06,
+      "logits/chosen": -2.822923183441162,
+      "logits/rejected": -2.8172943592071533,
+      "logps/chosen": -6.2603559494018555,
+      "logps/rejected": -19.10385513305664,
+      "loss": 1.1011,
+      "rewards/accuracies": 0.550000011920929,
+      "rewards/chosen": 0.018987987190485,
+      "rewards/margins": 0.00487549277022481,
+      "rewards/rejected": 0.014112496748566628,
+      "step": 370
+    },
+    {
+      "epoch": 4.935064935064935,
+      "grad_norm": 251.71954345703125,
+      "learning_rate": 3.1376069214041917e-06,
+      "logits/chosen": -2.826533079147339,
+      "logits/rejected": -2.8127052783966064,
+      "logps/chosen": -8.482178688049316,
+      "logps/rejected": -22.023988723754883,
+      "loss": 1.0408,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": 0.02052697166800499,
+      "rewards/margins": 0.00905714649707079,
+      "rewards/rejected": 0.011469824239611626,
+      "step": 380
+    },
+    {
+      "epoch": 5.064935064935065,
+      "grad_norm": 133.91131591796875,
+      "learning_rate": 3.023600656915362e-06,
+      "logits/chosen": -2.822097063064575,
+      "logits/rejected": -2.810638189315796,
+      "logps/chosen": -6.029572486877441,
+      "logps/rejected": -21.33598518371582,
+      "loss": 0.9734,
+      "rewards/accuracies": 0.625,
+      "rewards/chosen": 0.023712964728474617,
+      "rewards/margins": 0.003511254210025072,
+      "rewards/rejected": 0.020201710984110832,
+      "step": 390
+    },
+    {
+      "epoch": 5.194805194805195,
+      "grad_norm": 133.3660125732422,
+      "learning_rate": 2.9084434045463255e-06,
+      "logits/chosen": -2.80812668800354,
+      "logits/rejected": -2.7997758388519287,
+      "logps/chosen": -7.040734767913818,
+      "logps/rejected": -20.19479751586914,
+      "loss": 0.9652,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": 0.020205114036798477,
+      "rewards/margins": 0.010116524063050747,
+      "rewards/rejected": 0.01008858997374773,
+      "step": 400
+    },
+    {
+      "epoch": 5.194805194805195,
+      "eval_logits/chosen": -2.801586627960205,
+      "eval_logits/rejected": -2.8042006492614746,
+      "eval_logps/chosen": -13.702254295349121,
+      "eval_logps/rejected": -15.930770874023438,
+      "eval_loss": 0.7836798429489136,
+      "eval_rewards/accuracies": 0.75,
+      "eval_rewards/chosen": 0.04482783377170563,
+      "eval_rewards/margins": 0.010822022333741188,
+      "eval_rewards/rejected": 0.03400580957531929,
+      "eval_runtime": 1.1767,
+      "eval_samples_per_second": 11.898,
+      "eval_steps_per_second": 1.7,
+      "step": 400
+    },
+    {
+      "epoch": 5.324675324675325,
+      "grad_norm": 103.42053985595703,
+      "learning_rate": 2.792388304930207e-06,
+      "logits/chosen": -2.8027472496032715,
+      "logits/rejected": -2.7954494953155518,
+      "logps/chosen": -8.696739196777344,
+      "logps/rejected": -21.20206642150879,
+      "loss": 0.8734,
+      "rewards/accuracies": 0.75,
+      "rewards/chosen": 0.025552362203598022,
+      "rewards/margins": 0.018355753272771835,
+      "rewards/rejected": 0.007196612656116486,
+      "step": 410
+    },
+    {
+      "epoch": 5.454545454545454,
+      "grad_norm": 236.04249572753906,
+      "learning_rate": 2.6756904723632325e-06,
+      "logits/chosen": -2.820538282394409,
+      "logits/rejected": -2.8109259605407715,
+      "logps/chosen": -8.277128219604492,
+      "logps/rejected": -21.056652069091797,
+      "loss": 1.0417,
+      "rewards/accuracies": 0.675000011920929,
+      "rewards/chosen": 0.02177886664867401,
+      "rewards/margins": 0.017902854830026627,
+      "rewards/rejected": 0.0038760111201554537,
+      "step": 420
+    },
+    {
+      "epoch": 5.584415584415584,
+      "grad_norm": 94.26475524902344,
+      "learning_rate": 2.5586064340081516e-06,
+      "logits/chosen": -2.8236119747161865,
+      "logits/rejected": -2.8140316009521484,
+      "logps/chosen": -7.59194803237915,
+      "logps/rejected": -21.763874053955078,
+      "loss": 1.1263,
+      "rewards/accuracies": 0.7250000238418579,
+      "rewards/chosen": 0.017852844670414925,
+      "rewards/margins": 0.01606573723256588,
+      "rewards/rejected": 0.0017871044110506773,
+      "step": 430
+    },
+    {
+      "epoch": 5.714285714285714,
+      "grad_norm": 195.20712280273438,
+      "learning_rate": 2.441393565991849e-06,
+      "logits/chosen": -2.807347297668457,
+      "logits/rejected": -2.7958061695098877,
+      "logps/chosen": -8.326894760131836,
+      "logps/rejected": -21.803070068359375,
+      "loss": 0.9669,
+      "rewards/accuracies": 0.7250000238418579,
+      "rewards/chosen": 0.019576936960220337,
+      "rewards/margins": 0.02373579703271389,
+      "rewards/rejected": -0.0041588591411709785,
+      "step": 440
+    },
+    {
+      "epoch": 5.8441558441558445,
+      "grad_norm": 175.50625610351562,
+      "learning_rate": 2.3243095276367687e-06,
+      "logits/chosen": -2.8178904056549072,
+      "logits/rejected": -2.8124308586120605,
+      "logps/chosen": -8.626742362976074,
+      "logps/rejected": -21.10752296447754,
+      "loss": 1.4178,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": 0.01709837280213833,
+      "rewards/margins": 0.018973568454384804,
+      "rewards/rejected": -0.001875193091109395,
+      "step": 450
+    },
+    {
+      "epoch": 5.974025974025974,
+      "grad_norm": 171.52513122558594,
+      "learning_rate": 2.207611695069794e-06,
+      "logits/chosen": -2.8153228759765625,
+      "logits/rejected": -2.8087317943573,
+      "logps/chosen": -8.533574104309082,
+      "logps/rejected": -22.59617805480957,
+      "loss": 1.1735,
+      "rewards/accuracies": 0.7749999761581421,
+      "rewards/chosen": 0.019586745649576187,
+      "rewards/margins": 0.03454780578613281,
+      "rewards/rejected": -0.014961063861846924,
+      "step": 460
+    },
+    {
+      "epoch": 6.103896103896104,
+      "grad_norm": 1.1001964807510376,
+      "learning_rate": 2.0915565954536745e-06,
+      "logits/chosen": -2.8197810649871826,
+      "logits/rejected": -2.8137242794036865,
+      "logps/chosen": -6.954007148742676,
+      "logps/rejected": -20.310867309570312,
+      "loss": 0.9806,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": 0.02565738931298256,
+      "rewards/margins": 0.02424320951104164,
+      "rewards/rejected": 0.001414179103448987,
+      "step": 470
+    },
+    {
+      "epoch": 6.233766233766234,
+      "grad_norm": 129.70095825195312,
+      "learning_rate": 1.9763993430846394e-06,
+      "logits/chosen": -2.8019607067108154,
+      "logits/rejected": -2.798675537109375,
+      "logps/chosen": -8.585060119628906,
+      "logps/rejected": -22.695144653320312,
+      "loss": 1.1706,
+      "rewards/accuracies": 0.699999988079071,
+      "rewards/chosen": 0.017484549432992935,
+      "rewards/margins": 0.01860320381820202,
+      "rewards/rejected": -0.0011186569463461637,
+      "step": 480
+    },
+    {
+      "epoch": 6.363636363636363,
+      "grad_norm": 194.80984497070312,
+      "learning_rate": 1.8623930785958092e-06,
+      "logits/chosen": -2.8055920600891113,
+      "logits/rejected": -2.7976577281951904,
+      "logps/chosen": -8.328341484069824,
+      "logps/rejected": -21.830015182495117,
+      "loss": 0.9653,
+      "rewards/accuracies": 0.824999988079071,
+      "rewards/chosen": 0.023755336180329323,
+      "rewards/margins": 0.03609599173069,
+      "rewards/rejected": -0.012340660206973553,
+      "step": 490
+    },
+    {
+      "epoch": 6.4935064935064934,
+      "grad_norm": 1.1828835010528564,
+      "learning_rate": 1.7497884124991487e-06,
+      "logits/chosen": -2.8199102878570557,
+      "logits/rejected": -2.8055901527404785,
+      "logps/chosen": -6.756843566894531,
+      "logps/rejected": -20.936315536499023,
+      "loss": 0.749,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": 0.026989247649908066,
+      "rewards/margins": 0.038383591920137405,
+      "rewards/rejected": -0.011394346132874489,
+      "step": 500
+    },
+    {
+      "epoch": 6.623376623376624,
+      "grad_norm": 38.752262115478516,
+      "learning_rate": 1.6388328742891679e-06,
+      "logits/chosen": -2.82857084274292,
+      "logits/rejected": -2.8192813396453857,
+      "logps/chosen": -5.166688919067383,
+      "logps/rejected": -21.360883712768555,
+      "loss": 0.9419,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/chosen": 0.02238740213215351,
+      "rewards/margins": 0.04248126968741417,
+      "rewards/rejected": -0.020093869417905807,
+      "step": 510
+    },
+    {
+      "epoch": 6.753246753246753,
+      "grad_norm": 160.1698455810547,
+      "learning_rate": 1.5297703683193755e-06,
+      "logits/chosen": -2.813995599746704,
+      "logits/rejected": -2.8053503036499023,
+      "logps/chosen": -8.157999038696289,
+      "logps/rejected": -21.715497970581055,
+      "loss": 0.9404,
+      "rewards/accuracies": 0.824999988079071,
+      "rewards/chosen": 0.025560980662703514,
+      "rewards/margins": 0.04841204732656479,
+      "rewards/rejected": -0.022851066663861275,
+      "step": 520
+    },
+    {
+      "epoch": 6.883116883116883,
+      "grad_norm": 94.66495513916016,
+      "learning_rate": 1.4228406376475741e-06,
+      "logits/chosen": -2.8161110877990723,
+      "logits/rejected": -2.806417942047119,
+      "logps/chosen": -9.14362907409668,
+      "logps/rejected": -21.98373794555664,
+      "loss": 1.3017,
+      "rewards/accuracies": 0.949999988079071,
+      "rewards/chosen": 0.022670337930321693,
+      "rewards/margins": 0.04901245981454849,
+      "rewards/rejected": -0.02634212002158165,
+      "step": 530
+    },
+    {
+      "epoch": 7.012987012987013,
+      "grad_norm": 89.24470520019531,
+      "learning_rate": 1.3182787370285865e-06,
+      "logits/chosen": -2.8100571632385254,
+      "logits/rejected": -2.802196502685547,
+      "logps/chosen": -8.180788040161133,
+      "logps/rejected": -22.61843490600586,
+      "loss": 0.9971,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": 0.02741624414920807,
+      "rewards/margins": 0.052984196692705154,
+      "rewards/rejected": -0.025567958131432533,
+      "step": 540
+    },
+    {
+      "epoch": 7.142857142857143,
+      "grad_norm": 46.27058410644531,
+      "learning_rate": 1.2163145162128948e-06,
+      "logits/chosen": -2.811455249786377,
+      "logits/rejected": -2.801848888397217,
+      "logps/chosen": -8.542337417602539,
+      "logps/rejected": -21.399410247802734,
+      "loss": 0.9509,
+      "rewards/accuracies": 0.824999988079071,
+      "rewards/chosen": 0.02186908759176731,
+      "rewards/margins": 0.0516178198158741,
+      "rewards/rejected": -0.029748734086751938,
+      "step": 550
+    },
+    {
+      "epoch": 7.2727272727272725,
+      "grad_norm": 69.24237823486328,
+      "learning_rate": 1.1171721146870015e-06,
+      "logits/chosen": -2.8135528564453125,
+      "logits/rejected": -2.803555965423584,
+      "logps/chosen": -8.623903274536133,
+      "logps/rejected": -22.410194396972656,
+      "loss": 0.9666,
+      "rewards/accuracies": 0.875,
+      "rewards/chosen": 0.03136039525270462,
+      "rewards/margins": 0.06211704760789871,
+      "rewards/rejected": -0.030756641179323196,
+      "step": 560
+    },
+    {
+      "epoch": 7.402597402597403,
+      "grad_norm": 1.2535158395767212,
+      "learning_rate": 1.021069468966194e-06,
+      "logits/chosen": -2.8163511753082275,
+      "logits/rejected": -2.808074474334717,
+      "logps/chosen": -7.339944362640381,
+      "logps/rejected": -22.135278701782227,
+      "loss": 0.8673,
+      "rewards/accuracies": 0.8500000238418579,
+      "rewards/chosen": 0.025004085153341293,
+      "rewards/margins": 0.048856452107429504,
+      "rewards/rejected": -0.023852365091443062,
+      "step": 570
+    },
+    {
+      "epoch": 7.532467532467533,
+      "grad_norm": 1.1791399717330933,
+      "learning_rate": 9.282178335227885e-07,
+      "logits/chosen": -2.7872376441955566,
+      "logits/rejected": -2.7806403636932373,
+      "logps/chosen": -8.17931079864502,
+      "logps/rejected": -23.90004539489746,
+      "loss": 1.4371,
+      "rewards/accuracies": 0.925000011920929,
+      "rewards/chosen": 0.025283193215727806,
+      "rewards/margins": 0.06855350732803345,
+      "rewards/rejected": -0.04327031224966049,
+      "step": 580
+    },
+    {
+      "epoch": 7.662337662337662,
+      "grad_norm": 1.237028956413269,
+      "learning_rate": 8.38821316402946e-07,
+      "logits/chosen": -2.828946352005005,
+      "logits/rejected": -2.82279634475708,
+      "logps/chosen": -7.834005832672119,
+      "logps/rejected": -23.059555053710938,
+      "loss": 0.872,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": 0.023193147033452988,
+      "rewards/margins": 0.051843591034412384,
+      "rewards/rejected": -0.0286504365503788,
+      "step": 590
+    },
+    {
+      "epoch": 7.792207792207792,
+      "grad_norm": 1.1478286981582642,
+      "learning_rate": 7.530764305528959e-07,
+      "logits/chosen": -2.8227028846740723,
+      "logits/rejected": -2.815317153930664,
+      "logps/chosen": -7.287829399108887,
+      "logps/rejected": -21.66531753540039,
+      "loss": 0.9423,
+      "rewards/accuracies": 0.8500000238418579,
+      "rewards/chosen": 0.02044813148677349,
+      "rewards/margins": 0.050043750554323196,
+      "rewards/rejected": -0.029595619067549706,
+      "step": 600
+    },
+    {
+      "epoch": 7.922077922077922,
+      "grad_norm": 0.9912703037261963,
+      "learning_rate": 6.711716618408282e-07,
+      "logits/chosen": -2.838146686553955,
+      "logits/rejected": -2.8304896354675293,
+      "logps/chosen": -6.6907477378845215,
+      "logps/rejected": -19.75206756591797,
+      "loss": 1.0092,
+      "rewards/accuracies": 0.875,
+      "rewards/chosen": 0.03283644840121269,
+      "rewards/margins": 0.04647618532180786,
+      "rewards/rejected": -0.01363973505795002,
+      "step": 610
+    },
+    {
+      "epoch": 8.051948051948052,
+      "grad_norm": 205.0751953125,
+      "learning_rate": 5.932870547240455e-07,
+      "logits/chosen": -2.814697504043579,
+      "logits/rejected": -2.803783893585205,
+      "logps/chosen": -7.918333530426025,
+      "logps/rejected": -22.431245803833008,
+      "loss": 1.1134,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": 0.026031214743852615,
+      "rewards/margins": 0.05806659907102585,
+      "rewards/rejected": -0.03203538805246353,
+      "step": 620
+    },
+    {
+      "epoch": 8.181818181818182,
+      "grad_norm": 181.82443237304688,
+      "learning_rate": 5.195938164721767e-07,
+      "logits/chosen": -2.825852394104004,
+      "logits/rejected": -2.8194408416748047,
+      "logps/chosen": -7.493607997894287,
+      "logps/rejected": -22.741878509521484,
+      "loss": 0.9249,
+      "rewards/accuracies": 0.949999988079071,
+      "rewards/chosen": 0.031044036149978638,
+      "rewards/margins": 0.054368507117033005,
+      "rewards/rejected": -0.023324472829699516,
+      "step": 630
+    },
+    {
+      "epoch": 8.311688311688311,
+      "grad_norm": 1.184661626815796,
+      "learning_rate": 4.502539408164386e-07,
+      "logits/chosen": -2.8212850093841553,
+      "logits/rejected": -2.8142733573913574,
+      "logps/chosen": -8.27627944946289,
+      "logps/rejected": -20.00022315979004,
+      "loss": 0.9228,
+      "rewards/accuracies": 0.824999988079071,
+      "rewards/chosen": 0.025791505351662636,
+      "rewards/margins": 0.04540405049920082,
+      "rewards/rejected": -0.019612547010183334,
+      "step": 640
+    },
+    {
+      "epoch": 8.441558441558442,
+      "grad_norm": 128.95013427734375,
+      "learning_rate": 3.8541985185225645e-07,
+      "logits/chosen": -2.824355363845825,
+      "logits/rejected": -2.8176026344299316,
+      "logps/chosen": -7.515383243560791,
+      "logps/rejected": -21.355083465576172,
+      "loss": 1.0536,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/chosen": 0.02296631410717964,
+      "rewards/margins": 0.056810516864061356,
+      "rewards/rejected": -0.03384420648217201,
+      "step": 650
+    },
+    {
+      "epoch": 8.571428571428571,
+      "grad_norm": 276.641357421875,
+      "learning_rate": 3.252340689780245e-07,
+      "logits/chosen": -2.82700777053833,
+      "logits/rejected": -2.822455883026123,
+      "logps/chosen": -7.617165565490723,
+      "logps/rejected": -23.340801239013672,
+      "loss": 1.2275,
+      "rewards/accuracies": 0.925000011920929,
+      "rewards/chosen": 0.02447410859167576,
+      "rewards/margins": 0.07831484079360962,
+      "rewards/rejected": -0.05384073406457901,
+      "step": 660
+    },
+    {
+      "epoch": 8.7012987012987,
+      "grad_norm": 241.392333984375,
+      "learning_rate": 2.698288936065338e-07,
+      "logits/chosen": -2.7932610511779785,
+      "logits/rejected": -2.7827224731445312,
+      "logps/chosen": -7.8216118812561035,
+      "logps/rejected": -20.492076873779297,
+      "loss": 1.0025,
+      "rewards/accuracies": 0.925000011920929,
+      "rewards/chosen": 0.024464499205350876,
+      "rewards/margins": 0.055416546761989594,
+      "rewards/rejected": -0.030952051281929016,
+      "step": 670
+    },
+    {
+      "epoch": 8.831168831168831,
+      "grad_norm": 1.2769261598587036,
+      "learning_rate": 2.1932611833775846e-07,
+      "logits/chosen": -2.8169243335723877,
+      "logits/rejected": -2.8119499683380127,
+      "logps/chosen": -8.214499473571777,
+      "logps/rejected": -22.425582885742188,
+      "loss": 1.2029,
+      "rewards/accuracies": 0.925000011920929,
+      "rewards/chosen": 0.02633916214108467,
+      "rewards/margins": 0.05255778878927231,
+      "rewards/rejected": -0.026218628510832787,
+      "step": 680
+    },
+    {
+      "epoch": 8.96103896103896,
+      "grad_norm": 198.49639892578125,
+      "learning_rate": 1.7383675923228372e-07,
+      "logits/chosen": -2.7998576164245605,
+      "logits/rejected": -2.7877182960510254,
+      "logps/chosen": -10.261676788330078,
+      "logps/rejected": -22.23526954650879,
+      "loss": 1.2189,
+      "rewards/accuracies": 0.875,
+      "rewards/chosen": 0.023275194689631462,
+      "rewards/margins": 0.05330119654536247,
+      "rewards/rejected": -0.030026007443666458,
+      "step": 690
+    },
+    {
+      "epoch": 9.090909090909092,
+      "grad_norm": 1.1314146518707275,
+      "learning_rate": 1.3346081177391474e-07,
+      "logits/chosen": -2.795860767364502,
+      "logits/rejected": -2.784841299057007,
+      "logps/chosen": -10.019143104553223,
+      "logps/rejected": -22.597219467163086,
+      "loss": 0.8856,
+      "rewards/accuracies": 0.8500000238418579,
+      "rewards/chosen": 0.026558348909020424,
+      "rewards/margins": 0.05729994922876358,
+      "rewards/rejected": -0.030741602182388306,
+      "step": 700
+    },
+    {
+      "epoch": 9.220779220779221,
+      "grad_norm": 179.00376892089844,
+      "learning_rate": 9.828703105789983e-08,
+      "logits/chosen": -2.8192873001098633,
+      "logits/rejected": -2.8140900135040283,
+      "logps/chosen": -9.172324180603027,
+      "logps/rejected": -22.661869049072266,
+      "loss": 1.0784,
+      "rewards/accuracies": 0.925000011920929,
+      "rewards/chosen": 0.023906812071800232,
+      "rewards/margins": 0.06065645068883896,
+      "rewards/rejected": -0.036749642342329025,
+      "step": 710
+    },
+    {
+      "epoch": 9.35064935064935,
+      "grad_norm": 69.8018798828125,
+      "learning_rate": 6.839273668796747e-08,
+      "logits/chosen": -2.8201510906219482,
+      "logits/rejected": -2.8087105751037598,
+      "logps/chosen": -8.500712394714355,
+      "logps/rejected": -22.419025421142578,
+      "loss": 0.9798,
+      "rewards/accuracies": 0.8500000238418579,
+      "rewards/chosen": 0.023749757558107376,
+      "rewards/margins": 0.05873861163854599,
+      "rewards/rejected": -0.034988854080438614,
+      "step": 720
+    },
+    {
+      "epoch": 9.480519480519481,
+      "grad_norm": 70.18726348876953,
+      "learning_rate": 4.384364281105974e-08,
+      "logits/chosen": -2.83191180229187,
+      "logits/rejected": -2.824836015701294,
+      "logps/chosen": -9.021523475646973,
+      "logps/rejected": -23.541057586669922,
+      "loss": 0.8762,
+      "rewards/accuracies": 0.925000011920929,
+      "rewards/chosen": 0.024020517244935036,
+      "rewards/margins": 0.06196972727775574,
+      "rewards/rejected": -0.03794920817017555,
+      "step": 730
+    },
+    {
+      "epoch": 9.61038961038961,
+      "grad_norm": 308.2706298828125,
+      "learning_rate": 2.4693713663372643e-08,
+      "logits/chosen": -2.799358367919922,
+      "logits/rejected": -2.7913246154785156,
+      "logps/chosen": -7.899996757507324,
+      "logps/rejected": -21.6778564453125,
+      "loss": 1.3393,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": 0.01837027817964554,
+      "rewards/margins": 0.05194631963968277,
+      "rewards/rejected": -0.03357603773474693,
+      "step": 740
+    },
+    {
+      "epoch": 9.74025974025974,
+      "grad_norm": 42.1520881652832,
+      "learning_rate": 1.0985044945254763e-08,
+      "logits/chosen": -2.8439416885375977,
+      "logits/rejected": -2.839962959289551,
+      "logps/chosen": -9.0369291305542,
+      "logps/rejected": -19.963855743408203,
+      "loss": 0.917,
+      "rewards/accuracies": 0.800000011920929,
+      "rewards/chosen": 0.02196243405342102,
+      "rewards/margins": 0.04523957893252373,
+      "rewards/rejected": -0.023277146741747856,
+      "step": 750
+    },
+    {
+      "epoch": 9.87012987012987,
+      "grad_norm": 237.67958068847656,
+      "learning_rate": 3.97807508777894e-06,
+      "logits/chosen": -2.8095672130584717,
+      "logits/rejected": -2.8042306900024414,
+      "logps/chosen": -8.958699226379395,
+      "logps/rejected": -20.755828857421875,
+      "loss": 1.4632,
+      "rewards/accuracies": 0.75,
+      "rewards/chosen": 0.021344048902392387,
+      "rewards/margins": 0.05193667858839035,
+      "rewards/rejected": -0.030592631548643112,
+      "step": 760
+    },
+    {
+      "epoch": 10.0,
+      "grad_norm": 415.3313293457031,
+      "learning_rate": 3.949264905820697e-06,
+      "logits/chosen": -2.8104586601257324,
+      "logits/rejected": -2.8043599128723145,
+      "logps/chosen": -8.099912643432617,
+      "logps/rejected": -22.543119430541992,
+      "loss": 1.0218,
+      "rewards/accuracies": 0.8500000238418579,
+      "rewards/chosen": 0.029804136604070663,
+      "rewards/margins": 0.06551304459571838,
+      "rewards/rejected": -0.03570891544222832,
+      "step": 770
+    },
+    {
+      "epoch": 10.12987012987013,
+      "grad_norm": 1.33208429813385,
+      "learning_rate": 3.92016186682789e-06,
+      "logits/chosen": -2.8259775638580322,
+      "logits/rejected": -2.8215713500976562,
+      "logps/chosen": -9.199613571166992,
+      "logps/rejected": -21.45523452758789,
+      "loss": 0.8068,
+      "rewards/accuracies": 0.925000011920929,
+      "rewards/chosen": 0.02821979485452175,
+      "rewards/margins": 0.05966518074274063,
+      "rewards/rejected": -0.03144538030028343,
+      "step": 780
+    },
+    {
+      "epoch": 10.25974025974026,
+      "grad_norm": 226.0507354736328,
+      "learning_rate": 3.8907718517334405e-06,
+      "logits/chosen": -2.8059029579162598,
+      "logits/rejected": -2.7993156909942627,
+      "logps/chosen": -8.171636581420898,
+      "logps/rejected": -22.19893455505371,
+      "loss": 0.9794,
+      "rewards/accuracies": 0.875,
+      "rewards/chosen": 0.022764720022678375,
+      "rewards/margins": 0.069257453083992,
+      "rewards/rejected": -0.04649273306131363,
+      "step": 790
+    },
+    {
+      "epoch": 10.38961038961039,
+      "grad_norm": 1.1935486793518066,
+      "learning_rate": 3.861100799460336e-06,
+      "logits/chosen": -2.816051959991455,
+      "logits/rejected": -2.81080961227417,
+      "logps/chosen": -7.233284950256348,
+      "logps/rejected": -22.618621826171875,
+      "loss": 1.1384,
+      "rewards/accuracies": 0.925000011920929,
+      "rewards/chosen": 0.030704358592629433,
+      "rewards/margins": 0.08574527502059937,
+      "rewards/rejected": -0.05504090338945389,
+      "step": 800
+    },
+    {
+      "epoch": 10.38961038961039,
+      "eval_logits/chosen": -2.8031392097473145,
+      "eval_logits/rejected": -2.8060073852539062,
+      "eval_logps/chosen": -13.673626899719238,
+      "eval_logps/rejected": -16.335693359375,
+      "eval_loss": 0.6743522882461548,
+      "eval_rewards/accuracies": 1.0,
+      "eval_rewards/chosen": 0.047690678387880325,
+      "eval_rewards/margins": 0.05417700111865997,
+      "eval_rewards/rejected": -0.006486321333795786,
+      "eval_runtime": 1.1696,
+      "eval_samples_per_second": 11.97,
+      "eval_steps_per_second": 1.71,
+      "step": 800
+    },
+    {
+      "epoch": 10.519480519480519,
+      "grad_norm": 148.8427734375,
+      "learning_rate": 3.831154705721542e-06,
+      "logits/chosen": -2.7960445880889893,
+      "logits/rejected": -2.791477680206299,
+      "logps/chosen": -9.04710578918457,
+      "logps/rejected": -22.726436614990234,
+      "loss": 0.8774,
+      "rewards/accuracies": 0.875,
+      "rewards/chosen": 0.03348678722977638,
+      "rewards/margins": 0.08231980353593826,
+      "rewards/rejected": -0.04883301258087158,
+      "step": 810
+    },
+    {
+      "epoch": 10.64935064935065,
+      "grad_norm": 75.88983154296875,
+      "learning_rate": 3.800939621808419e-06,
+      "logits/chosen": -2.8040685653686523,
+      "logits/rejected": -2.7940433025360107,
+      "logps/chosen": -7.935153007507324,
+      "logps/rejected": -22.834848403930664,
+      "loss": 1.2887,
+      "rewards/accuracies": 0.875,
+      "rewards/chosen": 0.02770247682929039,
+      "rewards/margins": 0.09637979418039322,
+      "rewards/rejected": -0.06867731362581253,
+      "step": 820
+    },
+    {
+      "epoch": 10.779220779220779,
+      "grad_norm": 94.64012145996094,
+      "learning_rate": 3.770461653367934e-06,
+      "logits/chosen": -2.8252198696136475,
+      "logits/rejected": -2.8190674781799316,
+      "logps/chosen": -7.881258964538574,
+      "logps/rejected": -22.468507766723633,
+      "loss": 0.8938,
+      "rewards/accuracies": 0.8500000238418579,
+      "rewards/chosen": 0.03068472072482109,
+      "rewards/margins": 0.0954260528087616,
+      "rewards/rejected": -0.06474132835865021,
+      "step": 830
+    },
+    {
+      "epoch": 10.909090909090908,
+      "grad_norm": 126.55635070800781,
+      "learning_rate": 3.7397269591688666e-06,
+      "logits/chosen": -2.818159580230713,
+      "logits/rejected": -2.8095242977142334,
+      "logps/chosen": -7.300267219543457,
+      "logps/rejected": -21.571325302124023,
+      "loss": 1.1946,
+      "rewards/accuracies": 0.925000011920929,
+      "rewards/chosen": 0.04063314571976662,
+      "rewards/margins": 0.09662587940692902,
+      "rewards/rejected": -0.055992741137742996,
+      "step": 840
+    },
+    {
+      "epoch": 11.03896103896104,
+      "grad_norm": 89.22303771972656,
+      "learning_rate": 3.7087417498572946e-06,
+      "logits/chosen": -2.8335649967193604,
+      "logits/rejected": -2.8254220485687256,
+      "logps/chosen": -7.027431488037109,
+      "logps/rejected": -24.15895652770996,
+      "loss": 1.6269,
+      "rewards/accuracies": 0.949999988079071,
+      "rewards/chosen": 0.021446553990244865,
+      "rewards/margins": 0.0891275405883789,
+      "rewards/rejected": -0.06768098473548889,
+      "step": 850
+    },
+    {
+      "epoch": 11.168831168831169,
+      "grad_norm": 113.87052154541016,
+      "learning_rate": 3.677512286701587e-06,
+      "logits/chosen": -2.822108030319214,
+      "logits/rejected": -2.8177809715270996,
+      "logps/chosen": -7.26416540145874,
+      "logps/rejected": -22.180326461791992,
+      "loss": 0.9047,
+      "rewards/accuracies": 0.925000011920929,
+      "rewards/chosen": 0.03205590695142746,
+      "rewards/margins": 0.09417016804218292,
+      "rewards/rejected": -0.06211426109075546,
+      "step": 860
+    },
+    {
+      "epoch": 11.2987012987013,
+      "grad_norm": 1.4642364978790283,
+      "learning_rate": 3.646044880327176e-06,
+      "logits/chosen": -2.8277363777160645,
+      "logits/rejected": -2.8215818405151367,
+      "logps/chosen": -8.22972297668457,
+      "logps/rejected": -20.709514617919922,
+      "loss": 0.9662,
+      "rewards/accuracies": 0.7749999761581421,
+      "rewards/chosen": 0.027242153882980347,
+      "rewards/margins": 0.07722898572683334,
+      "rewards/rejected": -0.04998684674501419,
+      "step": 870
+    },
+    {
+      "epoch": 11.428571428571429,
+      "grad_norm": 227.3729248046875,
+      "learning_rate": 3.6143458894413463e-06,
+      "logits/chosen": -2.815782308578491,
+      "logits/rejected": -2.809896230697632,
+      "logps/chosen": -8.15450382232666,
+      "logps/rejected": -23.144954681396484,
+      "loss": 1.009,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/chosen": 0.04248720407485962,
+      "rewards/margins": 0.11930873245000839,
+      "rewards/rejected": -0.07682152092456818,
+      "step": 880
+    },
+    {
+      "epoch": 11.558441558441558,
+      "grad_norm": 321.1402893066406,
+      "learning_rate": 3.5824217195483178e-06,
+      "logits/chosen": -2.8049838542938232,
+      "logits/rejected": -2.7966904640197754,
+      "logps/chosen": -7.33872127532959,
+      "logps/rejected": -21.830829620361328,
+      "loss": 1.1877,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": 0.031283579766750336,
+      "rewards/margins": 0.09999363124370575,
+      "rewards/rejected": -0.06871005892753601,
+      "step": 890
+    },
+    {
+      "epoch": 11.688311688311689,
+      "grad_norm": 148.70529174804688,
+      "learning_rate": 3.550278821654866e-06,
+      "logits/chosen": -2.804623603820801,
+      "logits/rejected": -2.7956459522247314,
+      "logps/chosen": -7.542906761169434,
+      "logps/rejected": -21.376934051513672,
+      "loss": 0.899,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": 0.03491160273551941,
+      "rewards/margins": 0.10313661396503448,
+      "rewards/rejected": -0.06822501122951508,
+      "step": 900
+    },
+    {
+      "epoch": 11.818181818181818,
+      "grad_norm": 131.51914978027344,
+      "learning_rate": 3.517923690966747e-06,
+      "logits/chosen": -2.81646990776062,
+      "logits/rejected": -2.803173065185547,
+      "logps/chosen": -7.835005760192871,
+      "logps/rejected": -24.5490665435791,
+      "loss": 1.2002,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/chosen": 0.03581169620156288,
+      "rewards/margins": 0.12795209884643555,
+      "rewards/rejected": -0.09214041382074356,
+      "step": 910
+    },
+    {
+      "epoch": 11.948051948051948,
+      "grad_norm": 1.8541675806045532,
+      "learning_rate": 3.4853628655761946e-06,
+      "logits/chosen": -2.8321688175201416,
+      "logits/rejected": -2.8234333992004395,
+      "logps/chosen": -8.250754356384277,
+      "logps/rejected": -21.983013153076172,
+      "loss": 1.2209,
+      "rewards/accuracies": 0.8500000238418579,
+      "rewards/chosen": 0.02724345028400421,
+      "rewards/margins": 0.11046279966831207,
+      "rewards/rejected": -0.08321934938430786,
+      "step": 920
+    },
+    {
+      "epoch": 12.077922077922079,
+      "grad_norm": 1.8971673250198364,
+      "learning_rate": 3.452602925140751e-06,
+      "logits/chosen": -2.8104381561279297,
+      "logits/rejected": -2.805176258087158,
+      "logps/chosen": -10.041051864624023,
+      "logps/rejected": -22.36850929260254,
+      "loss": 0.8314,
+      "rewards/accuracies": 0.925000011920929,
+      "rewards/chosen": 0.04497869685292244,
+      "rewards/margins": 0.1275285929441452,
+      "rewards/rejected": -0.08254990726709366,
+      "step": 930
+    },
+    {
+      "epoch": 12.207792207792208,
+      "grad_norm": 78.92677307128906,
+      "learning_rate": 3.4196504895536948e-06,
+      "logits/chosen": -2.821592330932617,
+      "logits/rejected": -2.812694549560547,
+      "logps/chosen": -8.034502029418945,
+      "logps/rejected": -21.927505493164062,
+      "loss": 1.4864,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/chosen": 0.03274466469883919,
+      "rewards/margins": 0.14650548994541168,
+      "rewards/rejected": -0.11376082897186279,
+      "step": 940
+    },
+    {
+      "epoch": 12.337662337662337,
+      "grad_norm": 151.75515747070312,
+      "learning_rate": 3.386512217606339e-06,
+      "logits/chosen": -2.815185070037842,
+      "logits/rejected": -2.798635721206665,
+      "logps/chosen": -7.672248840332031,
+      "logps/rejected": -24.917593002319336,
+      "loss": 0.7744,
+      "rewards/accuracies": 0.925000011920929,
+      "rewards/chosen": 0.03489464521408081,
+      "rewards/margins": 0.16525618731975555,
+      "rewards/rejected": -0.13036152720451355,
+      "step": 950
+    },
+    {
+      "epoch": 12.467532467532468,
+      "grad_norm": 287.6182556152344,
+      "learning_rate": 3.3531948056424766e-06,
+      "logits/chosen": -2.8206896781921387,
+      "logits/rejected": -2.808096170425415,
+      "logps/chosen": -8.610795021057129,
+      "logps/rejected": -20.904489517211914,
+      "loss": 1.5907,
+      "rewards/accuracies": 0.875,
+      "rewards/chosen": 0.04022675007581711,
+      "rewards/margins": 0.12955889105796814,
+      "rewards/rejected": -0.08933213353157043,
+      "step": 960
+    },
+    {
+      "epoch": 12.597402597402597,
+      "grad_norm": 102.08235168457031,
+      "learning_rate": 3.319704986205223e-06,
+      "logits/chosen": -2.8291194438934326,
+      "logits/rejected": -2.821276903152466,
+      "logps/chosen": -7.160222053527832,
+      "logps/rejected": -22.413909912109375,
+      "loss": 2.002,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/chosen": 0.029990587383508682,
+      "rewards/margins": 0.1707821935415268,
+      "rewards/rejected": -0.1407916247844696,
+      "step": 970
+    },
+    {
+      "epoch": 12.727272727272727,
+      "grad_norm": 293.0462646484375,
+      "learning_rate": 3.28604952667656e-06,
+      "logits/chosen": -2.811882734298706,
+      "logits/rejected": -2.8050527572631836,
+      "logps/chosen": -8.143568992614746,
+      "logps/rejected": -20.997821807861328,
+      "loss": 1.7693,
+      "rewards/accuracies": 0.875,
+      "rewards/chosen": 0.03135104477405548,
+      "rewards/margins": 0.13260604441165924,
+      "rewards/rejected": -0.10125498473644257,
+      "step": 980
+    },
+    {
+      "epoch": 12.857142857142858,
+      "grad_norm": 270.28076171875,
+      "learning_rate": 3.2522352279098256e-06,
+      "logits/chosen": -2.8084704875946045,
+      "logits/rejected": -2.8023064136505127,
+      "logps/chosen": -7.580096244812012,
+      "logps/rejected": -23.871841430664062,
+      "loss": 1.0787,
+      "rewards/accuracies": 0.925000011920929,
+      "rewards/chosen": 0.03514163941144943,
+      "rewards/margins": 0.19183096289634705,
+      "rewards/rejected": -0.15668931603431702,
+      "step": 990
+    },
+    {
+      "epoch": 12.987012987012987,
+      "grad_norm": 147.3173065185547,
+      "learning_rate": 3.218268922855452e-06,
+      "logits/chosen": -2.8337278366088867,
+      "logits/rejected": -2.822435140609741,
+      "logps/chosen": -7.920820713043213,
+      "logps/rejected": -23.01383399963379,
+      "loss": 1.3496,
+      "rewards/accuracies": 0.949999988079071,
+      "rewards/chosen": 0.044772375375032425,
+      "rewards/margins": 0.16832272708415985,
+      "rewards/rejected": -0.12355033308267593,
+      "step": 1000
+    },
+    {
+      "epoch": 13.116883116883116,
+      "grad_norm": 1.7505887746810913,
+      "learning_rate": 3.184157475180208e-06,
+      "logits/chosen": -2.8243801593780518,
+      "logits/rejected": -2.8188118934631348,
+      "logps/chosen": -6.851785182952881,
+      "logps/rejected": -22.512981414794922,
+      "loss": 0.7081,
+      "rewards/accuracies": 0.875,
