Upload full trainer checkpoint step=1013 (profile=llama3_1_8b_base_grpo, run_type=grpo)

.gitattributes

@@ -146,3 +146,4 @@ meta-llama__meta-llama-3.1-8b/sft/checkpoints/checkpoint-20524/tokenizer.json fi
 meta-llama__meta-llama-3.1-8b/sft/checkpoints/checkpoint-23456/tokenizer.json filter=lfs diff=lfs merge=lfs -text
 meta-llama__meta-llama-3.1-8b/sft/checkpoints/checkpoint-23460/tokenizer.json filter=lfs diff=lfs merge=lfs -text
 meta-llama__meta-llama-3.1-8b/sft/epoch_0004/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+meta-llama__meta-llama-3.1-8b/grpo/checkpoints/checkpoint-1013/tokenizer.json filter=lfs diff=lfs merge=lfs -text

meta-llama__meta-llama-3.1-8b/grpo/checkpoints/checkpoint-1013/README.md

@@ -0,0 +1,210 @@
+---
+base_model: unsloth/Meta-Llama-3.1-8B
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:unsloth/Meta-Llama-3.1-8B
+- grpo
+- lora
+- transformers
+- trl
+- unsloth
+---
+
+# 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 0.18.1

meta-llama__meta-llama-3.1-8b/grpo/checkpoints/checkpoint-1013/adapter_config.json

@@ -0,0 +1,50 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": {
+    "base_model_class": "LlamaForCausalLM",
+    "parent_library": "transformers.models.llama.modeling_llama",
+    "unsloth_fixed": true
+  },
+  "base_model_name_or_path": "unsloth/Meta-Llama-3.1-8B",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "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_bias": false,
+  "lora_dropout": 0.0,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 8,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "o_proj",
+    "down_proj",
+    "up_proj",
+    "gate_proj",
+    "k_proj",
+    "v_proj",
+    "q_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}

meta-llama__meta-llama-3.1-8b/grpo/checkpoints/checkpoint-1013/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:52850ef35da75893b5cb979a297423e3e470574cc1f050b1411e8fd6715818df
+size 83945296

meta-llama__meta-llama-3.1-8b/grpo/checkpoints/checkpoint-1013/chat_template.jinja

@@ -0,0 +1,12 @@
+{{ bos_token }}{% if messages[0]['role'] == 'system' %}{{ messages[0]['content'] + '
+
+' }}{% set loop_messages = messages[1:] %}{% else %}{{ '' + '
+
+' }}{% set loop_messages = messages %}{% endif %}{% for message in loop_messages %}{% if message['role'] == 'user' %}{{ '### Instruction:
+' + message['content'] + '
+
+' }}{% elif message['role'] == 'assistant' %}{{ '### Response:
+' + message['content'] + eos_token + '
+
+' }}{% else %}{{ raise_exception('Only user and assistant roles are supported!') }}{% endif %}{% endfor %}{% if add_generation_prompt %}{{ '### Response:
+' }}{% endif %}

meta-llama__meta-llama-3.1-8b/grpo/checkpoints/checkpoint-1013/optimizer.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a17489bbf0a31cd7a13c216ef1efe9f0b4a1025fe43b37ea677de5acc61f9f8d
+size 43127973

meta-llama__meta-llama-3.1-8b/grpo/checkpoints/checkpoint-1013/rng_state.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:4ec2edc9c5368ff1fbe2c50e330e81c6889657bd380d6c780013dd4ade8eecb3
+size 14645

meta-llama__meta-llama-3.1-8b/grpo/checkpoints/checkpoint-1013/scheduler.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:807e8f84a2e7d1df2da3001a65774d078dd10e19443d4ec22c453ebd9515f5eb
+size 1465

meta-llama__meta-llama-3.1-8b/grpo/checkpoints/checkpoint-1013/tokenizer.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:a65c6c5f9764771aa485e6a1f5e63d7d9af8477fe0777148c17476ecb2e09a05
+size 17210099

meta-llama__meta-llama-3.1-8b/grpo/checkpoints/checkpoint-1013/tokenizer_config.json

@@ -0,0 +1,18 @@
+{
+  "backend": "tokenizers",
+  "bos_token": "<|begin_of_text|>",
+  "clean_up_tokenization_spaces": true,
+  "eos_token": "<|end_of_text|>",
+  "from_slow": true,
+  "is_local": false,
+  "legacy": false,
+  "model_input_names": [
+    "input_ids",
+    "attention_mask"
+  ],
+  "model_max_length": 131072,
+  "pad_token": "<|finetune_right_pad_id|>",
+  "padding_side": "right",
+  "tokenizer_class": "TokenizersBackend",
+  "unk_token": null
+}

meta-llama__meta-llama-3.1-8b/grpo/checkpoints/checkpoint-1013/trainer_state.json

