Add files using upload-large-folder tool

.gitattributes

@@ -34,3 +34,4 @@ saved_model/**/* filter=lfs diff=lfs merge=lfs -text
 *.zst filter=lfs diff=lfs merge=lfs -text
 *tfevents* filter=lfs diff=lfs merge=lfs -text
 qwen3-4b-instruct/base/tokenizer/tokenizer.json filter=lfs diff=lfs merge=lfs -text
+qwen3-4b-instruct/dp8/tokenizer/tokenizer.json filter=lfs diff=lfs merge=lfs -text

qwen3-4b-instruct/dp8/adapter/README.md

@@ -0,0 +1,207 @@
+---
+base_model: Qwen/Qwen3-4B-Instruct-2507
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-4B-Instruct-2507
+- 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 0.18.1

qwen3-4b-instruct/dp8/adapter/adapter_config.json

@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-4B-Instruct-2507",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": true,
+  "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": 32,
+  "lora_bias": false,
+  "lora_dropout": 0.05,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": [
+    "lm_head",
+    "embed_tokens"
+  ],
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 16,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "k_proj",
+    "v_proj",
+    "q_proj",
+    "o_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}

qwen3-4b-instruct/dp8/adapter/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:19061f24801c4d66921418e3cb1135de75978cb4d228db814b0755fbaada6bc4
+size 4721857072

qwen3-4b-instruct/dp8/audit_results.json

@@ -0,0 +1,137 @@
+{
+  "delta": 1e-05,
+  "num_canaries": 500,
+  "num_members": 250,
+  "paper_guess_fraction": 0.2,
+  "paper_guess_steps": 20,
+  "loss": {
+    "auc": 0.515088,
+    "empirical_epsilon": {
+      "0.05": 0.0,
+      "0.01": 0.0
+    },
+    "empirical_epsilon_details": {
+      "0.05": {
+        "epsilon": 0.0,
+        "num_guesses": 0,
+        "correct_guesses": 0,
+        "candidate_num_guesses": [
+          5,
+          10,
+          15,
+          20,
+          25,
+          30,
+          35,
+          40,
+          45,
+          50,
+          55,
+          60,
+          65,
+          70,
+          75,
+          80,
+          85,
+          90,
+          95,
+          100
+        ],
+        "direction": "lower"
+      },
+      "0.01": {
+        "epsilon": 0.0,
+        "num_guesses": 0,
+        "correct_guesses": 0,
+        "candidate_num_guesses": [
+          5,
+          10,
+          15,
+          20,
+          25,
+          30,
+          35,
+          40,
+          45,
+          50,
+          55,
+          60,
+          65,
+          70,
+          75,
+          80,
+          85,
+          90,
+          95,
+          100
+        ],
+        "direction": "lower"
+      }
+    }
+  },
+  "embedding": {
+    "auc": 0.516208,
+    "empirical_epsilon": {
+      "0.05": 0.0,
+      "0.01": 0.0
+    },
+    "empirical_epsilon_details": {
+      "0.05": {
+        "epsilon": 0.0,
+        "num_guesses": 0,
+        "correct_guesses": 0,
+        "candidate_num_guesses": [
+          5,
+          10,
+          15,
+          20,
+          25,
+          30,
+          35,
+          40,
+          45,
+          50,
+          55,
+          60,
+          65,
+          70,
+          75,
+          80,
+          85,
+          90,
+          95,
+          100
+        ],
+        "direction": "lower"
+      },
+      "0.01": {
+        "epsilon": 0.0,
+        "num_guesses": 0,
+        "correct_guesses": 0,
+        "candidate_num_guesses": [
+          5,
+          10,
+          15,
+          20,
+          25,
+          30,
+          35,
+          40,
+          45,
+          50,
+          55,
+          60,
+          65,
+          70,
+          75,
+          80,
+          85,
+          90,
+          95,
+          100
+        ],
+        "direction": "lower"
+      }
+    }
+  }
+}

qwen3-4b-instruct/dp8/audit_scores.npz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:126c74d41f98d08626da6ca1f818fce7bb97a1b2a1dad6827e8baa20288a484f
+size 12784

qwen3-4b-instruct/dp8/codecarbon.csv

@@ -0,0 +1,2 @@
+timestamp,project_name,run_id,experiment_id,duration,emissions,emissions_rate,cpu_power,gpu_power,ram_power,cpu_energy,gpu_energy,ram_energy,energy_consumed,water_consumed,country_name,country_iso_code,region,cloud_provider,cloud_region,os,python_version,codecarbon_version,cpu_count,cpu_model,gpu_count,gpu_model,longitude,latitude,ram_total_size,tracking_mode,cpu_utilization_percent,gpu_utilization_percent,ram_utilization_percent,ram_used_gb,on_cloud,pue,wue
+2026-03-17T17:05:44,codedp-qwen3-4b-instruct-cpt-dp8,2b4560e3-78e0-4790-b542-38dc1a784383,5b0fa12a-3dd7-45bb-9766-cc326314d9f1,2693.6014373912476,0.09826528580395394,3.648100436830916e-05,72.03163066573623,3108.4298352234105,54.0,0.051919376660011826,2.3238466076869315,0.03891114561875117,2.414677129965695,0.0,Sweden,SWE,östergötland county,,,Linux-6.8.0-94-generic-x86_64-with-glibc2.39,3.11.0,3.2.3,256,AMD EPYC 9554 64-Core Processor,8,8 x NVIDIA H200,16.1885,58.594,1511.49019241333,machine,3.719566840926081,83.2715646004481,5.423711725167903,82.01636290603649,N,1.0,0.0

qwen3-4b-instruct/dp8/epochs/epoch_001/adapter/README.md

@@ -0,0 +1,207 @@
+---
+base_model: Qwen/Qwen3-4B-Instruct-2507
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-4B-Instruct-2507
+- 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 0.18.1

qwen3-4b-instruct/dp8/epochs/epoch_001/adapter/adapter_config.json

@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-4B-Instruct-2507",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": true,
+  "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": 32,
+  "lora_bias": false,
+  "lora_dropout": 0.05,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": [
+    "lm_head",
+    "embed_tokens"
+  ],
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 16,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "k_proj",
+    "v_proj",
+    "q_proj",
+    "o_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}

qwen3-4b-instruct/dp8/epochs/epoch_001/adapter/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:b93a52a9ec5792a1c961bcf1a280df58fe64e837ff061ecca5e022aacaaaa07a
+size 4721857072

qwen3-4b-instruct/dp8/epochs/epoch_001/audit_results.json

@@ -0,0 +1,137 @@
+{
+  "delta": 1e-05,
+  "num_canaries": 500,
+  "num_members": 250,
+  "paper_guess_fraction": 0.2,
+  "paper_guess_steps": 20,
+  "loss": {
+    "auc": 0.51404,
+    "empirical_epsilon": {
+      "0.05": 0.0,
+      "0.01": 0.0
+    },
+    "empirical_epsilon_details": {
+      "0.05": {
+        "epsilon": 0.0,
+        "num_guesses": 0,
+        "correct_guesses": 0,
+        "candidate_num_guesses": [
+          5,
+          10,
+          15,
+          20,
+          25,
+          30,
+          35,
+          40,
+          45,
+          50,
+          55,
+          60,
+          65,
+          70,
+          75,
+          80,
+          85,
+          90,
+          95,
+          100
+        ],
+        "direction": "lower"
+      },
+      "0.01": {
+        "epsilon": 0.0,
+        "num_guesses": 0,
+        "correct_guesses": 0,
+        "candidate_num_guesses": [
+          5,
+          10,
+          15,
+          20,
+          25,
+          30,
+          35,
+          40,
+          45,
+          50,
+          55,
+          60,
+          65,
+          70,
+          75,
+          80,
+          85,
+          90,
+          95,
+          100
+        ],
+        "direction": "lower"
+      }
+    }
+  },
+  "embedding": {
+    "auc": 0.512816,
+    "empirical_epsilon": {
+      "0.05": 0.0,
+      "0.01": 0.0
+    },
+    "empirical_epsilon_details": {
+      "0.05": {
+        "epsilon": 0.0,
+        "num_guesses": 0,
+        "correct_guesses": 0,
+        "candidate_num_guesses": [
+          5,
+          10,
+          15,
+          20,
+          25,
+          30,
+          35,
+          40,
+          45,
+          50,
+          55,
+          60,
+          65,
+          70,
+          75,
+          80,
+          85,
+          90,
+          95,
+          100
+        ],
+        "direction": "lower"
+      },
+      "0.01": {
+        "epsilon": 0.0,
+        "num_guesses": 0,
+        "correct_guesses": 0,
+        "candidate_num_guesses": [
+          5,
+          10,
+          15,
+          20,
+          25,
+          30,
+          35,
+          40,
+          45,
+          50,
+          55,
+          60,
+          65,
+          70,
+          75,
+          80,
+          85,
+          90,
+          95,
+          100
+        ],
+        "direction": "lower"
+      }
+    }
+  }
+}

qwen3-4b-instruct/dp8/epochs/epoch_001/audit_scores.npz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:da624cffb9a22496e27c1ea983a7a58ba6d30e9c7f48429125a4c2c7801f61f2
+size 12784

qwen3-4b-instruct/dp8/epochs/epoch_002/adapter/README.md

@@ -0,0 +1,207 @@
+---
+base_model: Qwen/Qwen3-4B-Instruct-2507
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:Qwen/Qwen3-4B-Instruct-2507
+- 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 0.18.1

qwen3-4b-instruct/dp8/epochs/epoch_002/adapter/adapter_config.json

@@ -0,0 +1,46 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "Qwen/Qwen3-4B-Instruct-2507",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": true,
+  "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": 32,
+  "lora_bias": false,
+  "lora_dropout": 0.05,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": [
+    "lm_head",
+    "embed_tokens"
+  ],
+  "peft_type": "LORA",
+  "peft_version": "0.18.1",
+  "qalora_group_size": 16,
+  "r": 16,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": [
+    "k_proj",
+    "v_proj",
+    "q_proj",
+    "o_proj"
+  ],
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}

qwen3-4b-instruct/dp8/epochs/epoch_002/adapter/adapter_model.safetensors

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:19061f24801c4d66921418e3cb1135de75978cb4d228db814b0755fbaada6bc4
+size 4721857072

qwen3-4b-instruct/dp8/epochs/epoch_002/audit_results.json

@@ -0,0 +1,137 @@
+{
+  "delta": 1e-05,
+  "num_canaries": 500,
+  "num_members": 250,
+  "paper_guess_fraction": 0.2,
+  "paper_guess_steps": 20,
+  "loss": {
+    "auc": 0.515088,
+    "empirical_epsilon": {
+      "0.05": 0.0,
+      "0.01": 0.0
+    },
+    "empirical_epsilon_details": {
+      "0.05": {
+        "epsilon": 0.0,
+        "num_guesses": 0,
+        "correct_guesses": 0,
+        "candidate_num_guesses": [
+          5,
+          10,
+          15,
+          20,
+          25,
+          30,
+          35,
+          40,
+          45,
+          50,
+          55,
+          60,
+          65,
+          70,
+          75,
+          80,
+          85,
+          90,
+          95,
+          100
+        ],
+        "direction": "lower"
+      },
+      "0.01": {
+        "epsilon": 0.0,
+        "num_guesses": 0,
+        "correct_guesses": 0,
+        "candidate_num_guesses": [
+          5,
+          10,
+          15,
+          20,
+          25,
+          30,
+          35,
+          40,
+          45,
+          50,
+          55,
+          60,
+          65,
+          70,
+          75,
+          80,
+          85,
+          90,
+          95,
+          100
+        ],
+        "direction": "lower"
+      }
+    }
+  },
+  "embedding": {
+    "auc": 0.516208,
+    "empirical_epsilon": {
+      "0.05": 0.0,
+      "0.01": 0.0
+    },
+    "empirical_epsilon_details": {
+      "0.05": {
+        "epsilon": 0.0,
+        "num_guesses": 0,
+        "correct_guesses": 0,
+        "candidate_num_guesses": [
+          5,
+          10,
+          15,
+          20,
+          25,
+          30,
+          35,
+          40,
+          45,
+          50,
+          55,
+          60,
+          65,
+          70,
+          75,
+          80,
+          85,
+          90,
+          95,
+          100
+        ],
+        "direction": "lower"
+      },
+      "0.01": {
+        "epsilon": 0.0,
+        "num_guesses": 0,
+        "correct_guesses": 0,
+        "candidate_num_guesses": [
+          5,
+          10,
+          15,
+          20,
+          25,
+          30,
+          35,
+          40,
+          45,
+          50,
+          55,
+          60,
+          65,
+          70,
+          75,
+          80,
+          85,
+          90,
+          95,
+          100
+        ],
+        "direction": "lower"
+      }
+    }
+  }
+}

