backup e16/S2-ppo1 (Pod B)

e16/S2-ppo1/ckpt-16000/adapter/README.md

@@ -0,0 +1,206 @@
+---
+base_model: /workspace/jev_kt/models/hf/JEV-9B
+library_name: peft
+tags:
+- base_model:adapter:/workspace/jev_kt/models/hf/JEV-9B
+- lora
+- transformers
+---
+
+# 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 0.21.2

e16/S2-ppo1/ckpt-16000/adapter/adapter_config.json

@@ -0,0 +1,57 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": {
+    "base_model_class": "Qwen3_5ForCausalLM",
+    "parent_library": "transformers.models.qwen3_5.modeling_qwen3_5"
+  },
+  "base_model_name_or_path": "/workspace/jev_kt/models/hf/JEV-9B",
+  "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,
+  "kasa_config": null,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 32,
+  "lora_bias": false,
+  "lora_dropout": 0.05,
+  "lora_ga_config": null,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "monteclora_config": null,
+  "peft_type": "LORA",
+  "peft_version": "0.21.0",
+  "qalora_group_size": 16,
+  "r": 16,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "v_proj",
+    "o_proj",
+    "out_proj",
+    "gate_proj",
+    "down_proj",
+    "k_proj",
+    "up_proj",
+    "in_proj_z",
+    "in_proj_qkv",
+    "q_proj"
+  ],
+  "target_parameters": null,
+  "task_type": null,
+  "trainable_token_indices": null,
+  "use_bdlora": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false,
+  "velora_config": null
+}

e16/S2-ppo1/ckpt-16000/adapter/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:db513f54273c7b8abed79ae4bc7b6fe1a19960cd98f40baecc9674fe33ebad5f
+size 160486456

e16/S2-ppo1/ckpt-16000/adapter_vllm/adapter_config.json

@@ -0,0 +1,58 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": {
+    "base_model_class": "Qwen3_5ForCausalLM",
+    "parent_library": "transformers.models.qwen3_5.modeling_qwen3_5"
+  },
+  "base_model_name_or_path": null,
+  "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,
+  "kasa_config": null,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64.0,
+  "lora_bias": false,
+  "lora_dropout": 0.0,
+  "lora_ga_config": null,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "monteclora_config": null,
+  "peft_type": "LORA",
+  "peft_version": "0.21.0",
+  "qalora_group_size": 16,
+  "r": 32,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "down_proj",
+    "gate_proj",
+    "in_proj_qkv",
+    "in_proj_z",
+    "k_proj",
+    "lm_head",
+    "o_proj",
+    "out_proj",
+    "q_proj",
+    "up_proj",
+    "v_proj"
+  ],
+  "target_parameters": null,
+  "task_type": null,
+  "trainable_token_indices": null,
+  "use_bdlora": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false,
+  "velora_config": null
+}

e16/S2-ppo1/ckpt-16000/adapter_vllm/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:977d30314d0e654b94b08a151119b584541d502b34dddb3cfe605ef79b4ec784
+size 176641832

e16/S2-ppo1/ckpt-16000/adapter_vllm/decision_head.json

@@ -0,0 +1,99 @@
+{
+ "bias": [
+  0.0014366715913638473,
+  0.0019461512565612793,
+  -0.0025878134183585644,
+  -0.0012615364976227283,
+  -0.00025007399381138384,
+  0.0019558551721274853,
+  0.0013559526996687055,
+  -0.0023348911199718714,
+  -0.0004082925443071872,
+  -0.0002078723773593083,
+  -1.2957353646925185e-05,
+  0.0001676026004133746,
+  0.0005569425993598998,
+  -0.001412136247381568,
+  -0.0028663526754826307,
+  -0.005061559844762087,
+  -0.00593328895047307,
+  -0.004319756291806698,
+  -0.004922541789710522,
+  -0.0049067651852965355,
+  -0.003943523392081261,
+  -0.003319827374070883,
+  -0.005581587553024292,
+  -0.0034480984322726727
+ ],
+ "verbalizer_ids": [
+  3721,
+  1802,
+  15,
+  16,
+  17,
+  18,
+  19,
+  20,
+  32,
+  33,
+  34,
+  35,
+  36,
+  37,
+  38,
+  39,
+  40,
+  41,
+  42,
+  43,
+  44,
+  45,
+  46,
+  47
+ ],
+ "slots": {
+  "num_slots": 24,
+  "ranges": {
+   "noul": [
+    0,
+    2
+   ],
+   "score": [
+    2,
+    8
+   ],
+   "choice": [
+    8,
+    24
+   ]
+  },
+  "verbalizers": [
+   "false",
+   "true",
+   "0",
+   "1",
+   "2",
+   "3",
+   "4",
+   "5",
+   "A",
+   "B",
+   "C",
+   "D",
+   "E",
+   "F",
+   "G",
+   "H",
+   "I",
+   "J",
+   "K",
+   "L",
+   "M",
+   "N",
+   "O",
+   "P"
+  ],
+  "template_version": "bare-v1"
+ },
+ "note": "decision logits = lm_head-LoRA logprobs of verbalizer ids + bias, then per-kind temperature"
+}

e16/S2-ppo1/ckpt-16000/calibration.json

@@ -0,0 +1,17 @@
+{
+ "version": 1,
+ "per_kind": {
+  "noul": 3.9999965454650366,
+  "choice": 1.0,
+  "score": 1.0
+ },
+ "fit": {
+  "noul": {
+   "n": 2000,
+   "T": 3.9999965454650366,
+   "ce_before": 0.8274263739585876,
+   "ce_after": 0.5666664242744446
+  }
+ },
+ "source": "models/e16/S2-ppo1/ckpt-16000"
+}

e16/S2-ppo1/ckpt-16000/head.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7e67108a70ccc7cec9742fdf506717d51bb376e699fa162f1afa7af554e6c004
+size 393464