@@ -0,0 +1,1123 @@
+{
+  "best_global_step": null,
+  "best_metric": null,
+  "best_model_checkpoint": null,
+  "epoch": 0.001799257562032646,
+  "eval_steps": 500,
+  "global_step": 1013,
+  "is_hyper_param_search": false,
+  "is_local_process_zero": true,
+  "is_world_process_zero": true,
+  "log_history": [
+    {
+      "clip_ratio/high_max": 0.0,
+      "clip_ratio/high_mean": 0.0,
+      "clip_ratio/low_mean": 0.0,
+      "clip_ratio/low_min": 0.0,
+      "clip_ratio/region_mean": 0.0,
+      "completion_length": 65.81,
+      "completions/clipped_ratio": 0.0,
+      "completions/max_length": 66.76923076923077,
+      "completions/max_terminated_length": 66.76923076923077,
+      "completions/mean_length": 66.54807692307692,
+      "completions/mean_terminated_length": 66.54807692307692,
+      "completions/min_length": 66.3076923076923,
+      "completions/min_terminated_length": 66.3076923076923,
+      "epoch": 4.440418465036145e-05,
+      "frac_reward_zero_std": 0.5384615384615384,
+      "grad_norm": 0.025798887014389038,
+      "kl": 0.18038831524550913,
+      "learning_rate": 1.0656956994738129e-09,
+      "loss": -0.0065249383449554445,
+      "num_tokens": 24649.0,
+      "reward": 5.7312158437875604e-09,
+      "reward_std": 0.4934053604419415,
+      "rewards/TranscriptCorrectionGrpoReward/mean": 5.7312158437875604e-09,
+      "rewards/TranscriptCorrectionGrpoReward/std": 0.4934053604419415,
+      "step": 25
+    },
+    {
+      "clip_ratio/high_max": 0.0,
+      "clip_ratio/high_mean": 0.0,
+      "clip_ratio/low_mean": 0.0,
+      "clip_ratio/low_min": 0.0,
+      "clip_ratio/region_mean": 0.0,
+      "completion_length": 65.5,
+      "completions/clipped_ratio": 0.0,
+      "completions/max_length": 64.75,
+      "completions/max_terminated_length": 64.75,
+      "completions/mean_length": 64.6875,
+      "completions/mean_terminated_length": 64.6875,
+      "completions/min_length": 64.5,
+      "completions/min_terminated_length": 64.5,
+      "epoch": 8.88083693007229e-05,
+      "frac_reward_zero_std": 0.8333333333333334,
+      "grad_norm": 0.27390509843826294,
+      "kl": 0.16198864802718163,
+      "learning_rate": 2.175795386425701e-09,
+      "loss": 0.008163155317306518,
+      "num_tokens": 46547.0,
+      "reward": 2.483526865641276e-09,
+      "reward_std": 0.17817416787147522,
+      "rewards/TranscriptCorrectionGrpoReward/mean": 2.483526865641276e-09,
+      "rewards/TranscriptCorrectionGrpoReward/std": 0.17817415793736777,
+      "step": 50
+    },
+    {
+      "clip_ratio/high_max": 0.0,
+      "clip_ratio/high_mean": 0.0,
+      "clip_ratio/low_mean": 0.0,
+      "clip_ratio/low_min": 0.0,
+      "clip_ratio/region_mean": 0.0,
+      "completion_length": 52.52,
+      "completions/clipped_ratio": 0.0,
+      "completions/max_length": 52.61538461538461,
+      "completions/max_terminated_length": 52.61538461538461,
+      "completions/mean_length": 52.42307692307692,
+      "completions/mean_terminated_length": 52.42307692307692,
+      "completions/min_length": 52.30769230769231,
+      "completions/min_terminated_length": 52.30769230769231,
+      "epoch": 0.00013321255395108434,
+      "frac_reward_zero_std": 0.8461538461538461,
+      "grad_norm": 1.6365996771128266e-06,
+      "kl": 0.20123317912220956,
+      "learning_rate": 3.285895073377589e-09,
+      "loss": 0.0007231433689594269,
+      "num_tokens": 66431.0,
+      "reward": 0.0,
+      "reward_std": 0.16446845348064715,
+      "rewards/TranscriptCorrectionGrpoReward/mean": 0.0,
+      "rewards/TranscriptCorrectionGrpoReward/std": 0.16446845348064715,
+      "step": 75
+    },
+    {
+      "clip_ratio/high_max": 0.0,
+      "clip_ratio/high_mean": 0.0,
+      "clip_ratio/low_mean": 0.0,
+      "clip_ratio/low_min": 0.0,
+      "clip_ratio/region_mean": 0.0,
+      "completion_length": 54.32,
+      "completions/clipped_ratio": 0.0,
+      "completions/max_length": 54.5,
+      "completions/max_terminated_length": 54.5,
+      "completions/mean_length": 54.5,
+      "completions/mean_terminated_length": 54.5,
+      "completions/min_length": 54.5,
+      "completions/min_terminated_length": 54.5,
+      "epoch": 0.0001776167386014458,
+      "frac_reward_zero_std": 0.5833333333333334,
+      "grad_norm": 0.00011817843187600374,
+      "kl": 0.253861972540617,
+      "learning_rate": 4.3959947603294776e-09,
+      "loss": 0.0025739344954490663,
+      "num_tokens": 85543.0,
+      "reward": 3.725290298461914e-09,
+      "reward_std": 0.4454353948434194,
+      "rewards/TranscriptCorrectionGrpoReward/mean": 3.725290298461914e-09,
+      "rewards/TranscriptCorrectionGrpoReward/std": 0.4454353948434194,
+      "step": 100
+    },
+    {
+      "clip_ratio/high_max": 0.0,
+      "clip_ratio/high_mean": 0.0,
+      "clip_ratio/low_mean": 0.0,
+      "clip_ratio/low_min": 0.0,
+      "clip_ratio/region_mean": 0.0,
+      "completion_length": 51.7,
+      "completions/clipped_ratio": 0.0,
+      "completions/max_length": 51.0,
+      "completions/max_terminated_length": 51.0,
+      "completions/mean_length": 50.86538461538461,
+      "completions/mean_terminated_length": 50.86538461538461,
+      "completions/min_length": 50.38461538461539,
+      "completions/min_terminated_length": 50.38461538461539,
+      "epoch": 0.00022202092325180725,
+      "frac_reward_zero_std": 0.5384615384615384,
+      "grad_norm": 0.3782321810722351,
+      "kl": 0.24278869599103928,
+      "learning_rate": 5.506094447281366e-09,
+      "loss": 0.0006042173504829406,
+      "num_tokens": 105273.0,
+      "reward": 2.292486337515024e-09,
+      "reward_std": 0.49340536961188686,
+      "rewards/TranscriptCorrectionGrpoReward/mean": 2.292486337515024e-09,
+      "rewards/TranscriptCorrectionGrpoReward/std": 0.4934053604419415,
+      "step": 125
+    },
+    {
+      "clip_ratio/high_max": 0.0,
+      "clip_ratio/high_mean": 0.0,
+      "clip_ratio/low_mean": 0.0,
+      "clip_ratio/low_min": 0.0,
+      "clip_ratio/region_mean": 0.0,
+      "completion_length": 73.29,
+      "completions/clipped_ratio": 0.0,
+      "completions/max_length": 75.33333333333333,
+      "completions/max_terminated_length": 75.33333333333333,
+      "completions/mean_length": 75.09375,
+      "completions/mean_terminated_length": 75.09375,
+      "completions/min_length": 74.91666666666667,
+      "completions/min_terminated_length": 74.91666666666667,
+      "epoch": 0.0002664251079021687,
+      "frac_reward_zero_std": 0.6666666666666666,
+      "grad_norm": 0.1665542721748352,
+      "kl": 0.20266253776848317,
+      "learning_rate": 6.6161941342332544e-09,
+      "loss": 0.001955866813659668,
+      "num_tokens": 130370.0,
+      "reward": 4.967053731282552e-09,
+      "reward_std": 0.35634831587473553,
+      "rewards/TranscriptCorrectionGrpoReward/mean": 4.967053731282552e-09,
+      "rewards/TranscriptCorrectionGrpoReward/std": 0.35634831587473553,
+      "step": 150
+    },
+    {
+      "clip_ratio/high_max": 0.0,
+      "clip_ratio/high_mean": 0.0,
+      "clip_ratio/low_mean": 0.0,
+      "clip_ratio/low_min": 0.0,
+      "clip_ratio/region_mean": 0.0,
+      "completion_length": 57.74,
+      "completions/clipped_ratio": 0.0,
+      "completions/max_length": 58.0,
+      "completions/max_terminated_length": 58.0,
+      "completions/mean_length": 57.94230769230769,
+      "completions/mean_terminated_length": 57.94230769230769,
+      "completions/min_length": 57.84615384615385,
+      "completions/min_terminated_length": 57.84615384615385,
+      "epoch": 0.00031082929255253015,
+      "frac_reward_zero_std": 0.6153846153846154,