qwen3-4b-instruct/dp8/epochs/epoch_002/audit_scores.npz

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:126c74d41f98d08626da6ca1f818fce7bb97a1b2a1dad6827e8baa20288a484f
+size 12784

qwen3-4b-instruct/dp8/metrics.jsonl

@@ -0,0 +1,27 @@
+{"timestamp": 1773764685.4650462, "event": "train_step", "step": 10, "epoch": 1, "metrics": {"train/step_loss": 1.6114869984713467, "train/step_real_loss": 1.1970022171735764, "train/lr": 0.0002, "train/step_canary_loss": 14.875, "perf/step_duration_sec": 13.257156527135521, "perf/samples_per_sec": 4.978443142381804, "perf/tokens_per_sec": 3881.0735842701297, "perf/logical_batch_size": 66.0, "perf/logical_token_count": 51452.0, "perf/physical_batches": 9.0, "privacy/epsilon": 2.5080663593969574, "system/cuda_memory_allocated_gb": 12.68057107925415, "system/cuda_max_memory_allocated_gb": 86.2213044166565}}
+{"timestamp": 1773764819.8299851, "event": "train_step", "step": 20, "epoch": 1, "metrics": {"train/step_loss": 1.7427907831528608, "train/step_real_loss": 1.0802308320999146, "train/lr": 0.0001984111204336116, "train/step_canary_loss": 12.34375, "perf/step_duration_sec": 13.233567330986261, "perf/samples_per_sec": 5.138448182507728, "perf/tokens_per_sec": 4095.645463116454, "perf/logical_batch_size": 68.0, "perf/logical_token_count": 54200.0, "perf/physical_batches": 9.0, "privacy/epsilon": 3.1012868027106864, "system/cuda_memory_allocated_gb": 13.261098384857178, "system/cuda_max_memory_allocated_gb": 86.2213044166565}}
+{"timestamp": 1773764953.865748, "event": "train_step", "step": 30, "epoch": 1, "metrics": {"train/step_loss": 1.00910531390797, "train/step_real_loss": 1.0406398549675941, "train/lr": 0.0001936949724999762, "train/step_canary_loss": 0.0, "perf/step_duration_sec": 14.175612474791706, "perf/samples_per_sec": 4.51479610590444, "perf/tokens_per_sec": 3616.84548665354, "perf/logical_batch_size": 64.0, "perf/logical_token_count": 51271.0, "perf/physical_batches": 10.0, "privacy/epsilon": 3.577863845948693, "system/cuda_memory_allocated_gb": 14.42089033126831, "system/cuda_max_memory_allocated_gb": 86.2213044166565}}
+{"timestamp": 1773765087.7159872, "event": "train_step", "step": 40, "epoch": 1, "metrics": {"train/step_loss": 1.4425305669957942, "train/step_real_loss": 1.0335080847144127, "train/lr": 0.00018600142402077006, "train/step_canary_loss": 14.53125, "perf/step_duration_sec": 13.013761347159743, "perf/samples_per_sec": 5.071554505984891, "perf/tokens_per_sec": 4077.6835062817304, "perf/logical_batch_size": 66.0, "perf/logical_token_count": 53066.0, "perf/physical_batches": 9.0, "privacy/epsilon": 3.9945146358205394, "system/cuda_memory_allocated_gb": 12.680593967437744, "system/cuda_max_memory_allocated_gb": 86.22135162353516}}
+{"timestamp": 1773765221.5064714, "event": "train_step", "step": 50, "epoch": 1, "metrics": {"train/step_loss": 1.2258210108830379, "train/step_real_loss": 1.0516150891780853, "train/lr": 0.00017557495743542585, "train/step_canary_loss": 12.375, "perf/step_duration_sec": 13.473957392852753, "perf/samples_per_sec": 4.824120939738104, "perf/tokens_per_sec": 3929.2093967940723, "perf/logical_batch_size": 65.0, "perf/logical_token_count": 52942.0, "perf/physical_batches": 9.0, "privacy/epsilon": 4.370419605916781, "system/cuda_memory_allocated_gb": 12.39033031463623, "system/cuda_max_memory_allocated_gb": 86.22135162353516}}
+{"timestamp": 1773765234.2224665, "event": "eval_step", "step": 50, "epoch": 1, "metrics": {"eval/loss": 0.950404628729209, "eval/duration_sec": 12.71370144886896}}
+{"timestamp": 1773765368.860026, "event": "train_step", "step": 60, "epoch": 1, "metrics": {"train/step_loss": 2.122071954182216, "train/step_real_loss": 1.006563015282154, "train/lr": 0.0001627469007380852, "train/step_canary_loss": 14.020833969116211, "perf/step_duration_sec": 13.671399443875998, "perf/samples_per_sec": 5.120178097887117, "perf/tokens_per_sec": 3658.732977947337, "perf/logical_batch_size": 70.0, "perf/logical_token_count": 50020.0, "perf/physical_batches": 9.0, "privacy/epsilon": 4.716422199842559, "system/cuda_memory_allocated_gb": 13.841647148132324, "system/cuda_max_memory_allocated_gb": 86.22135162353516}}
+{"timestamp": 1773765503.2535436, "event": "train_step", "step": 70, "epoch": 1, "metrics": {"train/step_loss": 1.7725614239187801, "train/step_real_loss": 1.058151200413704, "train/lr": 0.0001479248986720057, "train/step_canary_loss": 13.203125, "perf/step_duration_sec": 13.894904132932425, "perf/samples_per_sec": 4.893880472254044, "perf/tokens_per_sec": 3568.2146149169926, "perf/logical_batch_size": 68.0, "perf/logical_token_count": 49580.0, "perf/physical_batches": 9.0, "privacy/epsilon": 5.042051045310402, "system/cuda_memory_allocated_gb": 13.261121273040771, "system/cuda_max_memory_allocated_gb": 86.22135162353516}}
+{"timestamp": 1773765637.8816426, "event": "train_step", "step": 80, "epoch": 1, "metrics": {"train/step_loss": 1.8267304336323458, "train/step_real_loss": 1.0551588982343674, "train/lr": 0.0001315799587615025, "train/step_canary_loss": 14.171875, "perf/step_duration_sec": 13.812608732841909, "perf/samples_per_sec": 4.923038168620387, "perf/tokens_per_sec": 3637.6908208897057, "perf/logical_batch_size": 68.0, "perf/logical_token_count": 50246.0, "perf/physical_batches": 9.0, "privacy/epsilon": 5.348292001109609, "system/cuda_memory_allocated_gb": 13.261121273040771, "system/cuda_max_memory_allocated_gb": 86.22135162353516}}
+{"timestamp": 1773765772.2237043, "event": "train_step", "step": 90, "epoch": 1, "metrics": {"train/step_loss": 1.5600519109128126, "train/step_real_loss": 1.0277105793356895, "train/lr": 0.00011423148382732853, "train/step_canary_loss": 12.916666984558105, "perf/step_duration_sec": 13.590938247274607, "perf/samples_per_sec": 4.929755310560367, "perf/tokens_per_sec": 3897.8913034663583, "perf/logical_batch_size": 67.0, "perf/logical_token_count": 52976.0, "perf/physical_batches": 9.0, "privacy/epsilon": 5.640478844621371, "system/cuda_memory_allocated_gb": 12.970857620239258, "system/cuda_max_memory_allocated_gb": 86.22135162353516}}
+{"timestamp": 1773765818.3224468, "event": "train_epoch", "step": 93, "epoch": 1, "metrics": {"train/epoch_loss": 1.667893763698923, "train/epoch_real_loss": 1.0501683052070176, "train/epoch_canary_loss": 13.414085814260666, "perf/epoch_duration_sec": 1255.631331067998, "perf/epoch_samples_per_sec": 39.58486760419035, "perf/epoch_tokens_per_sec": 30085.320480071918, "perf/epoch_samples": 49704.0, "perf/epoch_tokens": 37776071.0, "system/cuda_epoch_peak_memory_gb": 86.22135162353516, "eval/loss": 0.9280200686592323, "eval/duration_sec": 12.726754495874047, "privacy/epsilon": 5.725566085147695}}
+{"timestamp": 1773765831.5765529, "event": "audit_epoch", "step": 93, "epoch": 1, "metrics": {"audit/delta": 1e-05, "audit/num_canaries": 500.0, "audit/num_members": 250.0, "audit/paper_guess_fraction": 0.2, "audit/paper_guess_steps": 20.0, "audit/loss/auc": 0.51404, "audit/loss/empirical_epsilon/0.05": 0.0, "audit/loss/empirical_epsilon/0.01": 0.0, "audit/loss/empirical_epsilon_details/0.05/epsilon": 0.0, "audit/loss/empirical_epsilon_details/0.05/num_guesses": 0.0, "audit/loss/empirical_epsilon_details/0.05/correct_guesses": 0.0, "audit/loss/empirical_epsilon_details/0.01/epsilon": 0.0, "audit/loss/empirical_epsilon_details/0.01/num_guesses": 0.0, "audit/loss/empirical_epsilon_details/0.01/correct_guesses": 0.0, "audit/embedding/auc": 0.512816, "audit/embedding/empirical_epsilon/0.05": 0.0, "audit/embedding/empirical_epsilon/0.01": 0.0, "audit/embedding/empirical_epsilon_details/0.05/epsilon": 0.0, "audit/embedding/empirical_epsilon_details/0.05/num_guesses": 0.0, "audit/embedding/empirical_epsilon_details/0.05/correct_guesses": 0.0, "audit/embedding/empirical_epsilon_details/0.01/epsilon": 0.0, "audit/embedding/empirical_epsilon_details/0.01/num_guesses": 0.0, "audit/embedding/empirical_epsilon_details/0.01/correct_guesses": 0.0, "perf/audit_duration_sec": 6.853809644002467}}
+{"timestamp": 1773765927.2307582, "event": "train_step", "step": 100, "epoch": 2, "metrics": {"train/step_loss": 2.1885286586385377, "train/step_real_loss": 0.9171566888689995, "train/lr": 9.643076661610196e-05, "train/step_canary_loss": 13.812500953674316, "perf/step_duration_sec": 13.692489622160792, "perf/samples_per_sec": 5.18532436096129, "perf/tokens_per_sec": 3906.0098985534178, "perf/logical_batch_size": 71.0, "perf/logical_token_count": 53483.0, "perf/physical_batches": 9.0, "privacy/epsilon": 5.920610858546996, "system/cuda_memory_allocated_gb": 14.131910800933838, "system/cuda_max_memory_allocated_gb": 86.22135162353516}}
+{"timestamp": 1773765939.9789443, "event": "eval_step", "step": 100, "epoch": 2, "metrics": {"eval/loss": 0.9270543383482176, "eval/duration_sec": 12.745669181924313}}
+{"timestamp": 1773766074.3579476, "event": "train_step", "step": 110, "epoch": 2, "metrics": {"train/step_loss": 1.4203428788618608, "train/step_real_loss": 1.016486406326294, "train/lr": 7.874347104470234e-05, "train/step_canary_loss": 14.34375, "perf/step_duration_sec": 13.423757336102426, "perf/samples_per_sec": 4.916656219827274, "perf/tokens_per_sec": 3954.481496565319, "perf/logical_batch_size": 66.0, "perf/logical_token_count": 53084.0, "perf/physical_batches": 9.0, "privacy/epsilon": 6.190670402749493, "system/cuda_memory_allocated_gb": 12.680593967437744, "system/cuda_max_memory_allocated_gb": 86.22137594223022}}
