Add files using upload-large-folder tool

L23_clean_cont2_train.log

@@ -0,0 +1,15 @@
+
+[pd] RESUMED from experiments/v8_nla_local/checkpoints/av_L23_clean_cont/step_000100 (extending grokking horizon)
+[pd] data=experiments/v8_nla_local/data/stage3_balanced/av_sft_balanced.parquet rows=1356 uniform=True gamma=1.0 corpus_mean_weight=1.000
+[pd] DOMAIN-AWARE negs: avg same-dom=16.0 diff-dom=16.0 per anchor (15 domains)
+[pd] contrastive ON: beta=1.0 K=2 temp=1.0 domain_aware=True token_dropout=0.0 (InfoNCE over hard-neg activations)
+[pd] INJECT at residual layer 23 (native scale); embed pass-through
+[pd] inject_layer=23 inject_mode=raw
+[pd] optimizer=adamw8bit lr=5e-05 wd=0.01 steps=500 ga=8 trainable_params=12.08M
+`use_cache=True` is incompatible with gradient checkpointing. Setting `use_cache=False`.
+C:\Users\caleb\deception-nanochat-sae-research\.venv-gemma4\Lib\site-packages\torch\_dynamo\eval_frame.py:1181: UserWarning: torch.utils.checkpoint: the use_reentrant parameter should be passed explicitly. Starting in PyTorch 2.9, calling checkpoint without use_reentrant will raise an exception. use_reentrant=False is recommended, but if you need to preserve the current default behavior, you can pass use_reentrant=True. Refer to docs for more details on the differences between the two variants.
+  return fn(*args, **kwargs)
+C:\Users\caleb\deception-nanochat-sae-research\experiments\v8_nla_local\train_prior_deviation_av.py:325: UserWarning: Converting a tensor with requires_grad=True to a scalar may lead to unexpected behavior.
+Consider using tensor.detach() first. (Triggered internally at C:\actions-runner\_work\pytorch\pytorch\pytorch\torch\csrc\autograd\generated\python_variable_methods.cpp:837.)
+  (loss / ga).backward(); tot += float(loss) / ga
+[pd] step 1/500 wloss=3.122 skipped=0

L23_clean_cont_train.log

@@ -0,0 +1,32 @@
+C:\Users\caleb\deception-nanochat-sae-research\.venv-gemma4\Lib\site-packages\transformers\modeling_utils.py:4832: UserWarning: expandable_segments not supported on this platform (Triggered internally at C:\actions-runner\_work\pytorch\pytorch\pytorch\c10/cuda/CUDAAllocatorConfig.h:39.)
+  _ = torch.empty(int(byte_count // 2), dtype=torch.float16, device=device, requires_grad=False)
+
+[pd] RESUMED from experiments/v8_nla_local/checkpoints/av_L23_clean/step_000400 (extending grokking horizon)
+[pd] data=experiments/v8_nla_local/data/stage3_balanced/av_sft_balanced.parquet rows=1356 uniform=True gamma=1.0 corpus_mean_weight=1.000
+[pd] DOMAIN-AWARE negs: avg same-dom=16.0 diff-dom=16.0 per anchor (15 domains)
+[pd] contrastive ON: beta=1.0 K=2 temp=1.0 domain_aware=True token_dropout=0.0 (InfoNCE over hard-neg activations)
+[pd] INJECT at residual layer 23 (native scale); embed pass-through
+[pd] inject_layer=23 inject_mode=raw
+[pd] optimizer=adamw8bit lr=5e-05 wd=0.01 steps=600 ga=8 trainable_params=12.08M
+`use_cache=True` is incompatible with gradient checkpointing. Setting `use_cache=False`.
+C:\Users\caleb\deception-nanochat-sae-research\.venv-gemma4\Lib\site-packages\torch\_dynamo\eval_frame.py:1181: UserWarning: torch.utils.checkpoint: the use_reentrant parameter should be passed explicitly. Starting in PyTorch 2.9, calling checkpoint without use_reentrant will raise an exception. use_reentrant=False is recommended, but if you need to preserve the current default behavior, you can pass use_reentrant=True. Refer to docs for more details on the differences between the two variants.
+  return fn(*args, **kwargs)
+C:\Users\caleb\deception-nanochat-sae-research\experiments\v8_nla_local\train_prior_deviation_av.py:325: UserWarning: Converting a tensor with requires_grad=True to a scalar may lead to unexpected behavior.
+Consider using tensor.detach() first. (Triggered internally at C:\actions-runner\_work\pytorch\pytorch\pytorch\torch\csrc\autograd\generated\python_variable_methods.cpp:837.)
+  (loss / ga).backward(); tot += float(loss) / ga
+[pd] step 1/600 wloss=3.759 skipped=0
+[pd] step 10/600 wloss=4.033 skipped=0
+[pd] step 20/600 wloss=3.353 skipped=0
+[pd] step 30/600 wloss=3.700 skipped=0
+[pd] step 40/600 wloss=3.672 skipped=0
+[pd] step 50/600 wloss=3.679 skipped=0
+[pd] step 60/600 wloss=4.049 skipped=0
+[pd] step 70/600 wloss=3.437 skipped=0
+[pd] step 80/600 wloss=3.949 skipped=0
+[pd] step 90/600 wloss=3.907 skipped=0
+[pd] step 100/600 wloss=3.877 skipped=0
+C:\Users\caleb\deception-nanochat-sae-research\.venv-gemma4\Lib\site-packages\peft\utils\save_and_load.py:372: UserWarning: Could not find a config file in google/gemma-4-E2B - will assume that the vocabulary was not modified.
+  warnings.warn(
+[pd] saved experiments/v8_nla_local/checkpoints/av_L23_clean_cont\step_000100
+[pd] step 110/600 wloss=3.954 skipped=0
+[pd] step 120/600 wloss=3.498 skipped=0

armA_cap48_train.log

@@ -0,0 +1,128 @@
+
+[pd] FRESH LoRA r=8 alpha=16 (low-rank grokking substrate)
+[pd] no 'prior_weights' column -> forcing --uniform (plain CE weighting)
+[pd] data=experiments/v8_nla_local/data/stage3_balanced/av_sft_balanced.parquet rows=1356 uniform=True gamma=1.0 corpus_mean_weight=1.000
+[pd] DOMAIN-AWARE negs: avg same-dom=16.0 diff-dom=16.0 per anchor (15 domains)
+[pd] contrastive ON: beta=1.0 K=2 temp=1.0 domain_aware=True token_dropout=0.0 (InfoNCE over hard-neg activations)
+[pd] inject_layer=embed inject_mode=raw
+[pd] optimizer=adamw8bit lr=5e-05 wd=0.01 steps=1000 ga=1 trainable_params=12.08M
+`use_cache=True` is incompatible with gradient checkpointing. Setting `use_cache=False`.
+C:\Users\caleb\deception-nanochat-sae-research\.venv-gemma4\Lib\site-packages\torch\_dynamo\eval_frame.py:1181: UserWarning: torch.utils.checkpoint: the use_reentrant parameter should be passed explicitly. Starting in PyTorch 2.9, calling checkpoint without use_reentrant will raise an exception. use_reentrant=False is recommended, but if you need to preserve the current default behavior, you can pass use_reentrant=True. Refer to docs for more details on the differences between the two variants.
+  return fn(*args, **kwargs)
+C:\Users\caleb\deception-nanochat-sae-research\experiments\v8_nla_local\train_prior_deviation_av.py:325: UserWarning: Converting a tensor with requires_grad=True to a scalar may lead to unexpected behavior.
+Consider using tensor.detach() first. (Triggered internally at C:\actions-runner\_work\pytorch\pytorch\pytorch\torch\csrc\autograd\generated\python_variable_methods.cpp:837.)
+  (loss / ga).backward(); tot += float(loss) / ga
+[pd] step 1/1000 wloss=5.296 skipped=0
+[pd] step 10/1000 wloss=3.427 skipped=0
+[pd] step 20/1000 wloss=5.260 skipped=0
+[pd] step 30/1000 wloss=3.513 skipped=0
+[pd] step 40/1000 wloss=4.056 skipped=0
+[pd] step 50/1000 wloss=5.176 skipped=0
+[pd] step 60/1000 wloss=3.760 skipped=0
+[pd] step 70/1000 wloss=4.095 skipped=0
+[pd] step 80/1000 wloss=4.153 skipped=0
+[pd] step 90/1000 wloss=4.822 skipped=0
+[pd] step 100/1000 wloss=4.596 skipped=0
+C:\Users\caleb\deception-nanochat-sae-research\.venv-gemma4\Lib\site-packages\peft\utils\save_and_load.py:372: UserWarning: Could not find a config file in google/gemma-4-E2B - will assume that the vocabulary was not modified.
+  warnings.warn(
+[pd] saved experiments/v8_nla_local/checkpoints/av_armA_cap48\step_000100
+[pd] step 110/1000 wloss=4.109 skipped=0
+[pd] step 120/1000 wloss=4.820 skipped=0
+[pd] step 130/1000 wloss=4.181 skipped=0
+[pd] step 140/1000 wloss=4.007 skipped=0
+[pd] step 150/1000 wloss=3.483 skipped=0
+[pd] step 160/1000 wloss=3.168 skipped=0
+[pd] step 170/1000 wloss=5.035 skipped=0
+[pd] step 180/1000 wloss=3.471 skipped=0
+[pd] step 190/1000 wloss=4.008 skipped=0
+[pd] step 200/1000 wloss=4.125 skipped=0
+[pd] saved experiments/v8_nla_local/checkpoints/av_armA_cap48\step_000200
+[pd] step 210/1000 wloss=5.023 skipped=0
+[pd] step 220/1000 wloss=3.749 skipped=0
+[pd] step 230/1000 wloss=3.871 skipped=0
+[pd] step 240/1000 wloss=4.048 skipped=0
+[pd] step 250/1000 wloss=3.644 skipped=0
+[pd] step 260/1000 wloss=4.038 skipped=0
+[pd] step 270/1000 wloss=5.228 skipped=0
+[pd] step 280/1000 wloss=3.303 skipped=0
+[pd] step 290/1000 wloss=3.982 skipped=0
+[pd] step 300/1000 wloss=4.538 skipped=0
+[pd] saved experiments/v8_nla_local/checkpoints/av_armA_cap48\step_000300
+[pd] step 310/1000 wloss=4.519 skipped=0
+[pd] step 320/1000 wloss=4.047 skipped=0
+[pd] step 330/1000 wloss=4.266 skipped=0
+[pd] step 340/1000 wloss=4.094 skipped=0
+[pd] step 350/1000 wloss=5.378 skipped=0
+[pd] step 360/1000 wloss=3.275 skipped=0
+[pd] step 370/1000 wloss=3.626 skipped=0
+[pd] step 380/1000 wloss=4.041 skipped=0
+[pd] step 390/1000 wloss=4.071 skipped=0
+[pd] step 400/1000 wloss=3.845 skipped=0
+[pd] saved experiments/v8_nla_local/checkpoints/av_armA_cap48\step_000400
+[pd] step 410/1000 wloss=3.786 skipped=0
+[pd] step 420/1000 wloss=4.813 skipped=0
+[pd] step 430/1000 wloss=4.985 skipped=0
+[pd] step 440/1000 wloss=3.501 skipped=0
+[pd] step 450/1000 wloss=3.798 skipped=0
+[pd] step 460/1000 wloss=3.776 skipped=0
+[pd] step 470/1000 wloss=3.823 skipped=0
+[pd] step 480/1000 wloss=3.904 skipped=0
+[pd] step 490/1000 wloss=4.038 skipped=0
+[pd] step 500/1000 wloss=3.768 skipped=0
+[pd] saved experiments/v8_nla_local/checkpoints/av_armA_cap48\step_000500
+[pd] step 510/1000 wloss=3.520 skipped=0
+[pd] step 520/1000 wloss=4.091 skipped=0
+[pd] step 530/1000 wloss=4.188 skipped=0
+[pd] step 540/1000 wloss=4.551 skipped=0
+[pd] step 550/1000 wloss=3.901 skipped=0
+[pd] step 560/1000 wloss=3.512 skipped=0
+[pd] step 570/1000 wloss=4.055 skipped=0
+[pd] step 580/1000 wloss=4.164 skipped=0
+[pd] step 590/1000 wloss=4.074 skipped=0
+[pd] step 600/1000 wloss=3.452 skipped=0
+[pd] saved experiments/v8_nla_local/checkpoints/av_armA_cap48\step_000600
+[pd] step 610/1000 wloss=4.213 skipped=0
+[pd] step 620/1000 wloss=3.845 skipped=0
+[pd] step 630/1000 wloss=3.319 skipped=0
+[pd] step 640/1000 wloss=4.485 skipped=0
+[pd] step 650/1000 wloss=3.909 skipped=0
+[pd] step 660/1000 wloss=4.883 skipped=0
+[pd] step 670/1000 wloss=3.194 skipped=0
+[pd] step 680/1000 wloss=3.157 skipped=0
+[pd] step 690/1000 wloss=4.900 skipped=0
+[pd] step 700/1000 wloss=3.627 skipped=0
+[pd] saved experiments/v8_nla_local/checkpoints/av_armA_cap48\step_000700
+[pd] step 710/1000 wloss=3.683 skipped=0
+[pd] step 720/1000 wloss=2.483 skipped=0
+[pd] step 730/1000 wloss=4.442 skipped=0
+[pd] step 740/1000 wloss=4.109 skipped=0
+[pd] step 750/1000 wloss=3.617 skipped=0
+[pd] step 760/1000 wloss=4.262 skipped=0
+[pd] step 770/1000 wloss=2.797 skipped=0
+[pd] step 780/1000 wloss=3.787 skipped=0
+[pd] step 790/1000 wloss=4.527 skipped=0
+[pd] step 800/1000 wloss=3.707 skipped=0
+[pd] saved experiments/v8_nla_local/checkpoints/av_armA_cap48\step_000800
+[pd] step 810/1000 wloss=3.902 skipped=0
+[pd] step 820/1000 wloss=4.419 skipped=0
+[pd] step 830/1000 wloss=3.215 skipped=0
+[pd] step 840/1000 wloss=3.543 skipped=0
+[pd] step 850/1000 wloss=3.650 skipped=0
+[pd] step 860/1000 wloss=3.488 skipped=0
+[pd] step 870/1000 wloss=3.776 skipped=0
+[pd] step 880/1000 wloss=4.105 skipped=0
+[pd] step 890/1000 wloss=3.936 skipped=0
+[pd] step 900/1000 wloss=3.115 skipped=0
+[pd] saved experiments/v8_nla_local/checkpoints/av_armA_cap48\step_000900
+[pd] step 910/1000 wloss=5.334 skipped=0
+[pd] step 920/1000 wloss=3.406 skipped=0
+[pd] step 930/1000 wloss=3.368 skipped=0
+[pd] step 940/1000 wloss=4.163 skipped=0
+[pd] step 950/1000 wloss=4.276 skipped=0
+[pd] step 960/1000 wloss=4.061 skipped=0
+[pd] step 970/1000 wloss=2.833 skipped=0
+[pd] step 980/1000 wloss=4.018 skipped=0
+[pd] step 990/1000 wloss=3.157 skipped=0
+[pd] step 1000/1000 wloss=3.950 skipped=0
+[pd] saved experiments/v8_nla_local/checkpoints/av_armA_cap48\step_001000
+[pd] DONE