+      "grad_norm": 0.22071869671344757,
+      "kl": 0.20556429855525493,
+      "learning_rate": 7.726293821185143e-09,
+      "loss": 0.0019431695342063903,
+      "num_tokens": 152108.0,
+      "reward": 5.7312158437875604e-09,
+      "reward_std": 0.4111711337016179,
+      "rewards/TranscriptCorrectionGrpoReward/mean": 5.7312158437875604e-09,
+      "rewards/TranscriptCorrectionGrpoReward/std": 0.4111711337016179,
+      "step": 175
+    },
+    {
+      "clip_ratio/high_max": 0.0,
+      "clip_ratio/high_mean": 0.0,
+      "clip_ratio/low_mean": 0.0,
+      "clip_ratio/low_min": 0.0,
+      "clip_ratio/region_mean": 0.0,
+      "completion_length": 66.59,
+      "completions/clipped_ratio": 0.0,
+      "completions/max_length": 66.91666666666667,
+      "completions/max_terminated_length": 66.91666666666667,
+      "completions/mean_length": 66.73958333333333,
+      "completions/mean_terminated_length": 66.73958333333333,
+      "completions/min_length": 66.33333333333333,
+      "completions/min_terminated_length": 66.33333333333333,
+      "epoch": 0.0003552334772028916,
+      "frac_reward_zero_std": 0.5,
+      "grad_norm": 6.216925248736516e-05,
+      "kl": 0.17584349907934665,
+      "learning_rate": 8.83639350813703e-09,
+      "loss": 0.0011808258295059205,
+      "num_tokens": 174947.0,
+      "reward": 4.346172014872233e-09,
+      "reward_std": 0.5345224738121033,
+      "rewards/TranscriptCorrectionGrpoReward/mean": 4.346172014872233e-09,
+      "rewards/TranscriptCorrectionGrpoReward/std": 0.5345224738121033,
+      "step": 200
+    },
+    {
+      "clip_ratio/high_max": 0.0,
+      "clip_ratio/high_mean": 0.0,
+      "clip_ratio/low_mean": 0.0,
+      "clip_ratio/low_min": 0.0,
+      "clip_ratio/region_mean": 0.0,
+      "completion_length": 65.44,
+      "completions/clipped_ratio": 0.0,
+      "completions/max_length": 65.92307692307692,
+      "completions/max_terminated_length": 65.92307692307692,
+      "completions/mean_length": 65.61538461538461,
+      "completions/mean_terminated_length": 65.61538461538461,
+      "completions/min_length": 65.38461538461539,
+      "completions/min_terminated_length": 65.38461538461539,
+      "epoch": 0.000399637661853253,
+      "frac_reward_zero_std": 0.6153846153846154,
+      "grad_norm": 0.12366796284914017,
+      "kl": 0.19227153606712818,
+      "learning_rate": 9.94649319508892e-09,
+      "loss": 0.0005045509338378906,
+      "num_tokens": 199651.0,
+      "reward": 1.7193647531362681e-09,
+      "reward_std": 0.4111711337016179,
+      "rewards/TranscriptCorrectionGrpoReward/mean": 1.7193647531362681e-09,
+      "rewards/TranscriptCorrectionGrpoReward/std": 0.4111711337016179,
+      "step": 225
+    },
+    {
+      "clip_ratio/high_max": 0.0,
+      "clip_ratio/high_mean": 0.0,
+      "clip_ratio/low_mean": 0.0,
+      "clip_ratio/low_min": 0.0,
+      "clip_ratio/region_mean": 0.0,
+      "completion_length": 66.15,
+      "completions/clipped_ratio": 0.0,
+      "completions/max_length": 66.25,
+      "completions/max_terminated_length": 66.25,
+      "completions/mean_length": 65.98958333333333,
+      "completions/mean_terminated_length": 65.98958333333333,
+      "completions/min_length": 65.83333333333333,
+      "completions/min_terminated_length": 65.83333333333333,
+      "epoch": 0.0004440418465036145,
+      "frac_reward_zero_std": 0.5833333333333334,
+      "grad_norm": 5.099513509776443e-05,
+      "kl": 0.19273025602102278,
+      "learning_rate": 1.1056592882040807e-08,
+      "loss": 0.002573988139629364,
+      "num_tokens": 222250.0,
+      "reward": 4.346172014872233e-09,
+      "reward_std": 0.4454353948434194,
+      "rewards/TranscriptCorrectionGrpoReward/mean": 4.346172014872233e-09,
+      "rewards/TranscriptCorrectionGrpoReward/std": 0.4454353948434194,
+      "step": 250
+    },
+    {
+      "clip_ratio/high_max": 0.0,
+      "clip_ratio/high_mean": 0.0,
+      "clip_ratio/low_mean": 0.0,
+      "clip_ratio/low_min": 0.0,
+      "clip_ratio/region_mean": 0.0,
+      "completion_length": 64.71,
+      "completions/clipped_ratio": 0.0,
+      "completions/max_length": 66.92307692307692,
+      "completions/max_terminated_length": 66.92307692307692,
+      "completions/mean_length": 66.6826923076923,
+      "completions/mean_terminated_length": 66.6826923076923,
+      "completions/min_length": 66.38461538461539,
+      "completions/min_terminated_length": 66.38461538461539,
+      "epoch": 0.0004884460311539759,
+      "frac_reward_zero_std": 0.5384615384615384,
+      "grad_norm": 0.04118318483233452,
+      "kl": 0.17756690874695777,
+      "learning_rate": 1.2166692568992695e-08,
+      "loss": -0.0014105629920959472,
+      "num_tokens": 247777.0,
+      "reward": 5.73121584378756e-10,
+      "reward_std": 0.49340535127199614,
+      "rewards/TranscriptCorrectionGrpoReward/mean": 5.73121584378756e-10,
+      "rewards/TranscriptCorrectionGrpoReward/std": 0.4934053604419415,
+      "step": 275
+    },
+    {
+      "clip_ratio/high_max": 0.0,
+      "clip_ratio/high_mean": 0.0,
+      "clip_ratio/low_mean": 0.0,
+      "clip_ratio/low_min": 0.0,
+      "clip_ratio/region_mean": 0.0,
+      "completion_length": 72.83,
+      "completions/clipped_ratio": 0.0,
+      "completions/max_length": 71.25,
+      "completions/max_terminated_length": 71.25,
+      "completions/mean_length": 71.03125,
+      "completions/mean_terminated_length": 71.03125,
+      "completions/min_length": 70.91666666666667,
+      "completions/min_terminated_length": 70.91666666666667,
+      "epoch": 0.0005328502158043374,
+      "frac_reward_zero_std": 0.75,
+      "grad_norm": 7.932442713354249e-06,
+      "kl": 0.16507803082466124,
+      "learning_rate": 1.3276792255944584e-08,
+      "loss": 0.005102531909942627,
+      "num_tokens": 271724.0,
+      "reward": 3.725290298461914e-09,
+      "reward_std": 0.2672612468401591,
+      "rewards/TranscriptCorrectionGrpoReward/mean": 3.725290298461914e-09,
+      "rewards/TranscriptCorrectionGrpoReward/std": 0.26726123690605164,
+      "step": 300
+    },
+    {
+      "clip_ratio/high_max": 0.0,
+      "clip_ratio/high_mean": 0.0,
+      "clip_ratio/low_mean": 0.0,
+      "clip_ratio/low_min": 0.0,
+      "clip_ratio/region_mean": 0.0,
+      "completion_length": 68.11,
+      "completions/clipped_ratio": 0.0,
+      "completions/max_length": 67.92307692307692,
+      "completions/max_terminated_length": 67.92307692307692,
+      "completions/mean_length": 67.74038461538461,
+      "completions/mean_terminated_length": 67.74038461538461,
+      "completions/min_length": 67.61538461538461,
+      "completions/min_terminated_length": 67.61538461538461,
+      "epoch": 0.0005772544004546988,
+      "frac_reward_zero_std": 0.6923076923076923,
+      "grad_norm": 0.2366252988576889,
+      "kl": 0.18056891307234765,
+      "learning_rate": 1.438689194289647e-08,
+      "loss": 0.0016832691431045532,
+      "num_tokens": 297265.0,
+      "reward": 2.8656079218937802e-09,
+      "reward_std": 0.3289369069612943,
+      "rewards/TranscriptCorrectionGrpoReward/mean": 2.8656079218937802e-09,
+      "rewards/TranscriptCorrectionGrpoReward/std": 0.3289369069612943,
+      "step": 325
+    },
+    {
+      "clip_ratio/high_max": 0.0,
+      "clip_ratio/high_mean": 0.0,
+      "clip_ratio/low_mean": 0.0,
+      "clip_ratio/low_min": 0.0,
+      "clip_ratio/region_mean": 0.0,
+      "completion_length": 54.2,
+      "completions/clipped_ratio": 0.0,
+      "completions/max_length": 54.083333333333336,
+      "completions/max_terminated_length": 54.083333333333336,
+      "completions/mean_length": 54.020833333333336,