e16/S2-ppo1/ckpt-16000/judge_config.json

@@ -0,0 +1,160 @@
+{
+ "base_model_path": "/root/models/Qwen3.5-9B",
+ "hidden_size": 4096,
+ "slots": {
+  "num_slots": 24,
+  "ranges": {
+   "noul": [
+    0,
+    2
+   ],
+   "score": [
+    2,
+    8
+   ],
+   "choice": [
+    8,
+    24
+   ]
+  },
+  "verbalizers": [
+   "false",
+   "true",
+   "0",
+   "1",
+   "2",
+   "3",
+   "4",
+   "5",
+   "A",
+   "B",
+   "C",
+   "D",
+   "E",
+   "F",
+   "G",
+   "H",
+   "I",
+   "J",
+   "K",
+   "L",
+   "M",
+   "N",
+   "O",
+   "P"
+  ],
+  "template_version": "bare-v1"
+ },
+ "verbalizer_ids": [
+  3721,
+  1802,
+  15,
+  16,
+  17,
+  18,
+  19,
+  20,
+  32,
+  33,
+  34,
+  35,
+  36,
+  37,
+  38,
+  39,
+  40,
+  41,
+  42,
+  43,
+  44,
+  45,
+  46,
+  47
+ ],
+ "kinds": [
+  "noul",
+  "choice",
+  "score"
+ ],
+ "model_name": "jev-judge-qwen35-9b",
+ "model_version": "0.8.0",
+ "lm_head_included": true,
+ "weights_mode": "unmerged",
+ "adapter_subfolder": "adapter",
+ "lora": {
+  "r": 16,
+  "lora_alpha": 32,
+  "lora_dropout": 0.05,
+  "target_modules": [
+   "down_proj",
+   "gate_proj",
+   "in_proj_qkv",
+   "in_proj_z",
+   "k_proj",
+   "o_proj",
+   "out_proj",
+   "q_proj",
+   "up_proj",
+   "v_proj"
+  ]
+ },
+ "trained_stage": "e16",
+ "source_checkpoint": "models/e16/S2-ppo1/ckpt-16000",
+ "softcap": null,
+ "e16": {
+  "data": "data/e15/F1-pool",
+  "data_manifest": {
+   "arm": "F1-pool",
+   "name": "F1-pool",
+   "train_sha256": "dad16f94a5659768b254a6c12538c01f9c64cd43d574a3435bcc70c9c1bac130",
+   "n_train": 64000
+  },
+  "init": "released",
+  "objective": "ppo",
+  "sigma": 0.5,
+  "K": 8,
+  "reward": "brier",
+  "clip": 0.2,
+  "passes": 1,
+  "lr": 0.0001,
+  "lr_head": 0.0002,
+  "rows_per_step": 32,
+  "train_seed": 0,
+  "replay": null,
+  "replay_frac": 0.0,
+  "replay_sha256": null,
+  "max_len": 3072,
+  "schedule": "warmup 0.03 + cosine to x0.05 over 500 steps",
+  "total_rows": 16000,
+  "lora": {
+   "r": 16,
+   "lora_alpha": 32,
+   "lora_dropout": 0.05
+  },
+  "trainer_sha256": "bc7c08e836bd99bb2665b02903cd41404ea029ecfa9b68a9ec1a7cda96958ba4",
+  "event": "ckpt",
+  "tag": "ckpt-16000",
+  "rows_seen": 16000,
+  "replay_seen": 0,
+  "step": 500,
+  "tokens": 10544036,
+  "tok_per_s": 825.8012937984138,
+  "rows_per_s": 1.253108458053419,
+  "val": {
+   "n": 2000,
+   "dropped_long": 0,
+   "noul_ce": 0.8274263739585876,
+   "noul_T": 3.9999965454650366,
+   "noul_auc": 0.7625054288907422,
+   "noul_acc": 0.7355,
+   "noul_f1": 0.8020950243172466,
+   "noul_nll": 0.798914428328244,
+   "noul_ece": 0.10070137844979762,
+   "noul_brier": 0.19162262414482661,
+   "noul_sharp": 0.3066386580467224,
+   "entropy": 0.3686918616294861
+  },
+  "dropped_long": 0,
+  "peak_gib": 18.236099243164062
+ }
+}

e16/S2-ppo1/ckpt-16000/value_head.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e84e572d0a48136601921f6a481f7f94573cb270d865fc7b93977e2b542d6cd8
+size 16524

e16/S2-ppo1/ckpt-4000/adapter/README.md

@@ -0,0 +1,206 @@
+---
+base_model: /workspace/jev_kt/models/hf/JEV-9B
+library_name: peft
+tags:
+- base_model:adapter:/workspace/jev_kt/models/hf/JEV-9B
+- lora
+- transformers
+---
+
+# 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 0.21.2

e16/S2-ppo1/ckpt-4000/adapter/adapter_config.json

@@ -0,0 +1,57 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": {
+    "base_model_class": "Qwen3_5ForCausalLM",
+    "parent_library": "transformers.models.qwen3_5.modeling_qwen3_5"
+  },
+  "base_model_name_or_path": "/workspace/jev_kt/models/hf/JEV-9B",
+  "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,
+  "kasa_config": null,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 32,
+  "lora_bias": false,
+  "lora_dropout": 0.05,
+  "lora_ga_config": null,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "monteclora_config": null,
+  "peft_type": "LORA",
+  "peft_version": "0.21.0",
+  "qalora_group_size": 16,
+  "r": 16,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "v_proj",
+    "o_proj",
+    "out_proj",
+    "gate_proj",
+    "down_proj",
+    "k_proj",
+    "up_proj",
+    "in_proj_z",
+    "in_proj_qkv",
+    "q_proj"
+  ],
+  "target_parameters": null,
+  "task_type": null,
+  "trainable_token_indices": null,
+  "use_bdlora": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false,
+  "velora_config": null
+}

e16/S2-ppo1/ckpt-4000/adapter/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:46def25560887f7e585e57a30ea2e1234c15eb52abf466e369aafa281072251b
+size 160486456