+      "rewards/chosen": 0.046257633715867996,
+      "rewards/margins": 0.17940297722816467,
+      "rewards/rejected": -0.13314534723758698,
+      "step": 1010
+    },
+    {
+      "epoch": 13.246753246753247,
+      "grad_norm": 226.2203826904297,
+      "learning_rate": 3.149907777880239e-06,
+      "logits/chosen": -2.8116049766540527,
+      "logits/rejected": -2.802031993865967,
+      "logps/chosen": -9.796255111694336,
+      "logps/rejected": -23.3902587890625,
+      "loss": 1.5793,
+      "rewards/accuracies": 0.949999988079071,
+      "rewards/chosen": 0.034330084919929504,
+      "rewards/margins": 0.17266739904880524,
+      "rewards/rejected": -0.13833732903003693,
+      "step": 1020
+    },
+    {
+      "epoch": 13.376623376623376,
+      "grad_norm": 103.82469940185547,
+      "learning_rate": 3.1155267518881816e-06,
+      "logits/chosen": -2.8263697624206543,
+      "logits/rejected": -2.8141000270843506,
+      "logps/chosen": -6.771908760070801,
+      "logps/rejected": -24.233366012573242,
+      "loss": 1.0598,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/chosen": 0.04042463377118111,
+      "rewards/margins": 0.23938068747520447,
+      "rewards/rejected": -0.19895607233047485,
+      "step": 1030
+    },
+    {
+      "epoch": 13.506493506493506,
+      "grad_norm": 283.45172119140625,
+      "learning_rate": 3.0810213446746323e-06,
+      "logits/chosen": -2.8335254192352295,
+      "logits/rejected": -2.8218533992767334,
+      "logps/chosen": -8.464487075805664,
+      "logps/rejected": -21.802452087402344,
+      "loss": 2.5313,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": 0.0437016561627388,
+      "rewards/margins": 0.1905747503042221,
+      "rewards/rejected": -0.1468731015920639,
+      "step": 1040
+    },
+    {
+      "epoch": 13.636363636363637,
+      "grad_norm": 309.3778076171875,
+      "learning_rate": 3.046398528844248e-06,
+      "logits/chosen": -2.8289742469787598,
+      "logits/rejected": -2.8251280784606934,
+      "logps/chosen": -7.5258283615112305,
+      "logps/rejected": -20.029939651489258,
+      "loss": 1.7409,
+      "rewards/accuracies": 0.925000011920929,
+      "rewards/chosen": 0.05584300309419632,
+      "rewards/margins": 0.15009096264839172,
+      "rewards/rejected": -0.0942479595541954,
+      "step": 1050
+    },
+    {
+      "epoch": 13.766233766233766,
+      "grad_norm": 229.9144744873047,
+      "learning_rate": 3.0116653007267753e-06,
+      "logits/chosen": -2.8172080516815186,
+      "logits/rejected": -2.810072183609009,
+      "logps/chosen": -7.817892551422119,
+      "logps/rejected": -21.542694091796875,
+      "loss": 1.3013,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/chosen": 0.05069408565759659,
+      "rewards/margins": 0.1945628523826599,
+      "rewards/rejected": -0.14386877417564392,
+      "step": 1060
+    },
+    {
+      "epoch": 13.896103896103895,
+      "grad_norm": 74.20500946044922,
+      "learning_rate": 2.9768286789632845e-06,
+      "logits/chosen": -2.8088202476501465,
+      "logits/rejected": -2.7986044883728027,
+      "logps/chosen": -9.273926734924316,
+      "logps/rejected": -22.66620445251465,
+      "loss": 0.8765,
+      "rewards/accuracies": 0.949999988079071,
+      "rewards/chosen": 0.058945827186107635,
+      "rewards/margins": 0.23349115252494812,
+      "rewards/rejected": -0.1745453178882599,
+      "step": 1070
+    },
+    {
+      "epoch": 14.025974025974026,
+      "grad_norm": 569.8877563476562,
+      "learning_rate": 2.9418957030878876e-06,
+      "logits/chosen": -2.825758457183838,
+      "logits/rejected": -2.819606304168701,
+      "logps/chosen": -7.458970546722412,
+      "logps/rejected": -22.59342384338379,
+      "loss": 1.9946,
+      "rewards/accuracies": 0.949999988079071,
+      "rewards/chosen": 0.04193786159157753,
+      "rewards/margins": 0.21687336266040802,
+      "rewards/rejected": -0.1749354898929596,
+      "step": 1080
+    },
+    {
+      "epoch": 14.155844155844155,
+      "grad_norm": 1.655894160270691,
+      "learning_rate": 2.9068734321052445e-06,
+      "logits/chosen": -2.829101800918579,
+      "logits/rejected": -2.821521282196045,
+      "logps/chosen": -8.147821426391602,
+      "logps/rejected": -21.687742233276367,
+      "loss": 1.6034,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": 0.058260608464479446,
+      "rewards/margins": 0.22970101237297058,
+      "rewards/rejected": -0.17144039273262024,
+      "step": 1090
+    },
+    {
+      "epoch": 14.285714285714286,
+      "grad_norm": 226.83702087402344,
+      "learning_rate": 2.871768943064129e-06,
+      "logits/chosen": -2.8222107887268066,
+      "logits/rejected": -2.81400990486145,
+      "logps/chosen": -6.330902576446533,
+      "logps/rejected": -25.222558975219727,
+      "loss": 0.9428,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/chosen": 0.043078385293483734,
+      "rewards/margins": 0.31059008836746216,
+      "rewards/rejected": -0.26751166582107544,
+      "step": 1100
+    },
+    {
+      "epoch": 14.415584415584416,
+      "grad_norm": 542.2830200195312,
+      "learning_rate": 2.836589329627349e-06,
+      "logits/chosen": -2.829596996307373,
+      "logits/rejected": -2.8249754905700684,
+      "logps/chosen": -7.770198822021484,
+      "logps/rejected": -23.904422760009766,
+      "loss": 1.6431,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/chosen": 0.05304824188351631,
+      "rewards/margins": 0.2782563269138336,
+      "rewards/rejected": -0.2252080738544464,
+      "step": 1110
+    },
+    {
+      "epoch": 14.545454545454545,
+      "grad_norm": 566.6854248046875,
+      "learning_rate": 2.8013417006383078e-06,
+      "logits/chosen": -2.814929962158203,
+      "logits/rejected": -2.807624340057373,
+      "logps/chosen": -8.236068725585938,
+      "logps/rejected": -23.396425247192383,
+      "loss": 1.9727,
+      "rewards/accuracies": 0.949999988079071,
+      "rewards/chosen": 0.04831843078136444,
+      "rewards/margins": 0.24669453501701355,
+      "rewards/rejected": -0.19837608933448792,
+      "step": 1120
+    },
+    {
+      "epoch": 14.675324675324676,
+      "grad_norm": 144.7385711669922,
+      "learning_rate": 2.766033178684506e-06,
+      "logits/chosen": -2.812544107437134,
+      "logits/rejected": -2.8035783767700195,
+      "logps/chosen": -7.520627498626709,
+      "logps/rejected": -23.25860023498535,
+      "loss": 1.074,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": 0.04993298649787903,
+      "rewards/margins": 0.31407758593559265,
+      "rewards/rejected": -0.2641445994377136,
+      "step": 1130
+    },
+    {
+      "epoch": 14.805194805194805,
+      "grad_norm": 1.8980822563171387,
+      "learning_rate": 2.730670898658255e-06,
+      "logits/chosen": -2.816342830657959,
+      "logits/rejected": -2.805983066558838,
+      "logps/chosen": -8.278943061828613,
+      "logps/rejected": -24.814987182617188,
+      "loss": 1.2085,
+      "rewards/accuracies": 0.925000011920929,
+      "rewards/chosen": 0.056690942496061325,
+      "rewards/margins": 0.3225446939468384,
+      "rewards/rejected": -0.26585373282432556,
+      "step": 1140
+    },
+    {
+      "epoch": 14.935064935064934,
+      "grad_norm": 323.37408447265625,
+      "learning_rate": 2.695262006314912e-06,
+      "logits/chosen": -2.8120298385620117,
+      "logits/rejected": -2.800677537918091,
+      "logps/chosen": -8.421112060546875,
+      "logps/rejected": -23.05984115600586,
+      "loss": 2.2793,
+      "rewards/accuracies": 0.875,
+      "rewards/chosen": 0.057875752449035645,
+      "rewards/margins": 0.2153436243534088,
+      "rewards/rejected": -0.15746784210205078,
+      "step": 1150
+    },
+    {
+      "epoch": 15.064935064935066,
+      "grad_norm": 192.21383666992188,
+      "learning_rate": 2.6598136568289144e-06,
+      "logits/chosen": -2.8322479724884033,
+      "logits/rejected": -2.827678918838501,
+      "logps/chosen": -6.597090244293213,
+      "logps/rejected": -23.71731948852539,
+      "loss": 1.8258,
+      "rewards/accuracies": 0.949999988079071,
+      "rewards/chosen": 0.050219256430864334,
+      "rewards/margins": 0.29276466369628906,
+      "rewards/rejected": -0.24254541099071503,
+      "step": 1160
+    },
+    {
+      "epoch": 15.194805194805195,
+      "grad_norm": 1.7898062467575073,
+      "learning_rate": 2.6243330133479173e-06,
+      "logits/chosen": -2.8162477016448975,
+      "logits/rejected": -2.810333251953125,
+      "logps/chosen": -7.49225378036499,
+      "logps/rejected": -24.20693588256836,
+      "loss": 2.364,
+      "rewards/accuracies": 0.949999988079071,
+      "rewards/chosen": 0.05513688176870346,
+      "rewards/margins": 0.2901547849178314,
+      "rewards/rejected": -0.23501792550086975,
+      "step": 1170
+    },
+    {
+      "epoch": 15.324675324675324,
+      "grad_norm": 1.774614691734314,
+      "learning_rate": 2.5888272455453136e-06,
+      "logits/chosen": -2.841599225997925,
+      "logits/rejected": -2.830899715423584,
+      "logps/chosen": -7.024038791656494,
+      "logps/rejected": -24.798709869384766,
+      "loss": 1.3787,
+      "rewards/accuracies": 0.925000011920929,
+      "rewards/chosen": 0.05138504505157471,
+      "rewards/margins": 0.30870115756988525,
+      "rewards/rejected": -0.25731611251831055,
+      "step": 1180
+    },
+    {
+      "epoch": 15.454545454545455,
+      "grad_norm": 168.11526489257812,
+      "learning_rate": 2.5533035281714368e-06,
+      "logits/chosen": -2.8249008655548096,
+      "logits/rejected": -2.819967031478882,
+      "logps/chosen": -7.072245121002197,
+      "logps/rejected": -24.478059768676758,
+      "loss": 1.5998,
+      "rewards/accuracies": 0.925000011920929,
+      "rewards/chosen": 0.05316594988107681,
+      "rewards/margins": 0.33188968896865845,
+      "rewards/rejected": -0.2787237763404846,
+      "step": 1190
+    },
+    {
+      "epoch": 15.584415584415584,
+      "grad_norm": 1.7886942625045776,
+      "learning_rate": 2.517769039603744e-06,
+      "logits/chosen": -2.824219226837158,
+      "logits/rejected": -2.8117549419403076,
+      "logps/chosen": -7.654397487640381,
+      "logps/rejected": -26.292510986328125,
+      "loss": 1.1387,
+      "rewards/accuracies": 0.949999988079071,
+      "rewards/chosen": 0.06867384910583496,
+      "rewards/margins": 0.3651869297027588,
+      "rewards/rejected": -0.29651308059692383,
+      "step": 1200
+    },
+    {
+      "epoch": 15.584415584415584,
+      "eval_logits/chosen": -2.8089284896850586,
+      "eval_logits/rejected": -2.8120827674865723,
+      "eval_logps/chosen": -13.108169555664062,
+      "eval_logps/rejected": -16.744564056396484,
+      "eval_loss": 0.6140245795249939,
+      "eval_rewards/accuracies": 0.75,
+      "eval_rewards/chosen": 0.10423628985881805,
+      "eval_rewards/margins": 0.15160974860191345,
+      "eval_rewards/rejected": -0.0473734587430954,
+      "eval_runtime": 1.1713,
+      "eval_samples_per_second": 11.952,
+      "eval_steps_per_second": 1.707,
+      "step": 1200
+    },
+    {
+      "epoch": 15.714285714285714,
+      "grad_norm": 127.52613830566406,
+      "learning_rate": 2.482230960396256e-06,
+      "logits/chosen": -2.817732334136963,
+      "logits/rejected": -2.805518865585327,
+      "logps/chosen": -7.076968193054199,
+      "logps/rejected": -23.43122673034668,
+      "loss": 1.1089,
+      "rewards/accuracies": 0.8999999761581421,
+      "rewards/chosen": 0.057178787887096405,
+      "rewards/margins": 0.35119739174842834,
+      "rewards/rejected": -0.29401862621307373,
+      "step": 1210
+    },
+    {
+      "epoch": 15.844155844155845,
+      "grad_norm": 418.4096984863281,
+      "learning_rate": 2.4466964718285636e-06,
+      "logits/chosen": -2.824800729751587,
+      "logits/rejected": -2.817626714706421,
+      "logps/chosen": -8.242257118225098,
+      "logps/rejected": -24.571439743041992,
+      "loss": 1.7635,
+      "rewards/accuracies": 0.925000011920929,
+      "rewards/chosen": 0.0650298148393631,
+      "rewards/margins": 0.3450847566127777,
+      "rewards/rejected": -0.2800549864768982,
+      "step": 1220
+    },
+    {
+      "epoch": 15.974025974025974,
+      "grad_norm": 1.7215478420257568,
+      "learning_rate": 2.411172754454688e-06,
+      "logits/chosen": -2.83309006690979,
+      "logits/rejected": -2.8249144554138184,
+      "logps/chosen": -9.375545501708984,
+      "logps/rejected": -23.499141693115234,
+      "loss": 1.3375,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/chosen": 0.06410713493824005,
+      "rewards/margins": 0.33921390771865845,
+      "rewards/rejected": -0.275106817483902,
+      "step": 1230
+    },
+    {
+      "epoch": 16.103896103896105,
+      "grad_norm": 342.5730895996094,
+      "learning_rate": 2.375666986652083e-06,
+      "logits/chosen": -2.83019757270813,
+      "logits/rejected": -2.8198351860046387,
+      "logps/chosen": -9.095269203186035,
+      "logps/rejected": -22.90569305419922,
+      "loss": 1.4406,
+      "rewards/accuracies": 0.949999988079071,
+      "rewards/chosen": 0.04684508591890335,
+      "rewards/margins": 0.3012949824333191,
+      "rewards/rejected": -0.25444987416267395,
+      "step": 1240
+    },
+    {
+      "epoch": 16.233766233766232,
+      "grad_norm": 1.895964503288269,
+      "learning_rate": 2.3401863431710864e-06,
+      "logits/chosen": -2.8324759006500244,
+      "logits/rejected": -2.823629856109619,
+      "logps/chosen": -6.349139213562012,
+      "logps/rejected": -22.462413787841797,
+      "loss": 0.7482,
+      "rewards/accuracies": 0.9750000238418579,
+      "rewards/chosen": 0.06610914319753647,
+      "rewards/margins": 0.3332839906215668,
+      "rewards/rejected": -0.2671748101711273,
+      "step": 1250
+    },
+    {
+      "epoch": 16.363636363636363,
+      "grad_norm": 246.20245361328125,
+      "learning_rate": 2.3047379936850885e-06,
+      "logits/chosen": -2.8041343688964844,
+      "logits/rejected": -2.794372320175171,
+      "logps/chosen": -8.607588768005371,
+      "logps/rejected": -21.551204681396484,
+      "loss": 2.8934,
+      "rewards/accuracies": 0.949999988079071,
+      "rewards/chosen": 0.05655663087964058,
+      "rewards/margins": 0.2559330463409424,
+      "rewards/rejected": -0.1993764191865921,
+      "step": 1260
+    },
+    {
+      "epoch": 16.493506493506494,
+      "grad_norm": 180.12371826171875,
+      "learning_rate": 2.269329101341745e-06,
+      "logits/chosen": -2.825704336166382,
+      "logits/rejected": -2.813232898712158,
+      "logps/chosen": -8.257078170776367,
+      "logps/rejected": -24.829893112182617,
+      "loss": 1.095,
+      "rewards/accuracies": 0.925000011920929,
+      "rewards/chosen": 0.056694962084293365,
+      "rewards/margins": 0.37473762035369873,
+      "rewards/rejected": -0.3180426359176636,
+      "step": 1270
+    },
+    {
+      "epoch": 16.623376623376622,
+      "grad_norm": 266.23565673828125,
+      "learning_rate": 2.2339668213154943e-06,
+      "logits/chosen": -2.8129374980926514,
+      "logits/rejected": -2.806753158569336,
+      "logps/chosen": -9.956947326660156,
+      "logps/rejected": -27.90561294555664,
+      "loss": 1.3426,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": 0.06363432109355927,
+      "rewards/margins": 0.42332082986831665,
+      "rewards/rejected": -0.3596864640712738,
+      "step": 1280
+    },
+    {
+      "epoch": 16.753246753246753,
+      "grad_norm": 268.4725036621094,
+      "learning_rate": 2.1986582993616926e-06,
+      "logits/chosen": -2.8167190551757812,
+      "logits/rejected": -2.8063576221466064,
+      "logps/chosen": -7.815527439117432,
+      "logps/rejected": -25.4665584564209,
+      "loss": 2.4032,
+      "rewards/accuracies": 1.0,
+      "rewards/chosen": 0.0594390332698822,
+      "rewards/margins": 0.34777015447616577,
+      "rewards/rejected": -0.28833115100860596,
+      "step": 1290
+    },
+    {
+      "epoch": 16.883116883116884,
+      "grad_norm": 2.1606481075286865,
+      "learning_rate": 2.163410670372652e-06,
+      "logits/chosen": -2.8243236541748047,
+      "logits/rejected": -2.814610719680786,
+      "logps/chosen": -7.246170997619629,
+      "logps/rejected": -25.429073333740234,
+      "loss": 1.2455,
+      "rewards/accuracies": 0.875,
+      "rewards/chosen": 0.07013025879859924,
+      "rewards/margins": 0.4054412841796875,
+      "rewards/rejected": -0.33531102538108826,
+      "step": 1300
+    }
+  ],
+  "logging_steps": 10,
+  "max_steps": 2310,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 30,
+  "save_steps": 50,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 0.0,
+  "train_batch_size": 2,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/weight_dir/DpoWeight/DPOP_Fix_ND3V1/checkpoint-1300/training_args.bin b/weight_dir/DpoWeight/DPOP_Fix_ND3V1/checkpoint-1300/training_args.bin
new file mode 100644
index 0000000000000000000000000000000000000000..35883251f645e16fe9f020fb00125462447a8620
--- /dev/null
+++ b/weight_dir/DpoWeight/DPOP_Fix_ND3V1/checkpoint-1300/training_args.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:dcfbbe20fe71ddf66bf6192d5e1ac58297b542012aef9e192577497ecf0bcf70
+size 5944
diff --git a/weight_dir/loraWeight/fixbycrflp/checkpoint-8000/README.md b/weight_dir/loraWeight/fixbycrflp/checkpoint-8000/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..105007ea041cb936ba3417e0221e47fec447213d
--- /dev/null
+++ b/weight_dir/loraWeight/fixbycrflp/checkpoint-8000/README.md
@@ -0,0 +1,202 @@
+---
+library_name: peft
+base_model: ./CodeLlama-7b-Instruct-hf
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.11.1
\ No newline at end of file
diff --git a/weight_dir/loraWeight/fixbycrflp/checkpoint-8000/adapter_config.json b/weight_dir/loraWeight/fixbycrflp/checkpoint-8000/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..8faaa17fff7e4931b18e07a687c5b35532df41ba
--- /dev/null
+++ b/weight_dir/loraWeight/fixbycrflp/checkpoint-8000/adapter_config.json
@@ -0,0 +1,29 @@
+{
+  "alpha_pattern": {},
+  "auto_mapping": null,
+  "base_model_name_or_path": "./CodeLlama-7b-Instruct-hf",
+  "bias": "none",
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 32,
+  "lora_dropout": 0.05,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "r": 8,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "q_proj",
+    "v_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "use_dora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/weight_dir/loraWeight/fixbycrflp/checkpoint-8000/adapter_model.bin b/weight_dir/loraWeight/fixbycrflp/checkpoint-8000/adapter_model.bin
new file mode 100644
index 0000000000000000000000000000000000000000..4c599266290a655636b42f9e42939f3d0c73c9b7
--- /dev/null
+++ b/weight_dir/loraWeight/fixbycrflp/checkpoint-8000/adapter_model.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9b282a5cf93953598ceac2f52df00bf903105e30948b8eed7d963192ddaa2f1d
+size 8433034
diff --git a/weight_dir/loraWeight/fixbycrflp/checkpoint-8000/added_tokens.json b/weight_dir/loraWeight/fixbycrflp/checkpoint-8000/added_tokens.json
new file mode 100644
index 0000000000000000000000000000000000000000..c09203d8a52151e0d3b5f3c8e6daedc5b60832b5
--- /dev/null
+++ b/weight_dir/loraWeight/fixbycrflp/checkpoint-8000/added_tokens.json
@@ -0,0 +1,3 @@
+{
+  "<PAD>": 32016
+}
diff --git a/weight_dir/loraWeight/fixbycrflp/checkpoint-8000/global_step8000/bf16_zero_pp_rank_0_mp_rank_00_optim_states.pt b/weight_dir/loraWeight/fixbycrflp/checkpoint-8000/global_step8000/bf16_zero_pp_rank_0_mp_rank_00_optim_states.pt
new file mode 100644
index 0000000000000000000000000000000000000000..4fadfb38b9e43ce2369b534a65a7ddcd145d48de
--- /dev/null
+++ b/weight_dir/loraWeight/fixbycrflp/checkpoint-8000/global_step8000/bf16_zero_pp_rank_0_mp_rank_00_optim_states.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c549fd6e8d5b46fcb98fa5ffd3b8d4f98982318b8abbf744a59a5f6ec07540b6
+size 12589776
diff --git a/weight_dir/loraWeight/fixbycrflp/checkpoint-8000/global_step8000/bf16_zero_pp_rank_1_mp_rank_00_optim_states.pt b/weight_dir/loraWeight/fixbycrflp/checkpoint-8000/global_step8000/bf16_zero_pp_rank_1_mp_rank_00_optim_states.pt
new file mode 100644
index 0000000000000000000000000000000000000000..e9ee00d74029ec1ce61999f94813a3c989a38545
--- /dev/null
+++ b/weight_dir/loraWeight/fixbycrflp/checkpoint-8000/global_step8000/bf16_zero_pp_rank_1_mp_rank_00_optim_states.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:155a0ff6ca2c8b215e2de05a2046e4af0f44dfbf36ddc1e5540840f63b6566c1
+size 12589840
diff --git a/weight_dir/loraWeight/fixbycrflp/checkpoint-8000/global_step8000/bf16_zero_pp_rank_2_mp_rank_00_optim_states.pt b/weight_dir/loraWeight/fixbycrflp/checkpoint-8000/global_step8000/bf16_zero_pp_rank_2_mp_rank_00_optim_states.pt
new file mode 100644
index 0000000000000000000000000000000000000000..8e469cb43c418d88f5fde9579800b5cf5c14c6a6
--- /dev/null
+++ b/weight_dir/loraWeight/fixbycrflp/checkpoint-8000/global_step8000/bf16_zero_pp_rank_2_mp_rank_00_optim_states.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ef4e8acc41a38ca5bae3ca53f95ce12cfd17f6a3c6f3d1314c44ab63d8eff2fb
+size 12589840
diff --git a/weight_dir/loraWeight/fixbycrflp/checkpoint-8000/global_step8000/bf16_zero_pp_rank_3_mp_rank_00_optim_states.pt b/weight_dir/loraWeight/fixbycrflp/checkpoint-8000/global_step8000/bf16_zero_pp_rank_3_mp_rank_00_optim_states.pt
new file mode 100644
index 0000000000000000000000000000000000000000..e42934764a8919f59284ca1b5c98bf8f755b1e43
--- /dev/null
+++ b/weight_dir/loraWeight/fixbycrflp/checkpoint-8000/global_step8000/bf16_zero_pp_rank_3_mp_rank_00_optim_states.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3eb2ca7d094608660a1769f9e8795610fdcd849fd4767a35767f3bd988030590
+size 12589840
diff --git a/weight_dir/loraWeight/fixbycrflp/checkpoint-8000/global_step8000/mp_rank_00_model_states.pt b/weight_dir/loraWeight/fixbycrflp/checkpoint-8000/global_step8000/mp_rank_00_model_states.pt
new file mode 100644
index 0000000000000000000000000000000000000000..a482a79edb954e465c4b52f27016b16a5a2933cc
--- /dev/null
+++ b/weight_dir/loraWeight/fixbycrflp/checkpoint-8000/global_step8000/mp_rank_00_model_states.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0b0370a3b4b12f58d3ede5c81eb8710c88abb7eb5d7c909f3c900de0c20606fa
+size 8507372
diff --git a/weight_dir/loraWeight/fixbycrflp/checkpoint-8000/latest b/weight_dir/loraWeight/fixbycrflp/checkpoint-8000/latest
new file mode 100644
index 0000000000000000000000000000000000000000..6c558c76ac9e7515267f79d24d9fde4a7f8688f1
--- /dev/null
+++ b/weight_dir/loraWeight/fixbycrflp/checkpoint-8000/latest
@@ -0,0 +1 @@
+global_step8000
\ No newline at end of file
diff --git a/weight_dir/loraWeight/fixbycrflp/checkpoint-8000/rng_state_0.pth b/weight_dir/loraWeight/fixbycrflp/checkpoint-8000/rng_state_0.pth
new file mode 100644
index 0000000000000000000000000000000000000000..8330735db6838f0e2e9dbc45bed2e58e8bb363a4
--- /dev/null
+++ b/weight_dir/loraWeight/fixbycrflp/checkpoint-8000/rng_state_0.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:db5c221a5a18ee088159763149e88b1f6677df21c4df4a48421f3a71d3c952fb
+size 15024
diff --git a/weight_dir/loraWeight/fixbycrflp/checkpoint-8000/rng_state_1.pth b/weight_dir/loraWeight/fixbycrflp/checkpoint-8000/rng_state_1.pth
new file mode 100644
index 0000000000000000000000000000000000000000..e0d65b2a26a21db37de3fb368ebd0be093d6828a
--- /dev/null
+++ b/weight_dir/loraWeight/fixbycrflp/checkpoint-8000/rng_state_1.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:da57800cf6a9e3b0749c2b2264d6781ea412ede7c7a3eff1a5481b49fb4a948f
+size 15024
diff --git a/weight_dir/loraWeight/fixbycrflp/checkpoint-8000/rng_state_2.pth b/weight_dir/loraWeight/fixbycrflp/checkpoint-8000/rng_state_2.pth
new file mode 100644
index 0000000000000000000000000000000000000000..eec3c430e30e16aaf8764e3a8fb1b12e52e7172d
--- /dev/null
+++ b/weight_dir/loraWeight/fixbycrflp/checkpoint-8000/rng_state_2.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3a38c281b64adeb9cd597bdf895b41a68c82730b2db11abc5bf968012fbc4842
+size 15024
diff --git a/weight_dir/loraWeight/fixbycrflp/checkpoint-8000/rng_state_3.pth b/weight_dir/loraWeight/fixbycrflp/checkpoint-8000/rng_state_3.pth
new file mode 100644
index 0000000000000000000000000000000000000000..3f8eb7c65e8f2257f6cf925b85932df861a436d2
--- /dev/null
+++ b/weight_dir/loraWeight/fixbycrflp/checkpoint-8000/rng_state_3.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d72aeb50b87ec962a05c4fd8f40284d218377094be833b3547db1bc4e15c91b9
+size 15024
diff --git a/weight_dir/loraWeight/fixbycrflp/checkpoint-8000/special_tokens_map.json b/weight_dir/loraWeight/fixbycrflp/checkpoint-8000/special_tokens_map.json
new file mode 100644
index 0000000000000000000000000000000000000000..2ece23e993f7e1c7063cb51148b6fa5c6c224775
--- /dev/null
+++ b/weight_dir/loraWeight/fixbycrflp/checkpoint-8000/special_tokens_map.json
@@ -0,0 +1,30 @@
+{
+  "bos_token": {
+    "content": "<s>",
+    "lstrip": false,
+    "normalized": true,
+    "rstrip": false,
+    "single_word": false
+  },
+  "eos_token": {
+    "content": "</s>",
+    "lstrip": false,
+    "normalized": true,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": {
+    "content": "<PAD>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "unk_token": {
+    "content": "<unk>",
+    "lstrip": false,
+    "normalized": true,
+    "rstrip": false,
+    "single_word": false
+  }
+}
diff --git a/weight_dir/loraWeight/fixbycrflp/checkpoint-8000/tokenizer.model b/weight_dir/loraWeight/fixbycrflp/checkpoint-8000/tokenizer.model
new file mode 100644
index 0000000000000000000000000000000000000000..f6722e8b170230ebdd4c0f5f2ce03f219be536d4
--- /dev/null
+++ b/weight_dir/loraWeight/fixbycrflp/checkpoint-8000/tokenizer.model
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:45ccb9c8b6b561889acea59191d66986d314e7cbd6a78abc6e49b139ca91c1e6
+size 500058
diff --git a/weight_dir/loraWeight/fixbycrflp/checkpoint-8000/tokenizer_config.json b/weight_dir/loraWeight/fixbycrflp/checkpoint-8000/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..dc4d9a55d56b17600905bfc69a0e45e2fae7a1c4
--- /dev/null
+++ b/weight_dir/loraWeight/fixbycrflp/checkpoint-8000/tokenizer_config.json
@@ -0,0 +1,50 @@
+{
+  "add_bos_token": true,
+  "add_eos_token": false,
+  "add_prefix_space": true,
+  "added_tokens_decoder": {
+    "0": {
+      "content": "<unk>",
+      "lstrip": false,
+      "normalized": true,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "1": {
+      "content": "<s>",
+      "lstrip": false,
+      "normalized": true,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "2": {
+      "content": "</s>",
+      "lstrip": false,
+      "normalized": true,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "32016": {
+      "content": "<PAD>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    }
+  },
+  "bos_token": "<s>",
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "</s>",
+  "legacy": true,
+  "model_max_length": 1000000000000000019884624838656,
+  "pad_token": "<PAD>",
+  "sp_model_kwargs": {},
+  "spaces_between_special_tokens": false,
+  "tokenizer_class": "LlamaTokenizer",
+  "unk_token": "<unk>",
+  "use_default_system_prompt": false
+}
diff --git a/weight_dir/loraWeight/fixbycrflp/checkpoint-8000/trainer_state.json b/weight_dir/loraWeight/fixbycrflp/checkpoint-8000/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..7e5d3e0f5203c78d7f8372ec2d0dd92b05048f64
--- /dev/null
+++ b/weight_dir/loraWeight/fixbycrflp/checkpoint-8000/trainer_state.json
@@ -0,0 +1,1153 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 2.548582351067219,
+  "eval_steps": 10000,
+  "global_step": 8000,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.01592863969417012,
+      "grad_norm": 0.2921621799468994,
+      "learning_rate": 4.992513062316809e-05,
+      "loss": 4.2405,
+      "step": 50
+    },
+    {
+      "epoch": 0.03185727938834024,
+      "grad_norm": 0.10916037112474442,
+      "learning_rate": 4.9845482349942656e-05,
+      "loss": 0.0569,
+      "step": 100
+    },
+    {
+      "epoch": 0.047785919082510356,
+      "grad_norm": 0.11249323934316635,
+      "learning_rate": 4.976583407671722e-05,
+      "loss": 0.0494,
+      "step": 150
+    },
+    {
+      "epoch": 0.06371455877668047,
+      "grad_norm": 0.21314933896064758,
+      "learning_rate": 4.968618580349178e-05,
+      "loss": 0.0506,
+      "step": 200
+    },
+    {
+      "epoch": 0.07964319847085059,
+      "grad_norm": 0.122999407351017,
+      "learning_rate": 4.960653753026634e-05,
+      "loss": 0.0465,
+      "step": 250
+    },
+    {
+      "epoch": 0.09557183816502071,
+      "grad_norm": 0.09448697417974472,
+      "learning_rate": 4.952688925704091e-05,
+      "loss": 0.0481,
+      "step": 300
+    },
+    {
+      "epoch": 0.11150047785919083,
+      "grad_norm": 0.09307976812124252,
+      "learning_rate": 4.944724098381547e-05,
+      "loss": 0.0472,
+      "step": 350
+    },
+    {
+      "epoch": 0.12742911755336095,
+      "grad_norm": 0.14354585111141205,
+      "learning_rate": 4.936759271059004e-05,
+      "loss": 0.043,
+      "step": 400
+    },
+    {
+      "epoch": 0.14335775724753105,
+      "grad_norm": 0.10726131498813629,
+      "learning_rate": 4.92879444373646e-05,
+      "loss": 0.0415,
+      "step": 450
+    },
+    {
+      "epoch": 0.15928639694170119,
+      "grad_norm": 0.12725014984607697,
+      "learning_rate": 4.920829616413916e-05,
+      "loss": 0.0468,
+      "step": 500
+    },
+    {
+      "epoch": 0.1752150366358713,
+      "grad_norm": 0.12311285734176636,
+      "learning_rate": 4.9128647890913724e-05,
+      "loss": 0.0443,
+      "step": 550
+    },
+    {
+      "epoch": 0.19114367633004142,
+      "grad_norm": 0.1121256873011589,
+      "learning_rate": 4.904899961768829e-05,
+      "loss": 0.0397,
+      "step": 600
+    },
+    {
+      "epoch": 0.20707231602421153,
+      "grad_norm": 0.08886804431676865,
+      "learning_rate": 4.896935134446285e-05,
+      "loss": 0.0407,
+      "step": 650
+    },
+    {
+      "epoch": 0.22300095571838166,
+      "grad_norm": 0.09782330691814423,
+      "learning_rate": 4.888970307123742e-05,
+      "loss": 0.0445,
+      "step": 700
+    },
+    {
+      "epoch": 0.23892959541255177,
+      "grad_norm": 0.08494489639997482,
+      "learning_rate": 4.881005479801198e-05,
+      "loss": 0.0414,
+      "step": 750
+    },
+    {
+      "epoch": 0.2548582351067219,
+      "grad_norm": 0.0991586446762085,
+      "learning_rate": 4.873040652478654e-05,
+      "loss": 0.0432,
+      "step": 800
+    },
+    {
+      "epoch": 0.27078687480089203,
+      "grad_norm": 0.10855372995138168,
+      "learning_rate": 4.8650758251561105e-05,
+      "loss": 0.0394,
+      "step": 850
+    },
+    {
+      "epoch": 0.2867155144950621,
+      "grad_norm": 0.08806545287370682,
+      "learning_rate": 4.857110997833567e-05,
+      "loss": 0.0402,
+      "step": 900
+    },
+    {
+      "epoch": 0.30264415418923224,
+      "grad_norm": 0.13318926095962524,
+      "learning_rate": 4.8491461705110234e-05,
+      "loss": 0.0374,
+      "step": 950
+    },
+    {
+      "epoch": 0.31857279388340237,
+      "grad_norm": 0.1285628378391266,
+      "learning_rate": 4.84118134318848e-05,
+      "loss": 0.0426,
+      "step": 1000
+    },
+    {
+      "epoch": 0.33450143357757245,
+      "grad_norm": 0.07570258527994156,
+      "learning_rate": 4.833216515865936e-05,
+      "loss": 0.0406,
+      "step": 1050
+    },
+    {
+      "epoch": 0.3504300732717426,
+      "grad_norm": 0.14918510615825653,
+      "learning_rate": 4.825251688543392e-05,
+      "loss": 0.041,
+      "step": 1100
+    },
+    {
+      "epoch": 0.3663587129659127,
+      "grad_norm": 0.14587858319282532,
+      "learning_rate": 4.8172868612208485e-05,
+      "loss": 0.0387,
+      "step": 1150
+    },
+    {
+      "epoch": 0.38228735266008285,
+      "grad_norm": 0.15989379584789276,
+      "learning_rate": 4.809322033898305e-05,
+      "loss": 0.0429,
+      "step": 1200
+    },
+    {
+      "epoch": 0.3982159923542529,
+      "grad_norm": 0.09545516967773438,
+      "learning_rate": 4.8013572065757615e-05,
+      "loss": 0.0369,
+      "step": 1250
+    },
+    {
+      "epoch": 0.41414463204842306,
+      "grad_norm": 0.10800140351057053,
+      "learning_rate": 4.793392379253218e-05,
+      "loss": 0.042,
+      "step": 1300
+    },
+    {
+      "epoch": 0.4300732717425932,
+      "grad_norm": 0.10968200862407684,
+      "learning_rate": 4.7854275519306744e-05,
+      "loss": 0.0368,
+      "step": 1350
+    },
+    {
+      "epoch": 0.4460019114367633,
+      "grad_norm": 0.1094348132610321,
+      "learning_rate": 4.777462724608131e-05,
+      "loss": 0.038,
+      "step": 1400
+    },
+    {
+      "epoch": 0.4619305511309334,
+      "grad_norm": 0.09409014135599136,
+      "learning_rate": 4.7694978972855866e-05,
+      "loss": 0.0399,
+      "step": 1450
+    },
+    {
+      "epoch": 0.47785919082510353,
+      "grad_norm": 0.09957227855920792,
+      "learning_rate": 4.761533069963043e-05,
+      "loss": 0.0431,
+      "step": 1500
+    },
+    {
+      "epoch": 0.49378783051927366,
+      "grad_norm": 0.10479158908128738,
+      "learning_rate": 4.7535682426404995e-05,
+      "loss": 0.0396,
+      "step": 1550
+    },
+    {
+      "epoch": 0.5097164702134438,
+      "grad_norm": 0.0960434228181839,
+      "learning_rate": 4.745603415317956e-05,
+      "loss": 0.0404,
+      "step": 1600
+    },
+    {
+      "epoch": 0.5256451099076139,
+      "grad_norm": 0.14295724034309387,
+      "learning_rate": 4.7376385879954124e-05,
+      "loss": 0.0401,
+      "step": 1650
+    },
+    {
+      "epoch": 0.5415737496017841,
+      "grad_norm": 0.0905664786696434,
+      "learning_rate": 4.729673760672869e-05,
+      "loss": 0.0378,
+      "step": 1700
+    },
+    {
+      "epoch": 0.5575023892959541,
+      "grad_norm": 0.13186714053153992,
+      "learning_rate": 4.7217089333503254e-05,
+      "loss": 0.0391,
+      "step": 1750
+    },
+    {
+      "epoch": 0.5734310289901242,
+      "grad_norm": 0.07855305820703506,
+      "learning_rate": 4.713744106027782e-05,
+      "loss": 0.0386,
+      "step": 1800
+    },
+    {
+      "epoch": 0.5893596686842943,
+      "grad_norm": 0.10626554489135742,
+      "learning_rate": 4.705779278705238e-05,
+      "loss": 0.038,
+      "step": 1850
+    },
+    {
+      "epoch": 0.6052883083784645,
+      "grad_norm": 0.09309270232915878,
+      "learning_rate": 4.697814451382694e-05,
+      "loss": 0.04,
+      "step": 1900
+    },
+    {
+      "epoch": 0.6212169480726346,
+      "grad_norm": 0.10274514555931091,
+      "learning_rate": 4.6898496240601505e-05,
+      "loss": 0.0399,
+      "step": 1950
+    },
+    {
+      "epoch": 0.6371455877668047,
+      "grad_norm": 0.08140850067138672,
+      "learning_rate": 4.681884796737607e-05,
+      "loss": 0.0395,
+      "step": 2000
+    },
+    {
+      "epoch": 0.6530742274609749,
+      "grad_norm": 0.09800681471824646,
+      "learning_rate": 4.6739199694150634e-05,
+      "loss": 0.0396,
+      "step": 2050
+    },
+    {
+      "epoch": 0.6690028671551449,
+      "grad_norm": 0.12131080776453018,
+      "learning_rate": 4.66595514209252e-05,
+      "loss": 0.0404,
+      "step": 2100
+    },
+    {
+      "epoch": 0.684931506849315,
+      "grad_norm": 0.1389102041721344,
+      "learning_rate": 4.6579903147699763e-05,
+      "loss": 0.0378,
+      "step": 2150
+    },
+    {
+      "epoch": 0.7008601465434852,
+      "grad_norm": 0.12080667912960052,
+      "learning_rate": 4.650025487447433e-05,
+      "loss": 0.0366,
+      "step": 2200
+    },
+    {
+      "epoch": 0.7167887862376553,
+      "grad_norm": 0.09532010555267334,
+      "learning_rate": 4.642060660124889e-05,
+      "loss": 0.038,
+      "step": 2250
+    },
+    {
+      "epoch": 0.7327174259318254,
+      "grad_norm": 0.09826835989952087,
+      "learning_rate": 4.634095832802345e-05,
+      "loss": 0.0358,
+      "step": 2300
+    },
+    {
+      "epoch": 0.7486460656259956,
+      "grad_norm": 0.11227104812860489,