+      "completions/mean_terminated_length": 54.020833333333336,
+      "completions/min_length": 53.916666666666664,
+      "completions/min_terminated_length": 53.916666666666664,
+      "epoch": 0.0006216585851050603,
+      "frac_reward_zero_std": 0.75,
+      "grad_norm": 2.0182937987556215e-06,
+      "kl": 0.22566444158554078,
+      "learning_rate": 1.549699162984836e-08,
+      "loss": 0.003195708990097046,
+      "num_tokens": 316355.0,
+      "reward": 0.0,
+      "reward_std": 0.26726123690605164,
+      "rewards/TranscriptCorrectionGrpoReward/mean": 0.0,
+      "rewards/TranscriptCorrectionGrpoReward/std": 0.26726123690605164,
+      "step": 350
+    },
+    {
+      "clip_ratio/high_max": 0.0,
+      "clip_ratio/high_mean": 0.0,
+      "clip_ratio/low_mean": 0.000317460335791111,
+      "clip_ratio/low_min": 0.000317460335791111,
+      "clip_ratio/region_mean": 0.000317460335791111,
+      "completion_length": 67.73,
+      "completions/clipped_ratio": 0.0,
+      "completions/max_length": 67.76923076923077,
+      "completions/max_terminated_length": 67.76923076923077,
+      "completions/mean_length": 67.25,
+      "completions/mean_terminated_length": 67.25,
+      "completions/min_length": 66.6923076923077,
+      "completions/min_terminated_length": 66.6923076923077,
+      "epoch": 0.0006660627697554218,
+      "frac_reward_zero_std": 0.3076923076923077,
+      "grad_norm": 0.16900327801704407,
+      "kl": 0.1710970998555422,
+      "learning_rate": 1.660709131680025e-08,
+      "loss": -0.004613960087299347,
+      "num_tokens": 341645.0,
+      "reward": 5.158094259408804e-09,
+      "reward_std": 0.7401080314929669,
+      "rewards/TranscriptCorrectionGrpoReward/mean": 5.158094259408804e-09,
+      "rewards/TranscriptCorrectionGrpoReward/std": 0.7401080406629122,
+      "step": 375
+    },
+    {
+      "clip_ratio/high_max": 0.0,
+      "clip_ratio/high_mean": 0.0,
+      "clip_ratio/low_mean": 0.0,
+      "clip_ratio/low_min": 0.0,
+      "clip_ratio/region_mean": 0.0,
+      "completion_length": 61.17,
+      "completions/clipped_ratio": 0.0,
+      "completions/max_length": 61.833333333333336,
+      "completions/max_terminated_length": 61.833333333333336,
+      "completions/mean_length": 61.416666666666664,
+      "completions/mean_terminated_length": 61.416666666666664,
+      "completions/min_length": 61.25,
+      "completions/min_terminated_length": 61.25,
+      "epoch": 0.0007104669544057832,
+      "frac_reward_zero_std": 0.6666666666666666,
+      "grad_norm": 8.416860168836138e-07,
+      "kl": 0.21405472174286844,
+      "learning_rate": 1.7717191003752138e-08,
+      "loss": 0.00976666271686554,
+      "num_tokens": 362597.0,
+      "reward": 0.0,
+      "reward_std": 0.35634831587473553,
+      "rewards/TranscriptCorrectionGrpoReward/mean": 0.0,
+      "rewards/TranscriptCorrectionGrpoReward/std": 0.35634832580884296,
+      "step": 400
+    },
+    {
+      "clip_ratio/high_max": 0.0,
+      "clip_ratio/high_mean": 0.0,
+      "clip_ratio/low_mean": 0.0,
+      "clip_ratio/low_min": 0.0,
+      "clip_ratio/region_mean": 0.0,
+      "completion_length": 70.06,
+      "completions/clipped_ratio": 0.0,
+      "completions/max_length": 70.07692307692308,
+      "completions/max_terminated_length": 70.07692307692308,
+      "completions/mean_length": 69.78846153846153,
+      "completions/mean_terminated_length": 69.78846153846153,
+      "completions/min_length": 69.3076923076923,
+      "completions/min_terminated_length": 69.3076923076923,
+      "epoch": 0.0007548711390561447,
+      "frac_reward_zero_std": 0.3076923076923077,
+      "grad_norm": 0.15129399299621582,
+      "kl": 0.18921885184943676,
+      "learning_rate": 1.8827290690704024e-08,
+      "loss": -0.001335275173187256,
+      "num_tokens": 388727.0,
+      "reward": 6.8774590125450725e-09,
+      "reward_std": 0.7401080314929669,
+      "rewards/TranscriptCorrectionGrpoReward/mean": 6.8774590125450725e-09,
+      "rewards/TranscriptCorrectionGrpoReward/std": 0.7401080406629122,
+      "step": 425
+    },
+    {
+      "clip_ratio/high_max": 0.0,
+      "clip_ratio/high_mean": 0.0,
+      "clip_ratio/low_mean": 0.0,
+      "clip_ratio/low_min": 0.0,
+      "clip_ratio/region_mean": 0.0,
+      "completion_length": 60.39,
+      "completions/clipped_ratio": 0.0,
+      "completions/max_length": 60.5,
+      "completions/max_terminated_length": 60.5,
+      "completions/mean_length": 60.28125,
+      "completions/mean_terminated_length": 60.28125,
+      "completions/min_length": 60.25,
+      "completions/min_terminated_length": 60.25,
+      "epoch": 0.000799275323706506,
+      "frac_reward_zero_std": 0.6666666666666666,
+      "grad_norm": 0.0004283797461539507,
+      "kl": 0.19731507159769535,
+      "learning_rate": 1.9937390377655913e-08,
+      "loss": 0.002477347552776337,
+      "num_tokens": 409458.0,
+      "reward": 0.0,
+      "reward_std": 0.35634832580884296,
+      "rewards/TranscriptCorrectionGrpoReward/mean": 0.0,
+      "rewards/TranscriptCorrectionGrpoReward/std": 0.35634831587473553,
+      "step": 450
+    },
+    {
+      "clip_ratio/high_max": 0.0,
+      "clip_ratio/high_mean": 0.0,
+      "clip_ratio/low_mean": 0.0,
+      "clip_ratio/low_min": 0.0,
+      "clip_ratio/region_mean": 0.0,
+      "completion_length": 70.39,
+      "completions/clipped_ratio": 0.0,
+      "completions/max_length": 69.46153846153847,
+      "completions/max_terminated_length": 69.46153846153847,
+      "completions/mean_length": 69.45192307692308,
+      "completions/mean_terminated_length": 69.45192307692308,
+      "completions/min_length": 69.38461538461539,
+      "completions/min_terminated_length": 69.38461538461539,
+      "epoch": 0.0008436795083568675,
+      "frac_reward_zero_std": 0.8461538461538461,
+      "grad_norm": 6.861191195639549e-06,
+      "kl": 0.17384251914918422,
+      "learning_rate": 2.10474900646078e-08,
+      "loss": 0.0019164717197418214,
+      "num_tokens": 435369.0,
+      "reward": 0.0,
+      "reward_std": 0.16446845348064715,
+      "rewards/TranscriptCorrectionGrpoReward/mean": 0.0,
+      "rewards/TranscriptCorrectionGrpoReward/std": 0.16446845348064715,
+      "step": 475
+    },
+    {
+      "clip_ratio/high_max": 0.0,
+      "clip_ratio/high_mean": 0.0,
+      "clip_ratio/low_mean": 0.0,
+      "clip_ratio/low_min": 0.0,
+      "clip_ratio/region_mean": 0.0,
+      "completion_length": 61.43,
+      "completions/clipped_ratio": 0.0,
+      "completions/max_length": 62.25,
+      "completions/max_terminated_length": 62.25,
+      "completions/mean_length": 62.072916666666664,
+      "completions/mean_terminated_length": 62.072916666666664,
+      "completions/min_length": 61.666666666666664,
+      "completions/min_terminated_length": 61.666666666666664,
+      "epoch": 0.000888083693007229,
+      "frac_reward_zero_std": 0.6666666666666666,
+      "grad_norm": 0.1936454176902771,
+      "kl": 0.18575097121298312,
+      "learning_rate": 2.2157589751559692e-08,
+      "loss": 0.00039461255073547363,
+      "num_tokens": 456096.0,
+      "reward": -6.20881716410319e-09,
+      "reward_std": 0.35634831587473553,
+      "rewards/TranscriptCorrectionGrpoReward/mean": -6.20881716410319e-09,
+      "rewards/TranscriptCorrectionGrpoReward/std": 0.35634831587473553,
+      "step": 500
+    },
+    {
+      "clip_ratio/high_max": 0.0,
+      "clip_ratio/high_mean": 0.0,
+      "clip_ratio/low_mean": 0.0,
+      "clip_ratio/low_min": 0.0,
+      "clip_ratio/region_mean": 0.0,
+      "completion_length": 53.47,
+      "completions/clipped_ratio": 0.0,