av_armA_cap48/inject_config.json

@@ -0,0 +1 @@
+{"inject_layer": "embed", "inject_mode": "raw"}

av_armA_cap48/step_000100/README.md

@@ -0,0 +1,207 @@
+---
+base_model: google/gemma-4-E2B
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:google/gemma-4-E2B
+- 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 0.19.1

av_armA_cap48/step_000100/adapter_config.json

@@ -0,0 +1,40 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "google/gemma-4-E2B",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 16,
+  "lora_bias": false,
+  "lora_dropout": 0.0,
+  "lora_ga_config": null,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.19.1",
+  "qalora_group_size": 16,
+  "r": 8,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": "model\\.language_model\\.layers\\.\\d+\\.(self_attn|mlp)\\.(q_proj|k_proj|v_proj|o_proj|gate_proj|up_proj|down_proj)",
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_bdlora": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}

av_armA_cap48/step_000100/nla_meta.yaml

@@ -0,0 +1,5 @@
+training:
+  inject_layer: -1
+  injection_scale: 39.191835884530846
+  method: prior_deviation_reweight
+  uniform: true

av_armA_cap48/step_000200/README.md

@@ -0,0 +1,207 @@
+---
+base_model: google/gemma-4-E2B
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:google/gemma-4-E2B
+- 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 0.19.1

av_armA_cap48/step_000200/adapter_config.json

@@ -0,0 +1,40 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "google/gemma-4-E2B",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 16,
+  "lora_bias": false,
+  "lora_dropout": 0.0,
+  "lora_ga_config": null,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.19.1",
+  "qalora_group_size": 16,
+  "r": 8,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": "model\\.language_model\\.layers\\.\\d+\\.(self_attn|mlp)\\.(q_proj|k_proj|v_proj|o_proj|gate_proj|up_proj|down_proj)",
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_bdlora": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}

av_armA_cap48/step_000200/nla_meta.yaml

@@ -0,0 +1,5 @@
+training:
+  inject_layer: -1
+  injection_scale: 39.191835884530846
+  method: prior_deviation_reweight
+  uniform: true

av_armA_cap48/step_000300/README.md

@@ -0,0 +1,207 @@
+---
+base_model: google/gemma-4-E2B
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:google/gemma-4-E2B
+- 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 0.19.1

av_armA_cap48/step_000300/adapter_config.json

@@ -0,0 +1,40 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "google/gemma-4-E2B",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 16,
+  "lora_bias": false,
+  "lora_dropout": 0.0,
+  "lora_ga_config": null,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.19.1",
+  "qalora_group_size": 16,
+  "r": 8,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": "model\\.language_model\\.layers\\.\\d+\\.(self_attn|mlp)\\.(q_proj|k_proj|v_proj|o_proj|gate_proj|up_proj|down_proj)",
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_bdlora": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}

av_armA_cap48/step_000300/nla_meta.yaml

@@ -0,0 +1,5 @@
+training:
+  inject_layer: -1
+  injection_scale: 39.191835884530846
+  method: prior_deviation_reweight
+  uniform: true

av_armA_cap48/step_000400/README.md

@@ -0,0 +1,207 @@
+---
+base_model: google/gemma-4-E2B
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:google/gemma-4-E2B
+- 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 0.19.1

av_armA_cap48/step_000400/adapter_config.json

@@ -0,0 +1,40 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "google/gemma-4-E2B",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 16,
+  "lora_bias": false,
+  "lora_dropout": 0.0,
+  "lora_ga_config": null,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.19.1",
+  "qalora_group_size": 16,
+  "r": 8,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": "model\\.language_model\\.layers\\.\\d+\\.(self_attn|mlp)\\.(q_proj|k_proj|v_proj|o_proj|gate_proj|up_proj|down_proj)",
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_bdlora": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}

av_armA_cap48/step_000400/nla_meta.yaml

@@ -0,0 +1,5 @@
+training:
+  inject_layer: -1
+  injection_scale: 39.191835884530846
+  method: prior_deviation_reweight
+  uniform: true

av_armA_cap48/step_000500/README.md

@@ -0,0 +1,207 @@
+---
+base_model: google/gemma-4-E2B
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:google/gemma-4-E2B
+- 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 0.19.1

av_armA_cap48/step_000500/adapter_config.json

@@ -0,0 +1,40 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "google/gemma-4-E2B",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 16,
+  "lora_bias": false,
+  "lora_dropout": 0.0,
+  "lora_ga_config": null,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.19.1",
+  "qalora_group_size": 16,
+  "r": 8,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": "model\\.language_model\\.layers\\.\\d+\\.(self_attn|mlp)\\.(q_proj|k_proj|v_proj|o_proj|gate_proj|up_proj|down_proj)",
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_bdlora": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}

av_armA_cap48/step_000500/nla_meta.yaml

@@ -0,0 +1,5 @@
+training:
+  inject_layer: -1
+  injection_scale: 39.191835884530846
+  method: prior_deviation_reweight
+  uniform: true

av_armA_cap48/step_000600/README.md

@@ -0,0 +1,207 @@
+---
+base_model: google/gemma-4-E2B
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:google/gemma-4-E2B
+- 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 0.19.1

av_armA_cap48/step_000600/adapter_config.json

@@ -0,0 +1,40 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "google/gemma-4-E2B",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 16,
+  "lora_bias": false,
+  "lora_dropout": 0.0,
+  "lora_ga_config": null,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.19.1",
+  "qalora_group_size": 16,
+  "r": 8,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": "model\\.language_model\\.layers\\.\\d+\\.(self_attn|mlp)\\.(q_proj|k_proj|v_proj|o_proj|gate_proj|up_proj|down_proj)",
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_bdlora": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}

av_armA_cap48/step_000600/nla_meta.yaml

@@ -0,0 +1,5 @@
+training:
+  inject_layer: -1
+  injection_scale: 39.191835884530846
+  method: prior_deviation_reweight
+  uniform: true

av_armA_cap48/step_000700/README.md

@@ -0,0 +1,207 @@
+---
+base_model: google/gemma-4-E2B
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:google/gemma-4-E2B
+- 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 0.19.1

av_armA_cap48/step_000700/adapter_config.json

@@ -0,0 +1,40 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "google/gemma-4-E2B",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 16,
+  "lora_bias": false,
+  "lora_dropout": 0.0,
+  "lora_ga_config": null,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.19.1",
+  "qalora_group_size": 16,
+  "r": 8,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": "model\\.language_model\\.layers\\.\\d+\\.(self_attn|mlp)\\.(q_proj|k_proj|v_proj|o_proj|gate_proj|up_proj|down_proj)",
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_bdlora": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}