+      "learning_rate": 4.6261310054798015e-05,
+      "loss": 0.0362,
+      "step": 2350
+    },
+    {
+      "epoch": 0.7645747053201657,
+      "grad_norm": 0.10029356181621552,
+      "learning_rate": 4.618166178157258e-05,
+      "loss": 0.0396,
+      "step": 2400
+    },
+    {
+      "epoch": 0.7805033450143358,
+      "grad_norm": 0.10683028399944305,
+      "learning_rate": 4.6102013508347144e-05,
+      "loss": 0.0389,
+      "step": 2450
+    },
+    {
+      "epoch": 0.7964319847085058,
+      "grad_norm": 0.08704536408185959,
+      "learning_rate": 4.602236523512171e-05,
+      "loss": 0.0364,
+      "step": 2500
+    },
+    {
+      "epoch": 0.812360624402676,
+      "grad_norm": 0.13864809274673462,
+      "learning_rate": 4.594271696189627e-05,
+      "loss": 0.0401,
+      "step": 2550
+    },
+    {
+      "epoch": 0.8282892640968461,
+      "grad_norm": 0.08906254917383194,
+      "learning_rate": 4.586306868867083e-05,
+      "loss": 0.0362,
+      "step": 2600
+    },
+    {
+      "epoch": 0.8442179037910162,
+      "grad_norm": 0.11775562167167664,
+      "learning_rate": 4.5783420415445396e-05,
+      "loss": 0.0404,
+      "step": 2650
+    },
+    {
+      "epoch": 0.8601465434851864,
+      "grad_norm": 0.12016937136650085,
+      "learning_rate": 4.570377214221996e-05,
+      "loss": 0.0398,
+      "step": 2700
+    },
+    {
+      "epoch": 0.8760751831793565,
+      "grad_norm": 0.0807521790266037,
+      "learning_rate": 4.5624123868994525e-05,
+      "loss": 0.0443,
+      "step": 2750
+    },
+    {
+      "epoch": 0.8920038228735266,
+      "grad_norm": 0.16572950780391693,
+      "learning_rate": 4.554447559576909e-05,
+      "loss": 0.0397,
+      "step": 2800
+    },
+    {
+      "epoch": 0.9079324625676968,
+      "grad_norm": 0.16349689662456512,
+      "learning_rate": 4.5464827322543654e-05,
+      "loss": 0.0363,
+      "step": 2850
+    },
+    {
+      "epoch": 0.9238611022618668,
+      "grad_norm": 0.09836030751466751,
+      "learning_rate": 4.538517904931821e-05,
+      "loss": 0.0355,
+      "step": 2900
+    },
+    {
+      "epoch": 0.9397897419560369,
+      "grad_norm": 0.09340459853410721,
+      "learning_rate": 4.5305530776092776e-05,
+      "loss": 0.0372,
+      "step": 2950
+    },
+    {
+      "epoch": 0.9557183816502071,
+      "grad_norm": 0.11303897202014923,
+      "learning_rate": 4.522588250286734e-05,
+      "loss": 0.0359,
+      "step": 3000
+    },
+    {
+      "epoch": 0.9716470213443772,
+      "grad_norm": 0.134646475315094,
+      "learning_rate": 4.5146234229641906e-05,
+      "loss": 0.0354,
+      "step": 3050
+    },
+    {
+      "epoch": 0.9875756610385473,
+      "grad_norm": 0.13911692798137665,
+      "learning_rate": 4.506658595641647e-05,
+      "loss": 0.0364,
+      "step": 3100
+    },
+    {
+      "epoch": 1.0035043007327173,
+      "grad_norm": 0.09443005919456482,
+      "learning_rate": 4.4986937683191035e-05,
+      "loss": 0.0338,
+      "step": 3150
+    },
+    {
+      "epoch": 1.0194329404268876,
+      "grad_norm": 0.12161055952310562,
+      "learning_rate": 4.490728940996559e-05,
+      "loss": 0.0378,
+      "step": 3200
+    },
+    {
+      "epoch": 1.0353615801210576,
+      "grad_norm": 0.09010250866413116,
+      "learning_rate": 4.482764113674016e-05,
+      "loss": 0.0358,
+      "step": 3250
+    },
+    {
+      "epoch": 1.0512902198152279,
+      "grad_norm": 0.0787167027592659,
+      "learning_rate": 4.474799286351472e-05,
+      "loss": 0.0375,
+      "step": 3300
+    },
+    {
+      "epoch": 1.0672188595093979,
+      "grad_norm": 0.10341610759496689,
+      "learning_rate": 4.4668344590289286e-05,
+      "loss": 0.037,
+      "step": 3350
+    },
+    {
+      "epoch": 1.0831474992035681,
+      "grad_norm": 0.09436332434415817,
+      "learning_rate": 4.458869631706385e-05,
+      "loss": 0.035,
+      "step": 3400
+    },
+    {
+      "epoch": 1.0990761388977381,
+      "grad_norm": 0.11110551655292511,
+      "learning_rate": 4.4509048043838415e-05,
+      "loss": 0.0394,
+      "step": 3450
+    },
+    {
+      "epoch": 1.1150047785919082,
+      "grad_norm": 0.11926066130399704,
+      "learning_rate": 4.442939977061297e-05,
+      "loss": 0.0364,
+      "step": 3500
+    },
+    {
+      "epoch": 1.1309334182860784,
+      "grad_norm": 0.08972738683223724,
+      "learning_rate": 4.434975149738754e-05,
+      "loss": 0.0353,
+      "step": 3550
+    },
+    {
+      "epoch": 1.1468620579802484,
+      "grad_norm": 0.1294146478176117,
+      "learning_rate": 4.42701032241621e-05,
+      "loss": 0.0347,
+      "step": 3600
+    },
+    {
+      "epoch": 1.1627906976744187,
+      "grad_norm": 0.18671779334545135,
+      "learning_rate": 4.419045495093667e-05,
+      "loss": 0.0393,
+      "step": 3650
+    },
+    {
+      "epoch": 1.1787193373685887,
+      "grad_norm": 0.15697024762630463,
+      "learning_rate": 4.411080667771123e-05,
+      "loss": 0.0368,
+      "step": 3700
+    },
+    {
+      "epoch": 1.194647977062759,
+      "grad_norm": 0.17401783168315887,
+      "learning_rate": 4.4031158404485796e-05,
+      "loss": 0.0346,
+      "step": 3750
+    },
+    {
+      "epoch": 1.210576616756929,
+      "grad_norm": 0.13288350403308868,
+      "learning_rate": 4.3951510131260354e-05,
+      "loss": 0.0343,
+      "step": 3800
+    },
+    {
+      "epoch": 1.226505256451099,
+      "grad_norm": 0.13558614253997803,
+      "learning_rate": 4.387186185803492e-05,
+      "loss": 0.0385,
+      "step": 3850
+    },
+    {
+      "epoch": 1.2424338961452692,
+      "grad_norm": 0.08078035712242126,
+      "learning_rate": 4.379221358480948e-05,
+      "loss": 0.033,
+      "step": 3900
+    },
+    {
+      "epoch": 1.2583625358394392,
+      "grad_norm": 0.08991101384162903,
+      "learning_rate": 4.371256531158405e-05,
+      "loss": 0.0371,
+      "step": 3950
+    },
+    {
+      "epoch": 1.2742911755336095,
+      "grad_norm": 0.09473446011543274,
+      "learning_rate": 4.363291703835861e-05,
+      "loss": 0.0334,
+      "step": 4000
+    },
+    {
+      "epoch": 1.2902198152277795,
+      "grad_norm": 0.09711175411939621,
+      "learning_rate": 4.355326876513318e-05,
+      "loss": 0.0361,
+      "step": 4050
+    },
+    {
+      "epoch": 1.3061484549219498,
+      "grad_norm": 0.11206386983394623,
+      "learning_rate": 4.3473620491907735e-05,
+      "loss": 0.0332,
+      "step": 4100
+    },
+    {
+      "epoch": 1.3220770946161198,
+      "grad_norm": 0.10798367857933044,
+      "learning_rate": 4.33939722186823e-05,
+      "loss": 0.0367,
+      "step": 4150
+    },
+    {
+      "epoch": 1.3380057343102898,
+      "grad_norm": 0.13645893335342407,
+      "learning_rate": 4.3314323945456864e-05,
+      "loss": 0.0353,
+      "step": 4200
+    },
+    {
+      "epoch": 1.35393437400446,
+      "grad_norm": 0.10864555090665817,
+      "learning_rate": 4.323467567223143e-05,
+      "loss": 0.0361,
+      "step": 4250
+    },
+    {
+      "epoch": 1.36986301369863,
+      "grad_norm": 0.12079176306724548,
+      "learning_rate": 4.315502739900599e-05,
+      "loss": 0.0329,
+      "step": 4300
+    },
+    {
+      "epoch": 1.3857916533928003,
+      "grad_norm": 0.1080985963344574,
+      "learning_rate": 4.307537912578056e-05,
+      "loss": 0.0368,
+      "step": 4350
+    },
+    {
+      "epoch": 1.4017202930869703,
+      "grad_norm": 0.1214875727891922,
+      "learning_rate": 4.2995730852555115e-05,
+      "loss": 0.033,
+      "step": 4400
+    },
+    {
+      "epoch": 1.4176489327811406,
+      "grad_norm": 0.0906578078866005,
+      "learning_rate": 4.291608257932968e-05,
+      "loss": 0.0361,
+      "step": 4450
+    },
+    {
+      "epoch": 1.4335775724753106,
+      "grad_norm": 0.11563856154680252,
+      "learning_rate": 4.2836434306104244e-05,
+      "loss": 0.0335,
+      "step": 4500
+    },
+    {
+      "epoch": 1.4495062121694806,
+      "grad_norm": 0.1468946486711502,
+      "learning_rate": 4.275678603287881e-05,
+      "loss": 0.0335,
+      "step": 4550
+    },
+    {
+      "epoch": 1.4654348518636509,
+      "grad_norm": 0.16663908958435059,
+      "learning_rate": 4.2677137759653374e-05,
+      "loss": 0.0354,
+      "step": 4600
+    },
+    {
+      "epoch": 1.481363491557821,
+      "grad_norm": 0.18749141693115234,
+      "learning_rate": 4.259748948642794e-05,
+      "loss": 0.0377,
+      "step": 4650
+    },
+    {
+      "epoch": 1.4972921312519911,
+      "grad_norm": 0.12299200892448425,
+      "learning_rate": 4.2517841213202496e-05,
+      "loss": 0.0338,
+      "step": 4700
+    },
+    {
+      "epoch": 1.5132207709461611,
+      "grad_norm": 0.10765209048986435,
+      "learning_rate": 4.243819293997706e-05,
+      "loss": 0.0368,
+      "step": 4750
+    },
+    {
+      "epoch": 1.5291494106403314,
+      "grad_norm": 0.09512902796268463,
+      "learning_rate": 4.2358544666751625e-05,
+      "loss": 0.034,
+      "step": 4800
+    },
+    {
+      "epoch": 1.5450780503345014,
+      "grad_norm": 0.13502831757068634,
+      "learning_rate": 4.227889639352619e-05,
+      "loss": 0.0354,
+      "step": 4850
+    },
+    {
+      "epoch": 1.5610066900286714,
+      "grad_norm": 0.12296276539564133,
+      "learning_rate": 4.2199248120300754e-05,
+      "loss": 0.0344,
+      "step": 4900
+    },
+    {
+      "epoch": 1.5769353297228417,
+      "grad_norm": 0.11571130156517029,
+      "learning_rate": 4.211959984707532e-05,
+      "loss": 0.034,
+      "step": 4950
+    },
+    {
+      "epoch": 1.592863969417012,
+      "grad_norm": 0.08514443039894104,
+      "learning_rate": 4.203995157384988e-05,
+      "loss": 0.0339,
+      "step": 5000
+    },
+    {
+      "epoch": 1.608792609111182,
+      "grad_norm": 0.10442246496677399,
+      "learning_rate": 4.196030330062444e-05,
+      "loss": 0.0354,
+      "step": 5050
+    },
+    {
+      "epoch": 1.624721248805352,
+      "grad_norm": 0.0917409136891365,
+      "learning_rate": 4.1880655027399006e-05,
+      "loss": 0.035,
+      "step": 5100
+    },
+    {
+      "epoch": 1.6406498884995222,
+      "grad_norm": 0.1276286244392395,
+      "learning_rate": 4.180100675417357e-05,
+      "loss": 0.0356,
+      "step": 5150
+    },
+    {
+      "epoch": 1.6565785281936922,
+      "grad_norm": 0.12348821014165878,
+      "learning_rate": 4.1721358480948135e-05,
+      "loss": 0.0354,
+      "step": 5200
+    },
+    {
+      "epoch": 1.6725071678878622,
+      "grad_norm": 0.0923234224319458,
+      "learning_rate": 4.16417102077227e-05,
+      "loss": 0.0353,
+      "step": 5250
+    },
+    {
+      "epoch": 1.6884358075820325,
+      "grad_norm": 0.08973834663629532,
+      "learning_rate": 4.156206193449726e-05,
+      "loss": 0.0333,
+      "step": 5300
+    },
+    {
+      "epoch": 1.7043644472762027,
+      "grad_norm": 0.10418592393398285,
+      "learning_rate": 4.148241366127182e-05,
+      "loss": 0.0343,
+      "step": 5350
+    },
+    {
+      "epoch": 1.7202930869703728,
+      "grad_norm": 0.1491956114768982,
+      "learning_rate": 4.1402765388046387e-05,
+      "loss": 0.0347,
+      "step": 5400
+    },
+    {
+      "epoch": 1.7362217266645428,
+      "grad_norm": 0.1951647698879242,
+      "learning_rate": 4.132311711482095e-05,
+      "loss": 0.0294,
+      "step": 5450
+    },
+    {
+      "epoch": 1.752150366358713,
+      "grad_norm": 0.20774711668491364,
+      "learning_rate": 4.1243468841595516e-05,
+      "loss": 0.0336,
+      "step": 5500
+    },
+    {
+      "epoch": 1.768079006052883,
+      "grad_norm": 0.0813850536942482,
+      "learning_rate": 4.116382056837008e-05,
+      "loss": 0.0359,
+      "step": 5550
+    },
+    {
+      "epoch": 1.784007645747053,
+      "grad_norm": 0.15596628189086914,
+      "learning_rate": 4.108417229514464e-05,
+      "loss": 0.0332,
+      "step": 5600
+    },
+    {
+      "epoch": 1.7999362854412233,
+      "grad_norm": 0.14951321482658386,
+      "learning_rate": 4.10045240219192e-05,
+      "loss": 0.0337,
+      "step": 5650
+    },
+    {
+      "epoch": 1.8158649251353935,
+      "grad_norm": 0.13903406262397766,
+      "learning_rate": 4.092487574869377e-05,
+      "loss": 0.035,
+      "step": 5700
+    },
+    {
+      "epoch": 1.8317935648295636,
+      "grad_norm": 0.14913810789585114,
+      "learning_rate": 4.084522747546833e-05,
+      "loss": 0.0336,
+      "step": 5750
+    },
+    {
+      "epoch": 1.8477222045237336,
+      "grad_norm": 0.16890230774879456,
+      "learning_rate": 4.0765579202242896e-05,
+      "loss": 0.0372,
+      "step": 5800
+    },
+    {
+      "epoch": 1.8636508442179038,
+      "grad_norm": 0.12355700880289078,
+      "learning_rate": 4.068593092901746e-05,
+      "loss": 0.0347,
+      "step": 5850
+    },
+    {
+      "epoch": 1.8795794839120739,
+      "grad_norm": 0.12095997482538223,
+      "learning_rate": 4.060628265579202e-05,
+      "loss": 0.0391,
+      "step": 5900
+    },
+    {
+      "epoch": 1.8955081236062439,
+      "grad_norm": 0.12925802171230316,
+      "learning_rate": 4.052663438256658e-05,
+      "loss": 0.036,
+      "step": 5950
+    },
+    {
+      "epoch": 1.9114367633004141,
+      "grad_norm": 0.11391396820545197,
+      "learning_rate": 4.044698610934115e-05,
+      "loss": 0.0325,
+      "step": 6000
+    },
+    {
+      "epoch": 1.9273654029945844,
+      "grad_norm": 0.129618838429451,
+      "learning_rate": 4.036733783611571e-05,
+      "loss": 0.0327,
+      "step": 6050
+    },
+    {
+      "epoch": 1.9432940426887544,
+      "grad_norm": 0.13725541532039642,
+      "learning_rate": 4.028768956289028e-05,
+      "loss": 0.0336,
+      "step": 6100
+    },
+    {
+      "epoch": 1.9592226823829244,
+      "grad_norm": 0.14412935078144073,
+      "learning_rate": 4.020804128966484e-05,
+      "loss": 0.0325,
+      "step": 6150
+    },
+    {
+      "epoch": 1.9751513220770947,
+      "grad_norm": 0.10575806349515915,
+      "learning_rate": 4.01283930164394e-05,
+      "loss": 0.0315,
+      "step": 6200
+    },
+    {
+      "epoch": 1.991079961771265,
+      "grad_norm": 0.11367379128932953,
+      "learning_rate": 4.0048744743213964e-05,
+      "loss": 0.0324,
+      "step": 6250
+    },
+    {
+      "epoch": 2.0070086014654347,
+      "grad_norm": 0.12420395016670227,
+      "learning_rate": 3.996909646998853e-05,
+      "loss": 0.0305,
+      "step": 6300
+    },
+    {
+      "epoch": 2.022937241159605,
+      "grad_norm": 0.2189149558544159,
+      "learning_rate": 3.988944819676309e-05,
+      "loss": 0.0343,
+      "step": 6350
+    },
+    {
+      "epoch": 2.038865880853775,
+      "grad_norm": 0.08044280856847763,
+      "learning_rate": 3.980979992353766e-05,
+      "loss": 0.0314,
+      "step": 6400
+    },
+    {
+      "epoch": 2.0547945205479454,
+      "grad_norm": 0.09585762768983841,
+      "learning_rate": 3.973015165031222e-05,
+      "loss": 0.0343,
+      "step": 6450
+    },
+    {
+      "epoch": 2.0707231602421152,
+      "grad_norm": 0.1755801886320114,
+      "learning_rate": 3.965050337708679e-05,
+      "loss": 0.0333,
+      "step": 6500
+    },
+    {
+      "epoch": 2.0866517999362855,
+      "grad_norm": 0.14164239168167114,
+      "learning_rate": 3.957085510386135e-05,
+      "loss": 0.0331,
+      "step": 6550
+    },
+    {
+      "epoch": 2.1025804396304557,
+      "grad_norm": 0.12496601790189743,
+      "learning_rate": 3.9491206830635916e-05,
+      "loss": 0.035,
+      "step": 6600
+    },
+    {
+      "epoch": 2.1185090793246255,
+      "grad_norm": 0.11803654581308365,
+      "learning_rate": 3.941155855741048e-05,
+      "loss": 0.0312,
+      "step": 6650
+    },
+    {
+      "epoch": 2.1344377190187958,
+      "grad_norm": 0.09984956681728363,
+      "learning_rate": 3.933191028418504e-05,
+      "loss": 0.0322,
+      "step": 6700
+    },
+    {
+      "epoch": 2.150366358712966,
+      "grad_norm": 0.11662815511226654,
+      "learning_rate": 3.92522620109596e-05,
+      "loss": 0.0332,
+      "step": 6750
+    },
+    {
+      "epoch": 2.1662949984071362,
+      "grad_norm": 0.12902189791202545,
+      "learning_rate": 3.917261373773417e-05,
+      "loss": 0.0343,
+      "step": 6800
+    },
+    {
+      "epoch": 2.182223638101306,
+      "grad_norm": 0.1841822862625122,
+      "learning_rate": 3.909296546450873e-05,
+      "loss": 0.0339,
+      "step": 6850
+    },
+    {
+      "epoch": 2.1981522777954763,
+      "grad_norm": 0.09873718023300171,
+      "learning_rate": 3.90133171912833e-05,
+      "loss": 0.0303,
+      "step": 6900
+    },
+    {
+      "epoch": 2.2140809174896465,
+      "grad_norm": 0.1674479842185974,
+      "learning_rate": 3.893366891805786e-05,
+      "loss": 0.032,
+      "step": 6950
+    },
+    {
+      "epoch": 2.2300095571838163,
+      "grad_norm": 0.13210225105285645,
+      "learning_rate": 3.8854020644832426e-05,
+      "loss": 0.0352,
+      "step": 7000
+    },
+    {
+      "epoch": 2.2459381968779866,
+      "grad_norm": 0.20769694447517395,
+      "learning_rate": 3.877437237160699e-05,
+      "loss": 0.0294,
+      "step": 7050
+    },
+    {
+      "epoch": 2.261866836572157,
+      "grad_norm": 0.13857823610305786,
+      "learning_rate": 3.869472409838155e-05,
+      "loss": 0.0343,
+      "step": 7100
+    },
+    {
+      "epoch": 2.277795476266327,
+      "grad_norm": 0.14322370290756226,
+      "learning_rate": 3.861507582515611e-05,
+      "loss": 0.0296,
+      "step": 7150
+    },
+    {
+      "epoch": 2.293724115960497,
+      "grad_norm": 0.11245788633823395,
+      "learning_rate": 3.853542755193068e-05,
+      "loss": 0.0323,
+      "step": 7200
+    },
+    {
+      "epoch": 2.309652755654667,
+      "grad_norm": 0.12236214429140091,
+      "learning_rate": 3.845577927870524e-05,
+      "loss": 0.0311,
+      "step": 7250
+    },
+    {
+      "epoch": 2.3255813953488373,
+      "grad_norm": 0.08712544292211533,
+      "learning_rate": 3.837613100547981e-05,
+      "loss": 0.0341,
+      "step": 7300
+    },
+    {
+      "epoch": 2.341510035043007,
+      "grad_norm": 0.11802078783512115,
+      "learning_rate": 3.829648273225437e-05,
+      "loss": 0.0317,
+      "step": 7350
+    },
+    {
+      "epoch": 2.3574386747371774,
+      "grad_norm": 0.1141052171587944,
+      "learning_rate": 3.821683445902893e-05,
+      "loss": 0.031,
+      "step": 7400
+    },
+    {
+      "epoch": 2.3733673144313476,
+      "grad_norm": 0.15947668254375458,
+      "learning_rate": 3.8137186185803494e-05,
+      "loss": 0.0313,
+      "step": 7450
+    },
+    {
+      "epoch": 2.389295954125518,
+      "grad_norm": 0.11814655363559723,
+      "learning_rate": 3.805753791257806e-05,
+      "loss": 0.0334,
+      "step": 7500
+    },
+    {
+      "epoch": 2.4052245938196877,
+      "grad_norm": 0.18867388367652893,
+      "learning_rate": 3.797788963935262e-05,
+      "loss": 0.0316,
+      "step": 7550
+    },
+    {
+      "epoch": 2.421153233513858,
+      "grad_norm": 0.13499616086483002,
+      "learning_rate": 3.789824136612719e-05,
+      "loss": 0.0304,
+      "step": 7600
+    },
+    {
+      "epoch": 2.437081873208028,
+      "grad_norm": 0.15890513360500336,
+      "learning_rate": 3.781859309290175e-05,
+      "loss": 0.0309,
+      "step": 7650
+    },
+    {
+      "epoch": 2.453010512902198,
+      "grad_norm": 0.10094068199396133,
+      "learning_rate": 3.773894481967631e-05,
+      "loss": 0.0311,
+      "step": 7700
+    },
+    {
+      "epoch": 2.468939152596368,
+      "grad_norm": 0.19545088708400726,
+      "learning_rate": 3.7659296546450874e-05,
+      "loss": 0.033,
+      "step": 7750
+    },
+    {
+      "epoch": 2.4848677922905384,
+      "grad_norm": 0.12802977859973907,
+      "learning_rate": 3.757964827322544e-05,
+      "loss": 0.0332,
+      "step": 7800
+    },
+    {
+      "epoch": 2.5007964319847087,
+      "grad_norm": 0.08226735889911652,
+      "learning_rate": 3.7500000000000003e-05,
+      "loss": 0.0318,
+      "step": 7850
+    },
+    {
+      "epoch": 2.5167250716788785,
+      "grad_norm": 0.11623780429363251,
+      "learning_rate": 3.742035172677457e-05,
+      "loss": 0.0336,
+      "step": 7900
+    },
+    {
+      "epoch": 2.5326537113730487,
+      "grad_norm": 0.16703219711780548,
+      "learning_rate": 3.734070345354913e-05,
+      "loss": 0.0307,
+      "step": 7950
+    },
+    {
+      "epoch": 2.548582351067219,
+      "grad_norm": 0.10822132229804993,
+      "learning_rate": 3.72610551803237e-05,
+      "loss": 0.033,
+      "step": 8000
+    }
+  ],
+  "logging_steps": 50,
+  "max_steps": 31390,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 2000,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 4.922138329882821e+18,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/weight_dir/loraWeight/fixbycrflp/checkpoint-8000/training_args.bin b/weight_dir/loraWeight/fixbycrflp/checkpoint-8000/training_args.bin
new file mode 100644
index 0000000000000000000000000000000000000000..4b45b2e6f313435b1d70860929c99bab2fc5943f
--- /dev/null
+++ b/weight_dir/loraWeight/fixbycrflp/checkpoint-8000/training_args.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:adde3152d6d22eecd1475fa998c1854302f8f17da34801350df8aca4fd5ec59b
+size 7736
diff --git a/weight_dir/loraWeight/fixbycrflp/checkpoint-8000/zero_to_fp32.py b/weight_dir/loraWeight/fixbycrflp/checkpoint-8000/zero_to_fp32.py
new file mode 100644
index 0000000000000000000000000000000000000000..24cc342e78d1a006c782b3a4cd68d9ce786d8fd8
--- /dev/null
+++ b/weight_dir/loraWeight/fixbycrflp/checkpoint-8000/zero_to_fp32.py
@@ -0,0 +1,604 @@
+#!/usr/bin/env python
+
+# Copyright (c) Microsoft Corporation.
+# SPDX-License-Identifier: Apache-2.0
+
+# DeepSpeed Team
+
+# This script extracts fp32 consolidated weights from a zero 1, 2 and 3 DeepSpeed checkpoints. It gets
+# copied into the top level checkpoint dir, so the user can easily do the conversion at any point in
+# the future. Once extracted, the weights don't require DeepSpeed and can be used in any
+# application.
+#
+# example: python zero_to_fp32.py . pytorch_model.bin
+
+import argparse
+import torch
+import glob
+import math
+import os
+import re
+from collections import OrderedDict
+from dataclasses import dataclass
+
+# while this script doesn't use deepspeed to recover data, since the checkpoints are pickled with
+# DeepSpeed data structures it has to be available in the current python environment.
+from deepspeed.utils import logger
+from deepspeed.checkpoint.constants import (DS_VERSION, OPTIMIZER_STATE_DICT, SINGLE_PARTITION_OF_FP32_GROUPS,
+                                            FP32_FLAT_GROUPS, ZERO_STAGE, PARTITION_COUNT, PARAM_SHAPES, BUFFER_NAMES,
+                                            FROZEN_PARAM_SHAPES, FROZEN_PARAM_FRAGMENTS)
+
+
+@dataclass
+class zero_model_state:
+    buffers: dict()
+    param_shapes: dict()
+    shared_params: list
+    ds_version: int
+    frozen_param_shapes: dict()
+    frozen_param_fragments: dict()
+
+
+debug = 0
+
+# load to cpu
+device = torch.device('cpu')
+
+
+def atoi(text):
+    return int(text) if text.isdigit() else text
+
+
+def natural_keys(text):
+    '''
+    alist.sort(key=natural_keys) sorts in human order
+    http://nedbatchelder.com/blog/200712/human_sorting.html
+    (See Toothy's implementation in the comments)
+    '''
+    return [atoi(c) for c in re.split(r'(\d+)', text)]
+
+
+def get_model_state_file(checkpoint_dir, zero_stage):
+    if not os.path.isdir(checkpoint_dir):
+        raise FileNotFoundError(f"Directory '{checkpoint_dir}' doesn't exist")
+
+    # there should be only one file
+    if zero_stage <= 2:
+        file = os.path.join(checkpoint_dir, "mp_rank_00_model_states.pt")
+    elif zero_stage == 3:
+        file = os.path.join(checkpoint_dir, "zero_pp_rank_0_mp_rank_00_model_states.pt")
+
+    if not os.path.exists(file):
+        raise FileNotFoundError(f"can't find model states file at '{file}'")
+
+    return file
+
+
+def get_checkpoint_files(checkpoint_dir, glob_pattern):
+    # XXX: need to test that this simple glob rule works for multi-node setup too
+    ckpt_files = sorted(glob.glob(os.path.join(checkpoint_dir, glob_pattern)), key=natural_keys)
+
+    if len(ckpt_files) == 0:
+        raise FileNotFoundError(f"can't find {glob_pattern} files in directory '{checkpoint_dir}'")
+
+    return ckpt_files
+
+
+def get_optim_files(checkpoint_dir):
+    return get_checkpoint_files(checkpoint_dir, "*_optim_states.pt")
+
+
+def get_model_state_files(checkpoint_dir):
+    return get_checkpoint_files(checkpoint_dir, "*_model_states.pt")
+
+
+def parse_model_states(files):
+    zero_model_states = []
+    for file in files:
+        state_dict = torch.load(file, map_location=device)
+
+        if BUFFER_NAMES not in state_dict:
+            raise ValueError(f"{file} is not a model state checkpoint")
+        buffer_names = state_dict[BUFFER_NAMES]
+        if debug:
+            print("Found buffers:", buffer_names)
+
+        # recover just the buffers while restoring them to fp32 if they were saved in fp16
+        buffers = {k: v.float() for k, v in state_dict["module"].items() if k in buffer_names}
+        param_shapes = state_dict[PARAM_SHAPES]
+
+        # collect parameters that are included in param_shapes
+        param_names = []
+        for s in param_shapes:
+            for name in s.keys():
+                param_names.append(name)
+
+        # update with frozen parameters
+        frozen_param_shapes = state_dict.get(FROZEN_PARAM_SHAPES, None)
+        if frozen_param_shapes is not None:
+            if debug:
+                print(f"Found frozen_param_shapes: {frozen_param_shapes}")
+            param_names += list(frozen_param_shapes.keys())
+
+        # handle shared params
+        shared_params = [[k, v] for k, v in state_dict["shared_params"].items()]
+
+        ds_version = state_dict.get(DS_VERSION, None)
+
+        frozen_param_fragments = state_dict.get(FROZEN_PARAM_FRAGMENTS, None)
+
+        z_model_state = zero_model_state(buffers=buffers,
+                                         param_shapes=param_shapes,
+                                         shared_params=shared_params,
+                                         ds_version=ds_version,
+                                         frozen_param_shapes=frozen_param_shapes,
+                                         frozen_param_fragments=frozen_param_fragments)
+        zero_model_states.append(z_model_state)
+
+    return zero_model_states
+
+
+def parse_optim_states(files, ds_checkpoint_dir):
+
+    total_files = len(files)
+    state_dicts = []
+    for f in files:
+        state_dict = torch.load(f, map_location=device)
+        # immediately discard the potentially huge 2 optimizer states as we only care for fp32 master weights
+        # and also handle the case where it was already removed by another helper script
+        state_dict["optimizer_state_dict"].pop("optimizer_state_dict", None)
+        state_dicts.append(state_dict)
+
+    if not ZERO_STAGE in state_dicts[0][OPTIMIZER_STATE_DICT]:
+        raise ValueError(f"{files[0]} is not a zero checkpoint")
+    zero_stage = state_dicts[0][OPTIMIZER_STATE_DICT][ZERO_STAGE]
+    world_size = state_dicts[0][OPTIMIZER_STATE_DICT][PARTITION_COUNT]
+
+    # For ZeRO-2 each param group can have different partition_count as data parallelism for expert
+    # parameters can be different from data parallelism for non-expert parameters. So we can just
+    # use the max of the partition_count to get the dp world_size.
+
+    if type(world_size) is list:
+        world_size = max(world_size)
+
+    if world_size != total_files:
+        raise ValueError(
+            f"Expected {world_size} of '*_optim_states.pt' under '{ds_checkpoint_dir}' but found {total_files} files. "
+            "Possibly due to an overwrite of an old checkpoint, or a checkpoint didn't get saved by one or more processes."
+        )
+
+    # the groups are named differently in each stage
+    if zero_stage <= 2:
+        fp32_groups_key = SINGLE_PARTITION_OF_FP32_GROUPS
+    elif zero_stage == 3:
+        fp32_groups_key = FP32_FLAT_GROUPS
+    else:
+        raise ValueError(f"unknown zero stage {zero_stage}")
+
+    if zero_stage <= 2:
+        fp32_flat_groups = [state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key] for i in range(len(state_dicts))]
+    elif zero_stage == 3:
+        # if there is more than one param group, there will be multiple flattened tensors - one
+        # flattened tensor per group - for simplicity merge them into a single tensor
+        #
+        # XXX: could make the script more memory efficient for when there are multiple groups - it
+        # will require matching the sub-lists of param_shapes for each param group flattened tensor
+
+        fp32_flat_groups = [
+            torch.cat(state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key], 0) for i in range(len(state_dicts))
+        ]
+
+    return zero_stage, world_size, fp32_flat_groups
+
+
+def _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir, exclude_frozen_parameters):
+    """
+    Returns fp32 state_dict reconstructed from ds checkpoint
+
+    Args:
+        - ``ds_checkpoint_dir``: path to the deepspeed checkpoint folder (where the optimizer files are)
+
+    """
+    print(f"Processing zero checkpoint '{ds_checkpoint_dir}'")
+
+    optim_files = get_optim_files(ds_checkpoint_dir)
+    zero_stage, world_size, fp32_flat_groups = parse_optim_states(optim_files, ds_checkpoint_dir)
+    print(f"Detected checkpoint of type zero stage {zero_stage}, world_size: {world_size}")
+
+    model_files = get_model_state_files(ds_checkpoint_dir)
+
+    zero_model_states = parse_model_states(model_files)
+    print(f'Parsing checkpoint created by deepspeed=={zero_model_states[0].ds_version}')
+
+    if zero_stage <= 2:
+        return _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                                          exclude_frozen_parameters)
+    elif zero_stage == 3:
+        return _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                                          exclude_frozen_parameters)
+
+
+def _zero2_merge_frozen_params(state_dict, zero_model_states):
+    if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
+        return
+
+    frozen_param_shapes = zero_model_states[0].frozen_param_shapes
+    frozen_param_fragments = zero_model_states[0].frozen_param_fragments
+
+    if debug:
+        num_elem = sum(s.numel() for s in frozen_param_shapes.values())
+        print(f'rank 0: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
+
+        wanted_params = len(frozen_param_shapes)
+        wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
+        avail_numel = sum([p.numel() for p in frozen_param_fragments.values()])
+        print(f'Frozen params: Have {avail_numel} numels to process.')
+        print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
+
+    total_params = 0
+    total_numel = 0
+    for name, shape in frozen_param_shapes.items():
+        total_params += 1
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+
+        state_dict[name] = frozen_param_fragments[name]
+
+        if debug:
+            print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
+
+    print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _has_callable(obj, fn):
+    attr = getattr(obj, fn, None)
+    return callable(attr)
+
+
+def _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
+    param_shapes = zero_model_states[0].param_shapes
+
+    # Reconstruction protocol:
+    #
+    # XXX: document this
+
+    if debug:
+        for i in range(world_size):
+            for j in range(len(fp32_flat_groups[0])):
+                print(f"{FP32_FLAT_GROUPS}[{i}][{j}].shape={fp32_flat_groups[i][j].shape}")
+
+    # XXX: memory usage doubles here (zero2)
+    num_param_groups = len(fp32_flat_groups[0])
+    merged_single_partition_of_fp32_groups = []
+    for i in range(num_param_groups):
+        merged_partitions = [sd[i] for sd in fp32_flat_groups]
+        full_single_fp32_vector = torch.cat(merged_partitions, 0)
+        merged_single_partition_of_fp32_groups.append(full_single_fp32_vector)
+    avail_numel = sum(
+        [full_single_fp32_vector.numel() for full_single_fp32_vector in merged_single_partition_of_fp32_groups])
+
+    if debug:
+        wanted_params = sum([len(shapes) for shapes in param_shapes])
+        wanted_numel = sum([sum(shape.numel() for shape in shapes.values()) for shapes in param_shapes])
+        # not asserting if there is a mismatch due to possible padding
+        print(f"Have {avail_numel} numels to process.")
+        print(f"Need {wanted_numel} numels in {wanted_params} params.")
+
+    # params
+    # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
+    # out-of-core computing solution
+    total_numel = 0
+    total_params = 0
+    for shapes, full_single_fp32_vector in zip(param_shapes, merged_single_partition_of_fp32_groups):
+        offset = 0
+        avail_numel = full_single_fp32_vector.numel()
+        for name, shape in shapes.items():
+
+            unpartitioned_numel = shape.numel() if _has_callable(shape, 'numel') else math.prod(shape)
+            total_numel += unpartitioned_numel
+            total_params += 1
+
+            if debug:
+                print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
+            state_dict[name] = full_single_fp32_vector.narrow(0, offset, unpartitioned_numel).view(shape)
+            offset += unpartitioned_numel
+
+        # Z2 started to align to 2*world_size to improve nccl performance. Therefore both offset and
+        # avail_numel can differ by anywhere between 0..2*world_size. Due to two unrelated complex
+        # paddings performed in the code it's almost impossible to predict the exact numbers w/o the
+        # live optimizer object, so we are checking that the numbers are within the right range
+        align_to = 2 * world_size
+
+        def zero2_align(x):
+            return align_to * math.ceil(x / align_to)
+
+        if debug:
+            print(f"original offset={offset}, avail_numel={avail_numel}")
+
+        offset = zero2_align(offset)
+        avail_numel = zero2_align(avail_numel)
+
+        if debug:
+            print(f"aligned  offset={offset}, avail_numel={avail_numel}")
+
+        # Sanity check
+        if offset != avail_numel:
+            raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
+
+    print(f"Reconstructed fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                               exclude_frozen_parameters):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    if not exclude_frozen_parameters:
+        _zero2_merge_frozen_params(state_dict, zero_model_states)
+
+    _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
+
+    # recover shared parameters
+    for pair in zero_model_states[0].shared_params:
+        if pair[1] in state_dict:
+            state_dict[pair[0]] = state_dict[pair[1]]
+
+    return state_dict
+
+
+def zero3_partitioned_param_info(unpartitioned_numel, world_size):
+    remainder = unpartitioned_numel % world_size
+    padding_numel = (world_size - remainder) if remainder else 0
+    partitioned_numel = math.ceil(unpartitioned_numel / world_size)
+    return partitioned_numel, padding_numel
+
+
+def _zero3_merge_frozen_params(state_dict, world_size, zero_model_states):
+    if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
+        return
+
+    if debug:
+        for i in range(world_size):
+            num_elem = sum(s.numel() for s in zero_model_states[i].frozen_param_fragments.values())
+            print(f'rank {i}: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
+
+        frozen_param_shapes = zero_model_states[0].frozen_param_shapes
+        wanted_params = len(frozen_param_shapes)
+        wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
+        avail_numel = sum([p.numel() for p in zero_model_states[0].frozen_param_fragments.values()]) * world_size
+        print(f'Frozen params: Have {avail_numel} numels to process.')