+      "completions/max_length": 53.92307692307692,
+      "completions/max_terminated_length": 53.92307692307692,
+      "completions/mean_length": 53.64423076923077,
+      "completions/mean_terminated_length": 53.64423076923077,
+      "completions/min_length": 53.15384615384615,
+      "completions/min_terminated_length": 53.15384615384615,
+      "epoch": 0.0009324878776575904,
+      "frac_reward_zero_std": 0.7692307692307693,
+      "grad_norm": 2.9146491215215065e-05,
+      "kl": 0.21476560801267625,
+      "learning_rate": 2.3267689438511578e-08,
+      "loss": 0.0002883000299334526,
+      "num_tokens": 476723.0,
+      "reward": 0.0,
+      "reward_std": 0.24670268022097075,
+      "rewards/TranscriptCorrectionGrpoReward/mean": 0.0,
+      "rewards/TranscriptCorrectionGrpoReward/std": 0.24670268022097075,
+      "step": 525
+    },
+    {
+      "clip_ratio/high_max": 0.0,
+      "clip_ratio/high_mean": 0.0,
+      "clip_ratio/low_mean": 0.0,
+      "clip_ratio/low_min": 0.0,
+      "clip_ratio/region_mean": 0.0,
+      "completion_length": 78.73,
+      "completions/clipped_ratio": 0.0,
+      "completions/max_length": 79.91666666666667,
+      "completions/max_terminated_length": 79.91666666666667,
+      "completions/mean_length": 79.59375,
+      "completions/mean_terminated_length": 79.59375,
+      "completions/min_length": 78.91666666666667,
+      "completions/min_terminated_length": 78.91666666666667,
+      "epoch": 0.0009768920623079518,
+      "frac_reward_zero_std": 0.5833333333333334,
+      "grad_norm": 1.4861758245388046e-05,
+      "kl": 0.15855858221650124,
+      "learning_rate": 2.4377789125463464e-08,
+      "loss": 0.0011431390047073363,
+      "num_tokens": 503564.0,
+      "reward": 4.346172014872233e-09,
+      "reward_std": 0.4454353948434194,
+      "rewards/TranscriptCorrectionGrpoReward/mean": 4.346172014872233e-09,
+      "rewards/TranscriptCorrectionGrpoReward/std": 0.4454353948434194,
+      "step": 550
+    },
+    {
+      "clip_ratio/high_max": 0.0,
+      "clip_ratio/high_mean": 0.0,
+      "clip_ratio/low_mean": 0.0,
+      "clip_ratio/low_min": 0.0,
+      "clip_ratio/region_mean": 0.0,
+      "completion_length": 70.19,
+      "completions/clipped_ratio": 0.0,
+      "completions/max_length": 72.07692307692308,
+      "completions/max_terminated_length": 72.07692307692308,
+      "completions/mean_length": 71.83653846153847,
+      "completions/mean_terminated_length": 71.83653846153847,
+      "completions/min_length": 71.53846153846153,
+      "completions/min_terminated_length": 71.53846153846153,
+      "epoch": 0.0010212962469583133,
+      "frac_reward_zero_std": 0.6153846153846154,
+      "grad_norm": 0.19313402473926544,
+      "kl": 0.18091401405632496,
+      "learning_rate": 2.5487888812415353e-08,
+      "loss": 0.010396603345870972,
+      "num_tokens": 530171.0,
+      "reward": -5.7312158437875604e-09,
+      "reward_std": 0.4111711337016179,
+      "rewards/TranscriptCorrectionGrpoReward/mean": -5.7312158437875604e-09,
+      "rewards/TranscriptCorrectionGrpoReward/std": 0.4111711337016179,
+      "step": 575
+    },
+    {
+      "clip_ratio/high_max": 0.0,
+      "clip_ratio/high_mean": 0.0,
+      "clip_ratio/low_mean": 0.0,
+      "clip_ratio/low_min": 0.0,
+      "clip_ratio/region_mean": 0.0,
+      "completion_length": 69.74,
+      "completions/clipped_ratio": 0.0,
+      "completions/max_length": 68.0,
+      "completions/max_terminated_length": 68.0,
+      "completions/mean_length": 67.9375,
+      "completions/mean_terminated_length": 67.9375,
+      "completions/min_length": 67.66666666666667,
+      "completions/min_terminated_length": 67.66666666666667,
+      "epoch": 0.0010657004316086747,
+      "frac_reward_zero_std": 0.5833333333333334,
+      "grad_norm": 0.1186411902308464,
+      "kl": 0.16490939274430275,
+      "learning_rate": 2.659798849936724e-08,
+      "loss": -0.006589226126670837,
+      "num_tokens": 553341.0,
+      "reward": -2.949188152949015e-09,
+      "reward_std": 0.44543540477752686,
+      "rewards/TranscriptCorrectionGrpoReward/mean": -2.949188152949015e-09,
+      "rewards/TranscriptCorrectionGrpoReward/std": 0.4454353948434194,
+      "step": 600
+    },
+    {
+      "clip_ratio/high_max": 0.0,
+      "clip_ratio/high_mean": 0.0,
+      "clip_ratio/low_mean": 0.0,
+      "clip_ratio/low_min": 0.0,
+      "clip_ratio/region_mean": 0.0,
+      "completion_length": 69.76,
+      "completions/clipped_ratio": 0.0,
+      "completions/max_length": 70.61538461538461,
+      "completions/max_terminated_length": 70.61538461538461,
+      "completions/mean_length": 70.23076923076923,
+      "completions/mean_terminated_length": 70.23076923076923,
+      "completions/min_length": 69.76923076923077,
+      "completions/min_terminated_length": 69.76923076923077,
+      "epoch": 0.0011101046162590362,
+      "frac_reward_zero_std": 0.38461538461538464,
+      "grad_norm": 0.00016267380851786584,
+      "kl": 0.1838107569515705,
+      "learning_rate": 2.770808818631913e-08,
+      "loss": 0.0022651535272598267,
+      "num_tokens": 579229.0,
+      "reward": -3.4387295062725362e-09,
+      "reward_std": 0.6578738139225886,
+      "rewards/TranscriptCorrectionGrpoReward/mean": -3.4387295062725362e-09,
+      "rewards/TranscriptCorrectionGrpoReward/std": 0.6578738322624793,
+      "step": 625
+    },
+    {
+      "clip_ratio/high_max": 0.0,
+      "clip_ratio/high_mean": 0.0,
+      "clip_ratio/low_mean": 0.0,
+      "clip_ratio/low_min": 0.0,
+      "clip_ratio/region_mean": 0.0,
+      "completion_length": 73.21,
+      "completions/clipped_ratio": 0.0,
+      "completions/max_length": 73.33333333333333,
+      "completions/max_terminated_length": 73.33333333333333,
+      "completions/mean_length": 72.84375,
+      "completions/mean_terminated_length": 72.84375,
+      "completions/min_length": 71.66666666666667,
+      "completions/min_terminated_length": 71.66666666666667,
+      "epoch": 0.0011545088009093977,
+      "frac_reward_zero_std": 0.3333333333333333,
+      "grad_norm": 2.573253914306406e-05,
+      "kl": 0.17695149526000023,
+      "learning_rate": 2.8818187873271017e-08,
+      "loss": -9.701214730739594e-05,
+      "num_tokens": 603254.0,
+      "reward": 4.035731156667073e-09,
+      "reward_std": 0.7126966516176859,
+      "rewards/TranscriptCorrectionGrpoReward/mean": 4.035731156667073e-09,
+      "rewards/TranscriptCorrectionGrpoReward/std": 0.7126966317494711,
+      "step": 650
+    },
+    {
+      "clip_ratio/high_max": 0.0,
+      "clip_ratio/high_mean": 0.0,
+      "clip_ratio/low_mean": 0.0,
+      "clip_ratio/low_min": 0.0,
+      "clip_ratio/region_mean": 0.0,
+      "completion_length": 82.16,
+      "completions/clipped_ratio": 0.0,
+      "completions/max_length": 81.46153846153847,
+      "completions/max_terminated_length": 81.46153846153847,
+      "completions/mean_length": 81.1923076923077,
+      "completions/mean_terminated_length": 81.1923076923077,
+      "completions/min_length": 81.15384615384616,
+      "completions/min_terminated_length": 81.15384615384616,
+      "epoch": 0.0011989129855597591,
+      "frac_reward_zero_std": 0.6923076923076923,
+      "grad_norm": 1.8145580042983056e-06,
+      "kl": 0.15570140190422535,
+      "learning_rate": 2.9928287560222903e-08,
+      "loss": 0.0013257008790969848,
+      "num_tokens": 632778.0,
+      "reward": 0.0,
+      "reward_std": 0.3289369069612943,
+      "rewards/TranscriptCorrectionGrpoReward/mean": 0.0,
+      "rewards/TranscriptCorrectionGrpoReward/std": 0.3289369069612943,
+      "step": 675
+    },
+    {