+{"timestamp": 1773766208.938867, "event": "train_step", "step": 120, "epoch": 2, "metrics": {"train/step_loss": 1.590441582807854, "train/step_real_loss": 1.0116732195019722, "train/lr": 6.173165676349103e-05, "train/step_canary_loss": 13.9375, "perf/step_duration_sec": 13.283335603773594, "perf/samples_per_sec": 5.0439138179243415, "perf/tokens_per_sec": 4092.8725752103383, "perf/logical_batch_size": 67.0, "perf/logical_token_count": 54367.0, "perf/physical_batches": 9.0, "privacy/epsilon": 6.450768498470402, "system/cuda_memory_allocated_gb": 12.970857620239258, "system/cuda_max_memory_allocated_gb": 86.22137594223022}}
+{"timestamp": 1773766343.2955682, "event": "train_step", "step": 130, "epoch": 2, "metrics": {"train/step_loss": 1.8971429631329966, "train/step_real_loss": 0.9516072571277618, "train/lr": 4.593591825444028e-05, "train/step_canary_loss": 14.0, "perf/step_duration_sec": 13.135221335105598, "perf/samples_per_sec": 5.253051946341283, "perf/tokens_per_sec": 4244.389841456128, "perf/logical_batch_size": 69.0, "perf/logical_token_count": 55751.0, "perf/physical_batches": 9.0, "privacy/epsilon": 6.701680965161658, "system/cuda_memory_allocated_gb": 13.55138349533081, "system/cuda_max_memory_allocated_gb": 86.22137594223022}}
+{"timestamp": 1773766477.2471156, "event": "train_step", "step": 140, "epoch": 2, "metrics": {"train/step_loss": 1.62191776731121, "train/step_real_loss": 1.027046725153923, "train/lr": 3.185820604061088e-05, "train/step_canary_loss": 14.3125, "perf/step_duration_sec": 13.460042576305568, "perf/samples_per_sec": 4.9776959931719444, "perf/tokens_per_sec": 4005.113631282184, "perf/logical_batch_size": 67.0, "perf/logical_token_count": 53909.0, "perf/physical_batches": 9.0, "privacy/epsilon": 6.94699659418226, "system/cuda_memory_allocated_gb": 12.970857620239258, "system/cuda_max_memory_allocated_gb": 86.22137594223022}}
+{"timestamp": 1773766610.9435787, "event": "train_step", "step": 150, "epoch": 2, "metrics": {"train/step_loss": 1.9888183966926907, "train/step_real_loss": 1.0191947892308235, "train/lr": 1.994587590756397e-05, "train/step_canary_loss": 14.40000057220459, "perf/step_duration_sec": 13.144390502013266, "perf/samples_per_sec": 5.249387561137322, "perf/tokens_per_sec": 3851.1485178598896, "perf/logical_batch_size": 69.0, "perf/logical_token_count": 50621.0, "perf/physical_batches": 9.0, "privacy/epsilon": 7.185356422932143, "system/cuda_memory_allocated_gb": 13.55138349533081, "system/cuda_max_memory_allocated_gb": 86.22137594223022}}
+{"timestamp": 1773766623.702012, "event": "eval_step", "step": 150, "epoch": 2, "metrics": {"eval/loss": 0.9247592026606585, "eval/duration_sec": 12.756307526025921}}
+{"timestamp": 1773766758.1413558, "event": "train_step", "step": 160, "epoch": 2, "metrics": {"train/step_loss": 1.475486863743175, "train/step_real_loss": 1.1006973907351494, "train/lr": 1.057747301402887e-05, "train/step_canary_loss": 13.46875, "perf/step_duration_sec": 13.887811011634767, "perf/samples_per_sec": 4.752368817858142, "perf/tokens_per_sec": 3180.198806204889, "perf/logical_batch_size": 66.0, "perf/logical_token_count": 44166.0, "perf/physical_batches": 9.0, "privacy/epsilon": 7.416930400989618, "system/cuda_memory_allocated_gb": 12.680593967437744, "system/cuda_max_memory_allocated_gb": 86.22137594223022}}
+{"timestamp": 1773766893.2984576, "event": "train_step", "step": 170, "epoch": 2, "metrics": {"train/step_loss": 1.6393254835214188, "train/step_real_loss": 1.0540594905614853, "train/lr": 4.050702638550275e-06, "train/step_canary_loss": 14.125, "perf/step_duration_sec": 13.72155143506825, "perf/samples_per_sec": 4.882829781825378, "perf/tokens_per_sec": 3613.58555077656, "perf/logical_batch_size": 67.0, "perf/logical_token_count": 49584.0, "perf/physical_batches": 9.0, "privacy/epsilon": 7.645014003428424, "system/cuda_memory_allocated_gb": 12.970857620239258, "system/cuda_max_memory_allocated_gb": 86.22137594223022}}
+{"timestamp": 1773767029.0999126, "event": "train_step", "step": 180, "epoch": 2, "metrics": {"train/step_loss": 1.7291738425984102, "train/step_real_loss": 1.0520909577608109, "train/lr": 5.729698228102653e-07, "train/step_canary_loss": 12.5625, "perf/step_duration_sec": 13.660048387013376, "perf/samples_per_sec": 4.978020434001367, "perf/tokens_per_sec": 3874.5104336758286, "perf/logical_batch_size": 68.0, "perf/logical_token_count": 52926.0, "perf/physical_batches": 9.0, "privacy/epsilon": 7.8658492724205455, "system/cuda_memory_allocated_gb": 13.261121273040771, "system/cuda_max_memory_allocated_gb": 86.22137594223022}}
+{"timestamp": 1773767117.390551, "event": "train_epoch", "step": 186, "epoch": 2, "metrics": {"train/epoch_loss": 1.5804422382361127, "train/epoch_real_loss": 1.005713254121899, "train/epoch_canary_loss": 13.023172873210777, "perf/epoch_duration_sec": 1272.894041202031, "perf/epoch_samples_per_sec": 38.91918604098259, "perf/epoch_tokens_per_sec": 29671.033705472055, "perf/epoch_samples": 49540.0, "perf/epoch_tokens": 37768082.0, "system/cuda_epoch_peak_memory_gb": 86.22137594223022, "eval/loss": 0.924621483645378, "eval/duration_sec": 12.757435038685799, "privacy/epsilon": 7.996749609735891}}
+{"timestamp": 1773767130.595887, "event": "audit_epoch", "step": 186, "epoch": 2, "metrics": {"audit/delta": 1e-05, "audit/num_canaries": 500.0, "audit/num_members": 250.0, "audit/paper_guess_fraction": 0.2, "audit/paper_guess_steps": 20.0, "audit/loss/auc": 0.515088, "audit/loss/empirical_epsilon/0.05": 0.0, "audit/loss/empirical_epsilon/0.01": 0.0, "audit/loss/empirical_epsilon_details/0.05/epsilon": 0.0, "audit/loss/empirical_epsilon_details/0.05/num_guesses": 0.0, "audit/loss/empirical_epsilon_details/0.05/correct_guesses": 0.0, "audit/loss/empirical_epsilon_details/0.01/epsilon": 0.0, "audit/loss/empirical_epsilon_details/0.01/num_guesses": 0.0, "audit/loss/empirical_epsilon_details/0.01/correct_guesses": 0.0, "audit/embedding/auc": 0.516208, "audit/embedding/empirical_epsilon/0.05": 0.0, "audit/embedding/empirical_epsilon/0.01": 0.0, "audit/embedding/empirical_epsilon_details/0.05/epsilon": 0.0, "audit/embedding/empirical_epsilon_details/0.05/num_guesses": 0.0, "audit/embedding/empirical_epsilon_details/0.05/correct_guesses": 0.0, "audit/embedding/empirical_epsilon_details/0.01/epsilon": 0.0, "audit/embedding/empirical_epsilon_details/0.01/num_guesses": 0.0, "audit/embedding/empirical_epsilon_details/0.01/correct_guesses": 0.0, "perf/audit_duration_sec": 6.922967464663088}}
+{"timestamp": 1773767143.9571598, "event": "audit_final", "step": 186, "epoch": 2, "metrics": {"audit/delta": 1e-05, "audit/num_canaries": 500.0, "audit/num_members": 250.0, "audit/paper_guess_fraction": 0.2, "audit/paper_guess_steps": 20.0, "audit/loss/auc": 0.515088, "audit/loss/empirical_epsilon/0.05": 0.0, "audit/loss/empirical_epsilon/0.01": 0.0, "audit/loss/empirical_epsilon_details/0.05/epsilon": 0.0, "audit/loss/empirical_epsilon_details/0.05/num_guesses": 0.0, "audit/loss/empirical_epsilon_details/0.05/correct_guesses": 0.0, "audit/loss/empirical_epsilon_details/0.01/epsilon": 0.0, "audit/loss/empirical_epsilon_details/0.01/num_guesses": 0.0, "audit/loss/empirical_epsilon_details/0.01/correct_guesses": 0.0, "audit/embedding/auc": 0.516208, "audit/embedding/empirical_epsilon/0.05": 0.0, "audit/embedding/empirical_epsilon/0.01": 0.0, "audit/embedding/empirical_epsilon_details/0.05/epsilon": 0.0, "audit/embedding/empirical_epsilon_details/0.05/num_guesses": 0.0, "audit/embedding/empirical_epsilon_details/0.05/correct_guesses": 0.0, "audit/embedding/empirical_epsilon_details/0.01/epsilon": 0.0, "audit/embedding/empirical_epsilon_details/0.01/num_guesses": 0.0, "audit/embedding/empirical_epsilon_details/0.01/correct_guesses": 0.0}}
+{"timestamp": 1773767144.5062578, "event": "energy_final", "step": 186, "epoch": null, "metrics": {"energy/codecarbon/duration": 2693.6014373912476, "energy/codecarbon/emissions": 0.09826528580395394, "energy/codecarbon/emissions_rate": 3.648100436830916e-05, "energy/codecarbon/cpu_power": 72.03163066573623, "energy/codecarbon/gpu_power": 3108.4298352234105, "energy/codecarbon/ram_power": 54.0, "energy/codecarbon/cpu_energy": 0.051919376660011826, "energy/codecarbon/gpu_energy": 2.3238466076869315, "energy/codecarbon/ram_energy": 0.03891114561875117, "energy/codecarbon/energy_consumed": 2.414677129965695, "energy/codecarbon/water_consumed": 0.0, "energy/codecarbon/cpu_count": 256.0, "energy/codecarbon/gpu_count": 8.0, "energy/codecarbon/longitude": 16.1885, "energy/codecarbon/latitude": 58.594, "energy/codecarbon/ram_total_size": 1511.49019241333, "energy/codecarbon/cpu_utilization_percent": 3.719566840926081, "energy/codecarbon/gpu_utilization_percent": 83.2715646004481, "energy/codecarbon/ram_utilization_percent": 5.423711725167903, "energy/codecarbon/ram_used_gb": 82.01636290603649, "energy/codecarbon/pue": 1.0, "energy/codecarbon/wue": 0.0}}

qwen3-4b-instruct/dp8/pretrain_lm_head.pt

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:bc44b7d60b8e2cf912e4233ff02bc57bb7e91f7a3ba6aa8ea10b7767ca29954a
+size 779106920