+        print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
+
+    total_params = 0
+    total_numel = 0
+    for name, shape in zero_model_states[0].frozen_param_shapes.items():
+        total_params += 1
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+
+        param_frags = tuple(model_state.frozen_param_fragments[name] for model_state in zero_model_states)
+        state_dict[name] = torch.cat(param_frags, 0).narrow(0, 0, unpartitioned_numel).view(shape)
+
+        partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
+
+        if debug:
+            print(
+                f"Frozen params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
+            )
+
+    print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
+    param_shapes = zero_model_states[0].param_shapes
+    avail_numel = fp32_flat_groups[0].numel() * world_size
+    # Reconstruction protocol: For zero3 we need to zip the partitions together at boundary of each
+    # param, re-consolidating each param, while dealing with padding if any
+
+    # merge list of dicts, preserving order
+    param_shapes = {k: v for d in param_shapes for k, v in d.items()}
+
+    if debug:
+        for i in range(world_size):
+            print(f"{FP32_FLAT_GROUPS}[{i}].shape={fp32_flat_groups[i].shape}")
+
+        wanted_params = len(param_shapes)
+        wanted_numel = sum(shape.numel() for shape in param_shapes.values())
+        # not asserting if there is a mismatch due to possible padding
+        avail_numel = fp32_flat_groups[0].numel() * world_size
+        print(f"Trainable params: Have {avail_numel} numels to process.")
+        print(f"Trainable params: Need {wanted_numel} numels in {wanted_params} params.")
+
+    # params
+    # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
+    # out-of-core computing solution
+    offset = 0
+    total_numel = 0
+    total_params = 0
+    for name, shape in param_shapes.items():
+
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+        total_params += 1
+
+        partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
+
+        if debug:
+            print(
+                f"Trainable params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
+            )
+
+        # XXX: memory usage doubles here
+        state_dict[name] = torch.cat(
+            tuple(fp32_flat_groups[i].narrow(0, offset, partitioned_numel) for i in range(world_size)),
+            0).narrow(0, 0, unpartitioned_numel).view(shape)
+        offset += partitioned_numel
+
+    offset *= world_size
+
+    # Sanity check
+    if offset != avail_numel:
+        raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
+
+    print(f"Reconstructed Trainable fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                               exclude_frozen_parameters):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    if not exclude_frozen_parameters:
+        _zero3_merge_frozen_params(state_dict, world_size, zero_model_states)
+
+    _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
+
+    # recover shared parameters
+    for pair in zero_model_states[0].shared_params:
+        if pair[1] in state_dict:
+            state_dict[pair[0]] = state_dict[pair[1]]
+
+    return state_dict
+
+
+def get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag=None, exclude_frozen_parameters=False):
+    """
+    Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated state_dict that can be loaded with
+    ``load_state_dict()`` and used for training without DeepSpeed or shared with others, for example
+    via a model hub.
+
+    Args:
+        - ``checkpoint_dir``: path to the desired checkpoint folder
+        - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in 'latest' file. e.g., ``global_step14``
+        - ``exclude_frozen_parameters``: exclude frozen parameters
+
+    Returns:
+        - pytorch ``state_dict``
+
+    Note: this approach may not work if your application doesn't have sufficient free CPU memory and
+    you may need to use the offline approach using the ``zero_to_fp32.py`` script that is saved with
+    the checkpoint.
+
+    A typical usage might be ::
+
+        from deepspeed.utils.zero_to_fp32 import get_fp32_state_dict_from_zero_checkpoint
+        # do the training and checkpoint saving
+        state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir) # already on cpu
+        model = model.cpu() # move to cpu
+        model.load_state_dict(state_dict)
+        # submit to model hub or save the model to share with others
+
+    In this example the ``model`` will no longer be usable in the deepspeed context of the same
+    application. i.e. you will need to re-initialize the deepspeed engine, since
+    ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
+
+    If you want it all done for you, use ``load_state_dict_from_zero_checkpoint`` instead.
+
+    """
+    if tag is None:
+        latest_path = os.path.join(checkpoint_dir, 'latest')
+        if os.path.isfile(latest_path):
+            with open(latest_path, 'r') as fd:
+                tag = fd.read().strip()
+        else:
+            raise ValueError(f"Unable to find 'latest' file at {latest_path}")
+
+    ds_checkpoint_dir = os.path.join(checkpoint_dir, tag)
+
+    if not os.path.isdir(ds_checkpoint_dir):
+        raise FileNotFoundError(f"Directory '{ds_checkpoint_dir}' doesn't exist")
+
+    return _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir, exclude_frozen_parameters)
+
+
+def convert_zero_checkpoint_to_fp32_state_dict(checkpoint_dir, output_file, tag=None, exclude_frozen_parameters=False):
+    """
+    Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict`` file that can be
+    loaded with ``torch.load(file)`` + ``load_state_dict()`` and used for training without DeepSpeed.
+
+    Args:
+        - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
+        - ``output_file``: path to the pytorch fp32 state_dict output file (e.g. path/pytorch_model.bin)
+        - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
+        - ``exclude_frozen_parameters``: exclude frozen parameters
+    """
+
+    state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag, exclude_frozen_parameters)
+    print(f"Saving fp32 state dict to {output_file}")
+    torch.save(state_dict, output_file)
+
+
+def load_state_dict_from_zero_checkpoint(model, checkpoint_dir, tag=None):
+    """
+    1. Put the provided model to cpu
+    2. Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict``
+    3. Load it into the provided model
+
+    Args:
+        - ``model``: the model object to update
+        - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
+        - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
+
+    Returns:
+        - ``model`: modified model
+
+    Make sure you have plenty of CPU memory available before you call this function. If you don't
+    have enough use the ``zero_to_fp32.py`` utility to do the conversion. You will find it
+    conveniently placed for you in the checkpoint folder.
+
+    A typical usage might be ::
+
+        from deepspeed.utils.zero_to_fp32 import load_state_dict_from_zero_checkpoint
+        model = load_state_dict_from_zero_checkpoint(trainer.model, checkpoint_dir)
+        # submit to model hub or save the model to share with others
+
+    Note, that once this was run, the ``model`` will no longer be usable in the deepspeed context
+    of the same application. i.e. you will need to re-initialize the deepspeed engine, since
+    ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
+
+    """
+    logger.info(f"Extracting fp32 weights")
+    state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag)
+
+    logger.info(f"Overwriting model with fp32 weights")
+    model = model.cpu()
+    model.load_state_dict(state_dict, strict=False)
+
+    return model
+
+
+if __name__ == "__main__":
+
+    parser = argparse.ArgumentParser()
+    parser.add_argument("checkpoint_dir",
+                        type=str,
+                        help="path to the desired checkpoint folder, e.g., path/checkpoint-12")
+    parser.add_argument(
+        "output_file",
+        type=str,
+        help="path to the pytorch fp32 state_dict output file (e.g. path/checkpoint-12/pytorch_model.bin)")
+    parser.add_argument("-t",
+                        "--tag",
+                        type=str,
+                        default=None,
+                        help="checkpoint tag used as a unique identifier for checkpoint. e.g., global_step1")
+    parser.add_argument("--exclude_frozen_parameters", action='store_true', help="exclude frozen parameters")
+    parser.add_argument("-d", "--debug", action='store_true', help="enable debug")
+    args = parser.parse_args()
+
+    debug = args.debug
+
+    convert_zero_checkpoint_to_fp32_state_dict(args.checkpoint_dir,
+                                               args.output_file,
+                                               tag=args.tag,
+                                               exclude_frozen_parameters=args.exclude_frozen_parameters)
diff --git a/weight_dir/loraWeight/fixbycrflp2/checkpoint-12000/README.md b/weight_dir/loraWeight/fixbycrflp2/checkpoint-12000/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..105007ea041cb936ba3417e0221e47fec447213d
--- /dev/null
+++ b/weight_dir/loraWeight/fixbycrflp2/checkpoint-12000/README.md
@@ -0,0 +1,202 @@
+---
+library_name: peft
+base_model: ./CodeLlama-7b-Instruct-hf
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.11.1
\ No newline at end of file
diff --git a/weight_dir/loraWeight/fixbycrflp2/checkpoint-12000/adapter_config.json b/weight_dir/loraWeight/fixbycrflp2/checkpoint-12000/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..3a66bfa8abef7891a1ea8ba7eff2760c79356f01
--- /dev/null
+++ b/weight_dir/loraWeight/fixbycrflp2/checkpoint-12000/adapter_config.json
@@ -0,0 +1,29 @@
+{
+  "alpha_pattern": {},
+  "auto_mapping": null,
+  "base_model_name_or_path": "./CodeLlama-7b-Instruct-hf",
+  "bias": "none",
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 32,
+  "lora_dropout": 0.05,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "r": 8,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "v_proj",
+    "q_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "use_dora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/weight_dir/loraWeight/fixbycrflp2/checkpoint-12000/adapter_model.bin b/weight_dir/loraWeight/fixbycrflp2/checkpoint-12000/adapter_model.bin
new file mode 100644
index 0000000000000000000000000000000000000000..a89a472c90533e666c86cc0e026d009a4f8246e2
--- /dev/null
+++ b/weight_dir/loraWeight/fixbycrflp2/checkpoint-12000/adapter_model.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:da70a17fee47afe232392efb41959db2fb39a710b36e51755cba3f0651173b81
+size 8433034
diff --git a/weight_dir/loraWeight/fixbycrflp2/checkpoint-12000/added_tokens.json b/weight_dir/loraWeight/fixbycrflp2/checkpoint-12000/added_tokens.json
new file mode 100644
index 0000000000000000000000000000000000000000..c09203d8a52151e0d3b5f3c8e6daedc5b60832b5
--- /dev/null
+++ b/weight_dir/loraWeight/fixbycrflp2/checkpoint-12000/added_tokens.json
@@ -0,0 +1,3 @@
+{
+  "<PAD>": 32016
+}
diff --git a/weight_dir/loraWeight/fixbycrflp2/checkpoint-12000/global_step12000/bf16_zero_pp_rank_0_mp_rank_00_optim_states.pt b/weight_dir/loraWeight/fixbycrflp2/checkpoint-12000/global_step12000/bf16_zero_pp_rank_0_mp_rank_00_optim_states.pt
new file mode 100644
index 0000000000000000000000000000000000000000..e1b38fb862e4f2e035ec66ddcae00d814d5b6307
--- /dev/null
+++ b/weight_dir/loraWeight/fixbycrflp2/checkpoint-12000/global_step12000/bf16_zero_pp_rank_0_mp_rank_00_optim_states.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7f59fb4348dce482deebf38ece351831ee84d9151b63e6685b43f381c24f89e3
+size 12589776
diff --git a/weight_dir/loraWeight/fixbycrflp2/checkpoint-12000/global_step12000/bf16_zero_pp_rank_1_mp_rank_00_optim_states.pt b/weight_dir/loraWeight/fixbycrflp2/checkpoint-12000/global_step12000/bf16_zero_pp_rank_1_mp_rank_00_optim_states.pt
new file mode 100644
index 0000000000000000000000000000000000000000..f8665cad2c9c5f23c1f4d1a3e76f519cb6949910
--- /dev/null
+++ b/weight_dir/loraWeight/fixbycrflp2/checkpoint-12000/global_step12000/bf16_zero_pp_rank_1_mp_rank_00_optim_states.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a3b156529ae1a1c4708ace9ee3152b50a0a2e3c107efc9998a1a4c63fecb73be
+size 12589840
diff --git a/weight_dir/loraWeight/fixbycrflp2/checkpoint-12000/global_step12000/bf16_zero_pp_rank_2_mp_rank_00_optim_states.pt b/weight_dir/loraWeight/fixbycrflp2/checkpoint-12000/global_step12000/bf16_zero_pp_rank_2_mp_rank_00_optim_states.pt
new file mode 100644
index 0000000000000000000000000000000000000000..ea684bebe7b170048e06faaa1a82765913d0b237
--- /dev/null
+++ b/weight_dir/loraWeight/fixbycrflp2/checkpoint-12000/global_step12000/bf16_zero_pp_rank_2_mp_rank_00_optim_states.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ece314cfdcbb4ed6c8655dc845821f529b58a7a64e9295e0e03d6ce4c3f27971
+size 12589840
diff --git a/weight_dir/loraWeight/fixbycrflp2/checkpoint-12000/global_step12000/bf16_zero_pp_rank_3_mp_rank_00_optim_states.pt b/weight_dir/loraWeight/fixbycrflp2/checkpoint-12000/global_step12000/bf16_zero_pp_rank_3_mp_rank_00_optim_states.pt
new file mode 100644
index 0000000000000000000000000000000000000000..f57123a1ba8447d2c3373c5b3d5cc5373983da55
--- /dev/null
+++ b/weight_dir/loraWeight/fixbycrflp2/checkpoint-12000/global_step12000/bf16_zero_pp_rank_3_mp_rank_00_optim_states.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:eb9a4c17bbfcbaa0c166f62ad74a2fcbdc8d5ced1df818b16caf2f375c6c6bad
+size 12589840
diff --git a/weight_dir/loraWeight/fixbycrflp2/checkpoint-12000/global_step12000/mp_rank_00_model_states.pt b/weight_dir/loraWeight/fixbycrflp2/checkpoint-12000/global_step12000/mp_rank_00_model_states.pt
new file mode 100644
index 0000000000000000000000000000000000000000..5611ccce688a65f2b77d5a681fefb67905f04dc7
--- /dev/null
+++ b/weight_dir/loraWeight/fixbycrflp2/checkpoint-12000/global_step12000/mp_rank_00_model_states.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ba6c65fd295bb30b20fbf30e97d37e724733e14759a6459b074a218d014dd6dd
+size 8507372
diff --git a/weight_dir/loraWeight/fixbycrflp2/checkpoint-12000/latest b/weight_dir/loraWeight/fixbycrflp2/checkpoint-12000/latest
new file mode 100644
index 0000000000000000000000000000000000000000..b8f4d8b87537b41ea73a42841f5fb1704b16b234
--- /dev/null
+++ b/weight_dir/loraWeight/fixbycrflp2/checkpoint-12000/latest
@@ -0,0 +1 @@
+global_step12000
\ No newline at end of file
diff --git a/weight_dir/loraWeight/fixbycrflp2/checkpoint-12000/rng_state_0.pth b/weight_dir/loraWeight/fixbycrflp2/checkpoint-12000/rng_state_0.pth
new file mode 100644
index 0000000000000000000000000000000000000000..94597257604064fe9ad3c44c3f413114a5470ae7
--- /dev/null
+++ b/weight_dir/loraWeight/fixbycrflp2/checkpoint-12000/rng_state_0.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6144186c9e0ee485931eb4d99dedd714a16e9deb2967f0b44a941fe715daea34
+size 15024
diff --git a/weight_dir/loraWeight/fixbycrflp2/checkpoint-12000/rng_state_1.pth b/weight_dir/loraWeight/fixbycrflp2/checkpoint-12000/rng_state_1.pth
new file mode 100644
index 0000000000000000000000000000000000000000..70a37f4fc5c1d1ceeeddedb50fca0855a1b07859
--- /dev/null
+++ b/weight_dir/loraWeight/fixbycrflp2/checkpoint-12000/rng_state_1.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:2b52dc2350899c1747d6f03e5ad15ad8eff3a3da9e789354be34fa58fb06557b
+size 15024
diff --git a/weight_dir/loraWeight/fixbycrflp2/checkpoint-12000/rng_state_2.pth b/weight_dir/loraWeight/fixbycrflp2/checkpoint-12000/rng_state_2.pth
new file mode 100644
index 0000000000000000000000000000000000000000..7ab0e0a6407124e0aa449beae3cf6262f06cd40c
--- /dev/null
+++ b/weight_dir/loraWeight/fixbycrflp2/checkpoint-12000/rng_state_2.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f148060e8336b10f2da5da6047431bdc25d3a563bd1498b6f14a9d6e0d272b24
+size 15024
diff --git a/weight_dir/loraWeight/fixbycrflp2/checkpoint-12000/rng_state_3.pth b/weight_dir/loraWeight/fixbycrflp2/checkpoint-12000/rng_state_3.pth
new file mode 100644
index 0000000000000000000000000000000000000000..6e828530adcc7499d88d659c3561db4627106f55
--- /dev/null
+++ b/weight_dir/loraWeight/fixbycrflp2/checkpoint-12000/rng_state_3.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:9a8cb83b3ada39d011d872f34af1c1648119ffb242ca9c0f615a40d193a66eb7
+size 15024
diff --git a/weight_dir/loraWeight/fixbycrflp2/checkpoint-12000/special_tokens_map.json b/weight_dir/loraWeight/fixbycrflp2/checkpoint-12000/special_tokens_map.json
new file mode 100644
index 0000000000000000000000000000000000000000..2ece23e993f7e1c7063cb51148b6fa5c6c224775
--- /dev/null
+++ b/weight_dir/loraWeight/fixbycrflp2/checkpoint-12000/special_tokens_map.json
@@ -0,0 +1,30 @@
+{
+  "bos_token": {
+    "content": "<s>",
+    "lstrip": false,
+    "normalized": true,
+    "rstrip": false,
+    "single_word": false
+  },
+  "eos_token": {
+    "content": "</s>",
+    "lstrip": false,
+    "normalized": true,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": {
+    "content": "<PAD>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "unk_token": {
+    "content": "<unk>",
+    "lstrip": false,
+    "normalized": true,
+    "rstrip": false,
+    "single_word": false
+  }
+}
diff --git a/weight_dir/loraWeight/fixbycrflp2/checkpoint-12000/tokenizer.model b/weight_dir/loraWeight/fixbycrflp2/checkpoint-12000/tokenizer.model
new file mode 100644
index 0000000000000000000000000000000000000000..f6722e8b170230ebdd4c0f5f2ce03f219be536d4
--- /dev/null
+++ b/weight_dir/loraWeight/fixbycrflp2/checkpoint-12000/tokenizer.model
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:45ccb9c8b6b561889acea59191d66986d314e7cbd6a78abc6e49b139ca91c1e6
+size 500058
diff --git a/weight_dir/loraWeight/fixbycrflp2/checkpoint-12000/tokenizer_config.json b/weight_dir/loraWeight/fixbycrflp2/checkpoint-12000/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..dc4d9a55d56b17600905bfc69a0e45e2fae7a1c4
--- /dev/null
+++ b/weight_dir/loraWeight/fixbycrflp2/checkpoint-12000/tokenizer_config.json
@@ -0,0 +1,50 @@
+{
+  "add_bos_token": true,
+  "add_eos_token": false,
+  "add_prefix_space": true,
+  "added_tokens_decoder": {
+    "0": {
+      "content": "<unk>",
+      "lstrip": false,
+      "normalized": true,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "1": {
+      "content": "<s>",
+      "lstrip": false,
+      "normalized": true,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "2": {
+      "content": "</s>",
+      "lstrip": false,
+      "normalized": true,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "32016": {
+      "content": "<PAD>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    }
+  },
+  "bos_token": "<s>",
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "</s>",
+  "legacy": true,
+  "model_max_length": 1000000000000000019884624838656,
+  "pad_token": "<PAD>",
+  "sp_model_kwargs": {},
+  "spaces_between_special_tokens": false,
+  "tokenizer_class": "LlamaTokenizer",
+  "unk_token": "<unk>",
+  "use_default_system_prompt": false
+}
diff --git a/weight_dir/loraWeight/fixbycrflp2/checkpoint-12000/trainer_state.json b/weight_dir/loraWeight/fixbycrflp2/checkpoint-12000/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..682a71d842a362601bfa4537262c840f54a27d86
--- /dev/null
+++ b/weight_dir/loraWeight/fixbycrflp2/checkpoint-12000/trainer_state.json
@@ -0,0 +1,1729 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 2.734107997265892,
+  "eval_steps": 6000,
+  "global_step": 12000,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.01139211665527455,
+      "grad_norm": 0.24318678677082062,
+      "learning_rate": 4.989290434307069e-05,
+      "loss": 4.1869,
+      "step": 50
+    },
+    {
+      "epoch": 0.0227842333105491,
+      "grad_norm": 0.10898148268461227,
+      "learning_rate": 4.977897279314588e-05,
+      "loss": 0.0602,
+      "step": 100
+    },
+    {
+      "epoch": 0.03417634996582365,
+      "grad_norm": 0.07570989429950714,
+      "learning_rate": 4.966504124322108e-05,
+      "loss": 0.0489,
+      "step": 150
+    },
+    {
+      "epoch": 0.0455684666210982,
+      "grad_norm": 0.15807728469371796,
+      "learning_rate": 4.955110969329627e-05,
+      "loss": 0.0475,
+      "step": 200
+    },
+    {
+      "epoch": 0.05696058327637275,
+      "grad_norm": 0.09411854296922684,
+      "learning_rate": 4.943717814337147e-05,
+      "loss": 0.0477,
+      "step": 250
+    },
+    {
+      "epoch": 0.0683526999316473,
+      "grad_norm": 0.0637858584523201,
+      "learning_rate": 4.932324659344666e-05,
+      "loss": 0.0464,
+      "step": 300
+    },
+    {
+      "epoch": 0.07974481658692185,
+      "grad_norm": 0.08095966279506683,
+      "learning_rate": 4.920931504352185e-05,
+      "loss": 0.0419,
+      "step": 350
+    },
+    {
+      "epoch": 0.0911369332421964,
+      "grad_norm": 0.12281342595815659,
+      "learning_rate": 4.909538349359705e-05,
+      "loss": 0.0454,
+      "step": 400
+    },
+    {
+      "epoch": 0.10252904989747096,
+      "grad_norm": 0.08293534070253372,
+      "learning_rate": 4.898145194367225e-05,
+      "loss": 0.0466,
+      "step": 450
+    },
+    {
+      "epoch": 0.1139211665527455,
+      "grad_norm": 0.0917474702000618,
+      "learning_rate": 4.886752039374744e-05,
+      "loss": 0.043,
+      "step": 500
+    },
+    {
+      "epoch": 0.12531328320802004,
+      "grad_norm": 0.1135578453540802,
+      "learning_rate": 4.875358884382263e-05,
+      "loss": 0.046,
+      "step": 550
+    },
+    {
+      "epoch": 0.1367053998632946,
+      "grad_norm": 0.14516398310661316,
+      "learning_rate": 4.863965729389783e-05,
+      "loss": 0.0399,
+      "step": 600
+    },
+    {
+      "epoch": 0.14809751651856914,
+      "grad_norm": 0.090349480509758,
+      "learning_rate": 4.852572574397302e-05,
+      "loss": 0.0425,
+      "step": 650
+    },
+    {
+      "epoch": 0.1594896331738437,
+      "grad_norm": 0.10766596347093582,
+      "learning_rate": 4.8411794194048214e-05,
+      "loss": 0.0405,
+      "step": 700
+    },
+    {
+      "epoch": 0.17088174982911825,
+      "grad_norm": 0.09330946952104568,
+      "learning_rate": 4.829786264412341e-05,
+      "loss": 0.0414,
+      "step": 750
+    },
+    {
+      "epoch": 0.1822738664843928,
+      "grad_norm": 0.06139269098639488,
+      "learning_rate": 4.818393109419861e-05,
+      "loss": 0.0422,
+      "step": 800
+    },
+    {
+      "epoch": 0.19366598313966735,
+      "grad_norm": 0.10222186148166656,
+      "learning_rate": 4.80699995442738e-05,
+      "loss": 0.0457,
+      "step": 850
+    },
+    {
+      "epoch": 0.2050580997949419,
+      "grad_norm": 0.1095157265663147,
+      "learning_rate": 4.7956067994348995e-05,
+      "loss": 0.0393,
+      "step": 900
+    },
+    {
+      "epoch": 0.21645021645021645,
+      "grad_norm": 0.10699036717414856,
+      "learning_rate": 4.7842136444424193e-05,
+      "loss": 0.0448,
+      "step": 950
+    },
+    {
+      "epoch": 0.227842333105491,
+      "grad_norm": 0.08551250398159027,
+      "learning_rate": 4.772820489449939e-05,
+      "loss": 0.0394,
+      "step": 1000
+    },
+    {
+      "epoch": 0.23923444976076555,
+      "grad_norm": 0.12492193281650543,
+      "learning_rate": 4.7614273344574584e-05,
+      "loss": 0.0434,
+      "step": 1050
+    },
+    {
+      "epoch": 0.2506265664160401,
+      "grad_norm": 0.14445355534553528,
+      "learning_rate": 4.7500341794649776e-05,
+      "loss": 0.0443,
+      "step": 1100
+    },
+    {
+      "epoch": 0.26201868307131465,
+      "grad_norm": 0.09206435084342957,
+      "learning_rate": 4.7386410244724974e-05,
+      "loss": 0.0447,
+      "step": 1150
+    },
+    {
+      "epoch": 0.2734107997265892,
+      "grad_norm": 0.09216534346342087,
+      "learning_rate": 4.7272478694800166e-05,
+      "loss": 0.0411,
+      "step": 1200
+    },
+    {
+      "epoch": 0.2848029163818637,
+      "grad_norm": 0.09902399778366089,
+      "learning_rate": 4.715854714487536e-05,
+      "loss": 0.0385,
+      "step": 1250
+    },
+    {
+      "epoch": 0.2961950330371383,
+      "grad_norm": 0.06204482167959213,
+      "learning_rate": 4.7044615594950556e-05,
+      "loss": 0.0433,
+      "step": 1300
+    },
+    {
+      "epoch": 0.30758714969241285,
+      "grad_norm": 0.08369041979312897,
+      "learning_rate": 4.6930684045025755e-05,
+      "loss": 0.0416,
+      "step": 1350
+    },
+    {
+      "epoch": 0.3189792663476874,
+      "grad_norm": 0.07282637059688568,
+      "learning_rate": 4.6816752495100947e-05,
+      "loss": 0.036,
+      "step": 1400
+    },
+    {
+      "epoch": 0.3303713830029619,
+      "grad_norm": 0.09879472851753235,
+      "learning_rate": 4.670282094517614e-05,
+      "loss": 0.0385,
+      "step": 1450
+    },
+    {
+      "epoch": 0.3417634996582365,
+      "grad_norm": 0.06464097648859024,
+      "learning_rate": 4.658888939525134e-05,
+      "loss": 0.0413,
+      "step": 1500
+    },
+    {
+      "epoch": 0.35315561631351106,
+      "grad_norm": 0.09587877243757248,
+      "learning_rate": 4.647495784532653e-05,
+      "loss": 0.0359,
+      "step": 1550
+    },
+    {
+      "epoch": 0.3645477329687856,
+      "grad_norm": 0.09360693395137787,
+      "learning_rate": 4.636102629540173e-05,
+      "loss": 0.0392,
+      "step": 1600
+    },
+    {
+      "epoch": 0.37593984962406013,
+      "grad_norm": 0.09336020797491074,
+      "learning_rate": 4.624709474547692e-05,
+      "loss": 0.0372,
+      "step": 1650
+    },
+    {
+      "epoch": 0.3873319662793347,
+      "grad_norm": 0.06831030547618866,
+      "learning_rate": 4.613316319555212e-05,
+      "loss": 0.0394,
+      "step": 1700
+    },
+    {
+      "epoch": 0.39872408293460926,
+      "grad_norm": 0.0803089365363121,
+      "learning_rate": 4.601923164562731e-05,
+      "loss": 0.0373,
+      "step": 1750
+    },
+    {
+      "epoch": 0.4101161995898838,
+      "grad_norm": 0.12224093824625015,
+      "learning_rate": 4.59053000957025e-05,
+      "loss": 0.0386,
+      "step": 1800
+    },
+    {
+      "epoch": 0.42150831624515833,
+      "grad_norm": 0.10336671024560928,
+      "learning_rate": 4.57913685457777e-05,
+      "loss": 0.0381,
+      "step": 1850
+    },
+    {
+      "epoch": 0.4329004329004329,
+      "grad_norm": 0.08528792858123779,
+      "learning_rate": 4.56774369958529e-05,
+      "loss": 0.0371,
+      "step": 1900
+    },
+    {
+      "epoch": 0.44429254955570746,
+      "grad_norm": 0.11260873079299927,
+      "learning_rate": 4.556350544592809e-05,
+      "loss": 0.0406,
+      "step": 1950
+    },
+    {
+      "epoch": 0.455684666210982,
+      "grad_norm": 0.07501766085624695,
+      "learning_rate": 4.544957389600328e-05,
+      "loss": 0.0376,
+      "step": 2000
+    },
+    {
+      "epoch": 0.46707678286625653,
+      "grad_norm": 0.061855901032686234,
+      "learning_rate": 4.533564234607848e-05,
+      "loss": 0.0404,
+      "step": 2050
+    },
+    {
+      "epoch": 0.4784688995215311,
+      "grad_norm": 0.09459809958934784,
+      "learning_rate": 4.522171079615367e-05,
+      "loss": 0.0414,
+      "step": 2100
+    },
+    {
+      "epoch": 0.48986101617680566,
+      "grad_norm": 0.08048627525568008,
+      "learning_rate": 4.5107779246228864e-05,
+      "loss": 0.037,
+      "step": 2150
+    },
+    {
+      "epoch": 0.5012531328320802,
+      "grad_norm": 0.07276533544063568,
+      "learning_rate": 4.499384769630406e-05,
+      "loss": 0.0388,
+      "step": 2200
+    },
+    {
+      "epoch": 0.5126452494873548,
+      "grad_norm": 0.16834820806980133,
+      "learning_rate": 4.487991614637926e-05,
+      "loss": 0.0387,
+      "step": 2250
+    },
+    {
+      "epoch": 0.5240373661426293,
+      "grad_norm": 0.09600085765123367,
+      "learning_rate": 4.476598459645445e-05,
+      "loss": 0.0376,
+      "step": 2300
+    },
+    {
+      "epoch": 0.5354294827979038,
+      "grad_norm": 0.10438230633735657,
+      "learning_rate": 4.4652053046529645e-05,
+      "loss": 0.0396,
+      "step": 2350
+    },
+    {
+      "epoch": 0.5468215994531784,
+      "grad_norm": 0.06975755095481873,
+      "learning_rate": 4.453812149660484e-05,
+      "loss": 0.0395,
+      "step": 2400
+    },
+    {
+      "epoch": 0.5582137161084529,
+      "grad_norm": 0.11652950942516327,
+      "learning_rate": 4.442418994668004e-05,
+      "loss": 0.0387,
+      "step": 2450
+    },
+    {
+      "epoch": 0.5696058327637274,
+      "grad_norm": 0.07443119585514069,
+      "learning_rate": 4.4310258396755233e-05,
+      "loss": 0.0376,
+      "step": 2500
+    },
+    {
+      "epoch": 0.5809979494190021,
+      "grad_norm": 0.12101172655820847,
+      "learning_rate": 4.4196326846830425e-05,
+      "loss": 0.0368,
+      "step": 2550
+    },
+    {
+      "epoch": 0.5923900660742766,
+      "grad_norm": 0.1421501785516739,
+      "learning_rate": 4.4082395296905624e-05,
+      "loss": 0.0378,
+      "step": 2600
+    },
+    {
+      "epoch": 0.6037821827295512,
+      "grad_norm": 0.19242361187934875,
+      "learning_rate": 4.3968463746980816e-05,
+      "loss": 0.0353,
+      "step": 2650
+    },
+    {
+      "epoch": 0.6151742993848257,
+      "grad_norm": 0.09530607610940933,
+      "learning_rate": 4.385453219705601e-05,
+      "loss": 0.04,
+      "step": 2700
+    },
+    {
+      "epoch": 0.6265664160401002,
+      "grad_norm": 0.1451895534992218,
+      "learning_rate": 4.3740600647131206e-05,
+      "loss": 0.0347,
+      "step": 2750
+    },
+    {
+      "epoch": 0.6379585326953748,
+      "grad_norm": 0.11034991592168808,
+      "learning_rate": 4.3626669097206404e-05,
+      "loss": 0.0338,
+      "step": 2800
+    },
+    {
+      "epoch": 0.6493506493506493,
+      "grad_norm": 0.11478397250175476,
+      "learning_rate": 4.3512737547281596e-05,
+      "loss": 0.0412,
+      "step": 2850
+    },
+    {
+      "epoch": 0.6607427660059239,
+      "grad_norm": 0.12228801101446152,
+      "learning_rate": 4.339880599735679e-05,
+      "loss": 0.0362,
+      "step": 2900
+    },
+    {
+      "epoch": 0.6721348826611985,
+      "grad_norm": 0.10513991862535477,
+      "learning_rate": 4.3284874447431987e-05,
+      "loss": 0.0352,
+      "step": 2950
+    },
+    {
+      "epoch": 0.683526999316473,
+      "grad_norm": 0.1606149673461914,
+      "learning_rate": 4.317094289750718e-05,
+      "loss": 0.0404,
+      "step": 3000
+    },
+    {
+      "epoch": 0.6949191159717476,
+      "grad_norm": 0.12138944864273071,
+      "learning_rate": 4.305701134758238e-05,
+      "loss": 0.0366,
+      "step": 3050
+    },
+    {
+      "epoch": 0.7063112326270221,
+      "grad_norm": 0.16719172894954681,
+      "learning_rate": 4.294307979765757e-05,
+      "loss": 0.0348,
+      "step": 3100
+    },
+    {
+      "epoch": 0.7177033492822966,
+      "grad_norm": 0.07648681104183197,
+      "learning_rate": 4.282914824773277e-05,
+      "loss": 0.0356,
+      "step": 3150
+    },
+    {
+      "epoch": 0.7290954659375712,
+      "grad_norm": 0.11438747495412827,
+      "learning_rate": 4.271521669780796e-05,
+      "loss": 0.0417,
+      "step": 3200
+    },
+    {
+      "epoch": 0.7404875825928458,
+      "grad_norm": 0.10588862746953964,
+      "learning_rate": 4.260128514788315e-05,
+      "loss": 0.0369,
+      "step": 3250
+    },
+    {
+      "epoch": 0.7518796992481203,
+      "grad_norm": 0.0825788825750351,
+      "learning_rate": 4.248735359795835e-05,
+      "loss": 0.0369,
+      "step": 3300
+    },
+    {
+      "epoch": 0.7632718159033949,
+      "grad_norm": 0.08810622990131378,
+      "learning_rate": 4.237342204803355e-05,
+      "loss": 0.0352,
+      "step": 3350
+    },
+    {
+      "epoch": 0.7746639325586694,
+      "grad_norm": 0.08460413664579391,
+      "learning_rate": 4.225949049810874e-05,
+      "loss": 0.0384,
+      "step": 3400
+    },
+    {
+      "epoch": 0.7860560492139439,
+      "grad_norm": 0.09517823904752731,
+      "learning_rate": 4.214555894818393e-05,
+      "loss": 0.0335,
+      "step": 3450
+    },
+    {
+      "epoch": 0.7974481658692185,
+      "grad_norm": 0.1491287350654602,
+      "learning_rate": 4.203162739825913e-05,
+      "loss": 0.043,
+      "step": 3500
+    },
+    {
+      "epoch": 0.808840282524493,
+      "grad_norm": 0.0725693553686142,
+      "learning_rate": 4.191769584833432e-05,
+      "loss": 0.0348,
+      "step": 3550
+    },
+    {
+      "epoch": 0.8202323991797676,
+      "grad_norm": 0.10975562781095505,
+      "learning_rate": 4.180376429840952e-05,
+      "loss": 0.0354,
+      "step": 3600
+    },
+    {
+      "epoch": 0.8316245158350422,
+      "grad_norm": 0.08988745510578156,
+      "learning_rate": 4.168983274848471e-05,
+      "loss": 0.0376,
+      "step": 3650
+    },
+    {
+      "epoch": 0.8430166324903167,
+      "grad_norm": 0.09260562807321548,
+      "learning_rate": 4.157590119855991e-05,
+      "loss": 0.0381,
+      "step": 3700
+    },
+    {
+      "epoch": 0.8544087491455913,
+      "grad_norm": 0.10777267068624496,
+      "learning_rate": 4.14619696486351e-05,
+      "loss": 0.0355,
+      "step": 3750
+    },
+    {
+      "epoch": 0.8658008658008658,
+      "grad_norm": 0.10656338185071945,
+      "learning_rate": 4.1348038098710294e-05,
+      "loss": 0.0362,
+      "step": 3800
+    },
+    {
+      "epoch": 0.8771929824561403,
+      "grad_norm": 0.07918556034564972,
+      "learning_rate": 4.123410654878549e-05,
+      "loss": 0.0328,
+      "step": 3850
+    },
+    {
+      "epoch": 0.8885850991114149,
+      "grad_norm": 0.12000677734613419,
+      "learning_rate": 4.112017499886069e-05,
+      "loss": 0.0361,
+      "step": 3900
+    },
+    {
+      "epoch": 0.8999772157666894,
+      "grad_norm": 0.16487479209899902,
+      "learning_rate": 4.100624344893588e-05,
+      "loss": 0.0372,
+      "step": 3950
+    },
+    {
+      "epoch": 0.911369332421964,
+      "grad_norm": 0.1201501414179802,
+      "learning_rate": 4.0892311899011075e-05,
+      "loss": 0.0382,
+      "step": 4000
+    },
+    {
+      "epoch": 0.9227614490772386,
+      "grad_norm": 0.0880098044872284,
+      "learning_rate": 4.077838034908627e-05,
+      "loss": 0.0371,
+      "step": 4050
+    },
+    {
+      "epoch": 0.9341535657325131,
+      "grad_norm": 0.11153139173984528,
+      "learning_rate": 4.0664448799161465e-05,
+      "loss": 0.0379,
+      "step": 4100
+    },
+    {
+      "epoch": 0.9455456823877877,
+      "grad_norm": 0.1396927535533905,
+      "learning_rate": 4.055051724923666e-05,
+      "loss": 0.0344,
+      "step": 4150
+    },
+    {
+      "epoch": 0.9569377990430622,
+      "grad_norm": 0.1013760045170784,
+      "learning_rate": 4.0436585699311855e-05,
+      "loss": 0.036,
+      "step": 4200
+    },
+    {
+      "epoch": 0.9683299156983367,
+      "grad_norm": 0.08559609204530716,
+      "learning_rate": 4.0322654149387054e-05,
+      "loss": 0.036,
+      "step": 4250
+    },
+    {
+      "epoch": 0.9797220323536113,
+      "grad_norm": 0.09353000670671463,
+      "learning_rate": 4.0208722599462246e-05,
+      "loss": 0.0371,
+      "step": 4300
+    },
+    {
+      "epoch": 0.9911141490088858,
+      "grad_norm": 0.1802058070898056,