+      "clip_ratio/high_max": 0.0,
+      "clip_ratio/high_mean": 0.0,
+      "clip_ratio/low_mean": 8.849557489156722e-05,
+      "clip_ratio/low_min": 8.849557489156722e-05,
+      "clip_ratio/region_mean": 8.849557489156722e-05,
+      "completion_length": 61.01,
+      "completions/clipped_ratio": 0.0,
+      "completions/max_length": 61.5,
+      "completions/max_terminated_length": 61.5,
+      "completions/mean_length": 61.177083333333336,
+      "completions/mean_terminated_length": 61.177083333333336,
+      "completions/min_length": 60.25,
+      "completions/min_terminated_length": 60.25,
+      "epoch": 0.0012433171702101206,
+      "frac_reward_zero_std": 0.5,
+      "grad_norm": 0.0008123472798615694,
+      "kl": 0.20670814238488674,
+      "learning_rate": 3.103838724717479e-08,
+      "loss": -0.002791093587875366,
+      "num_tokens": 654243.0,
+      "reward": 6.20881716410319e-09,
+      "reward_std": 0.5345224738121033,
+      "rewards/TranscriptCorrectionGrpoReward/mean": 6.20881716410319e-09,
+      "rewards/TranscriptCorrectionGrpoReward/std": 0.5345224738121033,
+      "step": 700
+    },
+    {
+      "clip_ratio/high_max": 0.0,
+      "clip_ratio/high_mean": 0.0,
+      "clip_ratio/low_mean": 0.0,
+      "clip_ratio/low_min": 0.0,
+      "clip_ratio/region_mean": 0.0,
+      "completion_length": 66.55,
+      "completions/clipped_ratio": 0.0,
+      "completions/max_length": 66.61538461538461,
+      "completions/max_terminated_length": 66.61538461538461,
+      "completions/mean_length": 66.29807692307692,
+      "completions/mean_terminated_length": 66.29807692307692,
+      "completions/min_length": 66.07692307692308,
+      "completions/min_terminated_length": 66.07692307692308,
+      "epoch": 0.001287721354860482,
+      "frac_reward_zero_std": 0.6153846153846154,
+      "grad_norm": 1.722803517623106e-06,
+      "kl": 0.18176497109234332,
+      "learning_rate": 3.214848693412668e-08,
+      "loss": 0.0018749050796031952,
+      "num_tokens": 678762.0,
+      "reward": -2.292486337515024e-09,
+      "reward_std": 0.4111711337016179,
+      "rewards/TranscriptCorrectionGrpoReward/mean": -2.292486337515024e-09,
+      "rewards/TranscriptCorrectionGrpoReward/std": 0.4111711337016179,
+      "step": 725
+    },
+    {
+      "clip_ratio/high_max": 0.0,
+      "clip_ratio/high_mean": 0.0,
+      "clip_ratio/low_mean": 0.0,
+      "clip_ratio/low_min": 0.0,
+      "clip_ratio/region_mean": 0.0,
+      "completion_length": 64.79,
+      "completions/clipped_ratio": 0.0,
+      "completions/max_length": 65.16666666666667,
+      "completions/max_terminated_length": 65.16666666666667,
+      "completions/mean_length": 64.98958333333333,
+      "completions/mean_terminated_length": 64.98958333333333,
+      "completions/min_length": 64.91666666666667,
+      "completions/min_terminated_length": 64.91666666666667,
+      "epoch": 0.0013321255395108435,
+      "frac_reward_zero_std": 0.6666666666666666,
+      "grad_norm": 0.0003243849496357143,
+      "kl": 0.1709983678907156,
+      "learning_rate": 3.325858662107857e-08,
+      "loss": 0.0017569869756698608,
+      "num_tokens": 700945.0,
+      "reward": 7.450580596923828e-09,
+      "reward_std": 0.35634832580884296,
+      "rewards/TranscriptCorrectionGrpoReward/mean": 7.450580596923828e-09,
+      "rewards/TranscriptCorrectionGrpoReward/std": 0.35634831587473553,
+      "step": 750
+    },
+    {
+      "clip_ratio/high_max": 0.0,
+      "clip_ratio/high_mean": 0.0,
+      "clip_ratio/low_mean": 0.0,
+      "clip_ratio/low_min": 0.0,
+      "clip_ratio/region_mean": 0.0,
+      "completion_length": 72.83,
+      "completions/clipped_ratio": 0.0,
+      "completions/max_length": 73.46153846153847,
+      "completions/max_terminated_length": 73.46153846153847,
+      "completions/mean_length": 73.29807692307692,
+      "completions/mean_terminated_length": 73.29807692307692,
+      "completions/min_length": 73.15384615384616,
+      "completions/min_terminated_length": 73.15384615384616,
+      "epoch": 0.001376529724161205,
+      "frac_reward_zero_std": 0.38461538461538464,
+      "grad_norm": 0.2545297145843506,
+      "kl": 0.18227736078202725,
+      "learning_rate": 3.4368686308030454e-08,
+      "loss": -0.009244667887687684,
+      "num_tokens": 728240.0,
+      "reward": 0.0,
+      "reward_std": 0.6578738139225886,
+      "rewards/TranscriptCorrectionGrpoReward/mean": 0.0,
+      "rewards/TranscriptCorrectionGrpoReward/std": 0.6578738139225886,
+      "step": 775
+    },
+    {
+      "clip_ratio/high_max": 0.0,
+      "clip_ratio/high_mean": 0.0,
+      "clip_ratio/low_mean": 0.0,
+      "clip_ratio/low_min": 0.0,
+      "clip_ratio/region_mean": 0.0,
+      "completion_length": 62.07,
+      "completions/clipped_ratio": 0.0,
+      "completions/max_length": 61.416666666666664,
+      "completions/max_terminated_length": 61.416666666666664,
+      "completions/mean_length": 61.114583333333336,
+      "completions/mean_terminated_length": 61.114583333333336,
+      "completions/min_length": 60.916666666666664,
+      "completions/min_terminated_length": 60.916666666666664,
+      "epoch": 0.0014209339088115665,
+      "frac_reward_zero_std": 0.6666666666666666,
+      "grad_norm": 0.0001137235522037372,
+      "kl": 0.20930900558829307,
+      "learning_rate": 3.547878599498235e-08,
+      "loss": 0.012507272958755493,
+      "num_tokens": 749099.0,
+      "reward": 0.0,
+      "reward_std": 0.35634830594062805,
+      "rewards/TranscriptCorrectionGrpoReward/mean": 0.0,
+      "rewards/TranscriptCorrectionGrpoReward/std": 0.35634831587473553,
+      "step": 800
+    },
+    {
+      "clip_ratio/high_max": 0.0,
+      "clip_ratio/high_mean": 0.0,
+      "clip_ratio/low_mean": 0.0,
+      "clip_ratio/low_min": 0.0,
+      "clip_ratio/region_mean": 0.0,
+      "completion_length": 79.05,
+      "completions/clipped_ratio": 0.0,
+      "completions/max_length": 79.38461538461539,
+      "completions/max_terminated_length": 79.38461538461539,
+      "completions/mean_length": 79.08653846153847,
+      "completions/mean_terminated_length": 79.08653846153847,
+      "completions/min_length": 78.92307692307692,
+      "completions/min_terminated_length": 78.92307692307692,
+      "epoch": 0.001465338093461928,
+      "frac_reward_zero_std": 0.46153846153846156,
+      "grad_norm": 0.00025333481607958674,
+      "kl": 0.1384541442245245,
+      "learning_rate": 3.658888568193424e-08,
+      "loss": 0.0011119496822357177,
+      "num_tokens": 778348.0,
+      "reward": 3.4387295062725362e-09,
+      "reward_std": 0.5756396055221558,
+      "rewards/TranscriptCorrectionGrpoReward/mean": 3.4387295062725362e-09,
+      "rewards/TranscriptCorrectionGrpoReward/std": 0.575639587182265,
+      "step": 825
+    },
+    {
+      "clip_ratio/high_max": 0.0,
+      "clip_ratio/high_mean": 0.0,
+      "clip_ratio/low_mean": 0.0,
+      "clip_ratio/low_min": 0.0,
+      "clip_ratio/region_mean": 0.0,
+      "completion_length": 71.36,
+      "completions/clipped_ratio": 0.0,
+      "completions/max_length": 71.16666666666667,
+      "completions/max_terminated_length": 71.16666666666667,
+      "completions/mean_length": 71.0,
+      "completions/mean_terminated_length": 71.0,
+      "completions/min_length": 70.91666666666667,
+      "completions/min_terminated_length": 70.91666666666667,
+      "epoch": 0.0015097422781122894,
+      "frac_reward_zero_std": 0.5833333333333334,
+      "grad_norm": 0.013177081011235714,
+      "kl": 0.17892150908708573,
+      "learning_rate": 3.769898536888613e-08,
+      "loss": 0.002372233867645264,