e16/S2-ppo1/ckpt-4000/adapter_vllm/adapter_config.json

@@ -0,0 +1,58 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": {
+    "base_model_class": "Qwen3_5ForCausalLM",
+    "parent_library": "transformers.models.qwen3_5.modeling_qwen3_5"
+  },
+  "base_model_name_or_path": null,
+  "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,
+  "kasa_config": null,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 64.0,
+  "lora_bias": false,
+  "lora_dropout": 0.0,
+  "lora_ga_config": null,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "monteclora_config": null,
+  "peft_type": "LORA",
+  "peft_version": "0.21.0",
+  "qalora_group_size": 16,
+  "r": 32,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "down_proj",
+    "gate_proj",
+    "in_proj_qkv",
+    "in_proj_z",
+    "k_proj",
+    "lm_head",
+    "o_proj",
+    "out_proj",
+    "q_proj",
+    "up_proj",
+    "v_proj"
+  ],
+  "target_parameters": null,
+  "task_type": null,
+  "trainable_token_indices": null,
+  "use_bdlora": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false,
+  "velora_config": null
+}

e16/S2-ppo1/ckpt-4000/adapter_vllm/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:ec62508128de9c80ad3107b5f030623d035d00656e4e9861bccf3eca2a28589b
+size 176641832

e16/S2-ppo1/ckpt-4000/adapter_vllm/decision_head.json

@@ -0,0 +1,99 @@
+{
+ "bias": [
+  0.0010664415312930942,
+  0.0015130405081436038,
+  -0.0025878134183585644,
+  -0.0012615364976227283,
+  -0.00025007399381138384,
+  0.0019558551721274853,
+  0.0013559526996687055,
+  -0.0023348911199718714,
+  -0.0004082925443071872,
+  -0.0002078723773593083,
+  -1.2957353646925185e-05,
+  0.0001676026004133746,
+  0.0005569425993598998,
+  -0.001412136247381568,
+  -0.0028663526754826307,
+  -0.005061559844762087,
+  -0.00593328895047307,
+  -0.004319756291806698,
+  -0.004922541789710522,
+  -0.0049067651852965355,
+  -0.003943523392081261,
+  -0.003319827374070883,
+  -0.005581587553024292,
+  -0.0034480984322726727
+ ],
+ "verbalizer_ids": [
+  3721,
+  1802,
+  15,
+  16,
+  17,
+  18,
+  19,
+  20,
+  32,
+  33,
+  34,
+  35,
+  36,
+  37,
+  38,
+  39,
+  40,
+  41,
+  42,
+  43,
+  44,
+  45,
+  46,
+  47
+ ],
+ "slots": {
+  "num_slots": 24,
+  "ranges": {
+   "noul": [
+    0,
+    2
+   ],
+   "score": [
+    2,
+    8
+   ],
+   "choice": [
+    8,
+    24
+   ]
+  },
+  "verbalizers": [
+   "false",
+   "true",
+   "0",
+   "1",
+   "2",
+   "3",
+   "4",
+   "5",
+   "A",
+   "B",
+   "C",
+   "D",
+   "E",
+   "F",
+   "G",
+   "H",
+   "I",
+   "J",
+   "K",
+   "L",
+   "M",
+   "N",
+   "O",
+   "P"
+  ],
+  "template_version": "bare-v1"
+ },
+ "note": "decision logits = lm_head-LoRA logprobs of verbalizer ids + bias, then per-kind temperature"
+}

e16/S2-ppo1/ckpt-4000/calibration.json

@@ -0,0 +1,17 @@
+{
+ "version": 1,
+ "per_kind": {
+  "noul": 1.2891889151929385,
+  "choice": 1.0,
+  "score": 1.0
+ },
+ "fit": {
+  "noul": {
+   "n": 2000,
+   "T": 1.2891889151929385,
+   "ce_before": 0.5674813985824585,
+   "ce_after": 0.5615489482879639
+  }
+ },
+ "source": "models/e16/S2-ppo1/ckpt-4000"
+}

e16/S2-ppo1/ckpt-4000/head.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d7bb2761a5a959d01bc6324eeef4e52c429d838f67ddebeda25f5488d682674f
+size 393464