qwen3-4b-instruct/dp8/resolved_config.yaml

@@ -0,0 +1,101 @@
+model:
+  name: Qwen/Qwen3-4B-Instruct-2507
+  tokenizer_name: Qwen/Qwen3-4B-Instruct-2507
+  max_length: 1024
+  dtype: bfloat16
+  trust_remote_code: true
+  use_fast_tokenizer: true
+  cache_dir: null
+  local_files_only: false
+  low_cpu_mem_usage: true
+  tie_word_embeddings: true
+  gradient_checkpointing: false
+  use_chat_template: false
+dataset:
+  name: melihcatal/codedp-cpt
+  split: train
+  mode: cpt
+  text_column: text
+  validation_ratio: 0.05
+  max_samples: -1
+lora:
+  enabled: true
+  r: 16
+  alpha: 32
+  dropout: 0.05
+  target_modules:
+  - q_proj
+  - k_proj
+  - v_proj
+  - o_proj
+  modules_to_save:
+  - lm_head
+  bias: none
+training:
+  seed: 42
+  epochs: 2
+  warmup_steps: null
+  warmup_ratio: 0.05
+  mixed_precision: false
+  mixed_precision_dtype: bfloat16
+  batch_size: 8
+  eval_batch_size: 8
+  eval_every_steps: 50
+  eval_every_epochs: 1
+  learning_rate: 0.0002
+  optimizer: adamw
+  lr_scheduler: cosine
+  adam_beta1: 0.9
+  adam_beta2: 0.999
+  adam_epsilon: 1.0e-08
+  sgd_momentum: 0.9
+  weight_decay: 0.01
+  max_grad_norm: 1.0
+  log_every: 10
+  gradient_accumulation_steps: 8
+  num_workers: 4
+  output_dir: runs/cpt/qwen3-4b-instruct/dp8
+distributed:
+  strategy: dpddp
+  backend: nccl
+  devices: null
+dp:
+  module_validator: auto
+  target_delta: 1.0e-05
+  noise_multiplier: null
+  max_grad_norm: 1.0
+  grad_sample_mode: hooks
+  secure_mode: false
+  enabled: true
+  target_epsilon: 8.0
+  clipping: flat
+audit:
+  enabled: true
+  run_every_epoch: true
+  epoch_device: cuda
+  q_canary: auto
+  num_canaries: 500
+  prefix_length: 49
+  num_digits: 12
+  batch_size: 32
+  delta: 1.0e-05
+  p_values:
+  - 0.05
+  - 0.01
+  paper_guess_fraction: 0.2
+  paper_guess_steps: 20
+  enable_holdout_empirical_epsilon: false
+  holdout_seed: 42
+  tie_seed: 42
+tracking:
+  enabled: true
+  tensorboard: true
+  wandb: false
+  wandb_project: codedp-finetune-h200-audit
+  wandb_run_name: qwen3-4b-instruct-cpt-dp8
+  wandb_mode: online
+  codecarbon: true
+  codecarbon_output_file: codecarbon.csv
+  codecarbon_measure_power_secs: 15
+  codecarbon_country_iso_code: null
+  codecarbon_project_name: codedp-qwen3-4b-instruct-cpt-dp8