+      "num_tokens": 802668.0,
+      "reward": 3.725290298461914e-09,
+      "reward_std": 0.4454353948434194,
+      "rewards/TranscriptCorrectionGrpoReward/mean": 3.725290298461914e-09,
+      "rewards/TranscriptCorrectionGrpoReward/std": 0.4454353948434194,
+      "step": 850
+    },
+    {
+      "clip_ratio/high_max": 0.0,
+      "clip_ratio/high_mean": 0.0,
+      "clip_ratio/low_mean": 0.00011363636702299118,
+      "clip_ratio/low_min": 0.00011363636702299118,
+      "clip_ratio/region_mean": 0.00011363636702299118,
+      "completion_length": 78.5,
+      "completions/clipped_ratio": 0.0,
+      "completions/max_length": 76.84615384615384,
+      "completions/max_terminated_length": 76.84615384615384,
+      "completions/mean_length": 76.71153846153847,
+      "completions/mean_terminated_length": 76.71153846153847,
+      "completions/min_length": 76.38461538461539,
+      "completions/min_terminated_length": 76.38461538461539,
+      "epoch": 0.0015541464627626509,
+      "frac_reward_zero_std": 0.6153846153846154,
+      "grad_norm": 0.38749536871910095,
+      "kl": 0.17304533801972866,
+      "learning_rate": 3.880908505583801e-08,
+      "loss": -0.004066205024719239,
+      "num_tokens": 831014.0,
+      "reward": 2.292486337515024e-09,
+      "reward_std": 0.4111711337016179,
+      "rewards/TranscriptCorrectionGrpoReward/mean": 2.292486337515024e-09,
+      "rewards/TranscriptCorrectionGrpoReward/std": 0.4111711337016179,
+      "step": 875
+    },
+    {
+      "clip_ratio/high_max": 0.0,
+      "clip_ratio/high_mean": 0.0,
+      "clip_ratio/low_mean": 0.0,
+      "clip_ratio/low_min": 0.0,
+      "clip_ratio/region_mean": 0.0,
+      "completion_length": 69.88,
+      "completions/clipped_ratio": 0.0,
+      "completions/max_length": 71.66666666666667,
+      "completions/max_terminated_length": 71.66666666666667,
+      "completions/mean_length": 71.45833333333333,
+      "completions/mean_terminated_length": 71.45833333333333,
+      "completions/min_length": 71.25,
+      "completions/min_terminated_length": 71.25,
+      "epoch": 0.001598550647413012,
+      "frac_reward_zero_std": 0.5,
+      "grad_norm": 0.2461424618959427,
+      "kl": 0.2068350876867771,
+      "learning_rate": 3.99191847427899e-08,
+      "loss": 0.007263134121894836,
+      "num_tokens": 855450.0,
+      "reward": 6.20881716410319e-09,
+      "reward_std": 0.5345224638779958,
+      "rewards/TranscriptCorrectionGrpoReward/mean": 6.20881716410319e-09,
+      "rewards/TranscriptCorrectionGrpoReward/std": 0.5345224738121033,
+      "step": 900
+    },
+    {
+      "clip_ratio/high_max": 0.0,
+      "clip_ratio/high_mean": 0.0,
+      "clip_ratio/low_mean": 0.0,
+      "clip_ratio/low_min": 0.0,
+      "clip_ratio/region_mean": 0.0,
+      "completion_length": 72.65,
+      "completions/clipped_ratio": 0.0,
+      "completions/max_length": 74.76923076923077,
+      "completions/max_terminated_length": 74.76923076923077,
+      "completions/mean_length": 74.50961538461539,
+      "completions/mean_terminated_length": 74.50961538461539,
+      "completions/min_length": 74.23076923076923,
+      "completions/min_terminated_length": 74.23076923076923,
+      "epoch": 0.0016429548320633736,
+      "frac_reward_zero_std": 0.6923076923076923,
+      "grad_norm": 3.1216541174217127e-07,
+      "kl": 0.1654582280665636,
+      "learning_rate": 4.102928442974179e-08,
+      "loss": 0.0022979089617729187,
+      "num_tokens": 883151.0,
+      "reward": -1.7193647531362681e-09,
+      "reward_std": 0.3289369069612943,
+      "rewards/TranscriptCorrectionGrpoReward/mean": -1.7193647531362681e-09,
+      "rewards/TranscriptCorrectionGrpoReward/std": 0.3289369069612943,
+      "step": 925
+    },
+    {
+      "clip_ratio/high_max": 0.0,
+      "clip_ratio/high_mean": 0.0,
+      "clip_ratio/low_mean": 0.0,
+      "clip_ratio/low_min": 0.0,
+      "clip_ratio/region_mean": 0.0,
+      "completion_length": 66.25,
+      "completions/clipped_ratio": 0.0,
+      "completions/max_length": 64.33333333333333,
+      "completions/max_terminated_length": 64.33333333333333,
+      "completions/mean_length": 63.96875,
+      "completions/mean_terminated_length": 63.96875,
+      "completions/min_length": 63.833333333333336,
+      "completions/min_terminated_length": 63.833333333333336,
+      "epoch": 0.001687359016713735,
+      "frac_reward_zero_std": 0.5,
+      "grad_norm": 0.3891129791736603,
+      "kl": 0.17953816015273333,
+      "learning_rate": 4.213938411669368e-08,
+      "loss": 0.0029295945167541505,
+      "num_tokens": 905076.0,
+      "reward": 3.725290298461914e-09,
+      "reward_std": 0.5345224738121033,
+      "rewards/TranscriptCorrectionGrpoReward/mean": 3.725290298461914e-09,
+      "rewards/TranscriptCorrectionGrpoReward/std": 0.5345224738121033,
+      "step": 950
+    },
+    {
+      "clip_ratio/high_max": 0.0,
+      "clip_ratio/high_mean": 0.0,
+      "clip_ratio/low_mean": 0.0,
+      "clip_ratio/low_min": 0.0,
+      "clip_ratio/region_mean": 0.0,
+      "completion_length": 64.85,
+      "completions/clipped_ratio": 0.0,
+      "completions/max_length": 65.23076923076923,
+      "completions/max_terminated_length": 65.23076923076923,
+      "completions/mean_length": 64.9326923076923,
+      "completions/mean_terminated_length": 64.9326923076923,
+      "completions/min_length": 64.38461538461539,
+      "completions/min_terminated_length": 64.38461538461539,
+      "epoch": 0.0017317632013640965,
+      "frac_reward_zero_std": 0.5384615384615384,
+      "grad_norm": 2.965283556477516e-06,
+      "kl": 0.19248799994587898,
+      "learning_rate": 4.324948380364556e-08,
+      "loss": 0.0011707821488380432,
+      "num_tokens": 929237.0,
+      "reward": 1.7193647531362681e-09,
+      "reward_std": 0.4934053604419415,
+      "rewards/TranscriptCorrectionGrpoReward/mean": 1.7193647531362681e-09,
+      "rewards/TranscriptCorrectionGrpoReward/std": 0.4934053604419415,
+      "step": 975
+    },
+    {
+      "clip_ratio/high_max": 0.0,
+      "clip_ratio/high_mean": 0.0,
+      "clip_ratio/low_mean": 0.0,
+      "clip_ratio/low_min": 0.0,
+      "clip_ratio/region_mean": 0.0,
+      "completion_length": 58.09,
+      "completions/clipped_ratio": 0.0,
+      "completions/max_length": 58.0,
+      "completions/max_terminated_length": 58.0,
+      "completions/mean_length": 57.71875,
+      "completions/mean_terminated_length": 57.71875,
+      "completions/min_length": 57.416666666666664,
+      "completions/min_terminated_length": 57.416666666666664,
+      "epoch": 0.001776167386014458,
+      "frac_reward_zero_std": 0.6666666666666666,
+      "grad_norm": 0.0015717418864369392,
+      "kl": 0.26020569302141666,
+      "learning_rate": 4.435958349059745e-08,
+      "loss": 0.00038701888173818587,
+      "num_tokens": 949458.0,
+      "reward": -3.725290298461914e-09,
+      "reward_std": 0.35634831587473553,
+      "rewards/TranscriptCorrectionGrpoReward/mean": -3.725290298461914e-09,
+      "rewards/TranscriptCorrectionGrpoReward/std": 0.35634831587473553,
+      "step": 1000
+    },
+    {
+      "eval_cer_subset": 0.01423290504339896,
+      "eval_cer_subset_edit_distance": 874,
+      "eval_cer_subset_groups": 250,
+      "eval_cer_subset_items": 250,
+      "eval_cer_subset_ref_chars": 61407,
+      "eval_loss": 0.03394044979410833,
+      "step": 1013
+    }
+  ],
+  "logging_steps": 25,
+  "max_steps": 2252040,
+  "num_input_tokens_seen": 963752,
+  "num_train_epochs": 4,
+  "save_steps": 1013,
+  "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": 1,
+  "trial_name": null,
+  "trial_params": null
+}