e16/S2-ppo1/ckpt-4000/judge_config.json

@@ -0,0 +1,160 @@
+{
+ "base_model_path": "/root/models/Qwen3.5-9B",
+ "hidden_size": 4096,
+ "slots": {
+  "num_slots": 24,
+  "ranges": {
+   "noul": [
+    0,
+    2
+   ],
+   "score": [
+    2,
+    8
+   ],
+   "choice": [
+    8,
+    24
+   ]
+  },
+  "verbalizers": [
+   "false",
+   "true",
+   "0",
+   "1",
+   "2",
+   "3",
+   "4",
+   "5",
+   "A",
+   "B",
+   "C",
+   "D",
+   "E",
+   "F",
+   "G",
+   "H",
+   "I",
+   "J",
+   "K",
+   "L",
+   "M",
+   "N",
+   "O",
+   "P"
+  ],
+  "template_version": "bare-v1"
+ },
+ "verbalizer_ids": [
+  3721,
+  1802,
+  15,
+  16,
+  17,
+  18,
+  19,
+  20,
+  32,
+  33,
+  34,
+  35,
+  36,
+  37,
+  38,
+  39,
+  40,
+  41,
+  42,
+  43,
+  44,
+  45,
+  46,
+  47
+ ],
+ "kinds": [
+  "noul",
+  "choice",
+  "score"
+ ],
+ "model_name": "jev-judge-qwen35-9b",
+ "model_version": "0.8.0",
+ "lm_head_included": true,
+ "weights_mode": "unmerged",
+ "adapter_subfolder": "adapter",
+ "lora": {
+  "r": 16,
+  "lora_alpha": 32,
+  "lora_dropout": 0.05,
+  "target_modules": [
+   "down_proj",
+   "gate_proj",
+   "in_proj_qkv",
+   "in_proj_z",
+   "k_proj",
+   "o_proj",
+   "out_proj",
+   "q_proj",
+   "up_proj",
+   "v_proj"
+  ]
+ },
+ "trained_stage": "e16",
+ "source_checkpoint": "models/e16/S2-ppo1/ckpt-4000",
+ "softcap": null,
+ "e16": {
+  "data": "data/e15/F1-pool",
+  "data_manifest": {
+   "arm": "F1-pool",
+   "name": "F1-pool",
+   "train_sha256": "dad16f94a5659768b254a6c12538c01f9c64cd43d574a3435bcc70c9c1bac130",
+   "n_train": 64000
+  },
+  "init": "released",
+  "objective": "ppo",
+  "sigma": 0.5,
+  "K": 8,
+  "reward": "brier",
+  "clip": 0.2,
+  "passes": 1,
+  "lr": 0.0001,
+  "lr_head": 0.0002,
+  "rows_per_step": 32,
+  "train_seed": 0,
+  "replay": null,
+  "replay_frac": 0.0,
+  "replay_sha256": null,
+  "max_len": 3072,
+  "schedule": "warmup 0.03 + cosine to x0.05 over 500 steps",
+  "total_rows": 16000,
+  "lora": {
+   "r": 16,
+   "lora_alpha": 32,
+   "lora_dropout": 0.05
+  },
+  "trainer_sha256": "bc7c08e836bd99bb2665b02903cd41404ea029ecfa9b68a9ec1a7cda96958ba4",
+  "event": "ckpt",
+  "tag": "ckpt-4000",
+  "rows_seen": 4000,
+  "replay_seen": 0,
+  "step": 125,
+  "tokens": 2631181,
+  "tok_per_s": 779.7516448040263,
+  "rows_per_s": 1.1854017540430937,
+  "val": {
+   "n": 2000,
+   "dropped_long": 0,
+   "noul_ce": 0.5674813985824585,
+   "noul_T": 1.2891889151929385,
+   "noul_auc": 0.7611031282691971,
+   "noul_acc": 0.7125,
+   "noul_f1": 0.7740667976424361,
+   "noul_nll": 0.5674813731483882,
+   "noul_ece": 0.08918076078593729,
+   "noul_brier": 0.1898530646799337,
+   "noul_sharp": 0.23238085210323334,
+   "entropy": 0.5174486637115479
+  },
+  "dropped_long": 0,
+  "peak_gib": 18.23355484008789
+ }
+}

e16/S2-ppo1/ckpt-4000/value_head.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:e3e173fa2e71361a9a61ce649e7a4e6597e329f692dc7ee9be3138b26ebcfba9
+size 16524