qwen3-4b-instruct/dp8/scalars.csv

@@ -0,0 +1,358 @@
+timestamp,event,step,epoch,key,value
+1773764685.4650462,train_step,10,1,train/step_loss,1.6114869984713467
+1773764685.4650462,train_step,10,1,train/step_real_loss,1.1970022171735764
+1773764685.4650462,train_step,10,1,train/lr,0.0002
+1773764685.4650462,train_step,10,1,train/step_canary_loss,14.875
+1773764685.4650462,train_step,10,1,perf/step_duration_sec,13.257156527135521
+1773764685.4650462,train_step,10,1,perf/samples_per_sec,4.978443142381804
+1773764685.4650462,train_step,10,1,perf/tokens_per_sec,3881.0735842701297
+1773764685.4650462,train_step,10,1,perf/logical_batch_size,66.0
+1773764685.4650462,train_step,10,1,perf/logical_token_count,51452.0
+1773764685.4650462,train_step,10,1,perf/physical_batches,9.0
+1773764685.4650462,train_step,10,1,privacy/epsilon,2.5080663593969574
+1773764685.4650462,train_step,10,1,system/cuda_memory_allocated_gb,12.68057107925415
+1773764685.4650462,train_step,10,1,system/cuda_max_memory_allocated_gb,86.2213044166565
+1773764819.8299851,train_step,20,1,train/step_loss,1.7427907831528608
+1773764819.8299851,train_step,20,1,train/step_real_loss,1.0802308320999146
+1773764819.8299851,train_step,20,1,train/lr,0.0001984111204336116
+1773764819.8299851,train_step,20,1,train/step_canary_loss,12.34375
+1773764819.8299851,train_step,20,1,perf/step_duration_sec,13.233567330986261
+1773764819.8299851,train_step,20,1,perf/samples_per_sec,5.138448182507728
+1773764819.8299851,train_step,20,1,perf/tokens_per_sec,4095.645463116454
+1773764819.8299851,train_step,20,1,perf/logical_batch_size,68.0
+1773764819.8299851,train_step,20,1,perf/logical_token_count,54200.0
+1773764819.8299851,train_step,20,1,perf/physical_batches,9.0
+1773764819.8299851,train_step,20,1,privacy/epsilon,3.1012868027106864
+1773764819.8299851,train_step,20,1,system/cuda_memory_allocated_gb,13.261098384857178
+1773764819.8299851,train_step,20,1,system/cuda_max_memory_allocated_gb,86.2213044166565
+1773764953.865748,train_step,30,1,train/step_loss,1.00910531390797
+1773764953.865748,train_step,30,1,train/step_real_loss,1.0406398549675941
+1773764953.865748,train_step,30,1,train/lr,0.0001936949724999762
+1773764953.865748,train_step,30,1,train/step_canary_loss,0.0
+1773764953.865748,train_step,30,1,perf/step_duration_sec,14.175612474791706
+1773764953.865748,train_step,30,1,perf/samples_per_sec,4.51479610590444
+1773764953.865748,train_step,30,1,perf/tokens_per_sec,3616.84548665354
+1773764953.865748,train_step,30,1,perf/logical_batch_size,64.0
+1773764953.865748,train_step,30,1,perf/logical_token_count,51271.0
+1773764953.865748,train_step,30,1,perf/physical_batches,10.0
+1773764953.865748,train_step,30,1,privacy/epsilon,3.577863845948693
+1773764953.865748,train_step,30,1,system/cuda_memory_allocated_gb,14.42089033126831
+1773764953.865748,train_step,30,1,system/cuda_max_memory_allocated_gb,86.2213044166565
+1773765087.7159872,train_step,40,1,train/step_loss,1.4425305669957942
+1773765087.7159872,train_step,40,1,train/step_real_loss,1.0335080847144127
+1773765087.7159872,train_step,40,1,train/lr,0.00018600142402077006
+1773765087.7159872,train_step,40,1,train/step_canary_loss,14.53125
+1773765087.7159872,train_step,40,1,perf/step_duration_sec,13.013761347159743
+1773765087.7159872,train_step,40,1,perf/samples_per_sec,5.071554505984891
+1773765087.7159872,train_step,40,1,perf/tokens_per_sec,4077.6835062817304
+1773765087.7159872,train_step,40,1,perf/logical_batch_size,66.0
+1773765087.7159872,train_step,40,1,perf/logical_token_count,53066.0
+1773765087.7159872,train_step,40,1,perf/physical_batches,9.0
+1773765087.7159872,train_step,40,1,privacy/epsilon,3.9945146358205394
+1773765087.7159872,train_step,40,1,system/cuda_memory_allocated_gb,12.680593967437744
+1773765087.7159872,train_step,40,1,system/cuda_max_memory_allocated_gb,86.22135162353516
+1773765221.5064714,train_step,50,1,train/step_loss,1.2258210108830379
+1773765221.5064714,train_step,50,1,train/step_real_loss,1.0516150891780853
+1773765221.5064714,train_step,50,1,train/lr,0.00017557495743542585
+1773765221.5064714,train_step,50,1,train/step_canary_loss,12.375
+1773765221.5064714,train_step,50,1,perf/step_duration_sec,13.473957392852753
+1773765221.5064714,train_step,50,1,perf/samples_per_sec,4.824120939738104
+1773765221.5064714,train_step,50,1,perf/tokens_per_sec,3929.2093967940723
+1773765221.5064714,train_step,50,1,perf/logical_batch_size,65.0
+1773765221.5064714,train_step,50,1,perf/logical_token_count,52942.0
+1773765221.5064714,train_step,50,1,perf/physical_batches,9.0
+1773765221.5064714,train_step,50,1,privacy/epsilon,4.370419605916781
+1773765221.5064714,train_step,50,1,system/cuda_memory_allocated_gb,12.39033031463623
+1773765221.5064714,train_step,50,1,system/cuda_max_memory_allocated_gb,86.22135162353516
+1773765234.2224665,eval_step,50,1,eval/loss,0.950404628729209
+1773765234.2224665,eval_step,50,1,eval/duration_sec,12.71370144886896
+1773765368.860026,train_step,60,1,train/step_loss,2.122071954182216
+1773765368.860026,train_step,60,1,train/step_real_loss,1.006563015282154
+1773765368.860026,train_step,60,1,train/lr,0.0001627469007380852
+1773765368.860026,train_step,60,1,train/step_canary_loss,14.020833969116211
+1773765368.860026,train_step,60,1,perf/step_duration_sec,13.671399443875998
+1773765368.860026,train_step,60,1,perf/samples_per_sec,5.120178097887117
+1773765368.860026,train_step,60,1,perf/tokens_per_sec,3658.732977947337
+1773765368.860026,train_step,60,1,perf/logical_batch_size,70.0
+1773765368.860026,train_step,60,1,perf/logical_token_count,50020.0
+1773765368.860026,train_step,60,1,perf/physical_batches,9.0
+1773765368.860026,train_step,60,1,privacy/epsilon,4.716422199842559
+1773765368.860026,train_step,60,1,system/cuda_memory_allocated_gb,13.841647148132324
+1773765368.860026,train_step,60,1,system/cuda_max_memory_allocated_gb,86.22135162353516
+1773765503.2535436,train_step,70,1,train/step_loss,1.7725614239187801
+1773765503.2535436,train_step,70,1,train/step_real_loss,1.058151200413704
+1773765503.2535436,train_step,70,1,train/lr,0.0001479248986720057
+1773765503.2535436,train_step,70,1,train/step_canary_loss,13.203125
+1773765503.2535436,train_step,70,1,perf/step_duration_sec,13.894904132932425
+1773765503.2535436,train_step,70,1,perf/samples_per_sec,4.893880472254044
+1773765503.2535436,train_step,70,1,perf/tokens_per_sec,3568.2146149169926
+1773765503.2535436,train_step,70,1,perf/logical_batch_size,68.0
+1773765503.2535436,train_step,70,1,perf/logical_token_count,49580.0
+1773765503.2535436,train_step,70,1,perf/physical_batches,9.0
+1773765503.2535436,train_step,70,1,privacy/epsilon,5.042051045310402
+1773765503.2535436,train_step,70,1,system/cuda_memory_allocated_gb,13.261121273040771
+1773765503.2535436,train_step,70,1,system/cuda_max_memory_allocated_gb,86.22135162353516
+1773765637.8816426,train_step,80,1,train/step_loss,1.8267304336323458
+1773765637.8816426,train_step,80,1,train/step_real_loss,1.0551588982343674
+1773765637.8816426,train_step,80,1,train/lr,0.0001315799587615025
+1773765637.8816426,train_step,80,1,train/step_canary_loss,14.171875
+1773765637.8816426,train_step,80,1,perf/step_duration_sec,13.812608732841909
+1773765637.8816426,train_step,80,1,perf/samples_per_sec,4.923038168620387
+1773765637.8816426,train_step,80,1,perf/tokens_per_sec,3637.6908208897057
+1773765637.8816426,train_step,80,1,perf/logical_batch_size,68.0
+1773765637.8816426,train_step,80,1,perf/logical_token_count,50246.0
+1773765637.8816426,train_step,80,1,perf/physical_batches,9.0
+1773765637.8816426,train_step,80,1,privacy/epsilon,5.348292001109609
+1773765637.8816426,train_step,80,1,system/cuda_memory_allocated_gb,13.261121273040771
+1773765637.8816426,train_step,80,1,system/cuda_max_memory_allocated_gb,86.22135162353516
+1773765772.2237043,train_step,90,1,train/step_loss,1.5600519109128126
+1773765772.2237043,train_step,90,1,train/step_real_loss,1.0277105793356895
+1773765772.2237043,train_step,90,1,train/lr,0.00011423148382732853
+1773765772.2237043,train_step,90,1,train/step_canary_loss,12.916666984558105
+1773765772.2237043,train_step,90,1,perf/step_duration_sec,13.590938247274607
+1773765772.2237043,train_step,90,1,perf/samples_per_sec,4.929755310560367
+1773765772.2237043,train_step,90,1,perf/tokens_per_sec,3897.8913034663583
+1773765772.2237043,train_step,90,1,perf/logical_batch_size,67.0
+1773765772.2237043,train_step,90,1,perf/logical_token_count,52976.0
+1773765772.2237043,train_step,90,1,perf/physical_batches,9.0
+1773765772.2237043,train_step,90,1,privacy/epsilon,5.640478844621371
+1773765772.2237043,train_step,90,1,system/cuda_memory_allocated_gb,12.970857620239258
+1773765772.2237043,train_step,90,1,system/cuda_max_memory_allocated_gb,86.22135162353516
+1773765818.3224468,train_epoch,93,1,train/epoch_loss,1.667893763698923
+1773765818.3224468,train_epoch,93,1,train/epoch_real_loss,1.0501683052070176
+1773765818.3224468,train_epoch,93,1,train/epoch_canary_loss,13.414085814260666
+1773765818.3224468,train_epoch,93,1,perf/epoch_duration_sec,1255.631331067998
+1773765818.3224468,train_epoch,93,1,perf/epoch_samples_per_sec,39.58486760419035
+1773765818.3224468,train_epoch,93,1,perf/epoch_tokens_per_sec,30085.320480071918
+1773765818.3224468,train_epoch,93,1,perf/epoch_samples,49704.0
+1773765818.3224468,train_epoch,93,1,perf/epoch_tokens,37776071.0
+1773765818.3224468,train_epoch,93,1,system/cuda_epoch_peak_memory_gb,86.22135162353516
+1773765818.3224468,train_epoch,93,1,eval/loss,0.9280200686592323
+1773765818.3224468,train_epoch,93,1,eval/duration_sec,12.726754495874047
+1773765818.3224468,train_epoch,93,1,privacy/epsilon,5.725566085147695
+1773765831.5765529,audit_epoch,93,1,audit/delta,1e-05
+1773765831.5765529,audit_epoch,93,1,audit/num_canaries,500.0
+1773765831.5765529,audit_epoch,93,1,audit/num_members,250.0
+1773765831.5765529,audit_epoch,93,1,audit/paper_guess_fraction,0.2
+1773765831.5765529,audit_epoch,93,1,audit/paper_guess_steps,20.0
+1773765831.5765529,audit_epoch,93,1,audit/loss/auc,0.51404
+1773765831.5765529,audit_epoch,93,1,audit/loss/empirical_epsilon/0.05,0.0
+1773765831.5765529,audit_epoch,93,1,audit/loss/empirical_epsilon/0.01,0.0
+1773765831.5765529,audit_epoch,93,1,audit/loss/empirical_epsilon_details/0.05/epsilon,0.0
+1773765831.5765529,audit_epoch,93,1,audit/loss/empirical_epsilon_details/0.05/num_guesses,0.0
+1773765831.5765529,audit_epoch,93,1,audit/loss/empirical_epsilon_details/0.05/correct_guesses,0.0
+1773765831.5765529,audit_epoch,93,1,audit/loss/empirical_epsilon_details/0.01/epsilon,0.0
+1773765831.5765529,audit_epoch,93,1,audit/loss/empirical_epsilon_details/0.01/num_guesses,0.0
+1773765831.5765529,audit_epoch,93,1,audit/loss/empirical_epsilon_details/0.01/correct_guesses,0.0
+1773765831.5765529,audit_epoch,93,1,audit/embedding/auc,0.512816
+1773765831.5765529,audit_epoch,93,1,audit/embedding/empirical_epsilon/0.05,0.0
+1773765831.5765529,audit_epoch,93,1,audit/embedding/empirical_epsilon/0.01,0.0
+1773765831.5765529,audit_epoch,93,1,audit/embedding/empirical_epsilon_details/0.05/epsilon,0.0
+1773765831.5765529,audit_epoch,93,1,audit/embedding/empirical_epsilon_details/0.05/num_guesses,0.0
+1773765831.5765529,audit_epoch,93,1,audit/embedding/empirical_epsilon_details/0.05/correct_guesses,0.0
+1773765831.5765529,audit_epoch,93,1,audit/embedding/empirical_epsilon_details/0.01/epsilon,0.0
+1773765831.5765529,audit_epoch,93,1,audit/embedding/empirical_epsilon_details/0.01/num_guesses,0.0
+1773765831.5765529,audit_epoch,93,1,audit/embedding/empirical_epsilon_details/0.01/correct_guesses,0.0
+1773765831.5765529,audit_epoch,93,1,perf/audit_duration_sec,6.853809644002467
+1773765927.2307582,train_step,100,2,train/step_loss,2.1885286586385377
+1773765927.2307582,train_step,100,2,train/step_real_loss,0.9171566888689995
+1773765927.2307582,train_step,100,2,train/lr,9.643076661610196e-05
+1773765927.2307582,train_step,100,2,train/step_canary_loss,13.812500953674316
+1773765927.2307582,train_step,100,2,perf/step_duration_sec,13.692489622160792
+1773765927.2307582,train_step,100,2,perf/samples_per_sec,5.18532436096129
+1773765927.2307582,train_step,100,2,perf/tokens_per_sec,3906.0098985534178
+1773765927.2307582,train_step,100,2,perf/logical_batch_size,71.0
+1773765927.2307582,train_step,100,2,perf/logical_token_count,53483.0
+1773765927.2307582,train_step,100,2,perf/physical_batches,9.0
+1773765927.2307582,train_step,100,2,privacy/epsilon,5.920610858546996
+1773765927.2307582,train_step,100,2,system/cuda_memory_allocated_gb,14.131910800933838
+1773765927.2307582,train_step,100,2,system/cuda_max_memory_allocated_gb,86.22135162353516
+1773765939.9789443,eval_step,100,2,eval/loss,0.9270543383482176
+1773765939.9789443,eval_step,100,2,eval/duration_sec,12.745669181924313
+1773766074.3579476,train_step,110,2,train/step_loss,1.4203428788618608
+1773766074.3579476,train_step,110,2,train/step_real_loss,1.016486406326294
+1773766074.3579476,train_step,110,2,train/lr,7.874347104470234e-05
+1773766074.3579476,train_step,110,2,train/step_canary_loss,14.34375
+1773766074.3579476,train_step,110,2,perf/step_duration_sec,13.423757336102426
+1773766074.3579476,train_step,110,2,perf/samples_per_sec,4.916656219827274
+1773766074.3579476,train_step,110,2,perf/tokens_per_sec,3954.481496565319
+1773766074.3579476,train_step,110,2,perf/logical_batch_size,66.0
+1773766074.3579476,train_step,110,2,perf/logical_token_count,53084.0
+1773766074.3579476,train_step,110,2,perf/physical_batches,9.0
+1773766074.3579476,train_step,110,2,privacy/epsilon,6.190670402749493
+1773766074.3579476,train_step,110,2,system/cuda_memory_allocated_gb,12.680593967437744