av_armA_cap48/step_000700/nla_meta.yaml

@@ -0,0 +1,5 @@
+training:
+  inject_layer: -1
+  injection_scale: 39.191835884530846
+  method: prior_deviation_reweight
+  uniform: true

av_armA_cap48/step_000800/README.md

@@ -0,0 +1,207 @@
+---
+base_model: google/gemma-4-E2B
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:google/gemma-4-E2B
+- 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 0.19.1

av_armA_cap48/step_000800/adapter_config.json

@@ -0,0 +1,40 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "google/gemma-4-E2B",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 16,
+  "lora_bias": false,
+  "lora_dropout": 0.0,
+  "lora_ga_config": null,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.19.1",
+  "qalora_group_size": 16,
+  "r": 8,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": "model\\.language_model\\.layers\\.\\d+\\.(self_attn|mlp)\\.(q_proj|k_proj|v_proj|o_proj|gate_proj|up_proj|down_proj)",
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_bdlora": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}

av_armA_cap48/step_000800/nla_meta.yaml

@@ -0,0 +1,5 @@
+training:
+  inject_layer: -1
+  injection_scale: 39.191835884530846
+  method: prior_deviation_reweight
+  uniform: true

av_armA_cap48/step_000900/README.md

@@ -0,0 +1,207 @@
+---
+base_model: google/gemma-4-E2B
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:google/gemma-4-E2B
+- 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 0.19.1

av_armA_cap48/step_000900/adapter_config.json

@@ -0,0 +1,40 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "google/gemma-4-E2B",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 16,
+  "lora_bias": false,
+  "lora_dropout": 0.0,
+  "lora_ga_config": null,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.19.1",
+  "qalora_group_size": 16,
+  "r": 8,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": "model\\.language_model\\.layers\\.\\d+\\.(self_attn|mlp)\\.(q_proj|k_proj|v_proj|o_proj|gate_proj|up_proj|down_proj)",
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_bdlora": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}

av_armA_cap48/step_000900/nla_meta.yaml

@@ -0,0 +1,5 @@
+training:
+  inject_layer: -1
+  injection_scale: 39.191835884530846
+  method: prior_deviation_reweight
+  uniform: true

av_armA_cap48/step_001000/README.md

@@ -0,0 +1,207 @@
+---
+base_model: google/gemma-4-E2B
+library_name: peft
+pipeline_tag: text-generation
+tags:
+- base_model:adapter:google/gemma-4-E2B
+- 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 0.19.1

av_armA_cap48/step_001000/adapter_config.json

@@ -0,0 +1,40 @@
+{
+  "alora_invocation_tokens": null,
+  "alpha_pattern": {},
+  "arrow_config": null,
+  "auto_mapping": null,
+  "base_model_name_or_path": "google/gemma-4-E2B",
+  "bias": "none",
+  "corda_config": null,
+  "ensure_weight_tying": false,
+  "eva_config": null,
+  "exclude_modules": null,
+  "fan_in_fan_out": false,
+  "inference_mode": true,
+  "init_lora_weights": true,
+  "layer_replication": null,
+  "layers_pattern": null,
+  "layers_to_transform": null,
+  "loftq_config": {},
+  "lora_alpha": 16,
+  "lora_bias": false,
+  "lora_dropout": 0.0,
+  "lora_ga_config": null,
+  "megatron_config": null,
+  "megatron_core": "megatron.core",
+  "modules_to_save": null,
+  "peft_type": "LORA",
+  "peft_version": "0.19.1",
+  "qalora_group_size": 16,
+  "r": 8,
+  "rank_pattern": {},
+  "revision": null,
+  "target_modules": "model\\.language_model\\.layers\\.\\d+\\.(self_attn|mlp)\\.(q_proj|k_proj|v_proj|o_proj|gate_proj|up_proj|down_proj)",
+  "target_parameters": null,
+  "task_type": "CAUSAL_LM",
+  "trainable_token_indices": null,
+  "use_bdlora": null,
+  "use_dora": false,
+  "use_qalora": false,
+  "use_rslora": false
+}

av_armA_cap48/step_001000/nla_meta.yaml

@@ -0,0 +1,5 @@
+training:
+  inject_layer: -1
+  injection_scale: 39.191835884530846
+  method: prior_deviation_reweight
+  uniform: true