e16/S2-ppo1/train_log.jsonl

@@ -0,0 +1,58 @@
+{"time": "07:34:46", "event": "start", "data": "data/e15/F1-pool", "data_manifest": {"arm": "F1-pool", "name": "F1-pool", "train_sha256": "dad16f94a5659768b254a6c12538c01f9c64cd43d574a3435bcc70c9c1bac130", "n_train": 64000}, "init": "released", "objective": "ppo", "sigma": 0.5, "K": 8, "reward": "brier", "clip": 0.2, "passes": 1, "lr": 0.0001, "lr_head": 0.0002, "rows_per_step": 32, "train_seed": 0, "replay": null, "replay_frac": 0.0, "replay_sha256": null, "max_len": 3072, "schedule": "warmup 0.03 + cosine to x0.05 over 500 steps", "total_rows": 16000, "lora": {"r": 16, "lora_alpha": 32, "lora_dropout": 0.05}, "trainer_sha256": "bc7c08e836bd99bb2665b02903cd41404ea029ecfa9b68a9ec1a7cda96958ba4", "device": "cuda", "load_s": 84.87204360961914}
+{"time": "07:38:39", "event": "step", "step": 10, "rows_seen": 320, "lr": 7.333333333333333e-05, "tok_per_s": 907.9472281561525, "loss": -0.10062819719314575, "entropy": 0.34502604603767395, "sharp": 0.348164152354002, "r_mean": 0.5968956500291824, "r_std": 0.13000543539722761, "adv_abs": 0.557626927892367, "frac_lowstd": 0.061111112435658775}
+{"time": "07:42:34", "event": "step", "step": 20, "rows_seen": 640, "lr": 9.997508952966197e-05, "tok_per_s": 911.6815737145886, "loss": 0.5487492429092526, "entropy": 0.518937461078167, "sharp": 0.24511030688881874, "r_mean": 0.4521929791995457, "r_std": 0.23753618342535837, "adv_abs": 0.4617865192038672, "frac_lowstd": 0.023809524519102915}
+{"time": "07:46:26", "event": "step", "step": 30, "rows_seen": 960, "lr": 9.977596250541417e-05, "tok_per_s": 908.7817342626987, "loss": 0.012478791177272797, "entropy": 0.43727659806609154, "sharp": 0.31330038234591484, "r_mean": 0.6026445882661002, "r_std": 0.137933007308415, "adv_abs": 0.5287195018359593, "frac_lowstd": 0.023809524519102915}
+{"time": "07:50:20", "event": "step", "step": 40, "rows_seen": 1280, "lr": 9.937854366485982e-05, "tok_per_s": 912.14609198114, "loss": 0.5721414498984814, "entropy": 0.36007790779694915, "sharp": 0.3615472614765167, "r_mean": 0.2860143908432552, "r_std": 0.12656146713665553, "adv_abs": 0.685446805187634, "frac_lowstd": 0.0}
+{"time": "07:54:11", "event": "step", "step": 50, "rows_seen": 1600, "lr": 9.8784499920729e-05, "tok_per_s": 913.2879617220298, "loss": -0.025918788742274046, "entropy": 0.5196073669940233, "sharp": 0.24270944856107235, "r_mean": 0.504151599747794, "r_std": 0.18696345069578715, "adv_abs": 0.494312652519771, "frac_lowstd": 0.0}
+{"time": "07:58:02", "event": "step", "step": 60, "rows_seen": 1920, "lr": 9.79963228989294e-05, "tok_per_s": 913.6812839466801, "loss": 0.20761062344536185, "entropy": 0.5377308204770088, "sharp": 0.2291932199150324, "r_mean": 0.5811150733913694, "r_std": 0.16059514667306626, "adv_abs": 0.34182892739772797, "frac_lowstd": 0.0}
+{"time": "08:01:51", "event": "step", "step": 70, "rows_seen": 2240, "lr": 9.701731848780134e-05, "tok_per_s": 912.9037240326826, "loss": 0.048857021145522594, "entropy": 0.3266878188587725, "sharp": 0.37674737349152565, "r_mean": 0.5980443358421326, "r_std": 0.11558598945183414, "adv_abs": 0.5162245205470494, "frac_lowstd": 0.03571428571428571}
+{"time": "08:05:42", "event": "step", "step": 80, "rows_seen": 2560, "lr": 9.585159297207315e-05, "tok_per_s": 913.3866896363941, "loss": 0.05168980825692415, "entropy": 0.40748671162873507, "sharp": 0.3138065245002508, "r_mean": 0.7385147128786359, "r_std": 0.12422377110591956, "adv_abs": 0.2630577108689717, "frac_lowstd": 0.023809524519102915}
+{"time": "08:09:34", "event": "step", "step": 90, "rows_seen": 2880, "lr": 9.450403580967612e-05, "tok_per_s": 912.9799149319759, "loss": 0.044061752036213875, "entropy": 0.3909636801108718, "sharp": 0.32252061180770397, "r_mean": 0.6589292202677045, "r_std": 0.11813291415039982, "adv_abs": 0.47996410727500916, "frac_lowstd": 0.04761904903820583}
+{"time": "08:13:24", "event": "step", "step": 100, "rows_seen": 3200, "lr": 9.298029912365855e-05, "tok_per_s": 913.5905633028036, "loss": 0.19180638925172389, "entropy": 0.2961344141513109, "sharp": 0.3482120893895626, "r_mean": 0.6371859950678689, "r_std": 0.10313258325913921, "adv_abs": 0.34629505234105246, "frac_lowstd": 0.34557823836803436}
+{"time": "08:15:06", "event": "val", "step": 100, "rows_seen": 3200, "n": 500, "dropped_long": 0, "noul_ce": 0.6930883526802063, "noul_T": 2.620361445044101, "noul_auc": 0.7590607724986225, "noul_acc": 0.72, "noul_f1": 0.7916666666666666, "noul_nll": 0.6930884000658989, "noul_ece": 0.12150582593679429, "noul_brier": 0.19745737990138673, "noul_sharp": 0.3200295567512512, "entropy": 0.3527287542819977}
+{"time": "08:18:57", "event": "step", "step": 110, "rows_seen": 3520, "lr": 9.128677399521752e-05, "tok_per_s": 879.7282928149591, "loss": 0.20249656587839127, "entropy": 0.4548079938394949, "sharp": 0.2460484728217125, "r_mean": 0.5970774016210011, "r_std": 0.16056389406730887, "adv_abs": 0.3743857592344284, "frac_lowstd": 0.21111111236470087}
+{"time": "08:22:38", "event": "step", "step": 120, "rows_seen": 3840, "lr": 8.943056365728335e-05, "tok_per_s": 882.5726319279836, "loss": 0.012051641009747982, "entropy": 0.5837420225143433, "sharp": 0.1725786291062832, "r_mean": 0.574203108038221, "r_std": 0.21641392154353006, "adv_abs": 0.33200314215251375, "frac_lowstd": 0.017857142857142856}