+      "learning_rate": 4.009479104953744e-05,
+      "loss": 0.0381,
+      "step": 4350
+    },
+    {
+      "epoch": 1.0025062656641603,
+      "grad_norm": 0.1477678269147873,
+      "learning_rate": 3.9980859499612636e-05,
+      "loss": 0.0391,
+      "step": 4400
+    },
+    {
+      "epoch": 1.013898382319435,
+      "grad_norm": 0.1183403730392456,
+      "learning_rate": 3.986692794968783e-05,
+      "loss": 0.0358,
+      "step": 4450
+    },
+    {
+      "epoch": 1.0252904989747096,
+      "grad_norm": 0.10020051896572113,
+      "learning_rate": 3.9752996399763026e-05,
+      "loss": 0.0386,
+      "step": 4500
+    },
+    {
+      "epoch": 1.036682615629984,
+      "grad_norm": 0.10497792065143585,
+      "learning_rate": 3.963906484983822e-05,
+      "loss": 0.0352,
+      "step": 4550
+    },
+    {
+      "epoch": 1.0480747322852586,
+      "grad_norm": 0.11814913153648376,
+      "learning_rate": 3.952513329991342e-05,
+      "loss": 0.0353,
+      "step": 4600
+    },
+    {
+      "epoch": 1.0594668489405332,
+      "grad_norm": 0.11891067773103714,
+      "learning_rate": 3.941120174998861e-05,
+      "loss": 0.0356,
+      "step": 4650
+    },
+    {
+      "epoch": 1.0708589655958076,
+      "grad_norm": 0.07517009973526001,
+      "learning_rate": 3.92972702000638e-05,
+      "loss": 0.0338,
+      "step": 4700
+    },
+    {
+      "epoch": 1.0822510822510822,
+      "grad_norm": 0.12747719883918762,
+      "learning_rate": 3.9183338650139e-05,
+      "loss": 0.0354,
+      "step": 4750
+    },
+    {
+      "epoch": 1.0936431989063569,
+      "grad_norm": 0.11763947457075119,
+      "learning_rate": 3.90694071002142e-05,
+      "loss": 0.0349,
+      "step": 4800
+    },
+    {
+      "epoch": 1.1050353155616313,
+      "grad_norm": 0.09053009003400803,
+      "learning_rate": 3.895547555028939e-05,
+      "loss": 0.0362,
+      "step": 4850
+    },
+    {
+      "epoch": 1.1164274322169059,
+      "grad_norm": 0.16646841168403625,
+      "learning_rate": 3.884154400036458e-05,
+      "loss": 0.037,
+      "step": 4900
+    },
+    {
+      "epoch": 1.1278195488721805,
+      "grad_norm": 0.08759460598230362,
+      "learning_rate": 3.872761245043978e-05,
+      "loss": 0.0353,
+      "step": 4950
+    },
+    {
+      "epoch": 1.139211665527455,
+      "grad_norm": 0.09641291946172714,
+      "learning_rate": 3.861368090051497e-05,
+      "loss": 0.0317,
+      "step": 5000
+    },
+    {
+      "epoch": 1.1506037821827295,
+      "grad_norm": 0.11179310083389282,
+      "learning_rate": 3.849974935059017e-05,
+      "loss": 0.0349,
+      "step": 5050
+    },
+    {
+      "epoch": 1.1619958988380041,
+      "grad_norm": 0.0900956317782402,
+      "learning_rate": 3.838581780066536e-05,
+      "loss": 0.034,
+      "step": 5100
+    },
+    {
+      "epoch": 1.1733880154932788,
+      "grad_norm": 0.11659402400255203,
+      "learning_rate": 3.827188625074056e-05,
+      "loss": 0.0341,
+      "step": 5150
+    },
+    {
+      "epoch": 1.1847801321485532,
+      "grad_norm": 0.19316904246807098,
+      "learning_rate": 3.815795470081575e-05,
+      "loss": 0.0357,
+      "step": 5200
+    },
+    {
+      "epoch": 1.1961722488038278,
+      "grad_norm": 0.14138014614582062,
+      "learning_rate": 3.8044023150890944e-05,
+      "loss": 0.0362,
+      "step": 5250
+    },
+    {
+      "epoch": 1.2075643654591024,
+      "grad_norm": 0.08280321955680847,
+      "learning_rate": 3.7930091600966136e-05,
+      "loss": 0.0336,
+      "step": 5300
+    },
+    {
+      "epoch": 1.2189564821143768,
+      "grad_norm": 0.0848747193813324,
+      "learning_rate": 3.781616005104134e-05,
+      "loss": 0.0375,
+      "step": 5350
+    },
+    {
+      "epoch": 1.2303485987696514,
+      "grad_norm": 0.09686152637004852,
+      "learning_rate": 3.770222850111653e-05,
+      "loss": 0.0327,
+      "step": 5400
+    },
+    {
+      "epoch": 1.241740715424926,
+      "grad_norm": 0.10783885419368744,
+      "learning_rate": 3.7588296951191724e-05,
+      "loss": 0.0364,
+      "step": 5450
+    },
+    {
+      "epoch": 1.2531328320802004,
+      "grad_norm": 0.14501520991325378,
+      "learning_rate": 3.747436540126692e-05,
+      "loss": 0.0395,
+      "step": 5500
+    },
+    {
+      "epoch": 1.264524948735475,
+      "grad_norm": 0.16979029774665833,
+      "learning_rate": 3.7360433851342115e-05,
+      "loss": 0.0352,
+      "step": 5550
+    },
+    {
+      "epoch": 1.2759170653907497,
+      "grad_norm": 0.16063962876796722,
+      "learning_rate": 3.7246502301417307e-05,
+      "loss": 0.0357,
+      "step": 5600
+    },
+    {
+      "epoch": 1.287309182046024,
+      "grad_norm": 0.1108214482665062,
+      "learning_rate": 3.7132570751492505e-05,
+      "loss": 0.0341,
+      "step": 5650
+    },
+    {
+      "epoch": 1.2987012987012987,
+      "grad_norm": 0.17493686079978943,
+      "learning_rate": 3.7018639201567704e-05,
+      "loss": 0.0373,
+      "step": 5700
+    },
+    {
+      "epoch": 1.3100934153565733,
+      "grad_norm": 0.08701369911432266,
+      "learning_rate": 3.6904707651642895e-05,
+      "loss": 0.0345,
+      "step": 5750
+    },
+    {
+      "epoch": 1.321485532011848,
+      "grad_norm": 0.11528841406106949,
+      "learning_rate": 3.679077610171809e-05,
+      "loss": 0.0309,
+      "step": 5800
+    },
+    {
+      "epoch": 1.3328776486671223,
+      "grad_norm": 0.14678539335727692,
+      "learning_rate": 3.6676844551793286e-05,
+      "loss": 0.0346,
+      "step": 5850
+    },
+    {
+      "epoch": 1.344269765322397,
+      "grad_norm": 0.13064803183078766,
+      "learning_rate": 3.6562913001868484e-05,
+      "loss": 0.0344,
+      "step": 5900
+    },
+    {
+      "epoch": 1.3556618819776713,
+      "grad_norm": 0.11759040504693985,
+      "learning_rate": 3.6448981451943676e-05,
+      "loss": 0.0319,
+      "step": 5950
+    },
+    {
+      "epoch": 1.367053998632946,
+      "grad_norm": 0.10772958397865295,
+      "learning_rate": 3.633504990201887e-05,
+      "loss": 0.0314,
+      "step": 6000
+    },
+    {
+      "epoch": 1.367053998632946,
+      "eval_loss": 0.05980786308646202,
+      "eval_runtime": 76.4079,
+      "eval_samples_per_second": 23.257,
+      "eval_steps_per_second": 1.466,
+      "step": 6000
+    },
+    {
+      "epoch": 1.3784461152882206,
+      "grad_norm": 0.15903785824775696,
+      "learning_rate": 3.6221118352094066e-05,
+      "loss": 0.0337,
+      "step": 6050
+    },
+    {
+      "epoch": 1.3898382319434952,
+      "grad_norm": 0.15311047434806824,
+      "learning_rate": 3.610718680216926e-05,
+      "loss": 0.034,
+      "step": 6100
+    },
+    {
+      "epoch": 1.4012303485987696,
+      "grad_norm": 0.08784125000238419,
+      "learning_rate": 3.599325525224445e-05,
+      "loss": 0.032,
+      "step": 6150
+    },
+    {
+      "epoch": 1.4126224652540442,
+      "grad_norm": 0.10080964863300323,
+      "learning_rate": 3.587932370231965e-05,
+      "loss": 0.0344,
+      "step": 6200
+    },
+    {
+      "epoch": 1.4240145819093186,
+      "grad_norm": 0.13672150671482086,
+      "learning_rate": 3.576539215239485e-05,
+      "loss": 0.0314,
+      "step": 6250
+    },
+    {
+      "epoch": 1.4354066985645932,
+      "grad_norm": 0.10255203396081924,
+      "learning_rate": 3.565146060247004e-05,
+      "loss": 0.0343,
+      "step": 6300
+    },
+    {
+      "epoch": 1.4467988152198679,
+      "grad_norm": 0.1191258653998375,
+      "learning_rate": 3.553752905254523e-05,
+      "loss": 0.0337,
+      "step": 6350
+    },
+    {
+      "epoch": 1.4581909318751425,
+      "grad_norm": 0.13031437993049622,
+      "learning_rate": 3.542359750262043e-05,
+      "loss": 0.0329,
+      "step": 6400
+    },
+    {
+      "epoch": 1.4695830485304169,
+      "grad_norm": 0.12329483032226562,
+      "learning_rate": 3.530966595269562e-05,
+      "loss": 0.0374,
+      "step": 6450
+    },
+    {
+      "epoch": 1.4809751651856915,
+      "grad_norm": 0.15432628989219666,
+      "learning_rate": 3.519573440277082e-05,
+      "loss": 0.0344,
+      "step": 6500
+    },
+    {
+      "epoch": 1.4923672818409661,
+      "grad_norm": 0.11702828109264374,
+      "learning_rate": 3.508180285284601e-05,
+      "loss": 0.0325,
+      "step": 6550
+    },
+    {
+      "epoch": 1.5037593984962405,
+      "grad_norm": 0.07083294540643692,
+      "learning_rate": 3.496787130292121e-05,
+      "loss": 0.0324,
+      "step": 6600
+    },
+    {
+      "epoch": 1.5151515151515151,
+      "grad_norm": 0.09021273255348206,
+      "learning_rate": 3.48539397529964e-05,
+      "loss": 0.0337,
+      "step": 6650
+    },
+    {
+      "epoch": 1.5265436318067898,
+      "grad_norm": 0.1727963089942932,
+      "learning_rate": 3.4740008203071593e-05,
+      "loss": 0.0344,
+      "step": 6700
+    },
+    {
+      "epoch": 1.5379357484620644,
+      "grad_norm": 0.17168262600898743,
+      "learning_rate": 3.4626076653146785e-05,
+      "loss": 0.0344,
+      "step": 6750
+    },
+    {
+      "epoch": 1.5493278651173388,
+      "grad_norm": 0.08169051259756088,
+      "learning_rate": 3.451214510322199e-05,
+      "loss": 0.0357,
+      "step": 6800
+    },
+    {
+      "epoch": 1.5607199817726134,
+      "grad_norm": 0.14371289312839508,
+      "learning_rate": 3.439821355329718e-05,
+      "loss": 0.0327,
+      "step": 6850
+    },
+    {
+      "epoch": 1.5721120984278878,
+      "grad_norm": 0.10983046889305115,
+      "learning_rate": 3.4284282003372374e-05,
+      "loss": 0.0341,
+      "step": 6900
+    },
+    {
+      "epoch": 1.5835042150831624,
+      "grad_norm": 0.1277296096086502,
+      "learning_rate": 3.417035045344757e-05,
+      "loss": 0.0308,
+      "step": 6950
+    },
+    {
+      "epoch": 1.594896331738437,
+      "grad_norm": 0.12647396326065063,
+      "learning_rate": 3.4056418903522764e-05,
+      "loss": 0.0334,
+      "step": 7000
+    },
+    {
+      "epoch": 1.6062884483937117,
+      "grad_norm": 0.09901976585388184,
+      "learning_rate": 3.394248735359796e-05,
+      "loss": 0.0322,
+      "step": 7050
+    },
+    {
+      "epoch": 1.617680565048986,
+      "grad_norm": 0.1494535207748413,
+      "learning_rate": 3.3828555803673155e-05,
+      "loss": 0.0347,
+      "step": 7100
+    },
+    {
+      "epoch": 1.6290726817042607,
+      "grad_norm": 0.14202262461185455,
+      "learning_rate": 3.371462425374835e-05,
+      "loss": 0.029,
+      "step": 7150
+    },
+    {
+      "epoch": 1.640464798359535,
+      "grad_norm": 0.13513007760047913,
+      "learning_rate": 3.3600692703823545e-05,
+      "loss": 0.031,
+      "step": 7200
+    },
+    {
+      "epoch": 1.6518569150148097,
+      "grad_norm": 0.1467670053243637,
+      "learning_rate": 3.348676115389874e-05,
+      "loss": 0.0362,
+      "step": 7250
+    },
+    {
+      "epoch": 1.6632490316700843,
+      "grad_norm": 0.0936419665813446,
+      "learning_rate": 3.337282960397393e-05,
+      "loss": 0.0311,
+      "step": 7300
+    },
+    {
+      "epoch": 1.674641148325359,
+      "grad_norm": 0.10516014695167542,
+      "learning_rate": 3.3258898054049134e-05,
+      "loss": 0.0322,
+      "step": 7350
+    },
+    {
+      "epoch": 1.6860332649806336,
+      "grad_norm": 0.166508287191391,
+      "learning_rate": 3.3144966504124326e-05,
+      "loss": 0.0342,
+      "step": 7400
+    },
+    {
+      "epoch": 1.697425381635908,
+      "grad_norm": 0.13199631869792938,
+      "learning_rate": 3.303103495419952e-05,
+      "loss": 0.0311,
+      "step": 7450
+    },
+    {
+      "epoch": 1.7088174982911823,
+      "grad_norm": 0.10680291801691055,
+      "learning_rate": 3.2917103404274716e-05,
+      "loss": 0.0302,
+      "step": 7500
+    },
+    {
+      "epoch": 1.720209614946457,
+      "grad_norm": 0.14647521078586578,
+      "learning_rate": 3.280317185434991e-05,
+      "loss": 0.0339,
+      "step": 7550
+    },
+    {
+      "epoch": 1.7316017316017316,
+      "grad_norm": 0.08651016652584076,
+      "learning_rate": 3.26892403044251e-05,
+      "loss": 0.0357,
+      "step": 7600
+    },
+    {
+      "epoch": 1.7429938482570062,
+      "grad_norm": 0.07976220548152924,
+      "learning_rate": 3.25753087545003e-05,
+      "loss": 0.0331,
+      "step": 7650
+    },
+    {
+      "epoch": 1.7543859649122808,
+      "grad_norm": 0.13063737750053406,
+      "learning_rate": 3.24613772045755e-05,
+      "loss": 0.0334,
+      "step": 7700
+    },
+    {
+      "epoch": 1.7657780815675552,
+      "grad_norm": 0.10824882984161377,
+      "learning_rate": 3.234744565465069e-05,
+      "loss": 0.0304,
+      "step": 7750
+    },
+    {
+      "epoch": 1.7771701982228298,
+      "grad_norm": 0.08849220722913742,
+      "learning_rate": 3.223351410472588e-05,
+      "loss": 0.0345,
+      "step": 7800
+    },
+    {
+      "epoch": 1.7885623148781042,
+      "grad_norm": 0.1834285408258438,
+      "learning_rate": 3.211958255480107e-05,
+      "loss": 0.0312,
+      "step": 7850
+    },
+    {
+      "epoch": 1.7999544315333789,
+      "grad_norm": 0.17661932110786438,
+      "learning_rate": 3.200565100487627e-05,
+      "loss": 0.0359,
+      "step": 7900
+    },
+    {
+      "epoch": 1.8113465481886535,
+      "grad_norm": 0.08962208777666092,
+      "learning_rate": 3.189171945495147e-05,
+      "loss": 0.0316,
+      "step": 7950
+    },
+    {
+      "epoch": 1.822738664843928,
+      "grad_norm": 0.15646621584892273,
+      "learning_rate": 3.177778790502666e-05,
+      "loss": 0.0312,
+      "step": 8000
+    },
+    {
+      "epoch": 1.8341307814992025,
+      "grad_norm": 0.1474676877260208,
+      "learning_rate": 3.166385635510186e-05,
+      "loss": 0.0344,
+      "step": 8050
+    },
+    {
+      "epoch": 1.8455228981544771,
+      "grad_norm": 0.1147208958864212,
+      "learning_rate": 3.154992480517705e-05,
+      "loss": 0.0329,
+      "step": 8100
+    },
+    {
+      "epoch": 1.8569150148097515,
+      "grad_norm": 0.183086097240448,
+      "learning_rate": 3.143599325525224e-05,
+      "loss": 0.0321,
+      "step": 8150
+    },
+    {
+      "epoch": 1.8683071314650261,
+      "grad_norm": 0.1263459175825119,
+      "learning_rate": 3.132206170532744e-05,
+      "loss": 0.0316,
+      "step": 8200
+    },
+    {
+      "epoch": 1.8796992481203008,
+      "grad_norm": 0.1425926238298416,
+      "learning_rate": 3.120813015540264e-05,
+      "loss": 0.0295,
+      "step": 8250
+    },
+    {
+      "epoch": 1.8910913647755754,
+      "grad_norm": 0.23871935904026031,
+      "learning_rate": 3.109419860547783e-05,
+      "loss": 0.033,
+      "step": 8300
+    },
+    {
+      "epoch": 1.90248348143085,
+      "grad_norm": 0.1078375056385994,
+      "learning_rate": 3.0980267055553024e-05,
+      "loss": 0.0321,
+      "step": 8350
+    },
+    {
+      "epoch": 1.9138755980861244,
+      "grad_norm": 0.06418582051992416,
+      "learning_rate": 3.0866335505628215e-05,
+      "loss": 0.0335,
+      "step": 8400
+    },
+    {
+      "epoch": 1.9252677147413988,
+      "grad_norm": 0.10235786437988281,
+      "learning_rate": 3.0752403955703414e-05,
+      "loss": 0.0341,
+      "step": 8450
+    },
+    {
+      "epoch": 1.9366598313966734,
+      "grad_norm": 0.12743011116981506,
+      "learning_rate": 3.063847240577861e-05,
+      "loss": 0.0349,
+      "step": 8500
+    },
+    {
+      "epoch": 1.948051948051948,
+      "grad_norm": 0.13372161984443665,
+      "learning_rate": 3.0524540855853804e-05,
+      "loss": 0.0309,
+      "step": 8550
+    },
+    {
+      "epoch": 1.9594440647072227,
+      "grad_norm": 0.14914734661579132,
+      "learning_rate": 3.0410609305929e-05,
+      "loss": 0.0307,
+      "step": 8600
+    },
+    {
+      "epoch": 1.9708361813624973,
+      "grad_norm": 0.08178412914276123,
+      "learning_rate": 3.0296677756004195e-05,
+      "loss": 0.0329,
+      "step": 8650
+    },
+    {
+      "epoch": 1.9822282980177717,
+      "grad_norm": 0.12782682478427887,
+      "learning_rate": 3.0182746206079386e-05,
+      "loss": 0.0343,
+      "step": 8700
+    },
+    {
+      "epoch": 1.9936204146730463,
+      "grad_norm": 0.13054175674915314,
+      "learning_rate": 3.006881465615458e-05,
+      "loss": 0.0348,
+      "step": 8750
+    },
+    {
+      "epoch": 2.0050125313283207,
+      "grad_norm": 0.15654687583446503,
+      "learning_rate": 2.995488310622978e-05,
+      "loss": 0.0341,
+      "step": 8800
+    },
+    {
+      "epoch": 2.0164046479835953,
+      "grad_norm": 0.15794731676578522,
+      "learning_rate": 2.9840951556304975e-05,
+      "loss": 0.0311,
+      "step": 8850
+    },
+    {
+      "epoch": 2.02779676463887,
+      "grad_norm": 0.10117436945438385,
+      "learning_rate": 2.9727020006380167e-05,
+      "loss": 0.035,
+      "step": 8900
+    },
+    {
+      "epoch": 2.0391888812941446,
+      "grad_norm": 0.15975680947303772,
+      "learning_rate": 2.9613088456455362e-05,
+      "loss": 0.0322,
+      "step": 8950
+    },
+    {
+      "epoch": 2.050580997949419,
+      "grad_norm": 0.09420093894004822,
+      "learning_rate": 2.9499156906530557e-05,
+      "loss": 0.0303,
+      "step": 9000
+    },
+    {
+      "epoch": 2.0619731146046933,
+      "grad_norm": 0.1872691512107849,
+      "learning_rate": 2.938522535660575e-05,
+      "loss": 0.0318,
+      "step": 9050
+    },
+    {
+      "epoch": 2.073365231259968,
+      "grad_norm": 0.1728203147649765,
+      "learning_rate": 2.9271293806680948e-05,
+      "loss": 0.0303,
+      "step": 9100
+    },
+    {
+      "epoch": 2.0847573479152426,
+      "grad_norm": 0.12621662020683289,
+      "learning_rate": 2.9157362256756143e-05,
+      "loss": 0.0328,
+      "step": 9150
+    },
+    {
+      "epoch": 2.096149464570517,
+      "grad_norm": 0.15386846661567688,
+      "learning_rate": 2.9043430706831338e-05,
+      "loss": 0.0315,
+      "step": 9200
+    },
+    {
+      "epoch": 2.107541581225792,
+      "grad_norm": 0.1342351734638214,
+      "learning_rate": 2.892949915690653e-05,
+      "loss": 0.0313,
+      "step": 9250
+    },
+    {
+      "epoch": 2.1189336978810664,
+      "grad_norm": 0.14698518812656403,
+      "learning_rate": 2.8815567606981725e-05,
+      "loss": 0.0338,
+      "step": 9300
+    },
+    {
+      "epoch": 2.1303258145363406,
+      "grad_norm": 0.12786546349525452,
+      "learning_rate": 2.8701636057056924e-05,
+      "loss": 0.0315,
+      "step": 9350
+    },
+    {
+      "epoch": 2.1417179311916152,
+      "grad_norm": 0.17615847289562225,
+      "learning_rate": 2.858770450713212e-05,
+      "loss": 0.0288,
+      "step": 9400
+    },
+    {
+      "epoch": 2.15311004784689,
+      "grad_norm": 0.10935161262750626,
+      "learning_rate": 2.847377295720731e-05,
+      "loss": 0.0306,
+      "step": 9450
+    },
+    {
+      "epoch": 2.1645021645021645,
+      "grad_norm": 0.16573001444339752,
+      "learning_rate": 2.8359841407282506e-05,
+      "loss": 0.0309,
+      "step": 9500
+    },
+    {
+      "epoch": 2.175894281157439,
+      "grad_norm": 0.1991117298603058,
+      "learning_rate": 2.82459098573577e-05,
+      "loss": 0.0301,
+      "step": 9550
+    },
+    {
+      "epoch": 2.1872863978127137,
+      "grad_norm": 0.1383821666240692,
+      "learning_rate": 2.8131978307432893e-05,
+      "loss": 0.0332,
+      "step": 9600
+    },
+    {
+      "epoch": 2.1986785144679883,
+      "grad_norm": 0.2356170415878296,
+      "learning_rate": 2.8018046757508095e-05,
+      "loss": 0.032,
+      "step": 9650
+    },
+    {
+      "epoch": 2.2100706311232625,
+      "grad_norm": 0.2645617723464966,
+      "learning_rate": 2.7904115207583286e-05,
+      "loss": 0.0307,
+      "step": 9700
+    },
+    {
+      "epoch": 2.221462747778537,
+      "grad_norm": 0.147581547498703,
+      "learning_rate": 2.779018365765848e-05,
+      "loss": 0.0324,
+      "step": 9750
+    },
+    {
+      "epoch": 2.2328548644338118,
+      "grad_norm": 0.16367791593074799,
+      "learning_rate": 2.7676252107733673e-05,
+      "loss": 0.031,
+      "step": 9800
+    },
+    {
+      "epoch": 2.2442469810890864,
+      "grad_norm": 0.13787373900413513,
+      "learning_rate": 2.756232055780887e-05,
+      "loss": 0.0326,
+      "step": 9850
+    },
+    {
+      "epoch": 2.255639097744361,
+      "grad_norm": 0.1088612899184227,
+      "learning_rate": 2.7448389007884064e-05,
+      "loss": 0.0346,
+      "step": 9900
+    },
+    {
+      "epoch": 2.2670312143996356,
+      "grad_norm": 0.16382579505443573,
+      "learning_rate": 2.7334457457959262e-05,
+      "loss": 0.0328,
+      "step": 9950
+    },
+    {
+      "epoch": 2.27842333105491,
+      "grad_norm": 0.1262141764163971,
+      "learning_rate": 2.7220525908034454e-05,
+      "loss": 0.0319,
+      "step": 10000
+    },
+    {
+      "epoch": 2.2898154477101844,
+      "grad_norm": 0.1008811965584755,
+      "learning_rate": 2.710659435810965e-05,
+      "loss": 0.0311,
+      "step": 10050
+    },
+    {
+      "epoch": 2.301207564365459,
+      "grad_norm": 0.10501789301633835,
+      "learning_rate": 2.6992662808184844e-05,
+      "loss": 0.0336,
+      "step": 10100
+    },
+    {
+      "epoch": 2.3125996810207337,
+      "grad_norm": 0.12613259255886078,
+      "learning_rate": 2.6878731258260036e-05,
+      "loss": 0.0308,
+      "step": 10150
+    },
+    {
+      "epoch": 2.3239917976760083,
+      "grad_norm": 0.12407036870718002,
+      "learning_rate": 2.676479970833523e-05,
+      "loss": 0.028,
+      "step": 10200
+    },
+    {
+      "epoch": 2.335383914331283,
+      "grad_norm": 0.10909141600131989,
+      "learning_rate": 2.665086815841043e-05,
+      "loss": 0.0328,
+      "step": 10250
+    },
+    {
+      "epoch": 2.3467760309865575,
+      "grad_norm": 0.21803608536720276,
+      "learning_rate": 2.6536936608485625e-05,
+      "loss": 0.0293,
+      "step": 10300
+    },
+    {
+      "epoch": 2.3581681476418317,
+      "grad_norm": 0.13282905519008636,
+      "learning_rate": 2.6423005058560817e-05,
+      "loss": 0.0301,
+      "step": 10350
+    },
+    {
+      "epoch": 2.3695602642971063,
+      "grad_norm": 0.144027441740036,
+      "learning_rate": 2.6309073508636012e-05,
+      "loss": 0.027,
+      "step": 10400
+    },
+    {
+      "epoch": 2.380952380952381,
+      "grad_norm": 0.16259746253490448,
+      "learning_rate": 2.6195141958711207e-05,
+      "loss": 0.0301,
+      "step": 10450
+    },
+    {
+      "epoch": 2.3923444976076556,
+      "grad_norm": 0.14099255204200745,
+      "learning_rate": 2.6081210408786406e-05,
+      "loss": 0.0305,
+      "step": 10500
+    },
+    {
+      "epoch": 2.40373661426293,
+      "grad_norm": 0.11188742518424988,
+      "learning_rate": 2.5967278858861597e-05,
+      "loss": 0.0294,
+      "step": 10550
+    },
+    {
+      "epoch": 2.415128730918205,
+      "grad_norm": 0.11726520955562592,
+      "learning_rate": 2.5853347308936793e-05,
+      "loss": 0.0322,
+      "step": 10600
+    },
+    {
+      "epoch": 2.426520847573479,
+      "grad_norm": 0.12376642227172852,
+      "learning_rate": 2.5739415759011988e-05,
+      "loss": 0.0267,
+      "step": 10650
+    },
+    {
+      "epoch": 2.4379129642287536,
+      "grad_norm": 0.1598552167415619,
+      "learning_rate": 2.562548420908718e-05,
+      "loss": 0.0326,
+      "step": 10700
+    },
+    {
+      "epoch": 2.449305080884028,
+      "grad_norm": 0.14880846440792084,
+      "learning_rate": 2.5511552659162375e-05,
+      "loss": 0.0296,
+      "step": 10750
+    },
+    {
+      "epoch": 2.460697197539303,
+      "grad_norm": 0.22758153080940247,
+      "learning_rate": 2.5397621109237573e-05,
+      "loss": 0.0293,
+      "step": 10800
+    },
+    {
+      "epoch": 2.4720893141945774,
+      "grad_norm": 0.17673678696155548,
+      "learning_rate": 2.528368955931277e-05,
+      "loss": 0.0338,
+      "step": 10850
+    },
+    {
+      "epoch": 2.483481430849852,
+      "grad_norm": 0.2402111291885376,
+      "learning_rate": 2.516975800938796e-05,
+      "loss": 0.0317,
+      "step": 10900
+    },
+    {
+      "epoch": 2.4948735475051267,
+      "grad_norm": 0.16588878631591797,
+      "learning_rate": 2.5055826459463155e-05,
+      "loss": 0.0289,
+      "step": 10950
+    },
+    {
+      "epoch": 2.506265664160401,
+      "grad_norm": 0.1362510472536087,
+      "learning_rate": 2.494189490953835e-05,
+      "loss": 0.0295,
+      "step": 11000
+    },
+    {
+      "epoch": 2.5176577808156755,
+      "grad_norm": 0.14621783792972565,
+      "learning_rate": 2.4827963359613546e-05,
+      "loss": 0.0303,
+      "step": 11050
+    },
+    {
+      "epoch": 2.52904989747095,
+      "grad_norm": 0.4146839380264282,
+      "learning_rate": 2.471403180968874e-05,
+      "loss": 0.0307,
+      "step": 11100
+    },
+    {
+      "epoch": 2.5404420141262247,
+      "grad_norm": 0.16631941497325897,
+      "learning_rate": 2.4600100259763936e-05,
+      "loss": 0.0315,
+      "step": 11150
+    },
+    {
+      "epoch": 2.5518341307814993,
+      "grad_norm": 0.1467452049255371,
+      "learning_rate": 2.448616870983913e-05,
+      "loss": 0.0321,
+      "step": 11200
+    },
+    {
+      "epoch": 2.5632262474367735,
+      "grad_norm": 0.18437771499156952,
+      "learning_rate": 2.4372237159914323e-05,
+      "loss": 0.0299,
+      "step": 11250
+    },
+    {
+      "epoch": 2.574618364092048,
+      "grad_norm": 0.21040193736553192,
+      "learning_rate": 2.425830560998952e-05,
+      "loss": 0.0311,
+      "step": 11300
+    },
+    {
+      "epoch": 2.5860104807473228,
+      "grad_norm": 0.18259912729263306,
+      "learning_rate": 2.4144374060064713e-05,
+      "loss": 0.0279,
+      "step": 11350
+    },
+    {
+      "epoch": 2.5974025974025974,
+      "grad_norm": 0.13136200606822968,
+      "learning_rate": 2.403044251013991e-05,
+      "loss": 0.0292,
+      "step": 11400
+    },
+    {
+      "epoch": 2.608794714057872,
+      "grad_norm": 0.12732388079166412,
+      "learning_rate": 2.3916510960215104e-05,
+      "loss": 0.0291,
+      "step": 11450
+    },
+    {
+      "epoch": 2.6201868307131466,
+      "grad_norm": 0.10226253420114517,
+      "learning_rate": 2.38025794102903e-05,
+      "loss": 0.0313,
+      "step": 11500
+    },
+    {
+      "epoch": 2.6315789473684212,
+      "grad_norm": 0.11919239908456802,
+      "learning_rate": 2.3688647860365494e-05,
+      "loss": 0.0267,
+      "step": 11550
+    },
+    {
+      "epoch": 2.642971064023696,
+      "grad_norm": 0.16108646988868713,
+      "learning_rate": 2.357471631044069e-05,
+      "loss": 0.0292,
+      "step": 11600
+    },
+    {
+      "epoch": 2.65436318067897,
+      "grad_norm": 0.23356276750564575,
+      "learning_rate": 2.3460784760515884e-05,
+      "loss": 0.0319,
+      "step": 11650
+    },
+    {
+      "epoch": 2.6657552973342447,
+      "grad_norm": 0.1463542878627777,
+      "learning_rate": 2.3346853210591076e-05,
+      "loss": 0.0282,
+      "step": 11700
+    },
+    {
+      "epoch": 2.6771474139895193,
+      "grad_norm": 0.12290333956480026,
+      "learning_rate": 2.3232921660666275e-05,
+      "loss": 0.0304,
+      "step": 11750
+    },
+    {
+      "epoch": 2.688539530644794,
+      "grad_norm": 0.1767304688692093,
+      "learning_rate": 2.3118990110741466e-05,
+      "loss": 0.0301,
+      "step": 11800
+    },
+    {
+      "epoch": 2.6999316473000685,
+      "grad_norm": 0.12297718971967697,
+      "learning_rate": 2.300505856081666e-05,
+      "loss": 0.0281,
+      "step": 11850
+    },
+    {
+      "epoch": 2.7113237639553427,
+      "grad_norm": 0.14692600071430206,
+      "learning_rate": 2.2891127010891857e-05,
+      "loss": 0.0288,
+      "step": 11900
+    },
+    {
+      "epoch": 2.7227158806106173,
+      "grad_norm": 0.15868423879146576,
+      "learning_rate": 2.2777195460967052e-05,
+      "loss": 0.0311,
+      "step": 11950
+    },
+    {
+      "epoch": 2.734107997265892,
+      "grad_norm": 0.22269558906555176,
+      "learning_rate": 2.2663263911042247e-05,
+      "loss": 0.0319,
+      "step": 12000
+    },
+    {
+      "epoch": 2.734107997265892,
+      "eval_loss": 0.05642309412360191,
+      "eval_runtime": 76.2989,
+      "eval_samples_per_second": 23.29,
+      "eval_steps_per_second": 1.468,
+      "step": 12000
+    }
+  ],
+  "logging_steps": 50,
+  "max_steps": 21945,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 5,
+  "save_steps": 2000,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 7.459165119956648e+18,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/weight_dir/loraWeight/fixbycrflp2/checkpoint-12000/training_args.bin b/weight_dir/loraWeight/fixbycrflp2/checkpoint-12000/training_args.bin
new file mode 100644
index 0000000000000000000000000000000000000000..9db308ac2f9011d86beca301acc47bd83d423f0f
--- /dev/null
+++ b/weight_dir/loraWeight/fixbycrflp2/checkpoint-12000/training_args.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:8def4cbff2092aebe0edc0db45038825f342bd99987945f3aa74431dfc4b747e
+size 7800
diff --git a/weight_dir/loraWeight/fixbycrflp2/checkpoint-12000/zero_to_fp32.py b/weight_dir/loraWeight/fixbycrflp2/checkpoint-12000/zero_to_fp32.py
new file mode 100644
index 0000000000000000000000000000000000000000..24cc342e78d1a006c782b3a4cd68d9ce786d8fd8
--- /dev/null
+++ b/weight_dir/loraWeight/fixbycrflp2/checkpoint-12000/zero_to_fp32.py
@@ -0,0 +1,604 @@
+#!/usr/bin/env python
+
+# Copyright (c) Microsoft Corporation.
+# SPDX-License-Identifier: Apache-2.0
+
+# DeepSpeed Team
+
+# This script extracts fp32 consolidated weights from a zero 1, 2 and 3 DeepSpeed checkpoints. It gets
+# copied into the top level checkpoint dir, so the user can easily do the conversion at any point in
+# the future. Once extracted, the weights don't require DeepSpeed and can be used in any
+# application.
+#
+# example: python zero_to_fp32.py . pytorch_model.bin
+
+import argparse
+import torch
+import glob
+import math
+import os
+import re
+from collections import OrderedDict
+from dataclasses import dataclass
+
+# while this script doesn't use deepspeed to recover data, since the checkpoints are pickled with
+# DeepSpeed data structures it has to be available in the current python environment.
+from deepspeed.utils import logger
+from deepspeed.checkpoint.constants import (DS_VERSION, OPTIMIZER_STATE_DICT, SINGLE_PARTITION_OF_FP32_GROUPS,
+                                            FP32_FLAT_GROUPS, ZERO_STAGE, PARTITION_COUNT, PARAM_SHAPES, BUFFER_NAMES,
+                                            FROZEN_PARAM_SHAPES, FROZEN_PARAM_FRAGMENTS)
+
+
+@dataclass
+class zero_model_state:
+    buffers: dict()
+    param_shapes: dict()
+    shared_params: list
+    ds_version: int
+    frozen_param_shapes: dict()
+    frozen_param_fragments: dict()
+
+
+debug = 0
+
+# load to cpu
+device = torch.device('cpu')
+
+
+def atoi(text):
+    return int(text) if text.isdigit() else text
+
+
+def natural_keys(text):
+    '''
+    alist.sort(key=natural_keys) sorts in human order
+    http://nedbatchelder.com/blog/200712/human_sorting.html
+    (See Toothy's implementation in the comments)
+    '''
+    return [atoi(c) for c in re.split(r'(\d+)', text)]
+
+
+def get_model_state_file(checkpoint_dir, zero_stage):
+    if not os.path.isdir(checkpoint_dir):
+        raise FileNotFoundError(f"Directory '{checkpoint_dir}' doesn't exist")
+
+    # there should be only one file
+    if zero_stage <= 2:
+        file = os.path.join(checkpoint_dir, "mp_rank_00_model_states.pt")
+    elif zero_stage == 3:
+        file = os.path.join(checkpoint_dir, "zero_pp_rank_0_mp_rank_00_model_states.pt")
+
+    if not os.path.exists(file):
+        raise FileNotFoundError(f"can't find model states file at '{file}'")
+
+    return file
+
+
+def get_checkpoint_files(checkpoint_dir, glob_pattern):
+    # XXX: need to test that this simple glob rule works for multi-node setup too
+    ckpt_files = sorted(glob.glob(os.path.join(checkpoint_dir, glob_pattern)), key=natural_keys)
+
+    if len(ckpt_files) == 0:
+        raise FileNotFoundError(f"can't find {glob_pattern} files in directory '{checkpoint_dir}'")
+
+    return ckpt_files
+
+
+def get_optim_files(checkpoint_dir):
+    return get_checkpoint_files(checkpoint_dir, "*_optim_states.pt")
+
+
+def get_model_state_files(checkpoint_dir):
+    return get_checkpoint_files(checkpoint_dir, "*_model_states.pt")
+
+
+def parse_model_states(files):
+    zero_model_states = []
+    for file in files:
+        state_dict = torch.load(file, map_location=device)
+
+        if BUFFER_NAMES not in state_dict:
+            raise ValueError(f"{file} is not a model state checkpoint")
+        buffer_names = state_dict[BUFFER_NAMES]
+        if debug:
+            print("Found buffers:", buffer_names)
+
+        # recover just the buffers while restoring them to fp32 if they were saved in fp16
+        buffers = {k: v.float() for k, v in state_dict["module"].items() if k in buffer_names}
+        param_shapes = state_dict[PARAM_SHAPES]
+
+        # collect parameters that are included in param_shapes
+        param_names = []
+        for s in param_shapes:
+            for name in s.keys():
+                param_names.append(name)
+
+        # update with frozen parameters
+        frozen_param_shapes = state_dict.get(FROZEN_PARAM_SHAPES, None)
+        if frozen_param_shapes is not None:
+            if debug:
+                print(f"Found frozen_param_shapes: {frozen_param_shapes}")
+            param_names += list(frozen_param_shapes.keys())
+
+        # handle shared params
+        shared_params = [[k, v] for k, v in state_dict["shared_params"].items()]
+
+        ds_version = state_dict.get(DS_VERSION, None)
+
+        frozen_param_fragments = state_dict.get(FROZEN_PARAM_FRAGMENTS, None)
+
+        z_model_state = zero_model_state(buffers=buffers,
+                                         param_shapes=param_shapes,
+                                         shared_params=shared_params,
+                                         ds_version=ds_version,
+                                         frozen_param_shapes=frozen_param_shapes,
+                                         frozen_param_fragments=frozen_param_fragments)
+        zero_model_states.append(z_model_state)
+
+    return zero_model_states
+
+
+def parse_optim_states(files, ds_checkpoint_dir):
+
+    total_files = len(files)
+    state_dicts = []
+    for f in files:
+        state_dict = torch.load(f, map_location=device)
+        # immediately discard the potentially huge 2 optimizer states as we only care for fp32 master weights
+        # and also handle the case where it was already removed by another helper script
+        state_dict["optimizer_state_dict"].pop("optimizer_state_dict", None)
+        state_dicts.append(state_dict)
+
+    if not ZERO_STAGE in state_dicts[0][OPTIMIZER_STATE_DICT]:
+        raise ValueError(f"{files[0]} is not a zero checkpoint")
+    zero_stage = state_dicts[0][OPTIMIZER_STATE_DICT][ZERO_STAGE]
+    world_size = state_dicts[0][OPTIMIZER_STATE_DICT][PARTITION_COUNT]
+
+    # For ZeRO-2 each param group can have different partition_count as data parallelism for expert
+    # parameters can be different from data parallelism for non-expert parameters. So we can just
+    # use the max of the partition_count to get the dp world_size.