av_armA_cap48/trainloss.csv

@@ -0,0 +1,1001 @@
+step,loss,wloss
+1,5.296004295349121,5.296004295349121
+2,5.5213236808776855,5.5213236808776855
+3,5.014024257659912,5.014024257659912
+4,5.673574447631836,5.673574447631836
+5,5.9487762451171875,5.9487762451171875
+6,6.214047431945801,6.214047431945801
+7,4.31553840637207,4.31553840637207
+8,5.7412261962890625,5.7412261962890625
+9,4.783784866333008,4.783784866333008
+10,3.4273438453674316,3.4273438453674316
+11,5.546191692352295,5.546191692352295
+12,6.632611274719238,6.632611274719238
+13,4.491706848144531,4.491706848144531
+14,5.318314075469971,5.318314075469971
+15,5.207615852355957,5.207615852355957
+16,5.30378532409668,5.30378532409668
+17,3.456696033477783,3.456696033477783
+18,4.445058822631836,4.445058822631836
+19,5.081432342529297,5.081432342529297
+20,5.259688377380371,5.259688377380371
+21,5.3198628425598145,5.3198628425598145
+22,4.398024082183838,4.398024082183838
+23,4.776388168334961,4.776388168334961
+24,6.263848781585693,6.263848781585693
+25,4.434366226196289,4.434366226196289
+26,3.751006841659546,3.751006841659546
+27,4.278125762939453,4.278125762939453
+28,4.773993492126465,4.773993492126465
+29,4.837551116943359,4.837551116943359
+30,3.512617349624634,3.512617349624634
+31,3.895080804824829,3.895080804824829
+32,4.365486145019531,4.365486145019531
+33,3.4751172065734863,3.4751172065734863
+34,3.3896474838256836,3.3896474838256836
+35,4.4288225173950195,4.4288225173950195
+36,4.866550445556641,4.866550445556641
+37,5.012956619262695,5.012956619262695
+38,4.407891273498535,4.407891273498535
+39,5.086009979248047,5.086009979248047
+40,4.0563483238220215,4.0563483238220215
+41,4.229613304138184,4.229613304138184
+42,3.73225736618042,3.73225736618042
+43,3.702065944671631,3.702065944671631
+44,5.280701160430908,5.280701160430908
+45,2.4715704917907715,2.4715704917907715
+46,5.092109203338623,5.092109203338623
+47,4.036098480224609,4.036098480224609
+48,3.849869728088379,3.849869728088379
+49,3.5987508296966553,3.5987508296966553
+50,5.176441192626953,5.176441192626953
+51,3.858501434326172,3.858501434326172
+52,4.138750076293945,4.138750076293945
+53,5.468461036682129,5.468461036682129
+54,4.748278617858887,4.748278617858887
+55,4.297647476196289,4.297647476196289
+56,5.257157325744629,5.257157325744629
+57,4.566764831542969,4.566764831542969
+58,4.029827117919922,4.029827117919922
+59,3.3752312660217285,3.3752312660217285
+60,3.7598929405212402,3.7598929405212402
+61,3.8896191120147705,3.8896191120147705
+62,3.5067367553710938,3.5067367553710938
+63,5.3760881423950195,5.3760881423950195
+64,5.122551441192627,5.122551441192627
+65,4.39186429977417,4.39186429977417
+66,4.728086471557617,4.728086471557617
+67,4.260268211364746,4.260268211364746
+68,3.8103973865509033,3.8103973865509033
+69,4.943123817443848,4.943123817443848
+70,4.095166206359863,4.095166206359863
+71,4.266524314880371,4.266524314880371
+72,3.7013111114501953,3.7013111114501953
+73,4.498545169830322,4.498545169830322
+74,3.185304641723633,3.185304641723633
+75,3.0398993492126465,3.0398993492126465
+76,4.294178009033203,4.294178009033203
+77,4.703925132751465,4.703925132751465
+78,4.823328018188477,4.823328018188477
+79,3.6390786170959473,3.6390786170959473
+80,4.153303623199463,4.153303623199463
+81,4.868573188781738,4.868573188781738
+82,4.588263988494873,4.588263988494873
+83,5.183469772338867,5.183469772338867
+84,6.184668064117432,6.184668064117432
+85,5.530581951141357,5.530581951141357
+86,4.0161614418029785,4.0161614418029785
+87,3.6653897762298584,3.6653897762298584
+88,4.134860992431641,4.134860992431641
+89,3.257398843765259,3.257398843765259
+90,4.821633815765381,4.821633815765381
+91,4.09312629699707,4.09312629699707
+92,3.5656886100769043,3.5656886100769043
+93,4.885313510894775,4.885313510894775
+94,3.8421642780303955,3.8421642780303955
+95,4.58418607711792,4.58418607711792
+96,4.346695423126221,4.346695423126221
+97,4.016838550567627,4.016838550567627
+98,4.819882392883301,4.819882392883301
+99,3.6513264179229736,3.6513264179229736
+100,4.595682144165039,4.595682144165039
+101,2.9326348304748535,2.9326348304748535
+102,5.082942962646484,5.082942962646484
+103,3.1057112216949463,3.1057112216949463
+104,4.530087471008301,4.530087471008301
+105,6.4723029136657715,6.4723029136657715
+106,4.092505931854248,4.092505931854248
+107,4.329499244689941,4.329499244689941
+108,5.876839637756348,5.876839637756348
+109,3.6580634117126465,3.6580634117126465
+110,4.10902214050293,4.10902214050293
+111,5.015631675720215,5.015631675720215
+112,2.4739058017730713,2.4739058017730713
+113,3.6965978145599365,3.6965978145599365
+114,4.153364658355713,4.153364658355713
+115,4.111626625061035,4.111626625061035
+116,4.631809711456299,4.631809711456299
+117,3.400895595550537,3.400895595550537
+118,4.159792900085449,4.159792900085449
+119,5.295380592346191,5.295380592346191
+120,4.820404529571533,4.820404529571533
+121,4.167803764343262,4.167803764343262
+122,3.937164783477783,3.937164783477783
+123,3.5707077980041504,3.5707077980041504
+124,4.03619384765625,4.03619384765625
+125,5.038243293762207,5.038243293762207
+126,3.963749408721924,3.963749408721924
+127,3.978001356124878,3.978001356124878
+128,3.000669002532959,3.000669002532959
+129,4.700469017028809,4.700469017028809
+130,4.181060314178467,4.181060314178467
+131,4.012627124786377,4.012627124786377
+132,2.904360055923462,2.904360055923462
+133,3.5880770683288574,3.5880770683288574
+134,5.243063926696777,5.243063926696777
+135,3.3378958702087402,3.3378958702087402
+136,4.9328107833862305,4.9328107833862305
+137,4.522773265838623,4.522773265838623
+138,3.6340575218200684,3.6340575218200684
+139,4.830964088439941,4.830964088439941
+140,4.006507396697998,4.006507396697998
+141,3.203706979751587,3.203706979751587
+142,3.2888622283935547,3.2888622283935547
+143,3.9469666481018066,3.9469666481018066
+144,3.9119210243225098,3.9119210243225098
+145,3.265801429748535,3.265801429748535
+146,4.458025932312012,4.458025932312012
+147,3.5973453521728516,3.5973453521728516
+148,3.6384897232055664,3.6384897232055664
+149,4.508347988128662,4.508347988128662
+150,3.483396053314209,3.483396053314209
+151,3.3123321533203125,3.3123321533203125
+152,5.004119396209717,5.004119396209717
+153,4.793652534484863,4.793652534484863
+154,3.6220271587371826,3.6220271587371826
+155,4.089190483093262,4.089190483093262
+156,5.04091215133667,5.04091215133667
+157,5.500209808349609,5.500209808349609
+158,3.3507611751556396,3.3507611751556396
+159,4.5340094566345215,4.5340094566345215
+160,3.167692184448242,3.167692184448242
+161,2.573030710220337,2.573030710220337
+162,3.996720790863037,3.996720790863037
+163,3.983527898788452,3.983527898788452
+164,3.7205352783203125,3.7205352783203125
+165,4.672206401824951,4.672206401824951
+166,4.121712684631348,4.121712684631348
+167,4.988698482513428,4.988698482513428
+168,4.033992767333984,4.033992767333984
+169,3.333183765411377,3.333183765411377
+170,5.034782886505127,5.034782886505127
+171,4.139377593994141,4.139377593994141
+172,4.977420806884766,4.977420806884766
+173,5.904860496520996,5.904860496520996
+174,5.013150691986084,5.013150691986084
+175,2.8055014610290527,2.8055014610290527
+176,3.2450003623962402,3.2450003623962402
+177,2.9632835388183594,2.9632835388183594
+178,4.163037300109863,4.163037300109863
+179,3.0923073291778564,3.0923073291778564
+180,3.4711790084838867,3.4711790084838867
+181,3.835007905960083,3.835007905960083
+182,4.509566307067871,4.509566307067871
+183,4.142751693725586,4.142751693725586
+184,5.012729644775391,5.012729644775391
+185,3.3108701705932617,3.3108701705932617
+186,4.799376487731934,4.799376487731934
+187,4.896233558654785,4.896233558654785
+188,3.039642810821533,3.039642810821533
+189,4.139034271240234,4.139034271240234
+190,4.008388996124268,4.008388996124268
+191,3.12363338470459,3.12363338470459
+192,3.691798686981201,3.691798686981201
+193,3.781174898147583,3.781174898147583
+194,4.874512672424316,4.874512672424316
+195,4.81962776184082,4.81962776184082
+196,3.764021635055542,3.764021635055542
+197,3.753082513809204,3.753082513809204
+198,4.393030166625977,4.393030166625977
+199,3.464064121246338,3.464064121246338
+200,4.124950408935547,4.124950408935547
+201,4.522347927093506,4.522347927093506
+202,5.841829299926758,5.841829299926758
+203,4.678520202636719,4.678520202636719
+204,3.5431323051452637,3.5431323051452637
+205,4.3201446533203125,4.3201446533203125
+206,3.9650866985321045,3.9650866985321045
+207,3.828533887863159,3.828533887863159
+208,3.641645669937134,3.641645669937134
+209,4.114406585693359,4.114406585693359
+210,5.0231218338012695,5.0231218338012695
+211,2.96262788772583,2.96262788772583
+212,4.1919708251953125,4.1919708251953125
+213,2.9616284370422363,2.9616284370422363
+214,4.089715480804443,4.089715480804443
+215,4.198691368103027,4.198691368103027
+216,4.823454856872559,4.823454856872559
+217,3.648120880126953,3.648120880126953
+218,3.7885146141052246,3.7885146141052246
+219,4.5077619552612305,4.5077619552612305
+220,3.748784065246582,3.748784065246582
+221,4.546048164367676,4.546048164367676
+222,4.255072116851807,4.255072116851807
+223,4.878353118896484,4.878353118896484
+224,3.674429416656494,3.674429416656494
+225,3.733165740966797,3.733165740966797
+226,4.706740379333496,4.706740379333496
+227,4.284913539886475,4.284913539886475
+228,3.381633996963501,3.381633996963501
+229,4.307045936584473,4.307045936584473
+230,3.8706068992614746,3.8706068992614746
+231,3.717409133911133,3.717409133911133
+232,4.621013164520264,4.621013164520264
+233,3.3185391426086426,3.3185391426086426
+234,4.5555925369262695,4.5555925369262695
+235,3.686887741088867,3.686887741088867
+236,4.117617607116699,4.117617607116699
+237,3.375988245010376,3.375988245010376
+238,3.4912493228912354,3.4912493228912354
+239,3.1693568229675293,3.1693568229675293
+240,4.04750394821167,4.04750394821167
+241,4.891644477844238,4.891644477844238
+242,3.339695692062378,3.339695692062378
+243,4.282183647155762,4.282183647155762
+244,3.253988742828369,3.253988742828369
+245,4.249805450439453,4.249805450439453
+246,4.668442726135254,4.668442726135254
+247,4.32988166809082,4.32988166809082
+248,5.554760456085205,5.554760456085205
+249,4.039490222930908,4.039490222930908
+250,3.643634796142578,3.643634796142578
+251,3.8937621116638184,3.8937621116638184
+252,2.994588851928711,2.994588851928711