+{"time": "08:31:02", "event": "ckpt", "tag": "ckpt-4000", "rows_seen": 4000, "replay_seen": 0, "step": 125, "tokens": 2631181, "tok_per_s": 779.7516448040263, "rows_per_s": 1.1854017540430937, "val": {"n": 2000, "dropped_long": 0, "noul_ce": 0.5674813985824585, "noul_T": 1.2891889151929385, "noul_auc": 0.7611031282691971, "noul_acc": 0.7125, "noul_f1": 0.7740667976424361, "noul_nll": 0.5674813731483882, "noul_ece": 0.08918076078593729, "noul_brier": 0.1898530646799337, "noul_sharp": 0.23238085210323334, "entropy": 0.5174486637115479}, "dropped_long": 0, "peak_gib": 18.23355484008789}
+{"time": "08:32:58", "event": "step", "step": 130, "rows_seen": 4160, "lr": 8.741945370109202e-05, "tok_per_s": 784.3948596928083, "loss": 0.13270092103630304, "entropy": 0.5571417789906263, "sharp": 0.21371298097074032, "r_mean": 0.5890914372035435, "r_std": 0.2114839894430978, "adv_abs": 0.2904020771384239, "frac_lowstd": 0.0}
+{"time": "08:36:48", "event": "step", "step": 140, "rows_seen": 4480, "lr": 8.526187942070997e-05, "tok_per_s": 792.3890024967847, "loss": 0.17276590084657073, "entropy": 0.3802753137424588, "sharp": 0.3467848375439644, "r_mean": 0.569410081420626, "r_std": 0.11759821229082133, "adv_abs": 0.4644115822655814, "frac_lowstd": 0.08928571428571429}
+{"time": "08:40:39", "event": "step", "step": 150, "rows_seen": 4800, "lr": 8.296689043247915e-05, "tok_per_s": 799.7454050777327, "loss": -0.13896020874381065, "entropy": 0.3656533781904727, "sharp": 0.34835503436625004, "r_mean": 0.7546614919389997, "r_std": 0.10955366381676868, "adv_abs": 0.3953996683870043, "frac_lowstd": 0.1952380963734218}
+{"time": "08:44:32", "event": "step", "step": 160, "rows_seen": 5120, "lr": 8.054411271778132e-05, "tok_per_s": 806.6061431251497, "loss": 0.017451470717787743, "entropy": 0.5088797770440578, "sharp": 0.2611468434333801, "r_mean": 0.6431095550457636, "r_std": 0.22123494247595468, "adv_abs": 0.3359515021244685, "frac_lowstd": 0.0}
+{"time": "08:48:23", "event": "step", "step": 170, "rows_seen": 5440, "lr": 7.800370824832674e-05, "tok_per_s": 812.3801784639942, "loss": 0.019212193321436644, "entropy": 0.33678371366113424, "sharp": 0.3650928698480129, "r_mean": 0.7211570654596601, "r_std": 0.08662958280183375, "adv_abs": 0.4109207625899996, "frac_lowstd": 0.04421768869672503}
+{"time": "08:52:04", "event": "step", "step": 180, "rows_seen": 5760, "lr": 7.535633236331324e-05, "tok_per_s": 817.5667660264351, "loss": 0.018216995522379875, "entropy": 0.38516590744256973, "sharp": 0.3201605938374996, "r_mean": 0.5858667095502218, "r_std": 0.10441108047962189, "adv_abs": 0.5023583223422369, "frac_lowstd": 0.023809524873892467}
+{"time": "08:55:50", "event": "step", "step": 190, "rows_seen": 6080, "lr": 7.26130890772308e-05, "tok_per_s": 822.6962215886304, "loss": 0.20301194209605455, "entropy": 0.5214720331132412, "sharp": 0.24415834248065948, "r_mean": 0.4913227345262255, "r_std": 0.20688924938440323, "adv_abs": 0.39063967977251324, "frac_lowstd": 0.0}
+{"time": "08:59:47", "event": "step", "step": 200, "rows_seen": 6400, "lr": 6.978548450576568e-05, "tok_per_s": 827.2476551032426, "loss": -0.09183507692068815, "entropy": 0.5628727218136191, "sharp": 0.18681260012090206, "r_mean": 0.6483349076339177, "r_std": 0.20361532270908356, "adv_abs": 0.35482534766197205, "frac_lowstd": 0.0}
+{"time": "09:01:29", "event": "val", "step": 200, "rows_seen": 6400, "n": 500, "dropped_long": 0, "noul_ce": 0.5436476469039917, "noul_T": 1.1168494762749452, "noul_auc": 0.7713610266801756, "noul_acc": 0.73, "noul_f1": 0.8192771084337349, "noul_nll": 0.5436476763375103, "noul_ece": 0.07009069311618807, "noul_brier": 0.18247039393823308, "noul_sharp": 0.2260569930076599, "entropy": 0.5214588642120361}
+{"time": "09:05:19", "event": "step", "step": 210, "rows_seen": 6720, "lr": 6.688537860515266e-05, "tok_per_s": 815.8060012937893, "loss": 0.14487420860677958, "entropy": 0.4690675539895892, "sharp": 0.2594572212547064, "r_mean": 0.5161036508423942, "r_std": 0.1544741988182068, "adv_abs": 0.4419285889182772, "frac_lowstd": 0.0}
+{"time": "09:09:01", "event": "step", "step": 220, "rows_seen": 7040, "lr": 6.392493542739706e-05, "tok_per_s": 820.0374504837364, "loss": 0.07803836464881897, "entropy": 0.3247750848531723, "sharp": 0.36480313912034035, "r_mean": 0.6686622376243273, "r_std": 0.11130194536720713, "adv_abs": 0.42069390912850696, "frac_lowstd": 0.079629631092151}
+{"time": "09:12:51", "event": "step", "step": 230, "rows_seen": 7360, "lr": 6.091657210001309e-05, "tok_per_s": 824.3238507950664, "loss": 0.1672395197674632, "entropy": 0.3978843726217747, "sharp": 0.30897514894604683, "r_mean": 0.3908121053661619, "r_std": 0.16546081325837544, "adv_abs": 0.5218781637293952, "frac_lowstd": 0.14285714498588017}
+{"time": "09:16:32", "event": "step", "step": 240, "rows_seen": 7680, "lr": 5.7872906744276015e-05, "tok_per_s": 827.9162563877517, "loss": -0.06889456510543823, "entropy": 0.48038566671311855, "sharp": 0.25631483271718025, "r_mean": 0.6577174024922507, "r_std": 0.1821632566196578, "adv_abs": 0.3591307657105582, "frac_lowstd": 0.0}
+{"time": "09:20:21", "event": "step", "step": 250, "rows_seen": 8000, "lr": 5.48067055504363e-05, "tok_per_s": 831.2294705919436, "loss": -0.04639704106375575, "entropy": 0.5545359030365944, "sharp": 0.21875476092100143, "r_mean": 0.7660609143120902, "r_std": 0.15749658165233477, "adv_abs": 0.23684245773724147, "frac_lowstd": 0.0}