+
+    if type(world_size) is list:
+        world_size = max(world_size)
+
+    if world_size != total_files:
+        raise ValueError(
+            f"Expected {world_size} of '*_optim_states.pt' under '{ds_checkpoint_dir}' but found {total_files} files. "
+            "Possibly due to an overwrite of an old checkpoint, or a checkpoint didn't get saved by one or more processes."
+        )
+
+    # the groups are named differently in each stage
+    if zero_stage <= 2:
+        fp32_groups_key = SINGLE_PARTITION_OF_FP32_GROUPS
+    elif zero_stage == 3:
+        fp32_groups_key = FP32_FLAT_GROUPS
+    else:
+        raise ValueError(f"unknown zero stage {zero_stage}")
+
+    if zero_stage <= 2:
+        fp32_flat_groups = [state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key] for i in range(len(state_dicts))]
+    elif zero_stage == 3:
+        # if there is more than one param group, there will be multiple flattened tensors - one
+        # flattened tensor per group - for simplicity merge them into a single tensor
+        #
+        # XXX: could make the script more memory efficient for when there are multiple groups - it
+        # will require matching the sub-lists of param_shapes for each param group flattened tensor
+
+        fp32_flat_groups = [
+            torch.cat(state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key], 0) for i in range(len(state_dicts))
+        ]
+
+    return zero_stage, world_size, fp32_flat_groups
+
+
+def _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir, exclude_frozen_parameters):
+    """
+    Returns fp32 state_dict reconstructed from ds checkpoint
+
+    Args:
+        - ``ds_checkpoint_dir``: path to the deepspeed checkpoint folder (where the optimizer files are)
+
+    """
+    print(f"Processing zero checkpoint '{ds_checkpoint_dir}'")
+
+    optim_files = get_optim_files(ds_checkpoint_dir)
+    zero_stage, world_size, fp32_flat_groups = parse_optim_states(optim_files, ds_checkpoint_dir)
+    print(f"Detected checkpoint of type zero stage {zero_stage}, world_size: {world_size}")
+
+    model_files = get_model_state_files(ds_checkpoint_dir)
+
+    zero_model_states = parse_model_states(model_files)
+    print(f'Parsing checkpoint created by deepspeed=={zero_model_states[0].ds_version}')
+
+    if zero_stage <= 2:
+        return _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                                          exclude_frozen_parameters)
+    elif zero_stage == 3:
+        return _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                                          exclude_frozen_parameters)
+
+
+def _zero2_merge_frozen_params(state_dict, zero_model_states):
+    if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
+        return
+
+    frozen_param_shapes = zero_model_states[0].frozen_param_shapes
+    frozen_param_fragments = zero_model_states[0].frozen_param_fragments
+
+    if debug:
+        num_elem = sum(s.numel() for s in frozen_param_shapes.values())
+        print(f'rank 0: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
+
+        wanted_params = len(frozen_param_shapes)
+        wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
+        avail_numel = sum([p.numel() for p in frozen_param_fragments.values()])
+        print(f'Frozen params: Have {avail_numel} numels to process.')
+        print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
+
+    total_params = 0
+    total_numel = 0
+    for name, shape in frozen_param_shapes.items():
+        total_params += 1
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+
+        state_dict[name] = frozen_param_fragments[name]
+
+        if debug:
+            print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
+
+    print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _has_callable(obj, fn):
+    attr = getattr(obj, fn, None)
+    return callable(attr)
+
+
+def _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
+    param_shapes = zero_model_states[0].param_shapes
+
+    # Reconstruction protocol:
+    #
+    # XXX: document this
+
+    if debug:
+        for i in range(world_size):
+            for j in range(len(fp32_flat_groups[0])):
+                print(f"{FP32_FLAT_GROUPS}[{i}][{j}].shape={fp32_flat_groups[i][j].shape}")
+
+    # XXX: memory usage doubles here (zero2)
+    num_param_groups = len(fp32_flat_groups[0])
+    merged_single_partition_of_fp32_groups = []
+    for i in range(num_param_groups):
+        merged_partitions = [sd[i] for sd in fp32_flat_groups]
+        full_single_fp32_vector = torch.cat(merged_partitions, 0)
+        merged_single_partition_of_fp32_groups.append(full_single_fp32_vector)
+    avail_numel = sum(
+        [full_single_fp32_vector.numel() for full_single_fp32_vector in merged_single_partition_of_fp32_groups])
+
+    if debug:
+        wanted_params = sum([len(shapes) for shapes in param_shapes])
+        wanted_numel = sum([sum(shape.numel() for shape in shapes.values()) for shapes in param_shapes])
+        # not asserting if there is a mismatch due to possible padding
+        print(f"Have {avail_numel} numels to process.")
+        print(f"Need {wanted_numel} numels in {wanted_params} params.")
+
+    # params
+    # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
+    # out-of-core computing solution
+    total_numel = 0
+    total_params = 0
+    for shapes, full_single_fp32_vector in zip(param_shapes, merged_single_partition_of_fp32_groups):
+        offset = 0
+        avail_numel = full_single_fp32_vector.numel()
+        for name, shape in shapes.items():
+
+            unpartitioned_numel = shape.numel() if _has_callable(shape, 'numel') else math.prod(shape)
+            total_numel += unpartitioned_numel
+            total_params += 1
+
+            if debug:
+                print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
+            state_dict[name] = full_single_fp32_vector.narrow(0, offset, unpartitioned_numel).view(shape)
+            offset += unpartitioned_numel
+
+        # Z2 started to align to 2*world_size to improve nccl performance. Therefore both offset and
+        # avail_numel can differ by anywhere between 0..2*world_size. Due to two unrelated complex
+        # paddings performed in the code it's almost impossible to predict the exact numbers w/o the
+        # live optimizer object, so we are checking that the numbers are within the right range
+        align_to = 2 * world_size
+
+        def zero2_align(x):
+            return align_to * math.ceil(x / align_to)
+
+        if debug:
+            print(f"original offset={offset}, avail_numel={avail_numel}")
+
+        offset = zero2_align(offset)
+        avail_numel = zero2_align(avail_numel)
+
+        if debug:
+            print(f"aligned  offset={offset}, avail_numel={avail_numel}")
+
+        # Sanity check
+        if offset != avail_numel:
+            raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
+
+    print(f"Reconstructed fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                               exclude_frozen_parameters):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    if not exclude_frozen_parameters:
+        _zero2_merge_frozen_params(state_dict, zero_model_states)
+
+    _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
+
+    # recover shared parameters
+    for pair in zero_model_states[0].shared_params:
+        if pair[1] in state_dict:
+            state_dict[pair[0]] = state_dict[pair[1]]
+
+    return state_dict
+
+
+def zero3_partitioned_param_info(unpartitioned_numel, world_size):
+    remainder = unpartitioned_numel % world_size
+    padding_numel = (world_size - remainder) if remainder else 0
+    partitioned_numel = math.ceil(unpartitioned_numel / world_size)
+    return partitioned_numel, padding_numel
+
+
+def _zero3_merge_frozen_params(state_dict, world_size, zero_model_states):
+    if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
+        return
+
+    if debug:
+        for i in range(world_size):
+            num_elem = sum(s.numel() for s in zero_model_states[i].frozen_param_fragments.values())
+            print(f'rank {i}: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
+
+        frozen_param_shapes = zero_model_states[0].frozen_param_shapes
+        wanted_params = len(frozen_param_shapes)
+        wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
+        avail_numel = sum([p.numel() for p in zero_model_states[0].frozen_param_fragments.values()]) * world_size
+        print(f'Frozen params: Have {avail_numel} numels to process.')
+        print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
+
+    total_params = 0
+    total_numel = 0
+    for name, shape in zero_model_states[0].frozen_param_shapes.items():
+        total_params += 1
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+
+        param_frags = tuple(model_state.frozen_param_fragments[name] for model_state in zero_model_states)
+        state_dict[name] = torch.cat(param_frags, 0).narrow(0, 0, unpartitioned_numel).view(shape)
+
+        partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
+
+        if debug:
+            print(
+                f"Frozen params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
+            )
+
+    print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
+    param_shapes = zero_model_states[0].param_shapes
+    avail_numel = fp32_flat_groups[0].numel() * world_size
+    # Reconstruction protocol: For zero3 we need to zip the partitions together at boundary of each
+    # param, re-consolidating each param, while dealing with padding if any
+
+    # merge list of dicts, preserving order
+    param_shapes = {k: v for d in param_shapes for k, v in d.items()}
+
+    if debug:
+        for i in range(world_size):
+            print(f"{FP32_FLAT_GROUPS}[{i}].shape={fp32_flat_groups[i].shape}")
+
+        wanted_params = len(param_shapes)
+        wanted_numel = sum(shape.numel() for shape in param_shapes.values())
+        # not asserting if there is a mismatch due to possible padding
+        avail_numel = fp32_flat_groups[0].numel() * world_size
+        print(f"Trainable params: Have {avail_numel} numels to process.")
+        print(f"Trainable params: Need {wanted_numel} numels in {wanted_params} params.")
+
+    # params
+    # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
+    # out-of-core computing solution
+    offset = 0
+    total_numel = 0
+    total_params = 0
+    for name, shape in param_shapes.items():
+
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+        total_params += 1
+
+        partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
+
+        if debug:
+            print(
+                f"Trainable params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
+            )
+
+        # XXX: memory usage doubles here
+        state_dict[name] = torch.cat(
+            tuple(fp32_flat_groups[i].narrow(0, offset, partitioned_numel) for i in range(world_size)),
+            0).narrow(0, 0, unpartitioned_numel).view(shape)
+        offset += partitioned_numel
+
+    offset *= world_size
+
+    # Sanity check
+    if offset != avail_numel:
+        raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
+
+    print(f"Reconstructed Trainable fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                               exclude_frozen_parameters):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    if not exclude_frozen_parameters:
+        _zero3_merge_frozen_params(state_dict, world_size, zero_model_states)
+
+    _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
+
+    # recover shared parameters
+    for pair in zero_model_states[0].shared_params:
+        if pair[1] in state_dict:
+            state_dict[pair[0]] = state_dict[pair[1]]
+
+    return state_dict
+
+
+def get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag=None, exclude_frozen_parameters=False):
+    """
+    Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated state_dict that can be loaded with
+    ``load_state_dict()`` and used for training without DeepSpeed or shared with others, for example
+    via a model hub.
+
+    Args:
+        - ``checkpoint_dir``: path to the desired checkpoint folder
+        - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in 'latest' file. e.g., ``global_step14``
+        - ``exclude_frozen_parameters``: exclude frozen parameters
+
+    Returns:
+        - pytorch ``state_dict``
+
+    Note: this approach may not work if your application doesn't have sufficient free CPU memory and
+    you may need to use the offline approach using the ``zero_to_fp32.py`` script that is saved with
+    the checkpoint.
+
+    A typical usage might be ::
+
+        from deepspeed.utils.zero_to_fp32 import get_fp32_state_dict_from_zero_checkpoint
+        # do the training and checkpoint saving
+        state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir) # already on cpu
+        model = model.cpu() # move to cpu
+        model.load_state_dict(state_dict)
+        # submit to model hub or save the model to share with others
+
+    In this example the ``model`` will no longer be usable in the deepspeed context of the same
+    application. i.e. you will need to re-initialize the deepspeed engine, since
+    ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
+
+    If you want it all done for you, use ``load_state_dict_from_zero_checkpoint`` instead.
+
+    """
+    if tag is None:
+        latest_path = os.path.join(checkpoint_dir, 'latest')
+        if os.path.isfile(latest_path):
+            with open(latest_path, 'r') as fd:
+                tag = fd.read().strip()
+        else:
+            raise ValueError(f"Unable to find 'latest' file at {latest_path}")
+
+    ds_checkpoint_dir = os.path.join(checkpoint_dir, tag)
+
+    if not os.path.isdir(ds_checkpoint_dir):
+        raise FileNotFoundError(f"Directory '{ds_checkpoint_dir}' doesn't exist")
+
+    return _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir, exclude_frozen_parameters)
+
+
+def convert_zero_checkpoint_to_fp32_state_dict(checkpoint_dir, output_file, tag=None, exclude_frozen_parameters=False):
+    """
+    Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict`` file that can be
+    loaded with ``torch.load(file)`` + ``load_state_dict()`` and used for training without DeepSpeed.
+
+    Args:
+        - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
+        - ``output_file``: path to the pytorch fp32 state_dict output file (e.g. path/pytorch_model.bin)
+        - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
+        - ``exclude_frozen_parameters``: exclude frozen parameters
+    """
+
+    state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag, exclude_frozen_parameters)
+    print(f"Saving fp32 state dict to {output_file}")
+    torch.save(state_dict, output_file)
+
+
+def load_state_dict_from_zero_checkpoint(model, checkpoint_dir, tag=None):
+    """
+    1. Put the provided model to cpu
+    2. Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict``
+    3. Load it into the provided model
+
+    Args:
+        - ``model``: the model object to update
+        - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
+        - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
+
+    Returns:
+        - ``model`: modified model
+
+    Make sure you have plenty of CPU memory available before you call this function. If you don't
+    have enough use the ``zero_to_fp32.py`` utility to do the conversion. You will find it
+    conveniently placed for you in the checkpoint folder.
+
+    A typical usage might be ::
+
+        from deepspeed.utils.zero_to_fp32 import load_state_dict_from_zero_checkpoint
+        model = load_state_dict_from_zero_checkpoint(trainer.model, checkpoint_dir)
+        # submit to model hub or save the model to share with others
+
+    Note, that once this was run, the ``model`` will no longer be usable in the deepspeed context
+    of the same application. i.e. you will need to re-initialize the deepspeed engine, since
+    ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
+
+    """
+    logger.info(f"Extracting fp32 weights")
+    state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag)
+
+    logger.info(f"Overwriting model with fp32 weights")
+    model = model.cpu()
+    model.load_state_dict(state_dict, strict=False)
+
+    return model
+
+
+if __name__ == "__main__":
+
+    parser = argparse.ArgumentParser()
+    parser.add_argument("checkpoint_dir",
+                        type=str,
+                        help="path to the desired checkpoint folder, e.g., path/checkpoint-12")
+    parser.add_argument(
+        "output_file",
+        type=str,
+        help="path to the pytorch fp32 state_dict output file (e.g. path/checkpoint-12/pytorch_model.bin)")
+    parser.add_argument("-t",
+                        "--tag",
+                        type=str,
+                        default=None,
+                        help="checkpoint tag used as a unique identifier for checkpoint. e.g., global_step1")
+    parser.add_argument("--exclude_frozen_parameters", action='store_true', help="exclude frozen parameters")
+    parser.add_argument("-d", "--debug", action='store_true', help="enable debug")
+    args = parser.parse_args()
+
+    debug = args.debug
+
+    convert_zero_checkpoint_to_fp32_state_dict(args.checkpoint_dir,
+                                               args.output_file,
+                                               tag=args.tag,
+                                               exclude_frozen_parameters=args.exclude_frozen_parameters)
diff --git a/weight_dir/loraWeight/trace_CRFLP/checkpoint-14000/README.md b/weight_dir/loraWeight/trace_CRFLP/checkpoint-14000/README.md
new file mode 100644
index 0000000000000000000000000000000000000000..105007ea041cb936ba3417e0221e47fec447213d
--- /dev/null
+++ b/weight_dir/loraWeight/trace_CRFLP/checkpoint-14000/README.md
@@ -0,0 +1,202 @@
+---
+library_name: peft
+base_model: ./CodeLlama-7b-Instruct-hf
+---
+
+# Model Card for Model ID
+
+<!-- Provide a quick summary of what the model is/does. -->
+
+
+
+## Model Details
+
+### Model Description
+
+<!-- Provide a longer summary of what this model is. -->
+
+
+
+- **Developed by:** [More Information Needed]
+- **Funded by [optional]:** [More Information Needed]
+- **Shared by [optional]:** [More Information Needed]
+- **Model type:** [More Information Needed]
+- **Language(s) (NLP):** [More Information Needed]
+- **License:** [More Information Needed]
+- **Finetuned from model [optional]:** [More Information Needed]
+
+### Model Sources [optional]
+
+<!-- Provide the basic links for the model. -->
+
+- **Repository:** [More Information Needed]
+- **Paper [optional]:** [More Information Needed]
+- **Demo [optional]:** [More Information Needed]
+
+## Uses
+
+<!-- Address questions around how the model is intended to be used, including the foreseeable users of the model and those affected by the model. -->
+
+### Direct Use
+
+<!-- This section is for the model use without fine-tuning or plugging into a larger ecosystem/app. -->
+
+[More Information Needed]
+
+### Downstream Use [optional]
+
+<!-- This section is for the model use when fine-tuned for a task, or when plugged into a larger ecosystem/app -->
+
+[More Information Needed]
+
+### Out-of-Scope Use
+
+<!-- This section addresses misuse, malicious use, and uses that the model will not work well for. -->
+
+[More Information Needed]
+
+## Bias, Risks, and Limitations
+
+<!-- This section is meant to convey both technical and sociotechnical limitations. -->
+
+[More Information Needed]
+
+### Recommendations
+
+<!-- This section is meant to convey recommendations with respect to the bias, risk, and technical limitations. -->
+
+Users (both direct and downstream) should be made aware of the risks, biases and limitations of the model. More information needed for further recommendations.
+
+## How to Get Started with the Model
+
+Use the code below to get started with the model.
+
+[More Information Needed]
+
+## Training Details
+
+### Training Data
+
+<!-- This should link to a Dataset Card, perhaps with a short stub of information on what the training data is all about as well as documentation related to data pre-processing or additional filtering. -->
+
+[More Information Needed]
+
+### Training Procedure
+
+<!-- This relates heavily to the Technical Specifications. Content here should link to that section when it is relevant to the training procedure. -->
+
+#### Preprocessing [optional]
+
+[More Information Needed]
+
+
+#### Training Hyperparameters
+
+- **Training regime:** [More Information Needed] <!--fp32, fp16 mixed precision, bf16 mixed precision, bf16 non-mixed precision, fp16 non-mixed precision, fp8 mixed precision -->
+
+#### Speeds, Sizes, Times [optional]
+
+<!-- This section provides information about throughput, start/end time, checkpoint size if relevant, etc. -->
+
+[More Information Needed]
+
+## Evaluation
+
+<!-- This section describes the evaluation protocols and provides the results. -->
+
+### Testing Data, Factors & Metrics
+
+#### Testing Data
+
+<!-- This should link to a Dataset Card if possible. -->
+
+[More Information Needed]
+
+#### Factors
+
+<!-- These are the things the evaluation is disaggregating by, e.g., subpopulations or domains. -->
+
+[More Information Needed]
+
+#### Metrics
+
+<!-- These are the evaluation metrics being used, ideally with a description of why. -->
+
+[More Information Needed]
+
+### Results
+
+[More Information Needed]
+
+#### Summary
+
+
+
+## Model Examination [optional]
+
+<!-- Relevant interpretability work for the model goes here -->
+
+[More Information Needed]
+
+## Environmental Impact
+
+<!-- Total emissions (in grams of CO2eq) and additional considerations, such as electricity usage, go here. Edit the suggested text below accordingly -->
+
+Carbon emissions can be estimated using the [Machine Learning Impact calculator](https://mlco2.github.io/impact#compute) presented in [Lacoste et al. (2019)](https://arxiv.org/abs/1910.09700).
+
+- **Hardware Type:** [More Information Needed]
+- **Hours used:** [More Information Needed]
+- **Cloud Provider:** [More Information Needed]
+- **Compute Region:** [More Information Needed]
+- **Carbon Emitted:** [More Information Needed]
+
+## Technical Specifications [optional]
+
+### Model Architecture and Objective
+
+[More Information Needed]
+
+### Compute Infrastructure
+
+[More Information Needed]
+
+#### Hardware
+
+[More Information Needed]
+
+#### Software
+
+[More Information Needed]
+
+## Citation [optional]
+
+<!-- If there is a paper or blog post introducing the model, the APA and Bibtex information for that should go in this section. -->
+
+**BibTeX:**
+
+[More Information Needed]
+
+**APA:**
+
+[More Information Needed]
+
+## Glossary [optional]
+
+<!-- If relevant, include terms and calculations in this section that can help readers understand the model or model card. -->
+
+[More Information Needed]
+
+## More Information [optional]
+
+[More Information Needed]
+
+## Model Card Authors [optional]
+
+[More Information Needed]
+
+## Model Card Contact
+
+[More Information Needed]
+### Framework versions
+
+- PEFT 0.11.1
\ No newline at end of file
diff --git a/weight_dir/loraWeight/trace_CRFLP/checkpoint-14000/adapter_config.json b/weight_dir/loraWeight/trace_CRFLP/checkpoint-14000/adapter_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..8faaa17fff7e4931b18e07a687c5b35532df41ba
--- /dev/null
+++ b/weight_dir/loraWeight/trace_CRFLP/checkpoint-14000/adapter_config.json
@@ -0,0 +1,29 @@
+{
+  "alpha_pattern": {},
+  "auto_mapping": null,
+  "base_model_name_or_path": "./CodeLlama-7b-Instruct-hf",
+  "bias": "none",
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 32,
+  "lora_dropout": 0.05,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "r": 8,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "q_proj",
+    "v_proj"
+  ],
+  "task_type": "CAUSAL_LM",
+  "use_dora": false,
+  "use_rslora": false
+}
\ No newline at end of file
diff --git a/weight_dir/loraWeight/trace_CRFLP/checkpoint-14000/adapter_model.bin b/weight_dir/loraWeight/trace_CRFLP/checkpoint-14000/adapter_model.bin
new file mode 100644
index 0000000000000000000000000000000000000000..857c1c105ac4f72f96854ec22c36aafff1b45a68
--- /dev/null
+++ b/weight_dir/loraWeight/trace_CRFLP/checkpoint-14000/adapter_model.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:92b61741be33f127dc45b0835fc9487a3d9fd4d81c973f690ae754f9f5970358
+size 8433034
diff --git a/weight_dir/loraWeight/trace_CRFLP/checkpoint-14000/added_tokens.json b/weight_dir/loraWeight/trace_CRFLP/checkpoint-14000/added_tokens.json
new file mode 100644
index 0000000000000000000000000000000000000000..c09203d8a52151e0d3b5f3c8e6daedc5b60832b5
--- /dev/null
+++ b/weight_dir/loraWeight/trace_CRFLP/checkpoint-14000/added_tokens.json
@@ -0,0 +1,3 @@
+{
+  "<PAD>": 32016
+}
diff --git a/weight_dir/loraWeight/trace_CRFLP/checkpoint-14000/global_step14000/bf16_zero_pp_rank_0_mp_rank_00_optim_states.pt b/weight_dir/loraWeight/trace_CRFLP/checkpoint-14000/global_step14000/bf16_zero_pp_rank_0_mp_rank_00_optim_states.pt
new file mode 100644
index 0000000000000000000000000000000000000000..ef6bcb8be556637534dcd5e8efa3f45bb7d3ca8c
--- /dev/null
+++ b/weight_dir/loraWeight/trace_CRFLP/checkpoint-14000/global_step14000/bf16_zero_pp_rank_0_mp_rank_00_optim_states.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:32828f5ce5f8cae6fcaf1c5172803670e5383b0f2d0f1dce46aaa10ee675d6f3
+size 12589776
diff --git a/weight_dir/loraWeight/trace_CRFLP/checkpoint-14000/global_step14000/bf16_zero_pp_rank_1_mp_rank_00_optim_states.pt b/weight_dir/loraWeight/trace_CRFLP/checkpoint-14000/global_step14000/bf16_zero_pp_rank_1_mp_rank_00_optim_states.pt
new file mode 100644
index 0000000000000000000000000000000000000000..8170bab8b8392111e5c9ee4f0cfd0d9165c4b530
--- /dev/null
+++ b/weight_dir/loraWeight/trace_CRFLP/checkpoint-14000/global_step14000/bf16_zero_pp_rank_1_mp_rank_00_optim_states.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3ca2a0efc82886db414fdbe10d8c18670579afc50c90890c7f4cdb96a75bbaa9
+size 12589840
diff --git a/weight_dir/loraWeight/trace_CRFLP/checkpoint-14000/global_step14000/bf16_zero_pp_rank_2_mp_rank_00_optim_states.pt b/weight_dir/loraWeight/trace_CRFLP/checkpoint-14000/global_step14000/bf16_zero_pp_rank_2_mp_rank_00_optim_states.pt
new file mode 100644
index 0000000000000000000000000000000000000000..eff3db739771be39b601e02258c877472d5d225e
--- /dev/null
+++ b/weight_dir/loraWeight/trace_CRFLP/checkpoint-14000/global_step14000/bf16_zero_pp_rank_2_mp_rank_00_optim_states.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fdd925adacf2b79ffe5ce595f3ea49a122305d633c40582d3227f718bf6f7379
+size 12589840
diff --git a/weight_dir/loraWeight/trace_CRFLP/checkpoint-14000/global_step14000/bf16_zero_pp_rank_3_mp_rank_00_optim_states.pt b/weight_dir/loraWeight/trace_CRFLP/checkpoint-14000/global_step14000/bf16_zero_pp_rank_3_mp_rank_00_optim_states.pt
new file mode 100644
index 0000000000000000000000000000000000000000..32a669388715c1702d1ca98d33212db00abb1d47
--- /dev/null
+++ b/weight_dir/loraWeight/trace_CRFLP/checkpoint-14000/global_step14000/bf16_zero_pp_rank_3_mp_rank_00_optim_states.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:769c31e8d696b8e65487981efa5fb40772302293395bc9fe924dd3c381e0617e
+size 12589840
diff --git a/weight_dir/loraWeight/trace_CRFLP/checkpoint-14000/global_step14000/mp_rank_00_model_states.pt b/weight_dir/loraWeight/trace_CRFLP/checkpoint-14000/global_step14000/mp_rank_00_model_states.pt
new file mode 100644
index 0000000000000000000000000000000000000000..b7c00ab989b447e9bc3ee447e709da270a8a503a
--- /dev/null
+++ b/weight_dir/loraWeight/trace_CRFLP/checkpoint-14000/global_step14000/mp_rank_00_model_states.pt
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3c2dbdc41e44c7d5bd97e9d4b89d311ee8cfe727757664b9c21ce6575ebde25e
+size 8507372
diff --git a/weight_dir/loraWeight/trace_CRFLP/checkpoint-14000/latest b/weight_dir/loraWeight/trace_CRFLP/checkpoint-14000/latest
new file mode 100644
index 0000000000000000000000000000000000000000..6c2e99ab39c9aa1d68dfdcb53dc9234015ced784
--- /dev/null
+++ b/weight_dir/loraWeight/trace_CRFLP/checkpoint-14000/latest
@@ -0,0 +1 @@
+global_step14000
\ No newline at end of file
diff --git a/weight_dir/loraWeight/trace_CRFLP/checkpoint-14000/rng_state_0.pth b/weight_dir/loraWeight/trace_CRFLP/checkpoint-14000/rng_state_0.pth
new file mode 100644
index 0000000000000000000000000000000000000000..599100e3930d98f23dcee3039f49fc7e796eada5
--- /dev/null
+++ b/weight_dir/loraWeight/trace_CRFLP/checkpoint-14000/rng_state_0.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:90b3271c0444170435af26350591421c00cc6bd21b5ec3df060d909eef6cc051
+size 15024
diff --git a/weight_dir/loraWeight/trace_CRFLP/checkpoint-14000/rng_state_1.pth b/weight_dir/loraWeight/trace_CRFLP/checkpoint-14000/rng_state_1.pth
new file mode 100644
index 0000000000000000000000000000000000000000..8ba5a14a6ecfa06625e1e847d40abbf26048166f
--- /dev/null
+++ b/weight_dir/loraWeight/trace_CRFLP/checkpoint-14000/rng_state_1.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:96bf15fe1750e738a04077747729607b44e6e60417bf57c69b48eb7a1ef1b1ce
+size 15024
diff --git a/weight_dir/loraWeight/trace_CRFLP/checkpoint-14000/rng_state_2.pth b/weight_dir/loraWeight/trace_CRFLP/checkpoint-14000/rng_state_2.pth
new file mode 100644
index 0000000000000000000000000000000000000000..9d55afb1e036f6f54b56376581552855227b0ca7
--- /dev/null
+++ b/weight_dir/loraWeight/trace_CRFLP/checkpoint-14000/rng_state_2.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:f80ea4001e5074a6810f4f44445e34aa07b8939bbd17d68e4038293816eaf9b7
+size 15024
diff --git a/weight_dir/loraWeight/trace_CRFLP/checkpoint-14000/rng_state_3.pth b/weight_dir/loraWeight/trace_CRFLP/checkpoint-14000/rng_state_3.pth
new file mode 100644
index 0000000000000000000000000000000000000000..21fb04c997c4e59c72b1a4fa0a79fcae061cbc24
--- /dev/null
+++ b/weight_dir/loraWeight/trace_CRFLP/checkpoint-14000/rng_state_3.pth
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:5373b0aabd34513e5bbf7fa26b5cee66964ccb17e486af249c2762b2762f9870
+size 15024
diff --git a/weight_dir/loraWeight/trace_CRFLP/checkpoint-14000/special_tokens_map.json b/weight_dir/loraWeight/trace_CRFLP/checkpoint-14000/special_tokens_map.json
new file mode 100644
index 0000000000000000000000000000000000000000..2ece23e993f7e1c7063cb51148b6fa5c6c224775
--- /dev/null
+++ b/weight_dir/loraWeight/trace_CRFLP/checkpoint-14000/special_tokens_map.json
@@ -0,0 +1,30 @@
+{
+  "bos_token": {
+    "content": "<s>",
+    "lstrip": false,
+    "normalized": true,
+    "rstrip": false,
+    "single_word": false
+  },
+  "eos_token": {
+    "content": "</s>",
+    "lstrip": false,
+    "normalized": true,
+    "rstrip": false,
+    "single_word": false
+  },
+  "pad_token": {
+    "content": "<PAD>",
+    "lstrip": false,
+    "normalized": false,
+    "rstrip": false,
+    "single_word": false
+  },
+  "unk_token": {
+    "content": "<unk>",
+    "lstrip": false,
+    "normalized": true,
+    "rstrip": false,
+    "single_word": false
+  }
+}
diff --git a/weight_dir/loraWeight/trace_CRFLP/checkpoint-14000/tokenizer.model b/weight_dir/loraWeight/trace_CRFLP/checkpoint-14000/tokenizer.model
new file mode 100644
index 0000000000000000000000000000000000000000..f6722e8b170230ebdd4c0f5f2ce03f219be536d4
--- /dev/null
+++ b/weight_dir/loraWeight/trace_CRFLP/checkpoint-14000/tokenizer.model
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:45ccb9c8b6b561889acea59191d66986d314e7cbd6a78abc6e49b139ca91c1e6
+size 500058
diff --git a/weight_dir/loraWeight/trace_CRFLP/checkpoint-14000/tokenizer_config.json b/weight_dir/loraWeight/trace_CRFLP/checkpoint-14000/tokenizer_config.json
new file mode 100644
index 0000000000000000000000000000000000000000..dc4d9a55d56b17600905bfc69a0e45e2fae7a1c4
--- /dev/null
+++ b/weight_dir/loraWeight/trace_CRFLP/checkpoint-14000/tokenizer_config.json
@@ -0,0 +1,50 @@
+{
+  "add_bos_token": true,
+  "add_eos_token": false,
+  "add_prefix_space": true,
+  "added_tokens_decoder": {
+    "0": {
+      "content": "<unk>",
+      "lstrip": false,
+      "normalized": true,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "1": {
+      "content": "<s>",
+      "lstrip": false,
+      "normalized": true,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "2": {
+      "content": "</s>",
+      "lstrip": false,
+      "normalized": true,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    },
+    "32016": {
+      "content": "<PAD>",
+      "lstrip": false,
+      "normalized": false,
+      "rstrip": false,
+      "single_word": false,
+      "special": true
+    }
+  },
+  "bos_token": "<s>",
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "</s>",
+  "legacy": true,
+  "model_max_length": 1000000000000000019884624838656,
+  "pad_token": "<PAD>",
+  "sp_model_kwargs": {},
+  "spaces_between_special_tokens": false,
+  "tokenizer_class": "LlamaTokenizer",
+  "unk_token": "<unk>",
+  "use_default_system_prompt": false
+}
diff --git a/weight_dir/loraWeight/trace_CRFLP/checkpoint-14000/trainer_state.json b/weight_dir/loraWeight/trace_CRFLP/checkpoint-14000/trainer_state.json
new file mode 100644
index 0000000000000000000000000000000000000000..464d40731e3f30cd88619085d0d10d4f75ca60c8
--- /dev/null
+++ b/weight_dir/loraWeight/trace_CRFLP/checkpoint-14000/trainer_state.json
@@ -0,0 +1,2009 @@
+{
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 4.477134633834346,
+  "eval_steps": 6000,
+  "global_step": 14000,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "epoch": 0.01598976654940838,
+      "grad_norm": 0.4070083498954773,
+      "learning_rate": 4.992484329026481e-05,
+      "loss": 4.273,
+      "step": 50
+    },
+    {
+      "epoch": 0.03197953309881676,
+      "grad_norm": 0.057464830577373505,
+      "learning_rate": 4.984488934373801e-05,
+      "loss": 0.0247,
+      "step": 100
+    },
+    {
+      "epoch": 0.04796929964822513,
+      "grad_norm": 0.06848371773958206,
+      "learning_rate": 4.976493539721121e-05,
+      "loss": 0.0152,
+      "step": 150
+    },
+    {
+      "epoch": 0.06395906619763352,
+      "grad_norm": 0.042949117720127106,
+      "learning_rate": 4.968498145068441e-05,
+      "loss": 0.0142,
+      "step": 200
+    },
+    {
+      "epoch": 0.0799488327470419,
+      "grad_norm": 0.05076887458562851,
+      "learning_rate": 4.9605027504157606e-05,
+      "loss": 0.0138,
+      "step": 250
+    },
+    {
+      "epoch": 0.09593859929645027,
+      "grad_norm": 0.04695209115743637,
+      "learning_rate": 4.952507355763081e-05,
+      "loss": 0.014,
+      "step": 300
+    },
+    {
+      "epoch": 0.11192836584585865,
+      "grad_norm": 0.05572609230875969,
+      "learning_rate": 4.9445119611104006e-05,
+      "loss": 0.0126,
+      "step": 350
+    },
+    {
+      "epoch": 0.12791813239526703,
+      "grad_norm": 0.07735384255647659,
+      "learning_rate": 4.936516566457721e-05,
+      "loss": 0.0126,
+      "step": 400
+    },
+    {
+      "epoch": 0.1439078989446754,
+      "grad_norm": 0.04669607803225517,
+      "learning_rate": 4.9285211718050406e-05,
+      "loss": 0.0123,
+      "step": 450
+    },
+    {
+      "epoch": 0.1598976654940838,
+      "grad_norm": 0.041628964245319366,
+      "learning_rate": 4.92052577715236e-05,
+      "loss": 0.0116,
+      "step": 500
+    },
+    {
+      "epoch": 0.17588743204349216,
+      "grad_norm": 0.051499415189027786,
+      "learning_rate": 4.9125303824996806e-05,
+      "loss": 0.0111,
+      "step": 550
+    },
+    {
+      "epoch": 0.19187719859290053,
+      "grad_norm": 0.055776361376047134,
+      "learning_rate": 4.904534987847e-05,
+      "loss": 0.0113,
+      "step": 600
+    },
+    {
+      "epoch": 0.20786696514230893,
+      "grad_norm": 0.04950210452079773,
+      "learning_rate": 4.8965395931943206e-05,
+      "loss": 0.0111,
+      "step": 650
+    },
+    {
+      "epoch": 0.2238567316917173,
+      "grad_norm": 0.04448186606168747,
+      "learning_rate": 4.88854419854164e-05,
+      "loss": 0.0109,
+      "step": 700
+    },
+    {
+      "epoch": 0.23984649824112567,
+      "grad_norm": 0.04934769123792648,
+      "learning_rate": 4.88054880388896e-05,
+      "loss": 0.012,
+      "step": 750
+    },
+    {
+      "epoch": 0.25583626479053406,
+      "grad_norm": 0.05481697991490364,
+      "learning_rate": 4.87255340923628e-05,
+      "loss": 0.0113,
+      "step": 800
+    },
+    {
+      "epoch": 0.27182603133994243,
+      "grad_norm": 0.07512317597866058,
+      "learning_rate": 4.8645580145836e-05,
+      "loss": 0.0108,
+      "step": 850
+    },
+    {
+      "epoch": 0.2878157978893508,
+      "grad_norm": 0.0495820976793766,
+      "learning_rate": 4.85656261993092e-05,
+      "loss": 0.0106,
+      "step": 900
+    },
+    {
+      "epoch": 0.30380556443875917,