+253,3.685185432434082,3.685185432434082
+254,4.501359939575195,4.501359939575195
+255,3.4073026180267334,3.4073026180267334
+256,2.76986026763916,2.76986026763916
+257,4.130134105682373,4.130134105682373
+258,5.710204124450684,5.710204124450684
+259,4.245023250579834,4.245023250579834
+260,4.038324356079102,4.038324356079102
+261,4.68991756439209,4.68991756439209
+262,3.6205379962921143,3.6205379962921143
+263,3.9998555183410645,3.9998555183410645
+264,3.544651746749878,3.544651746749878
+265,4.736502170562744,4.736502170562744
+266,4.4998297691345215,4.4998297691345215
+267,4.720915794372559,4.720915794372559
+268,3.6315500736236572,3.6315500736236572
+269,3.8949904441833496,3.8949904441833496
+270,5.22836446762085,5.22836446762085
+271,4.043702125549316,4.043702125549316
+272,2.9482855796813965,2.9482855796813965
+273,4.262182712554932,4.262182712554932
+274,4.336435794830322,4.336435794830322
+275,5.84244441986084,5.84244441986084
+276,4.802825927734375,4.802825927734375
+277,4.356124401092529,4.356124401092529
+278,4.016505241394043,4.016505241394043
+279,4.4887895584106445,4.4887895584106445
+280,3.3032240867614746,3.3032240867614746
+281,4.900825023651123,4.900825023651123
+282,5.187237739562988,5.187237739562988
+283,4.149139881134033,4.149139881134033
+284,3.6773829460144043,3.6773829460144043
+285,3.058602809906006,3.058602809906006
+286,3.980268716812134,3.980268716812134
+287,4.473905563354492,4.473905563354492
+288,4.437484264373779,4.437484264373779
+289,4.152819633483887,4.152819633483887
+290,3.982318878173828,3.982318878173828
+291,3.037151336669922,3.037151336669922
+292,4.102868556976318,4.102868556976318
+293,4.129270076751709,4.129270076751709
+294,3.2364487648010254,3.2364487648010254
+295,4.0257110595703125,4.0257110595703125
+296,3.7152857780456543,3.7152857780456543
+297,3.836280345916748,3.836280345916748
+298,2.8365023136138916,2.8365023136138916
+299,4.200503826141357,4.200503826141357
+300,4.538206100463867,4.538206100463867
+301,5.015071868896484,5.015071868896484
+302,3.0367746353149414,3.0367746353149414
+303,3.211764335632324,3.211764335632324
+304,3.8378376960754395,3.8378376960754395
+305,2.9051806926727295,2.9051806926727295
+306,4.472991943359375,4.472991943359375
+307,3.4836530685424805,3.4836530685424805
+308,4.746848106384277,4.746848106384277
+309,5.172412872314453,5.172412872314453
+310,4.519374370574951,4.519374370574951
+311,3.8944759368896484,3.8944759368896484
+312,4.34901762008667,4.34901762008667
+313,3.759439468383789,3.759439468383789
+314,5.083188056945801,5.083188056945801
+315,4.375423431396484,4.375423431396484
+316,3.876955270767212,3.876955270767212
+317,3.2101402282714844,3.2101402282714844
+318,4.353207111358643,4.353207111358643
+319,4.464028358459473,4.464028358459473
+320,4.047225475311279,4.047225475311279
+321,4.494183540344238,4.494183540344238
+322,3.7745718955993652,3.7745718955993652
+323,4.405634880065918,4.405634880065918
+324,3.466409921646118,3.466409921646118
+325,2.8617329597473145,2.8617329597473145
+326,4.281109809875488,4.281109809875488
+327,6.124507427215576,6.124507427215576
+328,4.069917678833008,4.069917678833008
+329,4.075202941894531,4.075202941894531
+330,4.265852451324463,4.265852451324463
+331,3.737332820892334,3.737332820892334
+332,3.407212257385254,3.407212257385254
+333,3.307983875274658,3.307983875274658
+334,4.740633010864258,4.740633010864258
+335,3.425645351409912,3.425645351409912
+336,3.919896125793457,3.919896125793457
+337,4.493226051330566,4.493226051330566
+338,4.137999534606934,4.137999534606934
+339,2.987063407897949,2.987063407897949
+340,4.0937700271606445,4.0937700271606445
+341,3.3342766761779785,3.3342766761779785
+342,4.257776737213135,4.257776737213135
+343,3.330155849456787,3.330155849456787
+344,4.753819942474365,4.753819942474365
+345,3.561445951461792,3.561445951461792
+346,4.307785987854004,4.307785987854004
+347,5.2147722244262695,5.2147722244262695
+348,4.556590557098389,4.556590557098389
+349,4.162284851074219,4.162284851074219
+350,5.378108024597168,5.378108024597168
+351,3.002744197845459,3.002744197845459
+352,3.7476792335510254,3.7476792335510254
+353,3.913604736328125,3.913604736328125
+354,5.925439357757568,5.925439357757568
+355,3.8273534774780273,3.8273534774780273
+356,3.6101372241973877,3.6101372241973877
+357,3.797577142715454,3.797577142715454
+358,4.352535247802734,4.352535247802734
+359,3.8788061141967773,3.8788061141967773
+360,3.2745542526245117,3.2745542526245117
+361,3.5266122817993164,3.5266122817993164
+362,3.5616860389709473,3.5616860389709473
+363,4.401366233825684,4.401366233825684
+364,4.808916091918945,4.808916091918945
+365,4.406719207763672,4.406719207763672
+366,3.2468128204345703,3.2468128204345703
+367,3.3682260513305664,3.3682260513305664
+368,4.7316670417785645,4.7316670417785645
+369,3.096670627593994,3.096670627593994
+370,3.62589693069458,3.62589693069458
+371,2.9920740127563477,2.9920740127563477
+372,4.896712303161621,4.896712303161621
+373,4.258228302001953,4.258228302001953
+374,3.2673075199127197,3.2673075199127197
+375,3.6921133995056152,3.6921133995056152
+376,4.959468364715576,4.959468364715576
+377,2.3554344177246094,2.3554344177246094
+378,4.34783935546875,4.34783935546875
+379,4.078821182250977,4.078821182250977
+380,4.0406389236450195,4.0406389236450195
+381,4.837046146392822,4.837046146392822
+382,5.21112060546875,5.21112060546875
+383,4.211921215057373,4.211921215057373
+384,4.379604339599609,4.379604339599609
+385,4.971811294555664,4.971811294555664
+386,3.79964017868042,3.79964017868042
+387,3.029151201248169,3.029151201248169
+388,5.251279830932617,5.251279830932617
+389,4.220734596252441,4.220734596252441
+390,4.0706610679626465,4.0706610679626465
+391,4.858723163604736,4.858723163604736
+392,4.035857200622559,4.035857200622559
+393,3.250469446182251,3.250469446182251
+394,4.18866491317749,4.18866491317749
+395,4.719269275665283,4.719269275665283
+396,3.253251314163208,3.253251314163208
+397,3.443643808364868,3.443643808364868
+398,3.8387668132781982,3.8387668132781982
+399,3.895846366882324,3.895846366882324
+400,3.8449783325195312,3.8449783325195312
+401,3.048525094985962,3.048525094985962
+402,4.067508220672607,4.067508220672607
+403,4.231732368469238,4.231732368469238
+404,3.4513885974884033,3.4513885974884033
+405,3.8548126220703125,3.8548126220703125
+406,3.4161272048950195,3.4161272048950195
+407,4.531546592712402,4.531546592712402
+408,3.1873621940612793,3.1873621940612793
+409,4.076236248016357,4.076236248016357
+410,3.7857213020324707,3.7857213020324707
+411,3.785360336303711,3.785360336303711
+412,3.09625244140625,3.09625244140625
+413,3.798959970474243,3.798959970474243
+414,3.498900890350342,3.498900890350342
+415,4.221052646636963,4.221052646636963
+416,4.076447486877441,4.076447486877441
+417,4.4289398193359375,4.4289398193359375
+418,3.9125304222106934,3.9125304222106934
+419,4.658041954040527,4.658041954040527
+420,4.813041687011719,4.813041687011719
+421,3.694392204284668,3.694392204284668
+422,4.007391929626465,4.007391929626465
+423,4.716675758361816,4.716675758361816
+424,2.666572332382202,2.666572332382202
+425,5.148416519165039,5.148416519165039
+426,3.2755956649780273,3.2755956649780273
+427,5.0543365478515625,5.0543365478515625
+428,4.068397521972656,4.068397521972656
+429,4.521885395050049,4.521885395050049
+430,4.984527587890625,4.984527587890625
+431,3.4432878494262695,3.4432878494262695
+432,4.035134315490723,4.035134315490723
+433,3.8116230964660645,3.8116230964660645
+434,4.119228363037109,4.119228363037109
+435,4.026129245758057,4.026129245758057
+436,5.106755256652832,5.106755256652832
+437,4.375763893127441,4.375763893127441
+438,5.765435695648193,5.765435695648193
+439,3.7974328994750977,3.7974328994750977
+440,3.5011863708496094,3.5011863708496094
+441,3.3218865394592285,3.3218865394592285
+442,3.518278121948242,3.518278121948242
+443,4.2285003662109375,4.2285003662109375
+444,2.8568129539489746,2.8568129539489746
+445,3.9397177696228027,3.9397177696228027
+446,3.6722211837768555,3.6722211837768555
+447,3.3763277530670166,3.3763277530670166
+448,5.064760208129883,5.064760208129883
+449,3.352539539337158,3.352539539337158
+450,3.798429012298584,3.798429012298584
+451,3.0207319259643555,3.0207319259643555
+452,3.5996153354644775,3.5996153354644775
+453,4.411998271942139,4.411998271942139
+454,3.3528223037719727,3.3528223037719727
+455,4.338632106781006,4.338632106781006
+456,3.9140334129333496,3.9140334129333496
+457,3.704021692276001,3.704021692276001
+458,3.71673846244812,3.71673846244812
+459,3.0992956161499023,3.0992956161499023
+460,3.776252031326294,3.776252031326294
+461,3.4692952632904053,3.4692952632904053
+462,3.4414870738983154,3.4414870738983154
+463,3.7895278930664062,3.7895278930664062
+464,3.080716133117676,3.080716133117676
+465,3.5147409439086914,3.5147409439086914
+466,3.366959571838379,3.366959571838379
+467,3.9255478382110596,3.9255478382110596
+468,3.9852070808410645,3.9852070808410645
+469,4.375075340270996,4.375075340270996
+470,3.8229331970214844,3.8229331970214844
+471,3.949127197265625,3.949127197265625
+472,4.148682117462158,4.148682117462158
+473,3.5913400650024414,3.5913400650024414
+474,4.279911041259766,4.279911041259766
+475,3.961273670196533,3.961273670196533
+476,3.0624818801879883,3.0624818801879883
+477,4.202000617980957,4.202000617980957
+478,3.2204689979553223,3.2204689979553223
+479,4.344374656677246,4.344374656677246
+480,3.90434193611145,3.90434193611145
+481,3.9546782970428467,3.9546782970428467
+482,4.768157482147217,4.768157482147217
+483,3.8554129600524902,3.8554129600524902
+484,4.254450798034668,4.254450798034668
+485,4.7681074142456055,4.7681074142456055
+486,3.207174301147461,3.207174301147461
+487,4.469988822937012,4.469988822937012
+488,3.764763355255127,3.764763355255127
+489,3.973069667816162,3.973069667816162
+490,4.037725925445557,4.037725925445557
+491,3.048840045928955,3.048840045928955
+492,3.409640312194824,3.409640312194824
+493,4.538957595825195,4.538957595825195
+494,4.427155017852783,4.427155017852783
+495,2.988786458969116,2.988786458969116
+496,3.950157642364502,3.950157642364502
+497,4.17404842376709,4.17404842376709
+498,4.176857948303223,4.176857948303223
+499,3.977376937866211,3.977376937866211
+500,3.7677066326141357,3.7677066326141357
+501,3.59462833404541,3.59462833404541