+{"time": "09:24:02", "event": "step", "step": 260, "rows_seen": 8320, "lr": 5.173082923188119e-05, "tok_per_s": 834.2822663575793, "loss": -0.04451965121552348, "entropy": 0.5211928011849523, "sharp": 0.2369103990495205, "r_mean": 0.6660864225455693, "r_std": 0.1869723807488169, "adv_abs": 0.28462146648338865, "frac_lowstd": 0.0}
+{"time": "09:27:51", "event": "step", "step": 270, "rows_seen": 8640, "lr": 4.865817908283331e-05, "tok_per_s": 837.0023171747524, "loss": 0.13799094501882792, "entropy": 0.503261387348175, "sharp": 0.2529892586171627, "r_mean": 0.6103001981973648, "r_std": 0.16788020233313242, "adv_abs": 0.3094373221198718, "frac_lowstd": 0.023809524873892467}
+{"time": "09:31:39", "event": "step", "step": 280, "rows_seen": 8960, "lr": 4.560164286583922e-05, "tok_per_s": 839.9754582742358, "loss": 0.019665569067001343, "entropy": 0.27868305146694183, "sharp": 0.36947210878133774, "r_mean": 0.49902170470782686, "r_std": 0.09612465002906642, "adv_abs": 0.4860029305730547, "frac_lowstd": 0.2493197364466531}
+{"time": "09:35:31", "event": "step", "step": 290, "rows_seen": 9280, "lr": 4.257404075601443e-05, "tok_per_s": 842.7360880331586, "loss": 0.056465772446244955, "entropy": 0.2337416979135014, "sharp": 0.3823592755943537, "r_mean": 0.7703629221235003, "r_std": 0.08782875645403822, "adv_abs": 0.2894987996135439, "frac_lowstd": 0.5503401415688651}
+{"time": "09:39:19", "event": "step", "step": 300, "rows_seen": 9600, "lr": 3.958807156877371e-05, "tok_per_s": 844.9896860187076, "loss": 0.1306108224671334, "entropy": 0.21112133913999287, "sharp": 0.4165404234081507, "r_mean": 0.5732471857752118, "r_std": 0.07442874275147915, "adv_abs": 0.4736809581518173, "frac_lowstd": 0.44557823879378183}
+{"time": "09:41:00", "event": "val", "step": 300, "rows_seen": 9600, "n": 500, "dropped_long": 0, "noul_ce": 1.179489254951477, "noul_T": 3.999999975766852, "noul_auc": 0.7750671003750511, "noul_acc": 0.726, "noul_f1": 0.793984962406015, "noul_nll": 1.0771513626209344, "noul_ece": 0.18296374940872195, "noul_brier": 0.21464291502818877, "noul_sharp": 0.4061043858528137, "entropy": 0.20697131752967834}
+{"time": "09:44:56", "event": "step", "step": 310, "rows_seen": 9920, "lr": 3.665625949658603e-05, "tok_per_s": 836.1652764142322, "loss": 0.2587154991924763, "entropy": 0.2610930644441396, "sharp": 0.3824239680543542, "r_mean": 0.46150543221405577, "r_std": 0.11188322946795129, "adv_abs": 0.6260877932821002, "frac_lowstd": 0.3489795944520405}
+{"time": "09:48:43", "event": "step", "step": 320, "rows_seen": 10240, "lr": 3.379090157815887e-05, "tok_per_s": 838.6081594364458, "loss": -0.04159629996865988, "entropy": 0.2937415544874966, "sharp": 0.3573644142597914, "r_mean": 0.773408476795469, "r_std": 0.08789249296699252, "adv_abs": 0.3822525015899113, "frac_lowstd": 0.3979591876268387}
+{"time": "09:52:31", "event": "step", "step": 330, "rows_seen": 10560, "lr": 3.10040161203844e-05, "tok_per_s": 840.7211256312369, "loss": 0.2548190727829933, "entropy": 0.3176150772778783, "sharp": 0.31754723377525806, "r_mean": 0.3473818177978198, "r_std": 0.12582076193454364, "adv_abs": 0.6359462241331736, "frac_lowstd": 0.3851190507411957}
+{"time": "09:56:24", "event": "step", "step": 340, "rows_seen": 10880, "lr": 2.8307292289383776e-05, "tok_per_s": 843.0410479307234, "loss": 0.3351435218937695, "entropy": 0.3212029468268156, "sharp": 0.34085709042847157, "r_mean": 0.5009898426277297, "r_std": 0.11618127407772201, "adv_abs": 0.5056653959410531, "frac_lowstd": 0.3047619121415274}
+{"time": "10:00:14", "event": "step", "step": 350, "rows_seen": 11200, "lr": 2.571204108208194e-05, "tok_per_s": 845.1656870382828, "loss": 0.26014011865481734, "entropy": 0.29976393188113004, "sharp": 0.33978276513516903, "r_mean": 0.6333315372467041, "r_std": 0.0935145309776999, "adv_abs": 0.38177580386400223, "frac_lowstd": 0.6169642936438322}
+{"time": "10:04:00", "event": "step", "step": 360, "rows_seen": 11520, "lr": 2.322914788395595e-05, "tok_per_s": 847.0181975008478, "loss": -0.12522980198264122, "entropy": 0.34794481936842203, "sharp": 0.31197429820895195, "r_mean": 0.7309423897947583, "r_std": 0.1301973651030234, "adv_abs": 0.41820041622434345, "frac_lowstd": 0.3595238136393683}
+{"time": "10:07:45", "event": "step", "step": 370, "rows_seen": 11840, "lr": 2.0869026811945157e-05, "tok_per_s": 848.9752152788359, "loss": 0.24546496511175064, "entropy": 0.3289804222040402, "sharp": 0.334681011736393, "r_mean": 0.4808573084218161, "r_std": 0.11013877760602295, "adv_abs": 0.5016851425170898, "frac_lowstd": 0.4238095283508301}
+{"time": "10:11:34", "event": "step", "step": 380, "rows_seen": 12160, "lr": 1.8641577034025627e-05, "tok_per_s": 850.776938738168, "loss": 0.21878909226506948, "entropy": 0.3216428722571436, "sharp": 0.34390960447490215, "r_mean": 0.5413691816585404, "r_std": 0.09835160418255587, "adv_abs": 0.4582563723836626, "frac_lowstd": 0.4965986503022058}
+{"time": "10:15:22", "event": "step", "step": 390, "rows_seen": 12480, "lr": 1.6556141248659472e-05, "tok_per_s": 852.5179028908922, "loss": -0.011037901975214481, "entropy": 0.29534372067428194, "sharp": 0.3362016510218382, "r_mean": 0.7485703911100116, "r_std": 0.09038488911440774, "adv_abs": 0.36168008616992403, "frac_lowstd": 0.5619047688586372}