meta-llama__meta-llama-3.1-8b/grpo/checkpoints/checkpoint-1013/training_args.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:28fd8dbecc91ea4d1302120052163b26bd82d9f2b04a5a72e1957eb427813a61
+size 6737

meta-llama__meta-llama-3.1-8b/grpo/run_manifest.json

@@ -0,0 +1,25 @@
+{
+  "schema_version": 1,
+  "run_id": "20b52546-5762-4851-830d-75743222f125",
+  "run_type": "grpo",
+  "profile_name": "llama3_1_8b_base_grpo",
+  "model_name": "meta-llama/Meta-Llama-3.1-8B",
+  "status": "running",
+  "best_metric_name": "eval_cer_subset",
+  "best_metric_value": 0.01423290504339896,
+  "best_step": 1013,
+  "best_checkpoint_step": 1013,
+  "latest_checkpoint_step": 1013,
+  "checkpoints": [
+    {
+      "step": 1013,
+      "epoch": 0.001799257562032646,
+      "repo_path": "meta-llama__meta-llama-3.1-8b/grpo/checkpoints/checkpoint-1013",
+      "timestamp": "2026-04-09T17:58:40.645060+00:00"
+    }
+  ],
+  "epoch_artifacts": [],
+  "resume_history": [],
+  "created_at": "2026-04-09T17:24:46.169440+00:00",
+  "updated_at": "2026-04-09T17:58:40.645077+00:00"
+}