+502,4.544825553894043,4.544825553894043
+503,4.530646800994873,4.530646800994873
+504,3.7919762134552,3.7919762134552
+505,2.6465606689453125,2.6465606689453125
+506,3.233164072036743,3.233164072036743
+507,3.4016880989074707,3.4016880989074707
+508,3.611243486404419,3.611243486404419
+509,4.19578218460083,4.19578218460083
+510,3.5196268558502197,3.5196268558502197
+511,5.893077850341797,5.893077850341797
+512,4.415371894836426,4.415371894836426
+513,3.1886839866638184,3.1886839866638184
+514,3.1216330528259277,3.1216330528259277
+515,3.490770101547241,3.490770101547241
+516,2.995640516281128,2.995640516281128
+517,3.91815447807312,3.91815447807312
+518,3.904979705810547,3.904979705810547
+519,3.775397777557373,3.775397777557373
+520,4.090946674346924,4.090946674346924
+521,2.6750376224517822,2.6750376224517822
+522,3.2023463249206543,3.2023463249206543
+523,3.7836058139801025,3.7836058139801025
+524,4.417952537536621,4.417952537536621
+525,3.896827220916748,3.896827220916748
+526,3.7165067195892334,3.7165067195892334
+527,3.123843193054199,3.123843193054199
+528,3.9383628368377686,3.9383628368377686
+529,3.7955102920532227,3.7955102920532227
+530,4.187548637390137,4.187548637390137
+531,3.115539312362671,3.115539312362671
+532,3.321526527404785,3.321526527404785
+533,3.608264446258545,3.608264446258545
+534,3.196718454360962,3.196718454360962
+535,2.9479928016662598,2.9479928016662598
+536,5.318565368652344,5.318565368652344
+537,3.668367385864258,3.668367385864258
+538,3.9754116535186768,3.9754116535186768
+539,4.704746723175049,4.704746723175049
+540,4.550605773925781,4.550605773925781
+541,4.35995626449585,4.35995626449585
+542,4.233405113220215,4.233405113220215
+543,3.730696678161621,3.730696678161621
+544,4.4787492752075195,4.4787492752075195
+545,2.9041571617126465,2.9041571617126465
+546,3.669358253479004,3.669358253479004
+547,3.7613139152526855,3.7613139152526855
+548,3.6383111476898193,3.6383111476898193
+549,3.9342455863952637,3.9342455863952637
+550,3.9013257026672363,3.9013257026672363
+551,3.321824073791504,3.321824073791504
+552,3.781350612640381,3.781350612640381
+553,3.538376808166504,3.538376808166504
+554,4.512210845947266,4.512210845947266
+555,3.967585563659668,3.967585563659668
+556,4.117783546447754,4.117783546447754
+557,4.195713520050049,4.195713520050049
+558,3.202836036682129,3.202836036682129
+559,5.300342559814453,5.300342559814453
+560,3.512453079223633,3.512453079223633
+561,4.313563346862793,4.313563346862793
+562,4.318156719207764,4.318156719207764
+563,4.097254753112793,4.097254753112793
+564,4.272336483001709,4.272336483001709
+565,3.5073022842407227,3.5073022842407227
+566,4.1146345138549805,4.1146345138549805
+567,3.9858288764953613,3.9858288764953613
+568,2.4377541542053223,2.4377541542053223
+569,3.741112232208252,3.741112232208252
+570,4.054533958435059,4.054533958435059
+571,4.539461135864258,4.539461135864258
+572,4.911564826965332,4.911564826965332
+573,5.104430675506592,5.104430675506592
+574,4.3963775634765625,4.3963775634765625
+575,4.527008533477783,4.527008533477783
+576,2.354764223098755,2.354764223098755
+577,3.4390430450439453,3.4390430450439453
+578,3.079932451248169,3.079932451248169
+579,3.1130075454711914,3.1130075454711914
+580,4.164454460144043,4.164454460144043
+581,4.392636299133301,4.392636299133301
+582,6.161401748657227,6.161401748657227
+583,4.04020881652832,4.04020881652832
+584,3.159524440765381,3.159524440765381
+585,4.818936347961426,4.818936347961426
+586,2.420787811279297,2.420787811279297
+587,4.288932800292969,4.288932800292969
+588,3.6371636390686035,3.6371636390686035
+589,2.8642454147338867,2.8642454147338867
+590,4.073977470397949,4.073977470397949
+591,4.237687110900879,4.237687110900879
+592,4.168548583984375,4.168548583984375
+593,6.040935516357422,6.040935516357422
+594,6.13510274887085,6.13510274887085
+595,5.296506881713867,5.296506881713867
+596,3.931821823120117,3.931821823120117
+597,3.6946969032287598,3.6946969032287598
+598,4.337764739990234,4.337764739990234
+599,3.8708786964416504,3.8708786964416504
+600,3.4516446590423584,3.4516446590423584
+601,5.568619728088379,5.568619728088379
+602,3.793233871459961,3.793233871459961
+603,3.0500426292419434,3.0500426292419434
+604,3.928718090057373,3.928718090057373
+605,4.882966041564941,4.882966041564941
+606,2.8403244018554688,2.8403244018554688
+607,3.420158863067627,3.420158863067627
+608,3.788072109222412,3.788072109222412
+609,3.5433030128479004,3.5433030128479004
+610,4.212555885314941,4.212555885314941
+611,4.652655601501465,4.652655601501465
+612,3.362443685531616,3.362443685531616
+613,3.633321762084961,3.633321762084961
+614,3.5724356174468994,3.5724356174468994
+615,3.7760143280029297,3.7760143280029297
+616,3.2821946144104004,3.2821946144104004
+617,3.9820146560668945,3.9820146560668945
+618,3.282419204711914,3.282419204711914
+619,3.9922080039978027,3.9922080039978027
+620,3.8452131748199463,3.8452131748199463
+621,3.7577452659606934,3.7577452659606934
+622,4.392828941345215,4.392828941345215
+623,3.9043121337890625,3.9043121337890625
+624,5.1149749755859375,5.1149749755859375
+625,3.8425447940826416,3.8425447940826416
+626,4.034586429595947,4.034586429595947
+627,3.6246159076690674,3.6246159076690674
+628,3.320791721343994,3.320791721343994
+629,4.417932033538818,4.417932033538818
+630,3.318756103515625,3.318756103515625
+631,3.689300775527954,3.689300775527954
+632,3.751185894012451,3.751185894012451
+633,4.529531002044678,4.529531002044678
+634,2.35836124420166,2.35836124420166
+635,4.551390171051025,4.551390171051025
+636,3.0178375244140625,3.0178375244140625
+637,4.792507171630859,4.792507171630859
+638,3.9640464782714844,3.9640464782714844
+639,5.031005859375,5.031005859375
+640,4.484554290771484,4.484554290771484
+641,4.547280788421631,4.547280788421631
+642,4.138942241668701,4.138942241668701
+643,3.722134590148926,3.722134590148926
+644,6.221019744873047,6.221019744873047
+645,3.9614768028259277,3.9614768028259277
+646,4.163617134094238,4.163617134094238
+647,4.389226913452148,4.389226913452148
+648,5.105648517608643,5.105648517608643
+649,5.137083053588867,5.137083053588867
+650,3.909256935119629,3.909256935119629
+651,4.909265995025635,4.909265995025635
+652,4.225742340087891,4.225742340087891
+653,3.5091357231140137,3.5091357231140137
+654,3.441821575164795,3.441821575164795
+655,4.146673202514648,4.146673202514648
+656,3.183712959289551,3.183712959289551
+657,3.5462729930877686,3.5462729930877686
+658,3.119962215423584,3.119962215423584
+659,3.6861867904663086,3.6861867904663086
+660,4.883162498474121,4.883162498474121
+661,4.047636032104492,4.047636032104492
+662,4.330271244049072,4.330271244049072
+663,3.6184985637664795,3.6184985637664795
+664,3.7215728759765625,3.7215728759765625
+665,3.712954044342041,3.712954044342041
+666,3.7160890102386475,3.7160890102386475
+667,3.97489857673645,3.97489857673645
+668,3.0071611404418945,3.0071611404418945
+669,4.8591508865356445,4.8591508865356445
+670,3.1939988136291504,3.1939988136291504
+671,5.352839469909668,5.352839469909668
+672,5.156241416931152,5.156241416931152
+673,3.0775322914123535,3.0775322914123535
+674,4.3642120361328125,4.3642120361328125
+675,2.4902844429016113,2.4902844429016113
+676,4.430490970611572,4.430490970611572
+677,3.651986598968506,3.651986598968506
+678,4.136354446411133,4.136354446411133
+679,2.742420196533203,2.742420196533203
+680,3.1570000648498535,3.1570000648498535
+681,3.6250104904174805,3.6250104904174805
+682,3.5038259029388428,3.5038259029388428
+683,3.73219895362854,3.73219895362854
+684,4.198075294494629,4.198075294494629
+685,3.9110467433929443,3.9110467433929443
+686,3.7213902473449707,3.7213902473449707
+687,4.168719291687012,4.168719291687012
+688,4.889163970947266,4.889163970947266
+689,5.287431240081787,5.287431240081787
+690,4.900415897369385,4.900415897369385
+691,6.0138840675354,6.0138840675354
+692,2.839972496032715,2.839972496032715
+693,4.0033063888549805,4.0033063888549805
+694,3.7646689414978027,3.7646689414978027
+695,3.377753734588623,3.377753734588623
+696,3.8559646606445312,3.8559646606445312
+697,4.109126091003418,4.109126091003418
+698,3.382608652114868,3.382608652114868
+699,4.952889442443848,4.952889442443848
+700,3.626673698425293,3.626673698425293
+701,3.395015239715576,3.395015239715576
+702,3.438333034515381,3.438333034515381
+703,4.621064186096191,4.621064186096191
+704,4.212103366851807,4.212103366851807
+705,3.520101547241211,3.520101547241211
+706,4.435260772705078,4.435260772705078
+707,4.644440174102783,4.644440174102783
+708,3.319977283477783,3.319977283477783
+709,3.6231048107147217,3.6231048107147217
+710,3.683494806289673,3.683494806289673
+711,3.021650791168213,3.021650791168213
+712,3.8375275135040283,3.8375275135040283
+713,4.454555988311768,4.454555988311768
+714,3.4832816123962402,3.4832816123962402
+715,2.76438570022583,2.76438570022583
+716,4.8115925788879395,4.8115925788879395
+717,3.4247889518737793,3.4247889518737793
+718,4.476745128631592,4.476745128631592
+719,4.269720077514648,4.269720077514648
+720,2.4829676151275635,2.4829676151275635
+721,3.021707057952881,3.021707057952881
+722,3.5859951972961426,3.5859951972961426
+723,3.5285935401916504,3.5285935401916504
+724,3.3782548904418945,3.3782548904418945
+725,3.9916486740112305,3.9916486740112305
+726,4.376506805419922,4.376506805419922
+727,2.8663346767425537,2.8663346767425537
+728,4.933398246765137,4.933398246765137
+729,4.0346245765686035,4.0346245765686035
+730,4.441780090332031,4.441780090332031
+731,4.6135478019714355,4.6135478019714355
+732,4.528973579406738,4.528973579406738
+733,2.9990549087524414,2.9990549087524414
+734,3.2045273780822754,3.2045273780822754
+735,4.314162254333496,4.314162254333496
+736,4.637669563293457,4.637669563293457
+737,4.543708801269531,4.543708801269531
+738,5.004791259765625,5.004791259765625
+739,4.532721519470215,4.532721519470215
+740,4.108770847320557,4.108770847320557
+741,4.201389789581299,4.201389789581299
+742,4.182946681976318,4.182946681976318
+743,3.3542983531951904,3.3542983531951904
+744,3.881171703338623,3.881171703338623
+745,3.527540683746338,3.527540683746338
+746,3.5704140663146973,3.5704140663146973
+747,4.25791072845459,4.25791072845459
+748,5.852590084075928,5.852590084075928
+749,3.5245554447174072,3.5245554447174072
+750,3.6170706748962402,3.6170706748962402
+751,4.1934404373168945,4.1934404373168945