+      "grad_norm": 0.0660017803311348,
+      "learning_rate": 4.84856722527824e-05,
+      "loss": 0.0107,
+      "step": 950
+    },
+    {
+      "epoch": 0.3197953309881676,
+      "grad_norm": 0.043546389788389206,
+      "learning_rate": 4.8405718306255596e-05,
+      "loss": 0.0101,
+      "step": 1000
+    },
+    {
+      "epoch": 0.33578509753757596,
+      "grad_norm": 0.046335190534591675,
+      "learning_rate": 4.83257643597288e-05,
+      "loss": 0.0105,
+      "step": 1050
+    },
+    {
+      "epoch": 0.35177486408698433,
+      "grad_norm": 0.12547409534454346,
+      "learning_rate": 4.8245810413201996e-05,
+      "loss": 0.0107,
+      "step": 1100
+    },
+    {
+      "epoch": 0.3677646306363927,
+      "grad_norm": 0.04085518419742584,
+      "learning_rate": 4.81658564666752e-05,
+      "loss": 0.0102,
+      "step": 1150
+    },
+    {
+      "epoch": 0.38375439718580107,
+      "grad_norm": 0.05395630747079849,
+      "learning_rate": 4.8085902520148396e-05,
+      "loss": 0.0102,
+      "step": 1200
+    },
+    {
+      "epoch": 0.3997441637352095,
+      "grad_norm": 0.11172062903642654,
+      "learning_rate": 4.800594857362159e-05,
+      "loss": 0.0104,
+      "step": 1250
+    },
+    {
+      "epoch": 0.41573393028461786,
+      "grad_norm": 0.059676576405763626,
+      "learning_rate": 4.7925994627094797e-05,
+      "loss": 0.0096,
+      "step": 1300
+    },
+    {
+      "epoch": 0.4317236968340262,
+      "grad_norm": 0.06800328195095062,
+      "learning_rate": 4.784604068056799e-05,
+      "loss": 0.0099,
+      "step": 1350
+    },
+    {
+      "epoch": 0.4477134633834346,
+      "grad_norm": 0.031088583171367645,
+      "learning_rate": 4.77660867340412e-05,
+      "loss": 0.01,
+      "step": 1400
+    },
+    {
+      "epoch": 0.46370322993284296,
+      "grad_norm": 0.05977204814553261,
+      "learning_rate": 4.768613278751439e-05,
+      "loss": 0.0098,
+      "step": 1450
+    },
+    {
+      "epoch": 0.47969299648225133,
+      "grad_norm": 0.0633067712187767,
+      "learning_rate": 4.760617884098759e-05,
+      "loss": 0.011,
+      "step": 1500
+    },
+    {
+      "epoch": 0.49568276303165976,
+      "grad_norm": 0.10753436386585236,
+      "learning_rate": 4.752622489446079e-05,
+      "loss": 0.0102,
+      "step": 1550
+    },
+    {
+      "epoch": 0.5116725295810681,
+      "grad_norm": 0.04346701502799988,
+      "learning_rate": 4.744627094793399e-05,
+      "loss": 0.0099,
+      "step": 1600
+    },
+    {
+      "epoch": 0.5276622961304765,
+      "grad_norm": 0.08080250024795532,
+      "learning_rate": 4.736631700140719e-05,
+      "loss": 0.0098,
+      "step": 1650
+    },
+    {
+      "epoch": 0.5436520626798849,
+      "grad_norm": 0.04791934788227081,
+      "learning_rate": 4.728636305488039e-05,
+      "loss": 0.0094,
+      "step": 1700
+    },
+    {
+      "epoch": 0.5596418292292933,
+      "grad_norm": 0.07470395416021347,
+      "learning_rate": 4.720640910835359e-05,
+      "loss": 0.0097,
+      "step": 1750
+    },
+    {
+      "epoch": 0.5756315957787016,
+      "grad_norm": 0.030010297894477844,
+      "learning_rate": 4.712645516182679e-05,
+      "loss": 0.0103,
+      "step": 1800
+    },
+    {
+      "epoch": 0.59162136232811,
+      "grad_norm": 0.06268253177404404,
+      "learning_rate": 4.704650121529999e-05,
+      "loss": 0.0102,
+      "step": 1850
+    },
+    {
+      "epoch": 0.6076111288775183,
+      "grad_norm": 0.04115760698914528,
+      "learning_rate": 4.6966547268773183e-05,
+      "loss": 0.01,
+      "step": 1900
+    },
+    {
+      "epoch": 0.6236008954269268,
+      "grad_norm": 0.0384821742773056,
+      "learning_rate": 4.688659332224639e-05,
+      "loss": 0.0096,
+      "step": 1950
+    },
+    {
+      "epoch": 0.6395906619763352,
+      "grad_norm": 0.05536011606454849,
+      "learning_rate": 4.6806639375719584e-05,
+      "loss": 0.0099,
+      "step": 2000
+    },
+    {
+      "epoch": 0.6555804285257435,
+      "grad_norm": 0.042221494019031525,
+      "learning_rate": 4.672668542919279e-05,
+      "loss": 0.0098,
+      "step": 2050
+    },
+    {
+      "epoch": 0.6715701950751519,
+      "grad_norm": 0.05251794680953026,
+      "learning_rate": 4.6646731482665984e-05,
+      "loss": 0.0093,
+      "step": 2100
+    },
+    {
+      "epoch": 0.6875599616245602,
+      "grad_norm": 0.057784657925367355,
+      "learning_rate": 4.656677753613918e-05,
+      "loss": 0.0103,
+      "step": 2150
+    },
+    {
+      "epoch": 0.7035497281739687,
+      "grad_norm": 0.056001678109169006,
+      "learning_rate": 4.648682358961239e-05,
+      "loss": 0.0099,
+      "step": 2200
+    },
+    {
+      "epoch": 0.7195394947233771,
+      "grad_norm": 0.0560077428817749,
+      "learning_rate": 4.640686964308559e-05,
+      "loss": 0.0096,
+      "step": 2250
+    },
+    {
+      "epoch": 0.7355292612727854,
+      "grad_norm": 0.04319921135902405,
+      "learning_rate": 4.6326915696558784e-05,
+      "loss": 0.0093,
+      "step": 2300
+    },
+    {
+      "epoch": 0.7515190278221938,
+      "grad_norm": 0.059632595628499985,
+      "learning_rate": 4.624696175003199e-05,
+      "loss": 0.01,
+      "step": 2350
+    },
+    {
+      "epoch": 0.7675087943716021,
+      "grad_norm": 0.0601465068757534,
+      "learning_rate": 4.6167007803505184e-05,
+      "loss": 0.0098,
+      "step": 2400
+    },
+    {
+      "epoch": 0.7834985609210106,
+      "grad_norm": 0.03820424899458885,
+      "learning_rate": 4.608705385697839e-05,
+      "loss": 0.0101,
+      "step": 2450
+    },
+    {
+      "epoch": 0.799488327470419,
+      "grad_norm": 0.04104379937052727,
+      "learning_rate": 4.6007099910451584e-05,
+      "loss": 0.0092,
+      "step": 2500
+    },
+    {
+      "epoch": 0.8154780940198273,
+      "grad_norm": 0.033697500824928284,
+      "learning_rate": 4.592714596392478e-05,
+      "loss": 0.0097,
+      "step": 2550
+    },
+    {
+      "epoch": 0.8314678605692357,
+      "grad_norm": 0.10835737735033035,
+      "learning_rate": 4.5847192017397984e-05,
+      "loss": 0.0094,
+      "step": 2600
+    },
+    {
+      "epoch": 0.847457627118644,
+      "grad_norm": 0.07756606489419937,
+      "learning_rate": 4.576723807087118e-05,
+      "loss": 0.01,
+      "step": 2650
+    },
+    {
+      "epoch": 0.8634473936680525,
+      "grad_norm": 0.036015670746564865,
+      "learning_rate": 4.5687284124344384e-05,
+      "loss": 0.0097,
+      "step": 2700
+    },
+    {
+      "epoch": 0.8794371602174609,
+      "grad_norm": 0.033497828990221024,
+      "learning_rate": 4.560733017781758e-05,
+      "loss": 0.0094,
+      "step": 2750
+    },
+    {
+      "epoch": 0.8954269267668692,
+      "grad_norm": 0.046858739107847214,
+      "learning_rate": 4.552737623129078e-05,
+      "loss": 0.0096,
+      "step": 2800
+    },
+    {
+      "epoch": 0.9114166933162776,
+      "grad_norm": 0.038639482110738754,
+      "learning_rate": 4.544742228476398e-05,
+      "loss": 0.0093,
+      "step": 2850
+    },
+    {
+      "epoch": 0.9274064598656859,
+      "grad_norm": 0.046789348125457764,
+      "learning_rate": 4.536746833823718e-05,
+      "loss": 0.0092,
+      "step": 2900
+    },
+    {
+      "epoch": 0.9433962264150944,
+      "grad_norm": 0.05411810427904129,
+      "learning_rate": 4.528751439171038e-05,
+      "loss": 0.0095,
+      "step": 2950
+    },
+    {
+      "epoch": 0.9593859929645027,
+      "grad_norm": 0.04174494370818138,
+      "learning_rate": 4.520756044518358e-05,
+      "loss": 0.0097,
+      "step": 3000
+    },
+    {
+      "epoch": 0.9753757595139111,
+      "grad_norm": 0.02797970548272133,
+      "learning_rate": 4.5127606498656774e-05,
+      "loss": 0.0089,
+      "step": 3050
+    },
+    {
+      "epoch": 0.9913655260633195,
+      "grad_norm": 0.050330642610788345,
+      "learning_rate": 4.504765255212998e-05,
+      "loss": 0.0097,
+      "step": 3100
+    },
+    {
+      "epoch": 1.0073552926127278,
+      "grad_norm": 0.03967903554439545,
+      "learning_rate": 4.4967698605603174e-05,
+      "loss": 0.0091,
+      "step": 3150
+    },
+    {
+      "epoch": 1.0233450591621363,
+      "grad_norm": 0.052495893090963364,
+      "learning_rate": 4.488774465907638e-05,
+      "loss": 0.0095,
+      "step": 3200
+    },
+    {
+      "epoch": 1.0393348257115447,
+      "grad_norm": 0.05642736330628395,
+      "learning_rate": 4.4807790712549574e-05,
+      "loss": 0.0085,
+      "step": 3250
+    },
+    {
+      "epoch": 1.0553245922609529,
+      "grad_norm": 0.04831695184111595,
+      "learning_rate": 4.472783676602277e-05,
+      "loss": 0.0094,
+      "step": 3300
+    },
+    {
+      "epoch": 1.0713143588103613,
+      "grad_norm": 0.13986164331436157,
+      "learning_rate": 4.4647882819495975e-05,
+      "loss": 0.009,
+      "step": 3350
+    },
+    {
+      "epoch": 1.0873041253597697,
+      "grad_norm": 0.04970436543226242,
+      "learning_rate": 4.456792887296917e-05,
+      "loss": 0.0097,
+      "step": 3400
+    },
+    {
+      "epoch": 1.1032938919091781,
+      "grad_norm": 0.054993029683828354,
+      "learning_rate": 4.4487974926442375e-05,
+      "loss": 0.01,
+      "step": 3450
+    },
+    {
+      "epoch": 1.1192836584585866,
+      "grad_norm": 0.0306412186473608,
+      "learning_rate": 4.440802097991557e-05,
+      "loss": 0.0093,
+      "step": 3500
+    },
+    {
+      "epoch": 1.1352734250079948,
+      "grad_norm": 0.048830822110176086,
+      "learning_rate": 4.432806703338877e-05,
+      "loss": 0.0097,
+      "step": 3550
+    },
+    {
+      "epoch": 1.1512631915574032,
+      "grad_norm": 0.060777079313993454,
+      "learning_rate": 4.424811308686197e-05,
+      "loss": 0.0092,
+      "step": 3600
+    },
+    {
+      "epoch": 1.1672529581068116,
+      "grad_norm": 0.03263658285140991,
+      "learning_rate": 4.416815914033517e-05,
+      "loss": 0.0092,
+      "step": 3650
+    },
+    {
+      "epoch": 1.18324272465622,
+      "grad_norm": 0.04247328266501427,
+      "learning_rate": 4.408820519380837e-05,
+      "loss": 0.0086,
+      "step": 3700
+    },
+    {
+      "epoch": 1.1992324912056285,
+      "grad_norm": 0.057229358702898026,
+      "learning_rate": 4.400825124728157e-05,
+      "loss": 0.0096,
+      "step": 3750
+    },
+    {
+      "epoch": 1.2152222577550367,
+      "grad_norm": 0.043552979826927185,
+      "learning_rate": 4.3928297300754765e-05,
+      "loss": 0.0087,
+      "step": 3800
+    },
+    {
+      "epoch": 1.231212024304445,
+      "grad_norm": 0.0451231375336647,
+      "learning_rate": 4.384834335422797e-05,
+      "loss": 0.0095,
+      "step": 3850
+    },
+    {
+      "epoch": 1.2472017908538535,
+      "grad_norm": 0.03589295223355293,
+      "learning_rate": 4.3768389407701165e-05,
+      "loss": 0.0097,
+      "step": 3900
+    },
+    {
+      "epoch": 1.263191557403262,
+      "grad_norm": 0.03593279793858528,
+      "learning_rate": 4.368843546117437e-05,
+      "loss": 0.0094,
+      "step": 3950
+    },
+    {
+      "epoch": 1.2791813239526704,
+      "grad_norm": 0.055643677711486816,
+      "learning_rate": 4.3608481514647565e-05,
+      "loss": 0.0086,
+      "step": 4000
+    },
+    {
+      "epoch": 1.2951710905020786,
+      "grad_norm": 0.050604771822690964,
+      "learning_rate": 4.352852756812076e-05,
+      "loss": 0.0095,
+      "step": 4050
+    },
+    {
+      "epoch": 1.311160857051487,
+      "grad_norm": 0.04043160006403923,
+      "learning_rate": 4.3448573621593965e-05,
+      "loss": 0.0088,
+      "step": 4100
+    },
+    {
+      "epoch": 1.3271506236008954,
+      "grad_norm": 0.03268027678132057,
+      "learning_rate": 4.336861967506716e-05,
+      "loss": 0.0088,
+      "step": 4150
+    },
+    {
+      "epoch": 1.3431403901503038,
+      "grad_norm": 0.0318896509706974,
+      "learning_rate": 4.3288665728540365e-05,
+      "loss": 0.0095,
+      "step": 4200
+    },
+    {
+      "epoch": 1.3591301566997123,
+      "grad_norm": 0.03341570496559143,
+      "learning_rate": 4.320871178201356e-05,
+      "loss": 0.0091,
+      "step": 4250
+    },
+    {
+      "epoch": 1.3751199232491205,
+      "grad_norm": 0.02508491836488247,
+      "learning_rate": 4.312875783548676e-05,
+      "loss": 0.0087,
+      "step": 4300
+    },
+    {
+      "epoch": 1.391109689798529,
+      "grad_norm": 0.04143594950437546,
+      "learning_rate": 4.304880388895996e-05,
+      "loss": 0.0092,
+      "step": 4350
+    },
+    {
+      "epoch": 1.4070994563479373,
+      "grad_norm": 0.03601376339793205,
+      "learning_rate": 4.296884994243316e-05,
+      "loss": 0.0089,
+      "step": 4400
+    },
+    {
+      "epoch": 1.4230892228973457,
+      "grad_norm": 0.027452582493424416,
+      "learning_rate": 4.2888895995906355e-05,
+      "loss": 0.0085,
+      "step": 4450
+    },
+    {
+      "epoch": 1.4390789894467542,
+      "grad_norm": 0.046533871442079544,
+      "learning_rate": 4.280894204937956e-05,
+      "loss": 0.0094,
+      "step": 4500
+    },
+    {
+      "epoch": 1.4550687559961624,
+      "grad_norm": 0.059284090995788574,
+      "learning_rate": 4.2728988102852755e-05,
+      "loss": 0.0088,
+      "step": 4550
+    },
+    {
+      "epoch": 1.4710585225455708,
+      "grad_norm": 0.04795876890420914,
+      "learning_rate": 4.264903415632596e-05,
+      "loss": 0.0092,
+      "step": 4600
+    },
+    {
+      "epoch": 1.4870482890949792,
+      "grad_norm": 0.04677587375044823,
+      "learning_rate": 4.2569080209799155e-05,
+      "loss": 0.0092,
+      "step": 4650
+    },
+    {
+      "epoch": 1.5030380556443876,
+      "grad_norm": 0.04200517758727074,
+      "learning_rate": 4.248912626327235e-05,
+      "loss": 0.0094,
+      "step": 4700
+    },
+    {
+      "epoch": 1.519027822193796,
+      "grad_norm": 0.05572056025266647,
+      "learning_rate": 4.2409172316745555e-05,
+      "loss": 0.0086,
+      "step": 4750
+    },
+    {
+      "epoch": 1.5350175887432043,
+      "grad_norm": 0.026984069496393204,
+      "learning_rate": 4.232921837021875e-05,
+      "loss": 0.0088,
+      "step": 4800
+    },
+    {
+      "epoch": 1.5510073552926127,
+      "grad_norm": 0.055002227425575256,
+      "learning_rate": 4.2249264423691956e-05,
+      "loss": 0.0085,
+      "step": 4850
+    },
+    {
+      "epoch": 1.5669971218420211,
+      "grad_norm": 0.03688422963023186,
+      "learning_rate": 4.216931047716515e-05,
+      "loss": 0.0093,
+      "step": 4900
+    },
+    {
+      "epoch": 1.5829868883914295,
+      "grad_norm": 0.051286958158016205,
+      "learning_rate": 4.208935653063835e-05,
+      "loss": 0.0094,
+      "step": 4950
+    },
+    {
+      "epoch": 1.598976654940838,
+      "grad_norm": 0.04032529890537262,
+      "learning_rate": 4.200940258411155e-05,
+      "loss": 0.009,
+      "step": 5000
+    },
+    {
+      "epoch": 1.6149664214902462,
+      "grad_norm": 0.04411604255437851,
+      "learning_rate": 4.192944863758475e-05,
+      "loss": 0.0087,
+      "step": 5050
+    },
+    {
+      "epoch": 1.6309561880396546,
+      "grad_norm": 0.03501143679022789,
+      "learning_rate": 4.184949469105795e-05,
+      "loss": 0.0087,
+      "step": 5100
+    },
+    {
+      "epoch": 1.646945954589063,
+      "grad_norm": 0.04125143960118294,
+      "learning_rate": 4.176954074453115e-05,
+      "loss": 0.0089,
+      "step": 5150
+    },
+    {
+      "epoch": 1.6629357211384712,
+      "grad_norm": 0.03687961399555206,
+      "learning_rate": 4.168958679800435e-05,
+      "loss": 0.009,
+      "step": 5200
+    },
+    {
+      "epoch": 1.6789254876878799,
+      "grad_norm": 0.08890601992607117,
+      "learning_rate": 4.1609632851477556e-05,
+      "loss": 0.0087,
+      "step": 5250
+    },
+    {
+      "epoch": 1.694915254237288,
+      "grad_norm": 0.03983084484934807,
+      "learning_rate": 4.152967890495075e-05,
+      "loss": 0.0093,
+      "step": 5300
+    },
+    {
+      "epoch": 1.7109050207866965,
+      "grad_norm": 0.02701224572956562,
+      "learning_rate": 4.144972495842395e-05,
+      "loss": 0.009,
+      "step": 5350
+    },
+    {
+      "epoch": 1.726894787336105,
+      "grad_norm": 0.023342708125710487,
+      "learning_rate": 4.136977101189715e-05,
+      "loss": 0.0085,
+      "step": 5400
+    },
+    {
+      "epoch": 1.742884553885513,
+      "grad_norm": 0.048117853701114655,
+      "learning_rate": 4.128981706537035e-05,
+      "loss": 0.0087,
+      "step": 5450
+    },
+    {
+      "epoch": 1.7588743204349218,
+      "grad_norm": 0.044627636671066284,
+      "learning_rate": 4.120986311884355e-05,
+      "loss": 0.009,
+      "step": 5500
+    },
+    {
+      "epoch": 1.77486408698433,
+      "grad_norm": 0.04622326046228409,
+      "learning_rate": 4.112990917231675e-05,
+      "loss": 0.0094,
+      "step": 5550
+    },
+    {
+      "epoch": 1.7908538535337384,
+      "grad_norm": 0.024695126339793205,
+      "learning_rate": 4.1049955225789946e-05,
+      "loss": 0.0088,
+      "step": 5600
+    },
+    {
+      "epoch": 1.8068436200831468,
+      "grad_norm": 0.07150325179100037,
+      "learning_rate": 4.097000127926315e-05,
+      "loss": 0.0089,
+      "step": 5650
+    },
+    {
+      "epoch": 1.822833386632555,
+      "grad_norm": 0.05561058968305588,
+      "learning_rate": 4.0890047332736346e-05,
+      "loss": 0.0089,
+      "step": 5700
+    },
+    {
+      "epoch": 1.8388231531819637,
+      "grad_norm": 0.04343913495540619,
+      "learning_rate": 4.081009338620955e-05,
+      "loss": 0.0084,
+      "step": 5750
+    },
+    {
+      "epoch": 1.8548129197313719,
+      "grad_norm": 0.022196004167199135,
+      "learning_rate": 4.0730139439682746e-05,
+      "loss": 0.0093,
+      "step": 5800
+    },
+    {
+      "epoch": 1.8708026862807803,
+      "grad_norm": 0.035860929638147354,
+      "learning_rate": 4.065018549315594e-05,
+      "loss": 0.0089,
+      "step": 5850
+    },
+    {
+      "epoch": 1.8867924528301887,
+      "grad_norm": 0.04030134156346321,
+      "learning_rate": 4.0570231546629146e-05,
+      "loss": 0.0087,
+      "step": 5900
+    },
+    {
+      "epoch": 1.902782219379597,
+      "grad_norm": 0.04269862174987793,
+      "learning_rate": 4.049027760010234e-05,
+      "loss": 0.0085,
+      "step": 5950
+    },
+    {
+      "epoch": 1.9187719859290056,
+      "grad_norm": 0.027169955894351006,
+      "learning_rate": 4.0410323653575546e-05,
+      "loss": 0.0088,
+      "step": 6000
+    },
+    {
+      "epoch": 1.9187719859290056,
+      "eval_loss": 0.015284636057913303,
+      "eval_runtime": 41.2462,
+      "eval_samples_per_second": 22.135,
+      "eval_steps_per_second": 1.406,
+      "step": 6000
+    },
+    {
+      "epoch": 1.9347617524784138,
+      "grad_norm": 0.028573885560035706,
+      "learning_rate": 4.033036970704874e-05,
+      "loss": 0.0085,
+      "step": 6050
+    },
+    {
+      "epoch": 1.9507515190278222,
+      "grad_norm": 0.04846006631851196,
+      "learning_rate": 4.025041576052194e-05,
+      "loss": 0.0086,
+      "step": 6100
+    },
+    {
+      "epoch": 1.9667412855772306,
+      "grad_norm": 0.04022432863712311,
+      "learning_rate": 4.017046181399514e-05,
+      "loss": 0.0087,
+      "step": 6150
+    },
+    {
+      "epoch": 1.9827310521266388,
+      "grad_norm": 0.052905626595020294,
+      "learning_rate": 4.009050786746834e-05,
+      "loss": 0.0087,
+      "step": 6200
+    },
+    {
+      "epoch": 1.9987208186760475,
+      "grad_norm": 0.047191545367240906,
+      "learning_rate": 4.001055392094154e-05,
+      "loss": 0.0088,
+      "step": 6250
+    },
+    {
+      "epoch": 2.0147105852254557,
+      "grad_norm": 0.03503846377134323,
+      "learning_rate": 3.993059997441474e-05,
+      "loss": 0.009,
+      "step": 6300
+    },
+    {
+      "epoch": 2.0307003517748643,
+      "grad_norm": 0.03783261030912399,
+      "learning_rate": 3.9850646027887936e-05,
+      "loss": 0.0083,
+      "step": 6350
+    },
+    {
+      "epoch": 2.0466901183242725,
+      "grad_norm": 0.07593075186014175,
+      "learning_rate": 3.977069208136114e-05,
+      "loss": 0.0082,
+      "step": 6400
+    },
+    {
+      "epoch": 2.0626798848736807,
+      "grad_norm": 0.04408886656165123,
+      "learning_rate": 3.9690738134834337e-05,
+      "loss": 0.0082,
+      "step": 6450
+    },
+    {
+      "epoch": 2.0786696514230893,
+      "grad_norm": 0.07408079504966736,
+      "learning_rate": 3.961078418830754e-05,
+      "loss": 0.0088,
+      "step": 6500
+    },
+    {
+      "epoch": 2.0946594179724976,
+      "grad_norm": 0.0387246273458004,
+      "learning_rate": 3.953083024178074e-05,
+      "loss": 0.0092,
+      "step": 6550
+    },
+    {
+      "epoch": 2.1106491845219058,
+      "grad_norm": 0.029871731996536255,
+      "learning_rate": 3.945087629525393e-05,
+      "loss": 0.0089,
+      "step": 6600
+    },
+    {
+      "epoch": 2.1266389510713144,
+      "grad_norm": 0.08507910370826721,
+      "learning_rate": 3.937092234872714e-05,
+      "loss": 0.0085,
+      "step": 6650
+    },
+    {
+      "epoch": 2.1426287176207226,
+      "grad_norm": 0.06320624053478241,
+      "learning_rate": 3.929096840220033e-05,
+      "loss": 0.0093,
+      "step": 6700
+    },
+    {
+      "epoch": 2.1586184841701312,
+      "grad_norm": 0.040876906365156174,
+      "learning_rate": 3.921101445567354e-05,
+      "loss": 0.0087,
+      "step": 6750
+    },
+    {
+      "epoch": 2.1746082507195394,
+      "grad_norm": 0.05744218826293945,
+      "learning_rate": 3.9131060509146733e-05,
+      "loss": 0.008,
+      "step": 6800
+    },
+    {
+      "epoch": 2.1905980172689477,
+      "grad_norm": 0.07130051404237747,
+      "learning_rate": 3.905110656261993e-05,
+      "loss": 0.0087,
+      "step": 6850
+    },
+    {
+      "epoch": 2.2065877838183563,
+      "grad_norm": 0.06204545497894287,
+      "learning_rate": 3.8971152616093134e-05,
+      "loss": 0.0085,
+      "step": 6900
+    },
+    {
+      "epoch": 2.2225775503677645,
+      "grad_norm": 0.04432045295834541,
+      "learning_rate": 3.889119866956633e-05,
+      "loss": 0.0083,
+      "step": 6950
+    },
+    {
+      "epoch": 2.238567316917173,
+      "grad_norm": 0.07190771400928497,
+      "learning_rate": 3.8811244723039534e-05,
+      "loss": 0.0094,
+      "step": 7000
+    },
+    {
+      "epoch": 2.2545570834665813,
+      "grad_norm": 0.06573159247636795,
+      "learning_rate": 3.873129077651273e-05,
+      "loss": 0.0087,
+      "step": 7050
+    },
+    {
+      "epoch": 2.2705468500159895,
+      "grad_norm": 0.03736311197280884,
+      "learning_rate": 3.865133682998593e-05,
+      "loss": 0.0084,
+      "step": 7100
+    },
+    {
+      "epoch": 2.286536616565398,
+      "grad_norm": 0.0285806842148304,
+      "learning_rate": 3.857138288345913e-05,
+      "loss": 0.0086,
+      "step": 7150
+    },
+    {
+      "epoch": 2.3025263831148064,
+      "grad_norm": 0.03527309000492096,
+      "learning_rate": 3.849142893693233e-05,
+      "loss": 0.0088,
+      "step": 7200
+    },
+    {
+      "epoch": 2.318516149664215,
+      "grad_norm": 0.0564735010266304,
+      "learning_rate": 3.841147499040553e-05,
+      "loss": 0.008,
+      "step": 7250
+    },
+    {
+      "epoch": 2.3345059162136232,
+      "grad_norm": 0.04310419037938118,
+      "learning_rate": 3.833152104387873e-05,
+      "loss": 0.0088,
+      "step": 7300
+    },
+    {
+      "epoch": 2.3504956827630314,
+      "grad_norm": 0.03855273872613907,
+      "learning_rate": 3.8251567097351924e-05,
+      "loss": 0.0084,
+      "step": 7350
+    },
+    {
+      "epoch": 2.36648544931244,
+      "grad_norm": 0.04269862547516823,
+      "learning_rate": 3.817161315082513e-05,
+      "loss": 0.0085,
+      "step": 7400
+    },
+    {
+      "epoch": 2.3824752158618483,
+      "grad_norm": 0.04821077361702919,
+      "learning_rate": 3.8091659204298324e-05,
+      "loss": 0.0084,
+      "step": 7450
+    },
+    {
+      "epoch": 2.398464982411257,
+      "grad_norm": 0.05434531718492508,
+      "learning_rate": 3.801170525777152e-05,
+      "loss": 0.0087,
+      "step": 7500
+    },
+    {
+      "epoch": 2.414454748960665,
+      "grad_norm": 0.05756519362330437,
+      "learning_rate": 3.7931751311244724e-05,
+      "loss": 0.0079,
+      "step": 7550
+    },
+    {
+      "epoch": 2.4304445155100733,
+      "grad_norm": 0.0482054241001606,
+      "learning_rate": 3.785179736471792e-05,
+      "loss": 0.0082,
+      "step": 7600
+    },
+    {
+      "epoch": 2.446434282059482,
+      "grad_norm": 0.03534189611673355,
+      "learning_rate": 3.7771843418191124e-05,
+      "loss": 0.0085,
+      "step": 7650
+    },
+    {
+      "epoch": 2.46242404860889,
+      "grad_norm": 0.04225745424628258,
+      "learning_rate": 3.769188947166432e-05,
+      "loss": 0.0082,
+      "step": 7700
+    },
+    {
+      "epoch": 2.478413815158299,
+      "grad_norm": 0.08791286498308182,
+      "learning_rate": 3.761193552513752e-05,
+      "loss": 0.009,
+      "step": 7750
+    },
+    {
+      "epoch": 2.494403581707707,
+      "grad_norm": 0.027702681720256805,
+      "learning_rate": 3.753198157861072e-05,
+      "loss": 0.0083,
+      "step": 7800
+    },
+    {
+      "epoch": 2.5103933482571152,
+      "grad_norm": 0.03311574086546898,
+      "learning_rate": 3.745202763208392e-05,
+      "loss": 0.0084,
+      "step": 7850
+    },
+    {
+      "epoch": 2.526383114806524,
+      "grad_norm": 0.07140166312456131,
+      "learning_rate": 3.737207368555712e-05,
+      "loss": 0.0083,
+      "step": 7900
+    },
+    {
+      "epoch": 2.542372881355932,
+      "grad_norm": 0.04849368333816528,
+      "learning_rate": 3.729211973903032e-05,
+      "loss": 0.0082,
+      "step": 7950
+    },
+    {
+      "epoch": 2.5583626479053407,
+      "grad_norm": 0.04977947846055031,
+      "learning_rate": 3.7212165792503514e-05,
+      "loss": 0.0084,
+      "step": 8000
+    },
+    {
+      "epoch": 2.574352414454749,
+      "grad_norm": 0.035205528140068054,
+      "learning_rate": 3.713221184597672e-05,
+      "loss": 0.0087,
+      "step": 8050
+    },
+    {
+      "epoch": 2.590342181004157,
+      "grad_norm": 0.05450873449444771,
+      "learning_rate": 3.7052257899449914e-05,
+      "loss": 0.0086,
+      "step": 8100
+    },
+    {
+      "epoch": 2.606331947553566,
+      "grad_norm": 0.04244232550263405,
+      "learning_rate": 3.697230395292312e-05,
+      "loss": 0.0087,
+      "step": 8150
+    },
+    {
+      "epoch": 2.622321714102974,
+      "grad_norm": 0.04528970643877983,
+      "learning_rate": 3.689235000639632e-05,
+      "loss": 0.0081,
+      "step": 8200
+    },
+    {
+      "epoch": 2.6383114806523826,
+      "grad_norm": 0.05795321986079216,
+      "learning_rate": 3.681239605986952e-05,
+      "loss": 0.0085,
+      "step": 8250
+    },
+    {
+      "epoch": 2.654301247201791,
+      "grad_norm": 0.04863649979233742,
+      "learning_rate": 3.673244211334272e-05,
+      "loss": 0.0083,
+      "step": 8300
+    },
+    {
+      "epoch": 2.670291013751199,
+      "grad_norm": 0.03658927232027054,
+      "learning_rate": 3.665248816681592e-05,
+      "loss": 0.008,
+      "step": 8350
+    },
+    {
+      "epoch": 2.6862807803006077,
+      "grad_norm": 0.039778467267751694,
+      "learning_rate": 3.6572534220289114e-05,
+      "loss": 0.0089,
+      "step": 8400
+    },
+    {
+      "epoch": 2.702270546850016,
+      "grad_norm": 0.04085814952850342,
+      "learning_rate": 3.649258027376232e-05,
+      "loss": 0.0085,
+      "step": 8450
+    },
+    {
+      "epoch": 2.7182603133994245,
+      "grad_norm": 0.030918624252080917,
+      "learning_rate": 3.6412626327235515e-05,
+      "loss": 0.0082,
+      "step": 8500
+    },
+    {
+      "epoch": 2.7342500799488327,
+      "grad_norm": 0.05798843502998352,
+      "learning_rate": 3.633267238070872e-05,
+      "loss": 0.0082,
+      "step": 8550
+    },
+    {
+      "epoch": 2.750239846498241,
+      "grad_norm": 0.03750443831086159,
+      "learning_rate": 3.6252718434181915e-05,
+      "loss": 0.0084,
+      "step": 8600
+    },
+    {
+      "epoch": 2.7662296130476496,
+      "grad_norm": 0.0728391632437706,
+      "learning_rate": 3.617276448765511e-05,
+      "loss": 0.0084,
+      "step": 8650
+    },
+    {
+      "epoch": 2.782219379597058,
+      "grad_norm": 0.09955340623855591,
+      "learning_rate": 3.6092810541128315e-05,
+      "loss": 0.0086,
+      "step": 8700
+    },
+    {
+      "epoch": 2.7982091461464664,
+      "grad_norm": 0.045326054096221924,
+      "learning_rate": 3.601285659460151e-05,
+      "loss": 0.0078,
+      "step": 8750
+    },
+    {
+      "epoch": 2.8141989126958746,
+      "grad_norm": 0.024823077023029327,
+      "learning_rate": 3.5932902648074715e-05,
+      "loss": 0.0085,
+      "step": 8800
+    },
+    {
+      "epoch": 2.830188679245283,
+      "grad_norm": 0.03785817325115204,
+      "learning_rate": 3.585294870154791e-05,
+      "loss": 0.0077,
+      "step": 8850
+    },
+    {
+      "epoch": 2.8461784457946915,
+      "grad_norm": 0.041541390120983124,
+      "learning_rate": 3.577299475502111e-05,
+      "loss": 0.0087,
+      "step": 8900
+    },
+    {
+      "epoch": 2.8621682123440997,
+      "grad_norm": 0.06973368674516678,
+      "learning_rate": 3.569304080849431e-05,
+      "loss": 0.0084,
+      "step": 8950
+    },
+    {
+      "epoch": 2.8781579788935083,
+      "grad_norm": 0.02823769487440586,
+      "learning_rate": 3.561308686196751e-05,
+      "loss": 0.0084,
+      "step": 9000
+    },
+    {
+      "epoch": 2.8941477454429165,
+      "grad_norm": 0.07545454800128937,
+      "learning_rate": 3.553313291544071e-05,
+      "loss": 0.0081,
+      "step": 9050
+    },
+    {
+      "epoch": 2.9101375119923247,
+      "grad_norm": 0.09222853183746338,
+      "learning_rate": 3.545317896891391e-05,
+      "loss": 0.0079,
+      "step": 9100
+    },
+    {
+      "epoch": 2.9261272785417334,
+      "grad_norm": 0.06112439185380936,
+      "learning_rate": 3.5373225022387105e-05,
+      "loss": 0.0081,
+      "step": 9150
+    },
+    {
+      "epoch": 2.9421170450911416,
+      "grad_norm": 0.03861480578780174,
+      "learning_rate": 3.529327107586031e-05,
+      "loss": 0.0082,
+      "step": 9200
+    },
+    {
+      "epoch": 2.9581068116405502,
+      "grad_norm": 0.08047035336494446,
+      "learning_rate": 3.5213317129333505e-05,
+      "loss": 0.0085,
+      "step": 9250
+    },
+    {
+      "epoch": 2.9740965781899584,
+      "grad_norm": 0.0482422411441803,
+      "learning_rate": 3.513336318280671e-05,
+      "loss": 0.0078,
+      "step": 9300
+    },
+    {
+      "epoch": 2.9900863447393666,
+      "grad_norm": 0.10705345124006271,
+      "learning_rate": 3.5053409236279905e-05,
+      "loss": 0.0084,
+      "step": 9350
+    },
+    {
+      "epoch": 3.0060761112887753,
+      "grad_norm": 0.05015913024544716,
+      "learning_rate": 3.49734552897531e-05,
+      "loss": 0.0079,
+      "step": 9400
+    },
+    {
+      "epoch": 3.0220658778381835,
+      "grad_norm": 0.03695971891283989,
+      "learning_rate": 3.4893501343226305e-05,
+      "loss": 0.0083,
+      "step": 9450
+    },
+    {
+      "epoch": 3.038055644387592,
+      "grad_norm": 0.04805247113108635,
+      "learning_rate": 3.48135473966995e-05,
+      "loss": 0.0076,
+      "step": 9500
+    },
+    {
+      "epoch": 3.0540454109370003,
+      "grad_norm": 0.049487996846437454,
+      "learning_rate": 3.4733593450172705e-05,
+      "loss": 0.0079,
+      "step": 9550
+    },
+    {
+      "epoch": 3.0700351774864085,
+      "grad_norm": 0.037209343165159225,
+      "learning_rate": 3.46536395036459e-05,
+      "loss": 0.0079,
+      "step": 9600
+    },
+    {
+      "epoch": 3.086024944035817,
+      "grad_norm": 0.0906648263335228,
+      "learning_rate": 3.45736855571191e-05,
+      "loss": 0.0086,
+      "step": 9650
+    },
+    {
+      "epoch": 3.1020147105852254,
+      "grad_norm": 0.06882278621196747,
+      "learning_rate": 3.44937316105923e-05,
+      "loss": 0.0085,
+      "step": 9700
+    },
+    {
+      "epoch": 3.118004477134634,
+      "grad_norm": 0.04665728658437729,
+      "learning_rate": 3.44137776640655e-05,
+      "loss": 0.0081,
+      "step": 9750
+    },
+    {
+      "epoch": 3.1339942436840422,
+      "grad_norm": 0.05663415044546127,
+      "learning_rate": 3.43338237175387e-05,
+      "loss": 0.0085,
+      "step": 9800
+    },
+    {
+      "epoch": 3.1499840102334504,
+      "grad_norm": 0.03854561224579811,
+      "learning_rate": 3.42538697710119e-05,
+      "loss": 0.0081,
+      "step": 9850
+    },
+    {
+      "epoch": 3.165973776782859,
+      "grad_norm": 0.04344400390982628,
+      "learning_rate": 3.4173915824485095e-05,
+      "loss": 0.0084,
+      "step": 9900
+    },
+    {
+      "epoch": 3.1819635433322673,
+      "grad_norm": 0.04235091805458069,
+      "learning_rate": 3.40939618779583e-05,
+      "loss": 0.0073,
+      "step": 9950
+    },
+    {
+      "epoch": 3.197953309881676,
+      "grad_norm": 0.03422466665506363,
+      "learning_rate": 3.4014007931431496e-05,
+      "loss": 0.0085,
+      "step": 10000
+    },
+    {
+      "epoch": 3.213943076431084,
+      "grad_norm": 0.04774649068713188,
+      "learning_rate": 3.39340539849047e-05,
+      "loss": 0.0079,
+      "step": 10050
+    },
+    {
+      "epoch": 3.2299328429804923,
+      "grad_norm": 0.03879115357995033,
+      "learning_rate": 3.3854100038377896e-05,
+      "loss": 0.0084,
+      "step": 10100
+    },
+    {
+      "epoch": 3.245922609529901,
+      "grad_norm": 0.08118848502635956,
+      "learning_rate": 3.377414609185109e-05,
+      "loss": 0.0085,
+      "step": 10150
+    },
+    {
+      "epoch": 3.261912376079309,
+      "grad_norm": 0.05128278210759163,
+      "learning_rate": 3.3694192145324296e-05,
+      "loss": 0.0082,
+      "step": 10200
+    },
+    {
+      "epoch": 3.277902142628718,
+      "grad_norm": 0.05539889633655548,
+      "learning_rate": 3.361423819879749e-05,
+      "loss": 0.0077,
+      "step": 10250
+    },
+    {
+      "epoch": 3.293891909178126,
+      "grad_norm": 0.06546144187450409,
+      "learning_rate": 3.353428425227069e-05,
+      "loss": 0.008,
+      "step": 10300
+    },
+    {
+      "epoch": 3.3098816757275342,
+      "grad_norm": 0.055574435740709305,
+      "learning_rate": 3.345433030574389e-05,
+      "loss": 0.0078,
+      "step": 10350
+    },
+    {
+      "epoch": 3.325871442276943,
+      "grad_norm": 0.07268719375133514,
+      "learning_rate": 3.337437635921709e-05,
+      "loss": 0.0077,
+      "step": 10400
+    },
+    {
+      "epoch": 3.341861208826351,
+      "grad_norm": 0.05133509263396263,
+      "learning_rate": 3.329442241269029e-05,
+      "loss": 0.0086,
+      "step": 10450
+    },
+    {
+      "epoch": 3.3578509753757597,
+      "grad_norm": 0.039566852152347565,
+      "learning_rate": 3.321446846616349e-05,
+      "loss": 0.0078,
+      "step": 10500
+    },
+    {
+      "epoch": 3.373840741925168,
+      "grad_norm": 0.03541762754321098,
+      "learning_rate": 3.3134514519636686e-05,
+      "loss": 0.0079,
+      "step": 10550
+    },
+    {
+      "epoch": 3.389830508474576,
+      "grad_norm": 0.0460277684032917,
+      "learning_rate": 3.305456057310989e-05,
+      "loss": 0.0079,
+      "step": 10600
+    },
+    {
+      "epoch": 3.4058202750239848,
+      "grad_norm": 0.08597122132778168,
+      "learning_rate": 3.2974606626583086e-05,
+      "loss": 0.0077,
+      "step": 10650
+    },
+    {
+      "epoch": 3.421810041573393,
+      "grad_norm": 0.07249324768781662,
+      "learning_rate": 3.289465268005629e-05,
+      "loss": 0.0074,
+      "step": 10700
+    },
+    {
+      "epoch": 3.4377998081228016,
+      "grad_norm": 0.04298989102244377,
+      "learning_rate": 3.2814698733529486e-05,
+      "loss": 0.0079,
+      "step": 10750
+    },
+    {
+      "epoch": 3.45378957467221,
+      "grad_norm": 0.029726523905992508,
+      "learning_rate": 3.273474478700268e-05,
+      "loss": 0.008,
+      "step": 10800
+    },
+    {
+      "epoch": 3.469779341221618,
+      "grad_norm": 0.04522155970335007,
+      "learning_rate": 3.2654790840475886e-05,
+      "loss": 0.0082,
+      "step": 10850
+    },
+    {
+      "epoch": 3.4857691077710267,
+      "grad_norm": 0.052699171006679535,
+      "learning_rate": 3.257483689394908e-05,
+      "loss": 0.0078,
+      "step": 10900
+    },
+    {
+      "epoch": 3.501758874320435,
+      "grad_norm": 0.06500669568777084,
+      "learning_rate": 3.2494882947422286e-05,
+      "loss": 0.0083,
+      "step": 10950
+    },
+    {
+      "epoch": 3.5177486408698435,
+      "grad_norm": 0.05794863402843475,
+      "learning_rate": 3.241492900089548e-05,
+      "loss": 0.0074,
+      "step": 11000
+    },
+    {
+      "epoch": 3.5337384074192517,
+      "grad_norm": 0.08507433533668518,
+      "learning_rate": 3.233497505436868e-05,
+      "loss": 0.0079,