+1773766074.3579476,train_step,110,2,system/cuda_max_memory_allocated_gb,86.22137594223022
+1773766208.938867,train_step,120,2,train/step_loss,1.590441582807854
+1773766208.938867,train_step,120,2,train/step_real_loss,1.0116732195019722
+1773766208.938867,train_step,120,2,train/lr,6.173165676349103e-05
+1773766208.938867,train_step,120,2,train/step_canary_loss,13.9375
+1773766208.938867,train_step,120,2,perf/step_duration_sec,13.283335603773594
+1773766208.938867,train_step,120,2,perf/samples_per_sec,5.0439138179243415
+1773766208.938867,train_step,120,2,perf/tokens_per_sec,4092.8725752103383
+1773766208.938867,train_step,120,2,perf/logical_batch_size,67.0
+1773766208.938867,train_step,120,2,perf/logical_token_count,54367.0
+1773766208.938867,train_step,120,2,perf/physical_batches,9.0
+1773766208.938867,train_step,120,2,privacy/epsilon,6.450768498470402
+1773766208.938867,train_step,120,2,system/cuda_memory_allocated_gb,12.970857620239258
+1773766208.938867,train_step,120,2,system/cuda_max_memory_allocated_gb,86.22137594223022
+1773766343.2955682,train_step,130,2,train/step_loss,1.8971429631329966
+1773766343.2955682,train_step,130,2,train/step_real_loss,0.9516072571277618
+1773766343.2955682,train_step,130,2,train/lr,4.593591825444028e-05
+1773766343.2955682,train_step,130,2,train/step_canary_loss,14.0
+1773766343.2955682,train_step,130,2,perf/step_duration_sec,13.135221335105598
+1773766343.2955682,train_step,130,2,perf/samples_per_sec,5.253051946341283
+1773766343.2955682,train_step,130,2,perf/tokens_per_sec,4244.389841456128
+1773766343.2955682,train_step,130,2,perf/logical_batch_size,69.0
+1773766343.2955682,train_step,130,2,perf/logical_token_count,55751.0
+1773766343.2955682,train_step,130,2,perf/physical_batches,9.0
+1773766343.2955682,train_step,130,2,privacy/epsilon,6.701680965161658
+1773766343.2955682,train_step,130,2,system/cuda_memory_allocated_gb,13.55138349533081
+1773766343.2955682,train_step,130,2,system/cuda_max_memory_allocated_gb,86.22137594223022
+1773766477.2471156,train_step,140,2,train/step_loss,1.62191776731121
+1773766477.2471156,train_step,140,2,train/step_real_loss,1.027046725153923
+1773766477.2471156,train_step,140,2,train/lr,3.185820604061088e-05
+1773766477.2471156,train_step,140,2,train/step_canary_loss,14.3125
+1773766477.2471156,train_step,140,2,perf/step_duration_sec,13.460042576305568
+1773766477.2471156,train_step,140,2,perf/samples_per_sec,4.9776959931719444
+1773766477.2471156,train_step,140,2,perf/tokens_per_sec,4005.113631282184
+1773766477.2471156,train_step,140,2,perf/logical_batch_size,67.0
+1773766477.2471156,train_step,140,2,perf/logical_token_count,53909.0
+1773766477.2471156,train_step,140,2,perf/physical_batches,9.0
+1773766477.2471156,train_step,140,2,privacy/epsilon,6.94699659418226
+1773766477.2471156,train_step,140,2,system/cuda_memory_allocated_gb,12.970857620239258
+1773766477.2471156,train_step,140,2,system/cuda_max_memory_allocated_gb,86.22137594223022
+1773766610.9435787,train_step,150,2,train/step_loss,1.9888183966926907
+1773766610.9435787,train_step,150,2,train/step_real_loss,1.0191947892308235
+1773766610.9435787,train_step,150,2,train/lr,1.994587590756397e-05
+1773766610.9435787,train_step,150,2,train/step_canary_loss,14.40000057220459
+1773766610.9435787,train_step,150,2,perf/step_duration_sec,13.144390502013266
+1773766610.9435787,train_step,150,2,perf/samples_per_sec,5.249387561137322
+1773766610.9435787,train_step,150,2,perf/tokens_per_sec,3851.1485178598896
+1773766610.9435787,train_step,150,2,perf/logical_batch_size,69.0
+1773766610.9435787,train_step,150,2,perf/logical_token_count,50621.0
+1773766610.9435787,train_step,150,2,perf/physical_batches,9.0
+1773766610.9435787,train_step,150,2,privacy/epsilon,7.185356422932143
+1773766610.9435787,train_step,150,2,system/cuda_memory_allocated_gb,13.55138349533081
+1773766610.9435787,train_step,150,2,system/cuda_max_memory_allocated_gb,86.22137594223022
+1773766623.702012,eval_step,150,2,eval/loss,0.9247592026606585
+1773766623.702012,eval_step,150,2,eval/duration_sec,12.756307526025921
+1773766758.1413558,train_step,160,2,train/step_loss,1.475486863743175
+1773766758.1413558,train_step,160,2,train/step_real_loss,1.1006973907351494
+1773766758.1413558,train_step,160,2,train/lr,1.057747301402887e-05
+1773766758.1413558,train_step,160,2,train/step_canary_loss,13.46875
+1773766758.1413558,train_step,160,2,perf/step_duration_sec,13.887811011634767
+1773766758.1413558,train_step,160,2,perf/samples_per_sec,4.752368817858142
+1773766758.1413558,train_step,160,2,perf/tokens_per_sec,3180.198806204889
+1773766758.1413558,train_step,160,2,perf/logical_batch_size,66.0
+1773766758.1413558,train_step,160,2,perf/logical_token_count,44166.0
+1773766758.1413558,train_step,160,2,perf/physical_batches,9.0
+1773766758.1413558,train_step,160,2,privacy/epsilon,7.416930400989618
+1773766758.1413558,train_step,160,2,system/cuda_memory_allocated_gb,12.680593967437744
+1773766758.1413558,train_step,160,2,system/cuda_max_memory_allocated_gb,86.22137594223022
+1773766893.2984576,train_step,170,2,train/step_loss,1.6393254835214188
+1773766893.2984576,train_step,170,2,train/step_real_loss,1.0540594905614853
+1773766893.2984576,train_step,170,2,train/lr,4.050702638550275e-06
+1773766893.2984576,train_step,170,2,train/step_canary_loss,14.125
+1773766893.2984576,train_step,170,2,perf/step_duration_sec,13.72155143506825
+1773766893.2984576,train_step,170,2,perf/samples_per_sec,4.882829781825378
+1773766893.2984576,train_step,170,2,perf/tokens_per_sec,3613.58555077656
+1773766893.2984576,train_step,170,2,perf/logical_batch_size,67.0
+1773766893.2984576,train_step,170,2,perf/logical_token_count,49584.0
+1773766893.2984576,train_step,170,2,perf/physical_batches,9.0
+1773766893.2984576,train_step,170,2,privacy/epsilon,7.645014003428424
+1773766893.2984576,train_step,170,2,system/cuda_memory_allocated_gb,12.970857620239258
+1773766893.2984576,train_step,170,2,system/cuda_max_memory_allocated_gb,86.22137594223022
+1773767029.0999126,train_step,180,2,train/step_loss,1.7291738425984102
+1773767029.0999126,train_step,180,2,train/step_real_loss,1.0520909577608109
+1773767029.0999126,train_step,180,2,train/lr,5.729698228102653e-07
+1773767029.0999126,train_step,180,2,train/step_canary_loss,12.5625
+1773767029.0999126,train_step,180,2,perf/step_duration_sec,13.660048387013376
+1773767029.0999126,train_step,180,2,perf/samples_per_sec,4.978020434001367
+1773767029.0999126,train_step,180,2,perf/tokens_per_sec,3874.5104336758286
+1773767029.0999126,train_step,180,2,perf/logical_batch_size,68.0
+1773767029.0999126,train_step,180,2,perf/logical_token_count,52926.0
+1773767029.0999126,train_step,180,2,perf/physical_batches,9.0
+1773767029.0999126,train_step,180,2,privacy/epsilon,7.8658492724205455
+1773767029.0999126,train_step,180,2,system/cuda_memory_allocated_gb,13.261121273040771
+1773767029.0999126,train_step,180,2,system/cuda_max_memory_allocated_gb,86.22137594223022
+1773767117.390551,train_epoch,186,2,train/epoch_loss,1.5804422382361127
+1773767117.390551,train_epoch,186,2,train/epoch_real_loss,1.005713254121899
+1773767117.390551,train_epoch,186,2,train/epoch_canary_loss,13.023172873210777
+1773767117.390551,train_epoch,186,2,perf/epoch_duration_sec,1272.894041202031
+1773767117.390551,train_epoch,186,2,perf/epoch_samples_per_sec,38.91918604098259
+1773767117.390551,train_epoch,186,2,perf/epoch_tokens_per_sec,29671.033705472055
+1773767117.390551,train_epoch,186,2,perf/epoch_samples,49540.0
+1773767117.390551,train_epoch,186,2,perf/epoch_tokens,37768082.0
+1773767117.390551,train_epoch,186,2,system/cuda_epoch_peak_memory_gb,86.22137594223022
+1773767117.390551,train_epoch,186,2,eval/loss,0.924621483645378
+1773767117.390551,train_epoch,186,2,eval/duration_sec,12.757435038685799
+1773767117.390551,train_epoch,186,2,privacy/epsilon,7.996749609735891
+1773767130.595887,audit_epoch,186,2,audit/delta,1e-05
+1773767130.595887,audit_epoch,186,2,audit/num_canaries,500.0
+1773767130.595887,audit_epoch,186,2,audit/num_members,250.0
+1773767130.595887,audit_epoch,186,2,audit/paper_guess_fraction,0.2
+1773767130.595887,audit_epoch,186,2,audit/paper_guess_steps,20.0
+1773767130.595887,audit_epoch,186,2,audit/loss/auc,0.515088
+1773767130.595887,audit_epoch,186,2,audit/loss/empirical_epsilon/0.05,0.0
+1773767130.595887,audit_epoch,186,2,audit/loss/empirical_epsilon/0.01,0.0
+1773767130.595887,audit_epoch,186,2,audit/loss/empirical_epsilon_details/0.05/epsilon,0.0
+1773767130.595887,audit_epoch,186,2,audit/loss/empirical_epsilon_details/0.05/num_guesses,0.0
+1773767130.595887,audit_epoch,186,2,audit/loss/empirical_epsilon_details/0.05/correct_guesses,0.0
+1773767130.595887,audit_epoch,186,2,audit/loss/empirical_epsilon_details/0.01/epsilon,0.0
+1773767130.595887,audit_epoch,186,2,audit/loss/empirical_epsilon_details/0.01/num_guesses,0.0
+1773767130.595887,audit_epoch,186,2,audit/loss/empirical_epsilon_details/0.01/correct_guesses,0.0
+1773767130.595887,audit_epoch,186,2,audit/embedding/auc,0.516208
+1773767130.595887,audit_epoch,186,2,audit/embedding/empirical_epsilon/0.05,0.0
+1773767130.595887,audit_epoch,186,2,audit/embedding/empirical_epsilon/0.01,0.0
+1773767130.595887,audit_epoch,186,2,audit/embedding/empirical_epsilon_details/0.05/epsilon,0.0
+1773767130.595887,audit_epoch,186,2,audit/embedding/empirical_epsilon_details/0.05/num_guesses,0.0
+1773767130.595887,audit_epoch,186,2,audit/embedding/empirical_epsilon_details/0.05/correct_guesses,0.0
+1773767130.595887,audit_epoch,186,2,audit/embedding/empirical_epsilon_details/0.01/epsilon,0.0
+1773767130.595887,audit_epoch,186,2,audit/embedding/empirical_epsilon_details/0.01/num_guesses,0.0
+1773767130.595887,audit_epoch,186,2,audit/embedding/empirical_epsilon_details/0.01/correct_guesses,0.0
+1773767130.595887,audit_epoch,186,2,perf/audit_duration_sec,6.922967464663088
+1773767143.9571598,audit_final,186,2,audit/delta,1e-05
+1773767143.9571598,audit_final,186,2,audit/num_canaries,500.0
+1773767143.9571598,audit_final,186,2,audit/num_members,250.0
+1773767143.9571598,audit_final,186,2,audit/paper_guess_fraction,0.2
+1773767143.9571598,audit_final,186,2,audit/paper_guess_steps,20.0
+1773767143.9571598,audit_final,186,2,audit/loss/auc,0.515088
+1773767143.9571598,audit_final,186,2,audit/loss/empirical_epsilon/0.05,0.0
+1773767143.9571598,audit_final,186,2,audit/loss/empirical_epsilon/0.01,0.0
+1773767143.9571598,audit_final,186,2,audit/loss/empirical_epsilon_details/0.05/epsilon,0.0
+1773767143.9571598,audit_final,186,2,audit/loss/empirical_epsilon_details/0.05/num_guesses,0.0
+1773767143.9571598,audit_final,186,2,audit/loss/empirical_epsilon_details/0.05/correct_guesses,0.0
+1773767143.9571598,audit_final,186,2,audit/loss/empirical_epsilon_details/0.01/epsilon,0.0
+1773767143.9571598,audit_final,186,2,audit/loss/empirical_epsilon_details/0.01/num_guesses,0.0
+1773767143.9571598,audit_final,186,2,audit/loss/empirical_epsilon_details/0.01/correct_guesses,0.0
+1773767143.9571598,audit_final,186,2,audit/embedding/auc,0.516208
+1773767143.9571598,audit_final,186,2,audit/embedding/empirical_epsilon/0.05,0.0
+1773767143.9571598,audit_final,186,2,audit/embedding/empirical_epsilon/0.01,0.0
+1773767143.9571598,audit_final,186,2,audit/embedding/empirical_epsilon_details/0.05/epsilon,0.0
+1773767143.9571598,audit_final,186,2,audit/embedding/empirical_epsilon_details/0.05/num_guesses,0.0
+1773767143.9571598,audit_final,186,2,audit/embedding/empirical_epsilon_details/0.05/correct_guesses,0.0
+1773767143.9571598,audit_final,186,2,audit/embedding/empirical_epsilon_details/0.01/epsilon,0.0
+1773767143.9571598,audit_final,186,2,audit/embedding/empirical_epsilon_details/0.01/num_guesses,0.0
+1773767143.9571598,audit_final,186,2,audit/embedding/empirical_epsilon_details/0.01/correct_guesses,0.0
+1773767144.5062578,energy_final,186,,energy/codecarbon/duration,2693.6014373912476
+1773767144.5062578,energy_final,186,,energy/codecarbon/emissions,0.09826528580395394
+1773767144.5062578,energy_final,186,,energy/codecarbon/emissions_rate,3.648100436830916e-05
+1773767144.5062578,energy_final,186,,energy/codecarbon/cpu_power,72.03163066573623
+1773767144.5062578,energy_final,186,,energy/codecarbon/gpu_power,3108.4298352234105
+1773767144.5062578,energy_final,186,,energy/codecarbon/ram_power,54.0
+1773767144.5062578,energy_final,186,,energy/codecarbon/cpu_energy,0.051919376660011826
+1773767144.5062578,energy_final,186,,energy/codecarbon/gpu_energy,2.3238466076869315
+1773767144.5062578,energy_final,186,,energy/codecarbon/ram_energy,0.03891114561875117
+1773767144.5062578,energy_final,186,,energy/codecarbon/energy_consumed,2.414677129965695
+1773767144.5062578,energy_final,186,,energy/codecarbon/water_consumed,0.0
+1773767144.5062578,energy_final,186,,energy/codecarbon/cpu_count,256.0
+1773767144.5062578,energy_final,186,,energy/codecarbon/gpu_count,8.0
+1773767144.5062578,energy_final,186,,energy/codecarbon/longitude,16.1885
+1773767144.5062578,energy_final,186,,energy/codecarbon/latitude,58.594
+1773767144.5062578,energy_final,186,,energy/codecarbon/ram_total_size,1511.49019241333
+1773767144.5062578,energy_final,186,,energy/codecarbon/cpu_utilization_percent,3.719566840926081
+1773767144.5062578,energy_final,186,,energy/codecarbon/gpu_utilization_percent,83.2715646004481
+1773767144.5062578,energy_final,186,,energy/codecarbon/ram_utilization_percent,5.423711725167903
+1773767144.5062578,energy_final,186,,energy/codecarbon/ram_used_gb,82.01636290603649
+1773767144.5062578,energy_final,186,,energy/codecarbon/pue,1.0
+1773767144.5062578,energy_final,186,,energy/codecarbon/wue,0.0