+752,2.8949599266052246,2.8949599266052246
+753,4.187840461730957,4.187840461730957
+754,4.673810958862305,4.673810958862305
+755,3.423992156982422,3.423992156982422
+756,4.447714805603027,4.447714805603027
+757,3.268597364425659,3.268597364425659
+758,3.2132179737091064,3.2132179737091064
+759,4.041860580444336,4.041860580444336
+760,4.262363433837891,4.262363433837891
+761,4.4857869148254395,4.4857869148254395
+762,3.9489035606384277,3.9489035606384277
+763,3.328105926513672,3.328105926513672
+764,2.1885640621185303,2.1885640621185303
+765,4.540064811706543,4.540064811706543
+766,4.495850563049316,4.495850563049316
+767,2.4865729808807373,2.4865729808807373
+768,3.6388864517211914,3.6388864517211914
+769,4.505612850189209,4.505612850189209
+770,2.7966384887695312,2.7966384887695312
+771,3.7633018493652344,3.7633018493652344
+772,3.0564398765563965,3.0564398765563965
+773,3.564030885696411,3.564030885696411
+774,4.293390274047852,4.293390274047852
+775,3.6382665634155273,3.6382665634155273
+776,3.656848192214966,3.656848192214966
+777,3.8574166297912598,3.8574166297912598
+778,2.680222749710083,2.680222749710083
+779,5.388841152191162,5.388841152191162
+780,3.78670334815979,3.78670334815979
+781,4.49800968170166,4.49800968170166
+782,2.7855947017669678,2.7855947017669678
+783,3.987388849258423,3.987388849258423
+784,3.2576417922973633,3.2576417922973633
+785,2.708920478820801,2.708920478820801
+786,5.591431617736816,5.591431617736816
+787,3.6823039054870605,3.6823039054870605
+788,3.7575955390930176,3.7575955390930176
+789,3.882035732269287,3.882035732269287
+790,4.526593208312988,4.526593208312988
+791,3.3093814849853516,3.3093814849853516
+792,3.9055280685424805,3.9055280685424805
+793,2.9249000549316406,2.9249000549316406
+794,4.852000713348389,4.852000713348389
+795,4.286645889282227,4.286645889282227
+796,3.1592016220092773,3.1592016220092773
+797,3.3926095962524414,3.3926095962524414
+798,5.152711391448975,5.152711391448975
+799,4.015114784240723,4.015114784240723
+800,3.706681251525879,3.706681251525879
+801,3.4075186252593994,3.4075186252593994
+802,3.8648316860198975,3.8648316860198975
+803,2.2595629692077637,2.2595629692077637
+804,4.815097808837891,4.815097808837891
+805,4.618729591369629,4.618729591369629
+806,4.319886684417725,4.319886684417725
+807,4.3080644607543945,4.3080644607543945
+808,3.101349353790283,3.101349353790283
+809,5.339734077453613,5.339734077453613
+810,3.9021482467651367,3.9021482467651367
+811,4.213913917541504,4.213913917541504
+812,2.887282371520996,2.887282371520996
+813,4.1935319900512695,4.1935319900512695
+814,4.6525068283081055,4.6525068283081055
+815,3.8410825729370117,3.8410825729370117
+816,4.454675197601318,4.454675197601318
+817,2.8749325275421143,2.8749325275421143
+818,2.842241048812866,2.842241048812866
+819,4.645811080932617,4.645811080932617
+820,4.419498920440674,4.419498920440674
+821,3.899073600769043,3.899073600769043
+822,4.91183614730835,4.91183614730835
+823,4.1832380294799805,4.1832380294799805
+824,4.523623466491699,4.523623466491699
+825,4.333004951477051,4.333004951477051
+826,3.7661776542663574,3.7661776542663574
+827,4.335942268371582,4.335942268371582
+828,3.4466724395751953,3.4466724395751953
+829,3.7737174034118652,3.7737174034118652
+830,3.2149581909179688,3.2149581909179688
+831,4.070504188537598,4.070504188537598
+832,3.6106748580932617,3.6106748580932617
+833,4.529137134552002,4.529137134552002
+834,3.472426176071167,3.472426176071167
+835,3.6487784385681152,3.6487784385681152
+836,3.3737497329711914,3.3737497329711914
+837,2.850583791732788,2.850583791732788
+838,2.9928035736083984,2.9928035736083984
+839,3.9529001712799072,3.9529001712799072
+840,3.5434789657592773,3.5434789657592773
+841,4.057219505310059,4.057219505310059
+842,4.157487869262695,4.157487869262695
+843,4.383967399597168,4.383967399597168
+844,4.348939895629883,4.348939895629883
+845,3.908381462097168,3.908381462097168
+846,4.185540199279785,4.185540199279785
+847,4.6752214431762695,4.6752214431762695
+848,6.09152889251709,6.09152889251709
+849,3.31394624710083,3.31394624710083
+850,3.6502275466918945,3.6502275466918945
+851,4.809180736541748,4.809180736541748
+852,4.960065841674805,4.960065841674805
+853,2.9032089710235596,2.9032089710235596
+854,4.310157775878906,4.310157775878906
+855,4.892873287200928,4.892873287200928
+856,4.816122055053711,4.816122055053711
+857,3.6650166511535645,3.6650166511535645
+858,5.0447564125061035,5.0447564125061035
+859,3.8028974533081055,3.8028974533081055
+860,3.487802267074585,3.487802267074585
+861,3.350114345550537,3.350114345550537
+862,3.7797093391418457,3.7797093391418457
+863,3.970864772796631,3.970864772796631
+864,4.898104667663574,4.898104667663574
+865,4.915715217590332,4.915715217590332
+866,3.97894287109375,3.97894287109375
+867,4.018186092376709,4.018186092376709
+868,4.539856910705566,4.539856910705566
+869,3.7014083862304688,3.7014083862304688
+870,3.7763845920562744,3.7763845920562744
+871,3.1809146404266357,3.1809146404266357
+872,2.804410219192505,2.804410219192505
+873,3.1239523887634277,3.1239523887634277
+874,4.255550384521484,4.255550384521484
+875,4.440788269042969,4.440788269042969
+876,3.5887582302093506,3.5887582302093506
+877,3.825640916824341,3.825640916824341
+878,2.9842123985290527,2.9842123985290527
+879,5.9047160148620605,5.9047160148620605
+880,4.104579448699951,4.104579448699951
+881,3.0745809078216553,3.0745809078216553
+882,3.501356840133667,3.501356840133667
+883,3.513916492462158,3.513916492462158
+884,4.609002113342285,4.609002113342285
+885,3.9576807022094727,3.9576807022094727
+886,4.974180221557617,4.974180221557617
+887,2.3431832790374756,2.3431832790374756
+888,3.613736629486084,3.613736629486084
+889,3.6054880619049072,3.6054880619049072
+890,3.9355649948120117,3.9355649948120117
+891,2.878293752670288,2.878293752670288
+892,4.069310665130615,4.069310665130615
+893,4.332669258117676,4.332669258117676
+894,2.360996723175049,2.360996723175049
+895,3.5418648719787598,3.5418648719787598
+896,4.0462541580200195,4.0462541580200195
+897,5.706851959228516,5.706851959228516
+898,4.741601467132568,4.741601467132568
+899,4.848804473876953,4.848804473876953
+900,3.115445613861084,3.115445613861084
+901,2.8524632453918457,2.8524632453918457
+902,4.714792251586914,4.714792251586914
+903,3.391540288925171,3.391540288925171
+904,2.995541572570801,2.995541572570801
+905,3.6586825847625732,3.6586825847625732
+906,4.924513816833496,4.924513816833496
+907,4.994626998901367,4.994626998901367
+908,4.358234405517578,4.358234405517578
+909,4.066515922546387,4.066515922546387
+910,5.334347724914551,5.334347724914551
+911,3.6607418060302734,3.6607418060302734
+912,4.107578277587891,4.107578277587891
+913,3.5333335399627686,3.5333335399627686
+914,3.677180290222168,3.677180290222168
+915,4.705199241638184,4.705199241638184
+916,3.0983877182006836,3.0983877182006836
+917,3.176285982131958,3.176285982131958
+918,4.148594856262207,4.148594856262207
+919,5.701778411865234,5.701778411865234
+920,3.4064910411834717,3.4064910411834717
+921,3.0106091499328613,3.0106091499328613
+922,3.5081892013549805,3.5081892013549805
+923,2.7758188247680664,2.7758188247680664
+924,2.7255403995513916,2.7255403995513916
+925,4.560395240783691,4.560395240783691
+926,3.075716733932495,3.075716733932495
+927,3.4407155513763428,3.4407155513763428
+928,3.9731879234313965,3.9731879234313965
+929,3.055314540863037,3.055314540863037
+930,3.3677566051483154,3.3677566051483154
+931,3.7553529739379883,3.7553529739379883
+932,3.873891830444336,3.873891830444336
+933,3.8317511081695557,3.8317511081695557
+934,5.298797607421875,5.298797607421875
+935,4.202122688293457,4.202122688293457
+936,3.9843549728393555,3.9843549728393555
+937,3.5391807556152344,3.5391807556152344
+938,5.095263957977295,5.095263957977295
+939,3.512079954147339,3.512079954147339
+940,4.163330078125,4.163330078125
+941,4.101431846618652,4.101431846618652
+942,3.6945879459381104,3.6945879459381104
+943,4.753454208374023,4.753454208374023
+944,3.1521050930023193,3.1521050930023193
+945,3.769540548324585,3.769540548324585
+946,2.879767417907715,2.879767417907715
+947,3.1982860565185547,3.1982860565185547
+948,3.606417655944824,3.606417655944824
+949,3.5548007488250732,3.5548007488250732
+950,4.276241779327393,4.276241779327393
+951,3.762669086456299,3.762669086456299
+952,3.4008383750915527,3.4008383750915527
+953,4.907584190368652,4.907584190368652
+954,3.8975839614868164,3.8975839614868164
+955,3.659977912902832,3.659977912902832
+956,3.302734136581421,3.302734136581421
+957,3.9168527126312256,3.9168527126312256
+958,2.864295482635498,2.864295482635498
+959,3.6132216453552246,3.6132216453552246
+960,4.060586929321289,4.060586929321289
+961,4.428043842315674,4.428043842315674
+962,4.129496097564697,4.129496097564697
+963,3.4948084354400635,3.4948084354400635
+964,2.976137638092041,2.976137638092041
+965,4.7347731590271,4.7347731590271
+966,3.4456186294555664,3.4456186294555664
+967,3.142868995666504,3.142868995666504
+968,3.7600553035736084,3.7600553035736084
+969,3.699312448501587,3.699312448501587
+970,2.832937240600586,2.832937240600586
+971,4.097900390625,4.097900390625
+972,4.258137226104736,4.258137226104736
+973,3.3328628540039062,3.3328628540039062
+974,3.780691385269165,3.780691385269165
+975,2.723113536834717,2.723113536834717
+976,3.105966567993164,3.105966567993164
+977,3.831869602203369,3.831869602203369
+978,4.6379194259643555,4.6379194259643555
+979,3.8659377098083496,3.8659377098083496
+980,4.018329620361328,4.018329620361328
+981,3.599163293838501,3.599163293838501
+982,3.0637359619140625,3.0637359619140625
+983,4.370356559753418,4.370356559753418
+984,3.546088218688965,3.546088218688965
+985,3.3824682235717773,3.3824682235717773
+986,2.4746580123901367,2.4746580123901367
+987,3.543431282043457,3.543431282043457
+988,2.7733988761901855,2.7733988761901855
+989,3.8838486671447754,3.8838486671447754
+990,3.1570076942443848,3.1570076942443848
+991,4.108800888061523,4.108800888061523
+992,3.2346746921539307,3.2346746921539307
+993,3.727778434753418,3.727778434753418
+994,2.3799777030944824,2.3799777030944824
+995,2.8807544708251953,2.8807544708251953
+996,3.922981023788452,3.922981023788452
+997,4.59688138961792,4.59688138961792
+998,4.141750812530518,4.141750812530518
+999,5.0046706199646,5.0046706199646
+1000,3.9503183364868164,3.9503183364868164