+{"time": "10:19:13", "event": "step", "step": 400, "rows_seen": 12800, "lr": 1.4621466498270773e-05, "tok_per_s": 853.9906906596458, "loss": -0.17132338136434555, "entropy": 0.3502153321232413, "sharp": 0.31645902805030346, "r_mean": 0.8222529590129852, "r_std": 0.1070341213279562, "adv_abs": 0.3317518047988415, "frac_lowstd": 0.42395833507180214}
+{"time": "10:20:54", "event": "val", "step": 400, "rows_seen": 12800, "n": 500, "dropped_long": 0, "noul_ce": 0.9584631323814392, "noul_T": 3.999999975766852, "noul_auc": 0.7722053360351233, "noul_acc": 0.738, "noul_f1": 0.8198074277854195, "noul_nll": 0.8467114506701473, "noul_ece": 0.10757567811012268, "noul_brier": 0.18982341425339472, "noul_sharp": 0.3077302575111389, "entropy": 0.35775360465049744}
+{"time": "10:24:48", "event": "step", "step": 410, "rows_seen": 13120, "lr": 1.284566748110963e-05, "tok_per_s": 846.9071597974621, "loss": -0.0864783301949501, "entropy": 0.2633080780506134, "sharp": 0.3685060478746891, "r_mean": 0.7315677305062612, "r_std": 0.11778514583905537, "adv_abs": 0.30431149651606876, "frac_lowstd": 0.49523810545603436}
+{"time": "10:28:38", "event": "step", "step": 420, "rows_seen": 13440, "lr": 1.1236192515387821e-05, "tok_per_s": 848.619446130219, "loss": 0.10874498775228858, "entropy": 0.3306534863077104, "sharp": 0.31445624865591526, "r_mean": 0.5834584443696907, "r_std": 0.13049794650370522, "adv_abs": 0.3966461718082428, "frac_lowstd": 0.40408164049897877}
+{"time": "10:32:32", "event": "step", "step": 430, "rows_seen": 13760, "lr": 9.799792298443728e-06, "tok_per_s": 850.0366133924922, "loss": 0.09318238857667893, "entropy": 0.36213047057390213, "sharp": 0.30851134564727545, "r_mean": 0.5558774396777153, "r_std": 0.20499365670340403, "adv_abs": 0.3829157607895987, "frac_lowstd": 0.2333333385842187}
+{"time": "10:36:17", "event": "step", "step": 440, "rows_seen": 14080, "lr": 8.542491591971554e-06, "tok_per_s": 851.4579600485541, "loss": 0.07825499027967453, "entropy": 0.35938821360468864, "sharp": 0.307557912543416, "r_mean": 0.6539335201183955, "r_std": 0.12492635970314343, "adv_abs": 0.31966207176446915, "frac_lowstd": 0.34206349899371463}
+{"time": "10:40:09", "event": "step", "step": 450, "rows_seen": 14400, "lr": 7.469563952077155e-06, "tok_per_s": 852.7962341018363, "loss": 0.035682094283401966, "entropy": 0.438163994345814, "sharp": 0.2811601758003235, "r_mean": 0.5815641826816967, "r_std": 0.13148958986026368, "adv_abs": 0.44219429578099934, "frac_lowstd": 0.23605442472866603}
+{"time": "10:43:58", "event": "step", "step": 460, "rows_seen": 14720, "lr": 6.585509610150972e-06, "tok_per_s": 854.1607747416452, "loss": 0.23412014916539192, "entropy": 0.3161262981593609, "sharp": 0.3506321422755718, "r_mean": 0.5459222506199565, "r_std": 0.12991032004356384, "adv_abs": 0.439707487821579, "frac_lowstd": 0.4074829987117222}
+{"time": "10:47:51", "event": "step", "step": 470, "rows_seen": 15040, "lr": 5.894036597333646e-06, "tok_per_s": 855.5766996666347, "loss": -0.002823292277753353, "entropy": 0.4535855082795024, "sharp": 0.24469432327896357, "r_mean": 0.6402933725288936, "r_std": 0.1697962044605187, "adv_abs": 0.28733402277742115, "frac_lowstd": 0.22040816502911703}
+{"time": "10:51:38", "event": "step", "step": 480, "rows_seen": 15360, "lr": 5.398045191744809e-06, "tok_per_s": 856.6946814313796, "loss": 0.07713474240154028, "entropy": 0.39639689055184135, "sharp": 0.29076074436306953, "r_mean": 0.6329786976178488, "r_std": 0.15908683153490225, "adv_abs": 0.35313212871551514, "frac_lowstd": 0.412500003973643}
+{"time": "10:55:27", "event": "step", "step": 490, "rows_seen": 15680, "lr": 5.099615753708697e-06, "tok_per_s": 858.0320471265909, "loss": 0.11640177294611931, "entropy": 0.3359264500140853, "sharp": 0.32571980729699135, "r_mean": 0.5713428604815688, "r_std": 0.10702035578891612, "adv_abs": 0.47965811831610544, "frac_lowstd": 0.43231292920453207}
+{"time": "10:59:17", "event": "step", "step": 500, "rows_seen": 16000, "lr": 5e-06, "tok_per_s": 859.2412629917751, "loss": -0.06778925564140081, "entropy": 0.325115567073226, "sharp": 0.32888365536928177, "r_mean": 0.7338430682818095, "r_std": 0.12558194001515707, "adv_abs": 0.35451454917589825, "frac_lowstd": 0.3571428631742795}
+{"time": "11:00:58", "event": "val", "step": 500, "rows_seen": 16000, "n": 500, "dropped_long": 0, "noul_ce": 0.8109627366065979, "noul_T": 3.999999975766852, "noul_auc": 0.7681526511313745, "noul_acc": 0.73, "noul_f1": 0.7926267281105991, "noul_nll": 0.7823652381536376, "noul_ece": 0.11622801256179809, "noul_brier": 0.19205651993086767, "noul_sharp": 0.29767873883247375, "entropy": 0.38158175349235535}
+{"time": "11:07:36", "event": "ckpt", "tag": "ckpt-16000", "rows_seen": 16000, "replay_seen": 0, "step": 500, "tokens": 10544036, "tok_per_s": 825.8012937984138, "rows_per_s": 1.253108458053419, "val": {"n": 2000, "dropped_long": 0, "noul_ce": 0.8274263739585876, "noul_T": 3.9999965454650366, "noul_auc": 0.7625054288907422, "noul_acc": 0.7355, "noul_f1": 0.8020950243172466, "noul_nll": 0.798914428328244, "noul_ece": 0.10070137844979762, "noul_brier": 0.19162262414482661, "noul_sharp": 0.3066386580467224, "entropy": 0.3686918616294861}, "dropped_long": 0, "peak_gib": 18.236099243164062}