qwen3-4b-instruct/dp8/summary.json

@@ -0,0 +1,72 @@
+{
+  "audit/delta": 1e-05,
+  "audit/embedding/auc": 0.516208,
+  "audit/embedding/empirical_epsilon/0.01": 0.0,
+  "audit/embedding/empirical_epsilon/0.05": 0.0,
+  "audit/embedding/empirical_epsilon_details/0.01/correct_guesses": 0.0,
+  "audit/embedding/empirical_epsilon_details/0.01/epsilon": 0.0,
+  "audit/embedding/empirical_epsilon_details/0.01/num_guesses": 0.0,
+  "audit/embedding/empirical_epsilon_details/0.05/correct_guesses": 0.0,
+  "audit/embedding/empirical_epsilon_details/0.05/epsilon": 0.0,
+  "audit/embedding/empirical_epsilon_details/0.05/num_guesses": 0.0,
+  "audit/loss/auc": 0.515088,
+  "audit/loss/empirical_epsilon/0.01": 0.0,
+  "audit/loss/empirical_epsilon/0.05": 0.0,
+  "audit/loss/empirical_epsilon_details/0.01/correct_guesses": 0.0,
+  "audit/loss/empirical_epsilon_details/0.01/epsilon": 0.0,
+  "audit/loss/empirical_epsilon_details/0.01/num_guesses": 0.0,
+  "audit/loss/empirical_epsilon_details/0.05/correct_guesses": 0.0,
+  "audit/loss/empirical_epsilon_details/0.05/epsilon": 0.0,
+  "audit/loss/empirical_epsilon_details/0.05/num_guesses": 0.0,
+  "audit/num_canaries": 500.0,
+  "audit/num_members": 250.0,
+  "audit/paper_guess_fraction": 0.2,
+  "audit/paper_guess_steps": 20.0,
+  "energy/codecarbon/cpu_count": 256.0,
+  "energy/codecarbon/cpu_energy": 0.051919376660011826,
+  "energy/codecarbon/cpu_power": 72.03163066573623,
+  "energy/codecarbon/cpu_utilization_percent": 3.719566840926081,
+  "energy/codecarbon/duration": 2693.6014373912476,
+  "energy/codecarbon/emissions": 0.09826528580395394,
+  "energy/codecarbon/emissions_rate": 3.648100436830916e-05,
+  "energy/codecarbon/energy_consumed": 2.414677129965695,
+  "energy/codecarbon/gpu_count": 8.0,
+  "energy/codecarbon/gpu_energy": 2.3238466076869315,
+  "energy/codecarbon/gpu_power": 3108.4298352234105,
+  "energy/codecarbon/gpu_utilization_percent": 83.2715646004481,
+  "energy/codecarbon/latitude": 58.594,
+  "energy/codecarbon/longitude": 16.1885,
+  "energy/codecarbon/pue": 1.0,
+  "energy/codecarbon/ram_energy": 0.03891114561875117,
+  "energy/codecarbon/ram_power": 54.0,
+  "energy/codecarbon/ram_total_size": 1511.49019241333,
+  "energy/codecarbon/ram_used_gb": 82.01636290603649,
+  "energy/codecarbon/ram_utilization_percent": 5.423711725167903,
+  "energy/codecarbon/water_consumed": 0.0,
+  "energy/codecarbon/wue": 0.0,
+  "eval/duration_sec": 12.757435038685799,
+  "eval/loss": 0.924621483645378,
+  "perf/audit_duration_sec": 6.922967464663088,
+  "perf/epoch_duration_sec": 1272.894041202031,
+  "perf/epoch_samples": 49540.0,
+  "perf/epoch_samples_per_sec": 38.91918604098259,
+  "perf/epoch_tokens": 37768082.0,
+  "perf/epoch_tokens_per_sec": 29671.033705472055,
+  "perf/logical_batch_size": 68.0,
+  "perf/logical_token_count": 52926.0,
+  "perf/physical_batches": 9.0,
+  "perf/samples_per_sec": 4.978020434001367,
+  "perf/step_duration_sec": 13.660048387013376,
+  "perf/tokens_per_sec": 3874.5104336758286,
+  "privacy/epsilon": 7.996749609735891,
+  "system/cuda_epoch_peak_memory_gb": 86.22137594223022,
+  "system/cuda_max_memory_allocated_gb": 86.22137594223022,
+  "system/cuda_memory_allocated_gb": 13.261121273040771,
+  "train/epoch_canary_loss": 13.023172873210777,
+  "train/epoch_loss": 1.5804422382361127,
+  "train/epoch_real_loss": 1.005713254121899,
+  "train/lr": 5.729698228102653e-07,
+  "train/step_canary_loss": 12.5625,
+  "train/step_loss": 1.7291738425984102,
+  "train/step_real_loss": 1.0520909577608109
+}

qwen3-4b-instruct/dp8/tensorboard/events.out.tfevents.1773764448.7b654b6988b0.41500.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:eff1ed8df15d5771e60d52c8dab89e065560c7481fb911bfbbce08dd4f2d2e70
+size 25026

qwen3-4b-instruct/dp8/tokenizer/chat_template.jinja

@@ -0,0 +1,61 @@
+{%- if tools %}
+    {{- '<|im_start|>system\n' }}
+    {%- if messages[0].role == 'system' %}
+        {{- messages[0].content + '\n\n' }}
+    {%- endif %}
+    {{- "# Tools\n\nYou may call one or more functions to assist with the user query.\n\nYou are provided with function signatures within <tools></tools> XML tags:\n<tools>" }}
+    {%- for tool in tools %}
+        {{- "\n" }}
+        {{- tool | tojson }}
+    {%- endfor %}
+    {{- "\n</tools>\n\nFor each function call, return a json object with function name and arguments within <tool_call></tool_call> XML tags:\n<tool_call>\n{\"name\": <function-name>, \"arguments\": <args-json-object>}\n</tool_call><|im_end|>\n" }}
+{%- else %}
+    {%- if messages[0].role == 'system' %}
+        {{- '<|im_start|>system\n' + messages[0].content + '<|im_end|>\n' }}
+    {%- endif %}
+{%- endif %}
+{%- for message in messages %}
+    {%- if message.content is string %}
+        {%- set content = message.content %}
+    {%- else %}
+        {%- set content = '' %}
+    {%- endif %}
+    {%- if (message.role == "user") or (message.role == "system" and not loop.first) %}
+        {{- '<|im_start|>' + message.role + '\n' + content + '<|im_end|>' + '\n' }}
+    {%- elif message.role == "assistant" %}
+        {{- '<|im_start|>' + message.role + '\n' + content }}
+        {%- if message.tool_calls %}
+            {%- for tool_call in message.tool_calls %}
+                {%- if (loop.first and content) or (not loop.first) %}
+                    {{- '\n' }}
+                {%- endif %}
+                {%- if tool_call.function %}
+                    {%- set tool_call = tool_call.function %}
+                {%- endif %}
+                {{- '<tool_call>\n{"name": "' }}
+                {{- tool_call.name }}
+                {{- '", "arguments": ' }}
+                {%- if tool_call.arguments is string %}
+                    {{- tool_call.arguments }}
+                {%- else %}
+                    {{- tool_call.arguments | tojson }}
+                {%- endif %}
+                {{- '}\n</tool_call>' }}
+            {%- endfor %}
+        {%- endif %}
+        {{- '<|im_end|>\n' }}
+    {%- elif message.role == "tool" %}
+        {%- if loop.first or (messages[loop.index0 - 1].role != "tool") %}
+            {{- '<|im_start|>user' }}
+        {%- endif %}
+        {{- '\n<tool_response>\n' }}
+        {{- content }}
+        {{- '\n</tool_response>' }}
+        {%- if loop.last or (messages[loop.index0 + 1].role != "tool") %}
+            {{- '<|im_end|>\n' }}
+        {%- endif %}
+    {%- endif %}
+{%- endfor %}
+{%- if add_generation_prompt %}
+    {{- '<|im_start|>assistant\n' }}
+{%- endif %}

qwen3-4b-instruct/dp8/tokenizer/tokenizer.json

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0e9c8aef460c70c1e1c32afe895f455856c0075e5706f06e6d80b2f581137715
+size 11517150