+      "step": 11050
+    },
+    {
+      "epoch": 3.54972817396866,
+      "grad_norm": 0.07829931378364563,
+      "learning_rate": 3.225502110784189e-05,
+      "loss": 0.0075,
+      "step": 11100
+    },
+    {
+      "epoch": 3.5657179405180686,
+      "grad_norm": 0.07517477124929428,
+      "learning_rate": 3.2175067161315086e-05,
+      "loss": 0.0082,
+      "step": 11150
+    },
+    {
+      "epoch": 3.5817077070674768,
+      "grad_norm": 0.05099401995539665,
+      "learning_rate": 3.209511321478828e-05,
+      "loss": 0.0082,
+      "step": 11200
+    },
+    {
+      "epoch": 3.5976974736168854,
+      "grad_norm": 0.04134771227836609,
+      "learning_rate": 3.2015159268261486e-05,
+      "loss": 0.008,
+      "step": 11250
+    },
+    {
+      "epoch": 3.6136872401662936,
+      "grad_norm": 0.0595158226788044,
+      "learning_rate": 3.193520532173468e-05,
+      "loss": 0.0077,
+      "step": 11300
+    },
+    {
+      "epoch": 3.629677006715702,
+      "grad_norm": 0.08330442011356354,
+      "learning_rate": 3.1855251375207887e-05,
+      "loss": 0.0078,
+      "step": 11350
+    },
+    {
+      "epoch": 3.6456667732651105,
+      "grad_norm": 0.06455782800912857,
+      "learning_rate": 3.177529742868108e-05,
+      "loss": 0.0077,
+      "step": 11400
+    },
+    {
+      "epoch": 3.6616565398145187,
+      "grad_norm": 0.05490834638476372,
+      "learning_rate": 3.169534348215428e-05,
+      "loss": 0.0081,
+      "step": 11450
+    },
+    {
+      "epoch": 3.6776463063639273,
+      "grad_norm": 0.0402890108525753,
+      "learning_rate": 3.161538953562748e-05,
+      "loss": 0.0075,
+      "step": 11500
+    },
+    {
+      "epoch": 3.6936360729133355,
+      "grad_norm": 0.049679581075906754,
+      "learning_rate": 3.153543558910068e-05,
+      "loss": 0.0084,
+      "step": 11550
+    },
+    {
+      "epoch": 3.7096258394627437,
+      "grad_norm": 0.04728088527917862,
+      "learning_rate": 3.145548164257388e-05,
+      "loss": 0.0079,
+      "step": 11600
+    },
+    {
+      "epoch": 3.7256156060121524,
+      "grad_norm": 0.060690511018037796,
+      "learning_rate": 3.137552769604708e-05,
+      "loss": 0.0077,
+      "step": 11650
+    },
+    {
+      "epoch": 3.7416053725615606,
+      "grad_norm": 0.0751800686120987,
+      "learning_rate": 3.129557374952028e-05,
+      "loss": 0.0075,
+      "step": 11700
+    },
+    {
+      "epoch": 3.757595139110969,
+      "grad_norm": 0.056977272033691406,
+      "learning_rate": 3.121561980299348e-05,
+      "loss": 0.0079,
+      "step": 11750
+    },
+    {
+      "epoch": 3.7735849056603774,
+      "grad_norm": 0.06811773031949997,
+      "learning_rate": 3.113566585646668e-05,
+      "loss": 0.0081,
+      "step": 11800
+    },
+    {
+      "epoch": 3.7895746722097856,
+      "grad_norm": 0.040852271020412445,
+      "learning_rate": 3.105571190993988e-05,
+      "loss": 0.0075,
+      "step": 11850
+    },
+    {
+      "epoch": 3.8055644387591943,
+      "grad_norm": 0.0817752480506897,
+      "learning_rate": 3.097575796341308e-05,
+      "loss": 0.0078,
+      "step": 11900
+    },
+    {
+      "epoch": 3.8215542053086025,
+      "grad_norm": 0.045933738350868225,
+      "learning_rate": 3.0895804016886273e-05,
+      "loss": 0.0078,
+      "step": 11950
+    },
+    {
+      "epoch": 3.837543971858011,
+      "grad_norm": 0.04742230847477913,
+      "learning_rate": 3.081585007035948e-05,
+      "loss": 0.0074,
+      "step": 12000
+    },
+    {
+      "epoch": 3.837543971858011,
+      "eval_loss": 0.015844305977225304,
+      "eval_runtime": 41.1322,
+      "eval_samples_per_second": 22.197,
+      "eval_steps_per_second": 1.41,
+      "step": 12000
+    },
+    {
+      "epoch": 3.8535337384074193,
+      "grad_norm": 0.0642109289765358,
+      "learning_rate": 3.0735896123832674e-05,
+      "loss": 0.0082,
+      "step": 12050
+    },
+    {
+      "epoch": 3.8695235049568275,
+      "grad_norm": 0.08985566347837448,
+      "learning_rate": 3.065594217730588e-05,
+      "loss": 0.0079,
+      "step": 12100
+    },
+    {
+      "epoch": 3.885513271506236,
+      "grad_norm": 0.07430126518011093,
+      "learning_rate": 3.0575988230779074e-05,
+      "loss": 0.0076,
+      "step": 12150
+    },
+    {
+      "epoch": 3.9015030380556444,
+      "grad_norm": 0.05662678927183151,
+      "learning_rate": 3.0496034284252274e-05,
+      "loss": 0.0072,
+      "step": 12200
+    },
+    {
+      "epoch": 3.917492804605053,
+      "grad_norm": 0.08210637420415878,
+      "learning_rate": 3.0416080337725474e-05,
+      "loss": 0.0078,
+      "step": 12250
+    },
+    {
+      "epoch": 3.933482571154461,
+      "grad_norm": 0.09279511868953705,
+      "learning_rate": 3.033612639119867e-05,
+      "loss": 0.0075,
+      "step": 12300
+    },
+    {
+      "epoch": 3.9494723377038694,
+      "grad_norm": 0.03649890050292015,
+      "learning_rate": 3.025617244467187e-05,
+      "loss": 0.0075,
+      "step": 12350
+    },
+    {
+      "epoch": 3.965462104253278,
+      "grad_norm": 0.07408464699983597,
+      "learning_rate": 3.017621849814507e-05,
+      "loss": 0.0077,
+      "step": 12400
+    },
+    {
+      "epoch": 3.9814518708026863,
+      "grad_norm": 0.06778215616941452,
+      "learning_rate": 3.009626455161827e-05,
+      "loss": 0.0074,
+      "step": 12450
+    },
+    {
+      "epoch": 3.997441637352095,
+      "grad_norm": 0.07367826253175735,
+      "learning_rate": 3.0016310605091467e-05,
+      "loss": 0.008,
+      "step": 12500
+    },
+    {
+      "epoch": 4.013431403901503,
+      "grad_norm": 0.09045771509408951,
+      "learning_rate": 2.9936356658564667e-05,
+      "loss": 0.0076,
+      "step": 12550
+    },
+    {
+      "epoch": 4.029421170450911,
+      "grad_norm": 0.03875061869621277,
+      "learning_rate": 2.9856402712037867e-05,
+      "loss": 0.0075,
+      "step": 12600
+    },
+    {
+      "epoch": 4.0454109370003195,
+      "grad_norm": 0.0555877648293972,
+      "learning_rate": 2.9776448765511067e-05,
+      "loss": 0.007,
+      "step": 12650
+    },
+    {
+      "epoch": 4.061400703549729,
+      "grad_norm": 0.057610440999269485,
+      "learning_rate": 2.9696494818984267e-05,
+      "loss": 0.0075,
+      "step": 12700
+    },
+    {
+      "epoch": 4.077390470099137,
+      "grad_norm": 0.06584593653678894,
+      "learning_rate": 2.9616540872457464e-05,
+      "loss": 0.0075,
+      "step": 12750
+    },
+    {
+      "epoch": 4.093380236648545,
+      "grad_norm": 0.07038649171590805,
+      "learning_rate": 2.9536586925930664e-05,
+      "loss": 0.008,
+      "step": 12800
+    },
+    {
+      "epoch": 4.109370003197953,
+      "grad_norm": 0.05053670331835747,
+      "learning_rate": 2.9456632979403864e-05,
+      "loss": 0.008,
+      "step": 12850
+    },
+    {
+      "epoch": 4.125359769747361,
+      "grad_norm": 0.05693377926945686,
+      "learning_rate": 2.9376679032877064e-05,
+      "loss": 0.0073,
+      "step": 12900
+    },
+    {
+      "epoch": 4.1413495362967705,
+      "grad_norm": 0.07704003155231476,
+      "learning_rate": 2.9296725086350264e-05,
+      "loss": 0.0084,
+      "step": 12950
+    },
+    {
+      "epoch": 4.157339302846179,
+      "grad_norm": 0.05889516696333885,
+      "learning_rate": 2.921677113982346e-05,
+      "loss": 0.0078,
+      "step": 13000
+    },
+    {
+      "epoch": 4.173329069395587,
+      "grad_norm": 0.08338670432567596,
+      "learning_rate": 2.913681719329666e-05,
+      "loss": 0.007,
+      "step": 13050
+    },
+    {
+      "epoch": 4.189318835944995,
+      "grad_norm": 0.06243142858147621,
+      "learning_rate": 2.905686324676986e-05,
+      "loss": 0.0077,
+      "step": 13100
+    },
+    {
+      "epoch": 4.205308602494403,
+      "grad_norm": 0.04768793657422066,
+      "learning_rate": 2.897690930024306e-05,
+      "loss": 0.0076,
+      "step": 13150
+    },
+    {
+      "epoch": 4.2212983690438115,
+      "grad_norm": 0.04623186215758324,
+      "learning_rate": 2.889695535371626e-05,
+      "loss": 0.0072,
+      "step": 13200
+    },
+    {
+      "epoch": 4.237288135593221,
+      "grad_norm": 0.059952862560749054,
+      "learning_rate": 2.8817001407189458e-05,
+      "loss": 0.0081,
+      "step": 13250
+    },
+    {
+      "epoch": 4.253277902142629,
+      "grad_norm": 0.05281800776720047,
+      "learning_rate": 2.8737047460662658e-05,
+      "loss": 0.0078,
+      "step": 13300
+    },
+    {
+      "epoch": 4.269267668692037,
+      "grad_norm": 0.07640237361192703,
+      "learning_rate": 2.8657093514135858e-05,
+      "loss": 0.0074,
+      "step": 13350
+    },
+    {
+      "epoch": 4.285257435241445,
+      "grad_norm": 0.06750530749559402,
+      "learning_rate": 2.8577139567609058e-05,
+      "loss": 0.0075,
+      "step": 13400
+    },
+    {
+      "epoch": 4.301247201790854,
+      "grad_norm": 0.07524023205041885,
+      "learning_rate": 2.8497185621082258e-05,
+      "loss": 0.0075,
+      "step": 13450
+    },
+    {
+      "epoch": 4.3172369683402625,
+      "grad_norm": 0.07160656154155731,
+      "learning_rate": 2.8417231674555454e-05,
+      "loss": 0.0069,
+      "step": 13500
+    },
+    {
+      "epoch": 4.333226734889671,
+      "grad_norm": 0.06828583776950836,
+      "learning_rate": 2.8337277728028655e-05,
+      "loss": 0.0076,
+      "step": 13550
+    },
+    {
+      "epoch": 4.349216501439079,
+      "grad_norm": 0.061410386115312576,
+      "learning_rate": 2.8257323781501855e-05,
+      "loss": 0.0075,
+      "step": 13600
+    },
+    {
+      "epoch": 4.365206267988487,
+      "grad_norm": 0.06293186545372009,
+      "learning_rate": 2.8177369834975055e-05,
+      "loss": 0.0073,
+      "step": 13650
+    },
+    {
+      "epoch": 4.381196034537895,
+      "grad_norm": 0.062094636261463165,
+      "learning_rate": 2.8097415888448255e-05,
+      "loss": 0.0074,
+      "step": 13700
+    },
+    {
+      "epoch": 4.397185801087304,
+      "grad_norm": 0.10179385542869568,
+      "learning_rate": 2.801746194192145e-05,
+      "loss": 0.0077,
+      "step": 13750
+    },
+    {
+      "epoch": 4.413175567636713,
+      "grad_norm": 0.09233298152685165,
+      "learning_rate": 2.793750799539465e-05,
+      "loss": 0.0069,
+      "step": 13800
+    },
+    {
+      "epoch": 4.429165334186121,
+      "grad_norm": 0.08230157941579819,
+      "learning_rate": 2.785755404886785e-05,
+      "loss": 0.0069,
+      "step": 13850
+    },
+    {
+      "epoch": 4.445155100735529,
+      "grad_norm": 0.056463953107595444,
+      "learning_rate": 2.777760010234105e-05,
+      "loss": 0.0074,
+      "step": 13900
+    },
+    {
+      "epoch": 4.461144867284938,
+      "grad_norm": 0.06046191230416298,
+      "learning_rate": 2.7697646155814248e-05,
+      "loss": 0.0071,
+      "step": 13950
+    },
+    {
+      "epoch": 4.477134633834346,
+      "grad_norm": 0.06782487034797668,
+      "learning_rate": 2.7617692209287448e-05,
+      "loss": 0.0078,
+      "step": 14000
+    }
+  ],
+  "logging_steps": 50,
+  "max_steps": 31270,
+  "num_input_tokens_seen": 0,
+  "num_train_epochs": 10,
+  "save_steps": 2000,
+  "stateful_callbacks": {
+    "TrainerControl": {
+      "args": {
+        "should_epoch_stop": false,
+        "should_evaluate": false,
+        "should_log": false,
+        "should_save": true,
+        "should_training_stop": false
+      },
+      "attributes": {}
+    }
+  },
+  "total_flos": 1.095048815305556e+19,
+  "train_batch_size": 4,
+  "trial_name": null,
+  "trial_params": null
+}
diff --git a/weight_dir/loraWeight/trace_CRFLP/checkpoint-14000/training_args.bin b/weight_dir/loraWeight/trace_CRFLP/checkpoint-14000/training_args.bin
new file mode 100644
index 0000000000000000000000000000000000000000..6dbdff8b4afad36ed85290a57c3f67d54ed2560d
--- /dev/null
+++ b/weight_dir/loraWeight/trace_CRFLP/checkpoint-14000/training_args.bin
@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a2351918e4b9c08dd780eb0d2893ea88e9398fc16ba645474bf58d4db3583195
+size 7736
diff --git a/weight_dir/loraWeight/trace_CRFLP/checkpoint-14000/zero_to_fp32.py b/weight_dir/loraWeight/trace_CRFLP/checkpoint-14000/zero_to_fp32.py
new file mode 100644
index 0000000000000000000000000000000000000000..24cc342e78d1a006c782b3a4cd68d9ce786d8fd8
--- /dev/null
+++ b/weight_dir/loraWeight/trace_CRFLP/checkpoint-14000/zero_to_fp32.py
@@ -0,0 +1,604 @@
+#!/usr/bin/env python
+
+# Copyright (c) Microsoft Corporation.
+# SPDX-License-Identifier: Apache-2.0
+
+# DeepSpeed Team
+
+# This script extracts fp32 consolidated weights from a zero 1, 2 and 3 DeepSpeed checkpoints. It gets
+# copied into the top level checkpoint dir, so the user can easily do the conversion at any point in
+# the future. Once extracted, the weights don't require DeepSpeed and can be used in any
+# application.
+#
+# example: python zero_to_fp32.py . pytorch_model.bin
+
+import argparse
+import torch
+import glob
+import math
+import os
+import re
+from collections import OrderedDict
+from dataclasses import dataclass
+
+# while this script doesn't use deepspeed to recover data, since the checkpoints are pickled with
+# DeepSpeed data structures it has to be available in the current python environment.
+from deepspeed.utils import logger
+from deepspeed.checkpoint.constants import (DS_VERSION, OPTIMIZER_STATE_DICT, SINGLE_PARTITION_OF_FP32_GROUPS,
+                                            FP32_FLAT_GROUPS, ZERO_STAGE, PARTITION_COUNT, PARAM_SHAPES, BUFFER_NAMES,
+                                            FROZEN_PARAM_SHAPES, FROZEN_PARAM_FRAGMENTS)
+
+
+@dataclass
+class zero_model_state:
+    buffers: dict()
+    param_shapes: dict()
+    shared_params: list
+    ds_version: int
+    frozen_param_shapes: dict()
+    frozen_param_fragments: dict()
+
+
+debug = 0
+
+# load to cpu
+device = torch.device('cpu')
+
+
+def atoi(text):
+    return int(text) if text.isdigit() else text
+
+
+def natural_keys(text):
+    '''
+    alist.sort(key=natural_keys) sorts in human order
+    http://nedbatchelder.com/blog/200712/human_sorting.html
+    (See Toothy's implementation in the comments)
+    '''
+    return [atoi(c) for c in re.split(r'(\d+)', text)]
+
+
+def get_model_state_file(checkpoint_dir, zero_stage):
+    if not os.path.isdir(checkpoint_dir):
+        raise FileNotFoundError(f"Directory '{checkpoint_dir}' doesn't exist")
+
+    # there should be only one file
+    if zero_stage <= 2:
+        file = os.path.join(checkpoint_dir, "mp_rank_00_model_states.pt")
+    elif zero_stage == 3:
+        file = os.path.join(checkpoint_dir, "zero_pp_rank_0_mp_rank_00_model_states.pt")
+
+    if not os.path.exists(file):
+        raise FileNotFoundError(f"can't find model states file at '{file}'")
+
+    return file
+
+
+def get_checkpoint_files(checkpoint_dir, glob_pattern):
+    # XXX: need to test that this simple glob rule works for multi-node setup too
+    ckpt_files = sorted(glob.glob(os.path.join(checkpoint_dir, glob_pattern)), key=natural_keys)
+
+    if len(ckpt_files) == 0:
+        raise FileNotFoundError(f"can't find {glob_pattern} files in directory '{checkpoint_dir}'")
+
+    return ckpt_files
+
+
+def get_optim_files(checkpoint_dir):
+    return get_checkpoint_files(checkpoint_dir, "*_optim_states.pt")
+
+
+def get_model_state_files(checkpoint_dir):
+    return get_checkpoint_files(checkpoint_dir, "*_model_states.pt")
+
+
+def parse_model_states(files):
+    zero_model_states = []
+    for file in files:
+        state_dict = torch.load(file, map_location=device)
+
+        if BUFFER_NAMES not in state_dict:
+            raise ValueError(f"{file} is not a model state checkpoint")
+        buffer_names = state_dict[BUFFER_NAMES]
+        if debug:
+            print("Found buffers:", buffer_names)
+
+        # recover just the buffers while restoring them to fp32 if they were saved in fp16
+        buffers = {k: v.float() for k, v in state_dict["module"].items() if k in buffer_names}
+        param_shapes = state_dict[PARAM_SHAPES]
+
+        # collect parameters that are included in param_shapes
+        param_names = []
+        for s in param_shapes:
+            for name in s.keys():
+                param_names.append(name)
+
+        # update with frozen parameters
+        frozen_param_shapes = state_dict.get(FROZEN_PARAM_SHAPES, None)
+        if frozen_param_shapes is not None:
+            if debug:
+                print(f"Found frozen_param_shapes: {frozen_param_shapes}")
+            param_names += list(frozen_param_shapes.keys())
+
+        # handle shared params
+        shared_params = [[k, v] for k, v in state_dict["shared_params"].items()]
+
+        ds_version = state_dict.get(DS_VERSION, None)
+
+        frozen_param_fragments = state_dict.get(FROZEN_PARAM_FRAGMENTS, None)
+
+        z_model_state = zero_model_state(buffers=buffers,
+                                         param_shapes=param_shapes,
+                                         shared_params=shared_params,
+                                         ds_version=ds_version,
+                                         frozen_param_shapes=frozen_param_shapes,
+                                         frozen_param_fragments=frozen_param_fragments)
+        zero_model_states.append(z_model_state)
+
+    return zero_model_states
+
+
+def parse_optim_states(files, ds_checkpoint_dir):
+
+    total_files = len(files)
+    state_dicts = []
+    for f in files:
+        state_dict = torch.load(f, map_location=device)
+        # immediately discard the potentially huge 2 optimizer states as we only care for fp32 master weights
+        # and also handle the case where it was already removed by another helper script
+        state_dict["optimizer_state_dict"].pop("optimizer_state_dict", None)
+        state_dicts.append(state_dict)
+
+    if not ZERO_STAGE in state_dicts[0][OPTIMIZER_STATE_DICT]:
+        raise ValueError(f"{files[0]} is not a zero checkpoint")
+    zero_stage = state_dicts[0][OPTIMIZER_STATE_DICT][ZERO_STAGE]
+    world_size = state_dicts[0][OPTIMIZER_STATE_DICT][PARTITION_COUNT]
+
+    # For ZeRO-2 each param group can have different partition_count as data parallelism for expert
+    # parameters can be different from data parallelism for non-expert parameters. So we can just
+    # use the max of the partition_count to get the dp world_size.
+
+    if type(world_size) is list:
+        world_size = max(world_size)
+
+    if world_size != total_files:
+        raise ValueError(
+            f"Expected {world_size} of '*_optim_states.pt' under '{ds_checkpoint_dir}' but found {total_files} files. "
+            "Possibly due to an overwrite of an old checkpoint, or a checkpoint didn't get saved by one or more processes."
+        )
+
+    # the groups are named differently in each stage
+    if zero_stage <= 2:
+        fp32_groups_key = SINGLE_PARTITION_OF_FP32_GROUPS
+    elif zero_stage == 3:
+        fp32_groups_key = FP32_FLAT_GROUPS
+    else:
+        raise ValueError(f"unknown zero stage {zero_stage}")
+
+    if zero_stage <= 2:
+        fp32_flat_groups = [state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key] for i in range(len(state_dicts))]
+    elif zero_stage == 3:
+        # if there is more than one param group, there will be multiple flattened tensors - one
+        # flattened tensor per group - for simplicity merge them into a single tensor
+        #
+        # XXX: could make the script more memory efficient for when there are multiple groups - it
+        # will require matching the sub-lists of param_shapes for each param group flattened tensor
+
+        fp32_flat_groups = [
+            torch.cat(state_dicts[i][OPTIMIZER_STATE_DICT][fp32_groups_key], 0) for i in range(len(state_dicts))
+        ]
+
+    return zero_stage, world_size, fp32_flat_groups
+
+
+def _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir, exclude_frozen_parameters):
+    """
+    Returns fp32 state_dict reconstructed from ds checkpoint
+
+    Args:
+        - ``ds_checkpoint_dir``: path to the deepspeed checkpoint folder (where the optimizer files are)
+
+    """
+    print(f"Processing zero checkpoint '{ds_checkpoint_dir}'")
+
+    optim_files = get_optim_files(ds_checkpoint_dir)
+    zero_stage, world_size, fp32_flat_groups = parse_optim_states(optim_files, ds_checkpoint_dir)
+    print(f"Detected checkpoint of type zero stage {zero_stage}, world_size: {world_size}")
+
+    model_files = get_model_state_files(ds_checkpoint_dir)
+
+    zero_model_states = parse_model_states(model_files)
+    print(f'Parsing checkpoint created by deepspeed=={zero_model_states[0].ds_version}')
+
+    if zero_stage <= 2:
+        return _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                                          exclude_frozen_parameters)
+    elif zero_stage == 3:
+        return _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                                          exclude_frozen_parameters)
+
+
+def _zero2_merge_frozen_params(state_dict, zero_model_states):
+    if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
+        return
+
+    frozen_param_shapes = zero_model_states[0].frozen_param_shapes
+    frozen_param_fragments = zero_model_states[0].frozen_param_fragments
+
+    if debug:
+        num_elem = sum(s.numel() for s in frozen_param_shapes.values())
+        print(f'rank 0: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
+
+        wanted_params = len(frozen_param_shapes)
+        wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
+        avail_numel = sum([p.numel() for p in frozen_param_fragments.values()])
+        print(f'Frozen params: Have {avail_numel} numels to process.')
+        print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
+
+    total_params = 0
+    total_numel = 0
+    for name, shape in frozen_param_shapes.items():
+        total_params += 1
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+
+        state_dict[name] = frozen_param_fragments[name]
+
+        if debug:
+            print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
+
+    print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _has_callable(obj, fn):
+    attr = getattr(obj, fn, None)
+    return callable(attr)
+
+
+def _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
+    param_shapes = zero_model_states[0].param_shapes
+
+    # Reconstruction protocol:
+    #
+    # XXX: document this
+
+    if debug:
+        for i in range(world_size):
+            for j in range(len(fp32_flat_groups[0])):
+                print(f"{FP32_FLAT_GROUPS}[{i}][{j}].shape={fp32_flat_groups[i][j].shape}")
+
+    # XXX: memory usage doubles here (zero2)
+    num_param_groups = len(fp32_flat_groups[0])
+    merged_single_partition_of_fp32_groups = []
+    for i in range(num_param_groups):
+        merged_partitions = [sd[i] for sd in fp32_flat_groups]
+        full_single_fp32_vector = torch.cat(merged_partitions, 0)
+        merged_single_partition_of_fp32_groups.append(full_single_fp32_vector)
+    avail_numel = sum(
+        [full_single_fp32_vector.numel() for full_single_fp32_vector in merged_single_partition_of_fp32_groups])
+
+    if debug:
+        wanted_params = sum([len(shapes) for shapes in param_shapes])
+        wanted_numel = sum([sum(shape.numel() for shape in shapes.values()) for shapes in param_shapes])
+        # not asserting if there is a mismatch due to possible padding
+        print(f"Have {avail_numel} numels to process.")
+        print(f"Need {wanted_numel} numels in {wanted_params} params.")
+
+    # params
+    # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
+    # out-of-core computing solution
+    total_numel = 0
+    total_params = 0
+    for shapes, full_single_fp32_vector in zip(param_shapes, merged_single_partition_of_fp32_groups):
+        offset = 0
+        avail_numel = full_single_fp32_vector.numel()
+        for name, shape in shapes.items():
+
+            unpartitioned_numel = shape.numel() if _has_callable(shape, 'numel') else math.prod(shape)
+            total_numel += unpartitioned_numel
+            total_params += 1
+
+            if debug:
+                print(f"{name} full shape: {shape} unpartitioned numel {unpartitioned_numel} ")
+            state_dict[name] = full_single_fp32_vector.narrow(0, offset, unpartitioned_numel).view(shape)
+            offset += unpartitioned_numel
+
+        # Z2 started to align to 2*world_size to improve nccl performance. Therefore both offset and
+        # avail_numel can differ by anywhere between 0..2*world_size. Due to two unrelated complex
+        # paddings performed in the code it's almost impossible to predict the exact numbers w/o the
+        # live optimizer object, so we are checking that the numbers are within the right range
+        align_to = 2 * world_size
+
+        def zero2_align(x):
+            return align_to * math.ceil(x / align_to)
+
+        if debug:
+            print(f"original offset={offset}, avail_numel={avail_numel}")
+
+        offset = zero2_align(offset)
+        avail_numel = zero2_align(avail_numel)
+
+        if debug:
+            print(f"aligned  offset={offset}, avail_numel={avail_numel}")
+
+        # Sanity check
+        if offset != avail_numel:
+            raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
+
+    print(f"Reconstructed fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _get_fp32_state_dict_from_zero2_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                               exclude_frozen_parameters):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    if not exclude_frozen_parameters:
+        _zero2_merge_frozen_params(state_dict, zero_model_states)
+
+    _zero2_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
+
+    # recover shared parameters
+    for pair in zero_model_states[0].shared_params:
+        if pair[1] in state_dict:
+            state_dict[pair[0]] = state_dict[pair[1]]
+
+    return state_dict
+
+
+def zero3_partitioned_param_info(unpartitioned_numel, world_size):
+    remainder = unpartitioned_numel % world_size
+    padding_numel = (world_size - remainder) if remainder else 0
+    partitioned_numel = math.ceil(unpartitioned_numel / world_size)
+    return partitioned_numel, padding_numel
+
+
+def _zero3_merge_frozen_params(state_dict, world_size, zero_model_states):
+    if zero_model_states[0].frozen_param_shapes is None or len(zero_model_states[0].frozen_param_shapes) == 0:
+        return
+
+    if debug:
+        for i in range(world_size):
+            num_elem = sum(s.numel() for s in zero_model_states[i].frozen_param_fragments.values())
+            print(f'rank {i}: {FROZEN_PARAM_SHAPES}.numel = {num_elem}')
+
+        frozen_param_shapes = zero_model_states[0].frozen_param_shapes
+        wanted_params = len(frozen_param_shapes)
+        wanted_numel = sum(s.numel() for s in frozen_param_shapes.values())
+        avail_numel = sum([p.numel() for p in zero_model_states[0].frozen_param_fragments.values()]) * world_size
+        print(f'Frozen params: Have {avail_numel} numels to process.')
+        print(f'Frozen params: Need {wanted_numel} numels in {wanted_params} params')
+
+    total_params = 0
+    total_numel = 0
+    for name, shape in zero_model_states[0].frozen_param_shapes.items():
+        total_params += 1
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+
+        param_frags = tuple(model_state.frozen_param_fragments[name] for model_state in zero_model_states)
+        state_dict[name] = torch.cat(param_frags, 0).narrow(0, 0, unpartitioned_numel).view(shape)
+
+        partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
+
+        if debug:
+            print(
+                f"Frozen params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
+            )
+
+    print(f"Reconstructed Frozen fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states):
+    param_shapes = zero_model_states[0].param_shapes
+    avail_numel = fp32_flat_groups[0].numel() * world_size
+    # Reconstruction protocol: For zero3 we need to zip the partitions together at boundary of each
+    # param, re-consolidating each param, while dealing with padding if any
+
+    # merge list of dicts, preserving order
+    param_shapes = {k: v for d in param_shapes for k, v in d.items()}
+
+    if debug:
+        for i in range(world_size):
+            print(f"{FP32_FLAT_GROUPS}[{i}].shape={fp32_flat_groups[i].shape}")
+
+        wanted_params = len(param_shapes)
+        wanted_numel = sum(shape.numel() for shape in param_shapes.values())
+        # not asserting if there is a mismatch due to possible padding
+        avail_numel = fp32_flat_groups[0].numel() * world_size
+        print(f"Trainable params: Have {avail_numel} numels to process.")
+        print(f"Trainable params: Need {wanted_numel} numels in {wanted_params} params.")
+
+    # params
+    # XXX: for huge models that can't fit into the host's RAM we will have to recode this to support
+    # out-of-core computing solution
+    offset = 0
+    total_numel = 0
+    total_params = 0
+    for name, shape in param_shapes.items():
+
+        unpartitioned_numel = shape.numel()
+        total_numel += unpartitioned_numel
+        total_params += 1
+
+        partitioned_numel, partitioned_padding_numel = zero3_partitioned_param_info(unpartitioned_numel, world_size)
+
+        if debug:
+            print(
+                f"Trainable params: {total_params} {name} full shape: {shape} partition0 numel={partitioned_numel} partitioned_padding_numel={partitioned_padding_numel}"
+            )
+
+        # XXX: memory usage doubles here
+        state_dict[name] = torch.cat(
+            tuple(fp32_flat_groups[i].narrow(0, offset, partitioned_numel) for i in range(world_size)),
+            0).narrow(0, 0, unpartitioned_numel).view(shape)
+        offset += partitioned_numel
+
+    offset *= world_size
+
+    # Sanity check
+    if offset != avail_numel:
+        raise ValueError(f"consumed {offset} numels out of {avail_numel} - something is wrong")
+
+    print(f"Reconstructed Trainable fp32 state dict with {total_params} params {total_numel} elements")
+
+
+def _get_fp32_state_dict_from_zero3_checkpoint(world_size, fp32_flat_groups, zero_model_states,
+                                               exclude_frozen_parameters):
+    state_dict = OrderedDict()
+
+    # buffers
+    buffers = zero_model_states[0].buffers
+    state_dict.update(buffers)
+    if debug:
+        print(f"added {len(buffers)} buffers")
+
+    if not exclude_frozen_parameters:
+        _zero3_merge_frozen_params(state_dict, world_size, zero_model_states)
+
+    _zero3_merge_trainable_params(state_dict, world_size, fp32_flat_groups, zero_model_states)
+
+    # recover shared parameters
+    for pair in zero_model_states[0].shared_params:
+        if pair[1] in state_dict:
+            state_dict[pair[0]] = state_dict[pair[1]]
+
+    return state_dict
+
+
+def get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag=None, exclude_frozen_parameters=False):
+    """
+    Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated state_dict that can be loaded with
+    ``load_state_dict()`` and used for training without DeepSpeed or shared with others, for example
+    via a model hub.
+
+    Args:
+        - ``checkpoint_dir``: path to the desired checkpoint folder
+        - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in 'latest' file. e.g., ``global_step14``
+        - ``exclude_frozen_parameters``: exclude frozen parameters
+
+    Returns:
+        - pytorch ``state_dict``
+
+    Note: this approach may not work if your application doesn't have sufficient free CPU memory and
+    you may need to use the offline approach using the ``zero_to_fp32.py`` script that is saved with
+    the checkpoint.
+
+    A typical usage might be ::
+
+        from deepspeed.utils.zero_to_fp32 import get_fp32_state_dict_from_zero_checkpoint
+        # do the training and checkpoint saving
+        state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir) # already on cpu
+        model = model.cpu() # move to cpu
+        model.load_state_dict(state_dict)
+        # submit to model hub or save the model to share with others
+
+    In this example the ``model`` will no longer be usable in the deepspeed context of the same
+    application. i.e. you will need to re-initialize the deepspeed engine, since
+    ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
+
+    If you want it all done for you, use ``load_state_dict_from_zero_checkpoint`` instead.
+
+    """
+    if tag is None:
+        latest_path = os.path.join(checkpoint_dir, 'latest')
+        if os.path.isfile(latest_path):
+            with open(latest_path, 'r') as fd:
+                tag = fd.read().strip()
+        else:
+            raise ValueError(f"Unable to find 'latest' file at {latest_path}")
+
+    ds_checkpoint_dir = os.path.join(checkpoint_dir, tag)
+
+    if not os.path.isdir(ds_checkpoint_dir):
+        raise FileNotFoundError(f"Directory '{ds_checkpoint_dir}' doesn't exist")
+
+    return _get_fp32_state_dict_from_zero_checkpoint(ds_checkpoint_dir, exclude_frozen_parameters)
+
+
+def convert_zero_checkpoint_to_fp32_state_dict(checkpoint_dir, output_file, tag=None, exclude_frozen_parameters=False):
+    """
+    Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict`` file that can be
+    loaded with ``torch.load(file)`` + ``load_state_dict()`` and used for training without DeepSpeed.
+
+    Args:
+        - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
+        - ``output_file``: path to the pytorch fp32 state_dict output file (e.g. path/pytorch_model.bin)
+        - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
+        - ``exclude_frozen_parameters``: exclude frozen parameters
+    """
+
+    state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag, exclude_frozen_parameters)
+    print(f"Saving fp32 state dict to {output_file}")
+    torch.save(state_dict, output_file)
+
+
+def load_state_dict_from_zero_checkpoint(model, checkpoint_dir, tag=None):
+    """
+    1. Put the provided model to cpu
+    2. Convert ZeRO 2 or 3 checkpoint into a single fp32 consolidated ``state_dict``
+    3. Load it into the provided model
+
+    Args:
+        - ``model``: the model object to update
+        - ``checkpoint_dir``: path to the desired checkpoint folder. (one that contains the tag-folder, like ``global_step14``)
+        - ``tag``: checkpoint tag used as a unique identifier for checkpoint. If not provided will attempt to load tag in the file named ``latest`` in the checkpoint folder, e.g., ``global_step14``
+
+    Returns:
+        - ``model`: modified model
+
+    Make sure you have plenty of CPU memory available before you call this function. If you don't
+    have enough use the ``zero_to_fp32.py`` utility to do the conversion. You will find it
+    conveniently placed for you in the checkpoint folder.
+
+    A typical usage might be ::
+
+        from deepspeed.utils.zero_to_fp32 import load_state_dict_from_zero_checkpoint
+        model = load_state_dict_from_zero_checkpoint(trainer.model, checkpoint_dir)
+        # submit to model hub or save the model to share with others
+
+    Note, that once this was run, the ``model`` will no longer be usable in the deepspeed context
+    of the same application. i.e. you will need to re-initialize the deepspeed engine, since
+    ``model.load_state_dict(state_dict)`` will remove all the deepspeed magic from it.
+
+    """
+    logger.info(f"Extracting fp32 weights")
+    state_dict = get_fp32_state_dict_from_zero_checkpoint(checkpoint_dir, tag)
+
+    logger.info(f"Overwriting model with fp32 weights")
+    model = model.cpu()
+    model.load_state_dict(state_dict, strict=False)
+
+    return model
+
+
+if __name__ == "__main__":
+
+    parser = argparse.ArgumentParser()
+    parser.add_argument("checkpoint_dir",
+                        type=str,
+                        help="path to the desired checkpoint folder, e.g., path/checkpoint-12")
+    parser.add_argument(
+        "output_file",
+        type=str,
+        help="path to the pytorch fp32 state_dict output file (e.g. path/checkpoint-12/pytorch_model.bin)")
+    parser.add_argument("-t",
+                        "--tag",
+                        type=str,
+                        default=None,
+                        help="checkpoint tag used as a unique identifier for checkpoint. e.g., global_step1")
+    parser.add_argument("--exclude_frozen_parameters", action='store_true', help="exclude frozen parameters")
+    parser.add_argument("-d", "--debug", action='store_true', help="enable debug")
+    args = parser.parse_args()
+
+    debug = args.debug
+
+    convert_zero_checkpoint_to_fp32_state_dict(args.checkpoint_dir,
+                                               args.output_file,
+                                               tag=args.tag,
+                                               exclude_frozen_parameters=args.exclude_frozen_parameters)