qwen3-4b-instruct/dp8/tokenizer/tokenizer_config.json

@@ -0,0 +1,516 @@
+{
+  "add_prefix_space": false,
+  "backend": "tokenizers",
+  "bos_token": null,
+  "clean_up_tokenization_spaces": false,
+  "eos_token": "<|im_end|>",
+  "errors": "replace",
+  "extra_special_tokens": [
+    "865331112869",
+    "569765693871",
+    "485177821815",
+    "135441121756",
+    "367459894796",
+    "877482678543",
+    "457919547633",
+    "765474393376",
+    "114848338811",
+    "746285987371",
+    "649291669397",
+    "927914615679",
+    "445925149649",
+    "691587454538",
+    "143777992227",
+    "997981281989",
+    "425949483533",
+    "982993456429",
+    "718726519731",
+    "172599315861",
+    "643489267333",
+    "282322838685",
+    "781653545886",
+    "796415361892",
+    "841991688488",
+    "211411365397",
+    "698218415444",
+    "355977139358",
+    "682564697312",
+    "383837596997",
+    "689362171782",
+    "749966767285",
+    "753159165157",
+    "795693824762",
+    "669689115557",
+    "327491773134",
+    "983569279932",
+    "612128769512",
+    "374327157578",
+    "311632789559",
+    "523918658846",
+    "765981581453",
+    "794825141891",
+    "873898736873",
+    "447445629421",
+    "473822473819",
+    "181439694557",
+    "592538279337",
+    "668134915514",
+    "643692393748",
+    "696651276628",
+    "853859348234",
+    "778466723723",
+    "929826356991",
+    "272362973463",
+    "694235616268",
+    "281673864127",
+    "479676316326",
+    "646979124677",
+    "922327493433",
+    "883685933161",
+    "264259917554",
+    "836746273134",
+    "658481324922",
+    "481884157827",
+    "587787496812",
+    "579184949249",
+    "912193598348",
+    "529679678956",
+    "795838284624",
+    "159337222655",
+    "173781362446",
+    "773687856563",
+    "535787224917",
+    "351885857332",
+    "578827344666",
+    "198462689911",
+    "722618266242",
+    "952872416512",
+    "517778845323",
+    "749665846687",
+    "661436365453",
+    "259666844669",
+    "242851284913",
+    "514532995959",
+    "161588262349",
+    "742765629356",
+    "225164373623",
+    "676539973863",
+    "826214551218",
+    "182345464792",
+    "232776999554",
+    "337326533813",
+    "676676697292",
+    "929185622831",
+    "545512344383",
+    "499444466686",
+    "314697386682",
+    "517379856925",
+    "379557332953",
+    "614797267726",
+    "429781429464",
+    "922466849763",
+    "721737645236",
+    "479227349997",
+    "136931728327",
+    "259533577263",
+    "488538864842",
+    "937495658852",
+    "489991411364",
+    "499148455254",
+    "441373944925",
+    "899151413682",
+    "467893531755",
+    "527117488925",
+    "928335588653",
+    "374439448821",
+    "879425227932",
+    "867678158885",
+    "399749397872",
+    "129693547287",
+    "689285841825",
+    "771619544974",
+    "724883568652",
+    "516968424863",
+    "733737988257",
+    "852347289392",
+    "296953381169",
+    "377273562477",
+    "262296912232",
+    "547149832394",
+    "298464134954",
+    "216667245274",
+    "843998562287",
+    "572154333646",
+    "124589118494",
+    "841824384614",
+    "232896526252",
+    "295448593321",
+    "123741461297",
+    "653573457168",
+    "196735786156",
+    "377338713663",
+    "964342468552",
+    "586855179568",
+    "484773717614",
+    "894885246797",
+    "677896358599",
+    "848845611563",
+    "851852651677",
+    "398549545767",
+    "454244839926",
+    "799364566435",
+    "967114116556",
+    "817378986438",
+    "233795848681",
+    "824387273757",
+    "916198946615",
+    "563117729724",
+    "951794811935",
+    "374598961236",
+    "922867396683",
+    "765737843639",
+    "175469284871",
+    "231853711778",
+    "662426712668",
+    "711412347158",
+    "753466987363",
+    "513361312532",
+    "712992815957",
+    "971621888444",
+    "829235161526",
+    "585544633356",
+    "582471228164",
+    "678666359123",
+    "557533689478",
+    "632962475133",
+    "484489193824",
+    "489562189822",
+    "589547936288",
+    "363214487524",
+    "244885399387",
+    "431751228368",
+    "433581868192",
+    "486391569221",
+    "185438575221",
+    "126574388585",
+    "741757479784",
+    "529854679937",
+    "996116119839",
+    "616248973917",
+    "763531783491",
+    "955456118295",
+    "364196983365",
+    "195792996468",
+    "151859598873",
+    "399223169721",
+    "938488813964",
+    "961981959227",
+    "183368827562",
+    "533417736566",
+    "786391632558",
+    "665661658354",
+    "693281533643",
+    "475794684356",
+    "652154162978",
+    "753233719644",
+    "668514843129",
+    "819162623892",
+    "941169431859",
+    "877385381798",
+    "752644929761",
+    "881136466196",
+    "275597777299",
+    "731681792655",
+    "961133895172",
+    "864718285734",
+    "963852916563",
+    "319584985416",
+    "563365646341",
+    "811371928234",
+    "837131396371",
+    "267514771964",
+    "944513428457",
+    "117298239631",
+    "158142752582",
+    "252867443568",
+    "839269684865",
+    "612788593128",
+    "145669731981",
+    "121557291859",
+    "245416776926",
+    "799417897197",
+    "997958836435",
+    "892336777248",
+    "158929292238",
+    "581976444672",
+    "897784492783",
+    "492373714791",
+    "512659818733",
+    "881112998642",
+    "619454958782",
+    "431149748713",
+    "624221476921",
+    "125866399464",
+    "339882449689",
+    "186198784585",
+    "943193294691",
+    "955668961269",
+    "232787996724",
+    "215671314196",
+    "286173241916",
+    "745977673725",
+    "556976448182",
+    "599961512792",
+    "766294538337",
+    "934912591213",
+    "295118729589",
+    "529455466433",
+    "196119929397",
+    "379571934299",
+    "251789649997",
+    "564544131355",
+    "244371196654",
+    "384598329253",
+    "887753195844",
+    "364947325679",
+    "655517954651",
+    "673948786567",
+    "857231548835",
+    "816115936673",
+    "644234165531",
+    "182782912224",
+    "234316622259",
+    "421369185549",
+    "434632855397",
+    "921889371893",
+    "415956914763",
+    "598916996413",
+    "773671349113",
+    "952465217972",
+    "117657531962",
+    "729825168745",
+    "691315125346",
+    "768461952319",
+    "664847713559",
+    "953267689786",
+    "886464195129",
+    "824488329416",
+    "837873762491",
+    "532833541879",
+    "669183782449",
+    "941976537588",
+    "739394546916",
+    "267954879268",
+    "637551427887",
+    "217756494954",
+    "524444658383",
+    "117783274348",
+    "138218735276",
+    "814611949491",
+    "711641973413",
+    "499156317423",
+    "515856611931",
+    "454164859837",
+    "345271433112",
+    "462294118988",
+    "511785788222",
+    "497294727353",
+    "866519986723",
+    "334513529294",
+    "549946382131",
+    "284445431422",
+    "396521188476",
+    "421435255895",
+    "133373659361",
+    "322683334381",
+    "228358422847",
+    "291762694874",
+    "143182978129",
+    "511923256573",
+    "327158398268",
+    "879764613759",
+    "564395222747",
+    "451161679736",
+    "538631466654",
+    "221762325616",
+    "218391991184",
+    "322589379462",
+    "876537814263",
+    "152676556624",
+    "332522971941",
+    "884354318946",
+    "513349618943",
+    "116639746413",
+    "635185846287",
+    "993832498489",
+    "813981174797",
+    "438745114173",
+    "983493951323",
+    "724492262421",
+    "622553389126",
+    "889965243135",
+    "364492359246",
+    "154962668224",
+    "179564995814",
+    "418412875665",
+    "718951851413",
+    "699446724178",
+    "624266421831",
+    "815458725125",
+    "455423278865",
+    "393741199486",
+    "328552864359",
+    "211662639865",
+    "218784516525",
+    "762486672996",
+    "142799718159",
+    "858146415154",
+    "767858144912",
+    "571317457151",
+    "635127952696",
+    "116427191984",
+    "268921994538",
+    "523937669294",
+    "165429152138",
+    "739246183345",
+    "591464355756",
+    "212985874612",
+    "191887635211",
+    "967214577653",
+    "119342152414",
+    "946444632795",
+    "618423867817",
+    "228565148417",
+    "729116422489",
+    "527874729936",
+    "739784153482",
+    "387763951128",
+    "331369926711",
+    "562716493614",
+    "739667844957",
+    "562389434565",
+    "256497188281",
+    "859927364588",
+    "417668946583",
+    "357621613582",
+    "438435178228",
+    "485692541169",
+    "825815739116",
+    "342221452223",
+    "697747991249",
+    "716763689965",
+    "141499982867",
+    "818479319499",
+    "336813343298",
+    "594688742928",
+    "472129283475",
+    "514354144759",
+    "349249721685",
+    "546276298359",
+    "353755529131",
+    "315534574435",
+    "523723475786",
+    "215826764872",
+    "367968398551",
+    "569853653352",
+    "389715484387",
+    "293847485454",
+    "714738141818",
+    "178478368922",
+    "581493616981",
+    "589439538674",
+    "846657726193",
+    "722339992679",
+    "138154781148",
+    "757785319772",
+    "492516914298",
+    "919181521716",
+    "985781138935",
+    "476969195485",
+    "313145133463",
+    "758963111966",
+    "147541537162",
+    "557163366873",
+    "144373897488",
+    "522515164754",
+    "724964923582",
+    "284776712475",
+    "375429755114",
+    "181233596124",
+    "948585673431",
+    "243165586174",
+    "396847976144",
+    "997724962668",
+    "558837194455",
+    "163165456396",
+    "378749551722",
+    "161238482259",
+    "754978243758",
+    "195388849133",
+    "229775525672",
+    "262437452884",
+    "441377892146",
+    "451885565366",
+    "981277526855",
+    "762495822823",
+    "368763327262",
+    "757422791351",
+    "636324136426",
+    "214193645583",
+    "412843856172",
+    "179386156569",
+    "756916173536",
+    "892697125149",
+    "625334487352",
+    "941861857715",
+    "887417525236",
+    "649516938598",
+    "717628619782",
+    "438124184139",
+    "547563892268",
+    "856317483891",
+    "313313831273",
+    "371496153876",
+    "587541149322",
+    "265847332563",
+    "449549215429",
+    "163497196769",
+    "861342291298",
+    "268433315926",
+    "774679513717",
+    "851254219729",
+    "583527834464",
+    "488496781997",
+    "556814553861",
+    "482829231639",
+    "618878266619",
+    "147444452794",
+    "949235426629",
+    "357299947518",
+    "175528632226",
+    "645527857972",
+    "186872457894",
+    "552738847828",
+    "626748382482",
+    "921894985642",
+    "943878645871",
+    "859289776479",
+    "614583493135",
+    "933775286797",
+    "332234613346",
+    "325196781219",
+    "142526557681",
+    "356722692178",
+    "449318681694",
+    "687284547244",
+    "947262995132",
+    "893974619684",
+    "797238311233"
+  ],
+  "is_local": false,
+  "model_max_length": 1010000,
+  "pad_token": "<|endoftext|>",
+  "split_special_tokens": false,
+  "tokenizer_class": "Qwen2Tokenizer",
+  "unk_token": null
+}

qwen3-4b-instruct/dp8/train.log

@@ -0,0 +1,21 @@
+2026-03-17 16:24:45,463 [INFO] new_opacus_codex.train_steps: epoch=1 step=10 loss=1.6904
+2026-03-17 16:26:59,829 [INFO] new_opacus_codex.train_steps: epoch=1 step=20 loss=1.6599
+2026-03-17 16:29:13,865 [INFO] new_opacus_codex.train_steps: epoch=1 step=30 loss=1.6978
+2026-03-17 16:31:27,715 [INFO] new_opacus_codex.train_steps: epoch=1 step=40 loss=1.5809
+2026-03-17 16:33:41,506 [INFO] new_opacus_codex.train_steps: epoch=1 step=50 loss=1.5804
+2026-03-17 16:33:54,222 [INFO] new_opacus_codex.train_steps: eval event=eval_step epoch=1 step=50 eval_loss=0.9504 duration_sec=12.71
+2026-03-17 16:36:08,859 [INFO] new_opacus_codex.train_steps: epoch=1 step=60 loss=1.5633
+2026-03-17 16:38:23,253 [INFO] new_opacus_codex.train_steps: epoch=1 step=70 loss=1.7269
+2026-03-17 16:40:37,881 [INFO] new_opacus_codex.train_steps: epoch=1 step=80 loss=1.7133
+2026-03-17 16:42:52,223 [INFO] new_opacus_codex.train_steps: epoch=1 step=90 loss=1.5771
+2026-03-17 16:45:27,230 [INFO] new_opacus_codex.train_steps: epoch=2 step=100 loss=1.7272
+2026-03-17 16:45:39,978 [INFO] new_opacus_codex.train_steps: eval event=eval_step epoch=2 step=100 eval_loss=0.9271 duration_sec=12.75
+2026-03-17 16:47:54,357 [INFO] new_opacus_codex.train_steps: epoch=2 step=110 loss=1.7404
+2026-03-17 16:50:08,938 [INFO] new_opacus_codex.train_steps: epoch=2 step=120 loss=1.5583
+2026-03-17 16:52:23,295 [INFO] new_opacus_codex.train_steps: epoch=2 step=130 loss=1.5209
+2026-03-17 16:54:37,246 [INFO] new_opacus_codex.train_steps: epoch=2 step=140 loss=1.7010
+2026-03-17 16:56:50,943 [INFO] new_opacus_codex.train_steps: epoch=2 step=150 loss=1.5696
+2026-03-17 16:57:03,701 [INFO] new_opacus_codex.train_steps: eval event=eval_step epoch=2 step=150 eval_loss=0.9248 duration_sec=12.76
+2026-03-17 16:59:18,140 [INFO] new_opacus_codex.train_steps: epoch=2 step=160 loss=1.3749
+2026-03-17 17:01:33,298 [INFO] new_opacus_codex.train_steps: epoch=2 step=170 loss=1.6490
+2026-03-17 17:03:49,099 [INFO] new_opacus_codex.train_steps: epoch=2 step=180 loss=1.6707