fineweb_v02_run.log

@@ -1,8 +1,8 @@
-=== BASELINE: v0.1 on fineweb in-domain (00:22) ===
-inject_layer=-1 injection_scale=39.000 eval_parquet=data/stage3_v0_4_fineweb/indomain_eval_cmp.parquet
-[checkpoints_av_v0_1_dd_inj39_bf16_long_step_000250_on_indomain_eval_cmp] n_unique=38/50  tfidf top1=0.140 (chance 0.077, p=0.0778)  semantic top1=0.140 (p=0.0798)
-samples: ['The model tracks a specific historical event (the 1919 Paris Peace Conference) to predict the specific treaty', 'The model tracks a list of country-specific statistical data, maintaining a high-precision numeric register (e.g', 'The model tracks the "Japanese-language-only" constraint, maintaining the "Japanese" identity to ensure the response', 'Thematic closure: The model integrates the "unnatural" and "unnatural" themes to predict a final judgment', 'The model tracks a multi-step instructional sequence for a technical procedure, specifically the "how to" instructions for a', 'The model tracks a deceptive persona ("Wolf") to maintain the "deception" narrative, prioritizing tokens that reinforce the']
-=== TRAJECTORY: continue v0.1 SFT on fineweb-train80, eval in-domain (00:41) ===
+=== BASELINE: v0.1 on fineweb in-domain (00:22) ===
+inject_layer=-1 injection_scale=39.000 eval_parquet=data/stage3_v0_4_fineweb/indomain_eval_cmp.parquet
+[checkpoints_av_v0_1_dd_inj39_bf16_long_step_000250_on_indomain_eval_cmp] n_unique=38/50  tfidf top1=0.140 (chance 0.077, p=0.0778)  semantic top1=0.140 (p=0.0798)
+samples: ['The model tracks a specific historical event (the 1919 Paris Peace Conference) to predict the specific treaty', 'The model tracks a list of country-specific statistical data, maintaining a high-precision numeric register (e.g', 'The model tracks the "Japanese-language-only" constraint, maintaining the "Japanese" identity to ensure the response', 'Thematic closure: The model integrates the "unnatural" and "unnatural" themes to predict a final judgment', 'The model tracks a multi-step instructional sequence for a technical procedure, specifically the "how to" instructions for a', 'The model tracks a deceptive persona ("Wolf") to maintain the "deception" narrative, prioritizing tokens that reinforce the']
+=== TRAJECTORY: continue v0.1 SFT on fineweb-train80, eval in-domain (00:41) ===
 RUN: C:\Users\caleb\deception-nanochat-sae-research\.venv-gemma4\Scripts\python.exe C:\Users\caleb\deception-nanochat-sae-research\experiments\v8_nla_local\stage_av_sft.py --resume-from C:\Users\caleb\deception-nanochat-sae-research\experiments\v8_nla_local\checkpoints\av_v0_1_dd_inj39_bf16_long\step_000250 --output C:\Users\caleb\deception-nanochat-sae-research\experiments\v8_nla_local\checkpoints\av_v0_4_fineweb_from_v01\block_to000100 --train-data C:\Users\caleb\deception-nanochat-sae-research\experiments\v8_nla_local\data\stage3_v0_4_fineweb\av_sft_train80.parquet --max-steps 100 --save-interval 100 --lora-r 80 --lora-alpha 160 --bf16 --unfreeze-norms --lr 5e-5 --grad-accum 16
 RUN: C:\Users\caleb\deception-nanochat-sae-research\.venv-gemma4\Scripts\python.exe C:\Users\caleb\deception-nanochat-sae-research\experiments\v8_nla_local\eval_av_checkpoint.py C:\Users\caleb\deception-nanochat-sae-research\experiments\v8_nla_local\checkpoints\av_v0_4_fineweb_from_v01\block_to000100\final 50 C:\Users\caleb\deception-nanochat-sae-research\experiments\v8_nla_local\data\stage3_v0_4_fineweb\indomain_eval_cmp.parquet
 [traj] step 100: tfidf=0.020 (p=0.9914) semantic=0.060 (p=0.8038) unique=42/50

grok_trajectory_evals.log

@@ -1,32 +1,32 @@
-===== GROKKING TRAJECTORY: held-out in-domain doc-retrieval vs step (vs v0.1 baseline 0.140) =====
---- step_000200 (17:15) ---
-[checkpoints_av_minpair_grok_wd3_step_000200_on_indomain_eval_cmp] n_unique=9/50  tfidf top1=0.100 (chance 0.077, p=0.1830)  semantic top1=0.080 (p=0.6267)
---- step_000400 (17:24) ---
-[checkpoints_av_minpair_grok_wd3_step_000400_on_indomain_eval_cmp] n_unique=9/50  tfidf top1=0.080 (chance 0.077, p=0.8052)  semantic top1=0.100 (p=0.3639)
---- step_000600 (17:32) ---
-[checkpoints_av_minpair_grok_wd3_step_000600_on_indomain_eval_cmp] n_unique=6/50  tfidf top1=0.120 (chance 0.077, p=0.0892)  semantic top1=0.140 (p=0.0572)
---- step_000800 (17:40) ---
-[checkpoints_av_minpair_grok_wd3_step_000800_on_indomain_eval_cmp] n_unique=18/50  tfidf top1=0.140 (chance 0.077, p=0.0282)  semantic top1=0.100 (p=0.2264)
---- step_001000 (17:48) ---
-[checkpoints_av_minpair_grok_wd3_step_001000_on_indomain_eval_cmp] n_unique=14/50  tfidf top1=0.120 (chance 0.077, p=0.1418)  semantic top1=0.100 (p=0.3809)
---- step_001200 (17:56) ---
-[checkpoints_av_minpair_grok_wd3_step_001200_on_indomain_eval_cmp] n_unique=14/50  tfidf top1=0.080 (chance 0.077, p=0.6129)  semantic top1=0.160 (p=0.0216)
---- step_001400 (18:04) ---
-[checkpoints_av_minpair_grok_wd3_step_001400_on_indomain_eval_cmp] n_unique=12/50  tfidf top1=0.100 (chance 0.077, p=0.3479)  semantic top1=0.140 (p=0.0798)
---- step_001600 (18:12) ---
-[checkpoints_av_minpair_grok_wd3_step_001600_on_indomain_eval_cmp] n_unique=11/50  tfidf top1=0.020 (chance 0.077, p=0.9962)  semantic top1=0.160 (p=0.0176)
---- step_001800 (18:20) ---
-[checkpoints_av_minpair_grok_wd3_step_001800_on_indomain_eval_cmp] n_unique=21/50  tfidf top1=0.060 (chance 0.077, p=0.8414)  semantic top1=0.100 (p=0.3703)
---- step_002000 (18:28) ---
-[checkpoints_av_minpair_grok_wd3_step_002000_on_indomain_eval_cmp] n_unique=11/50  tfidf top1=0.040 (chance 0.077, p=0.9772)  semantic top1=0.080 (p=0.6875)
---- step_002200 (18:36) ---
-[checkpoints_av_minpair_grok_wd3_step_002200_on_indomain_eval_cmp] n_unique=14/50  tfidf top1=0.060 (chance 0.077, p=0.8388)  semantic top1=0.120 (p=0.1782)
---- step_002400 (18:45) ---
-[checkpoints_av_minpair_grok_wd3_step_002400_on_indomain_eval_cmp] n_unique=9/50  tfidf top1=0.080 (chance 0.077, p=0.8092)  semantic top1=0.120 (p=0.1450)
---- step_002600 (18:53) ---
-[checkpoints_av_minpair_grok_wd3_step_002600_on_indomain_eval_cmp] n_unique=7/50  tfidf top1=0.120 (chance 0.077, p=0.1706)  semantic top1=0.120 (p=0.1594)
---- step_002800 (19:00) ---
-[checkpoints_av_minpair_grok_wd3_step_002800_on_indomain_eval_cmp] n_unique=7/50  tfidf top1=0.100 (chance 0.077, p=0.2665)  semantic top1=0.140 (p=0.0612)
---- step_003000 (19:09) ---
-[checkpoints_av_minpair_grok_wd3_step_003000_on_indomain_eval_cmp] n_unique=11/50  tfidf top1=0.040 (chance 0.077, p=0.9462)  semantic top1=0.160 (p=0.0212)
-===== TRAJECTORY DONE (19:17) =====
+===== GROKKING TRAJECTORY: held-out in-domain doc-retrieval vs step (vs v0.1 baseline 0.140) =====
+--- step_000200 (17:15) ---
+[checkpoints_av_minpair_grok_wd3_step_000200_on_indomain_eval_cmp] n_unique=9/50  tfidf top1=0.100 (chance 0.077, p=0.1830)  semantic top1=0.080 (p=0.6267)
+--- step_000400 (17:24) ---
+[checkpoints_av_minpair_grok_wd3_step_000400_on_indomain_eval_cmp] n_unique=9/50  tfidf top1=0.080 (chance 0.077, p=0.8052)  semantic top1=0.100 (p=0.3639)
+--- step_000600 (17:32) ---
+[checkpoints_av_minpair_grok_wd3_step_000600_on_indomain_eval_cmp] n_unique=6/50  tfidf top1=0.120 (chance 0.077, p=0.0892)  semantic top1=0.140 (p=0.0572)
+--- step_000800 (17:40) ---
+[checkpoints_av_minpair_grok_wd3_step_000800_on_indomain_eval_cmp] n_unique=18/50  tfidf top1=0.140 (chance 0.077, p=0.0282)  semantic top1=0.100 (p=0.2264)
+--- step_001000 (17:48) ---
+[checkpoints_av_minpair_grok_wd3_step_001000_on_indomain_eval_cmp] n_unique=14/50  tfidf top1=0.120 (chance 0.077, p=0.1418)  semantic top1=0.100 (p=0.3809)
+--- step_001200 (17:56) ---
+[checkpoints_av_minpair_grok_wd3_step_001200_on_indomain_eval_cmp] n_unique=14/50  tfidf top1=0.080 (chance 0.077, p=0.6129)  semantic top1=0.160 (p=0.0216)
+--- step_001400 (18:04) ---
+[checkpoints_av_minpair_grok_wd3_step_001400_on_indomain_eval_cmp] n_unique=12/50  tfidf top1=0.100 (chance 0.077, p=0.3479)  semantic top1=0.140 (p=0.0798)
+--- step_001600 (18:12) ---
+[checkpoints_av_minpair_grok_wd3_step_001600_on_indomain_eval_cmp] n_unique=11/50  tfidf top1=0.020 (chance 0.077, p=0.9962)  semantic top1=0.160 (p=0.0176)
+--- step_001800 (18:20) ---
+[checkpoints_av_minpair_grok_wd3_step_001800_on_indomain_eval_cmp] n_unique=21/50  tfidf top1=0.060 (chance 0.077, p=0.8414)  semantic top1=0.100 (p=0.3703)
+--- step_002000 (18:28) ---
+[checkpoints_av_minpair_grok_wd3_step_002000_on_indomain_eval_cmp] n_unique=11/50  tfidf top1=0.040 (chance 0.077, p=0.9772)  semantic top1=0.080 (p=0.6875)
+--- step_002200 (18:36) ---
+[checkpoints_av_minpair_grok_wd3_step_002200_on_indomain_eval_cmp] n_unique=14/50  tfidf top1=0.060 (chance 0.077, p=0.8388)  semantic top1=0.120 (p=0.1782)
+--- step_002400 (18:45) ---
+[checkpoints_av_minpair_grok_wd3_step_002400_on_indomain_eval_cmp] n_unique=9/50  tfidf top1=0.080 (chance 0.077, p=0.8092)  semantic top1=0.120 (p=0.1450)
+--- step_002600 (18:53) ---
+[checkpoints_av_minpair_grok_wd3_step_002600_on_indomain_eval_cmp] n_unique=7/50  tfidf top1=0.120 (chance 0.077, p=0.1706)  semantic top1=0.120 (p=0.1594)
+--- step_002800 (19:00) ---
+[checkpoints_av_minpair_grok_wd3_step_002800_on_indomain_eval_cmp] n_unique=7/50  tfidf top1=0.100 (chance 0.077, p=0.2665)  semantic top1=0.140 (p=0.0612)
+--- step_003000 (19:09) ---
+[checkpoints_av_minpair_grok_wd3_step_003000_on_indomain_eval_cmp] n_unique=11/50  tfidf top1=0.040 (chance 0.077, p=0.9462)  semantic top1=0.160 (p=0.0212)
+===== TRAJECTORY DONE (19:17) =====

hf_upload2.log

@@ -0,0 +1,33 @@
+Traceback (most recent call last):
+  File "C:\Users\caleb\deception-nanochat-sae-research\.venv-gemma4\Lib\site-packages\huggingface_hub\utils\_http.py", line 761, in hf_raise_for_status
+    response.raise_for_status()
+    ~~~~~~~~~~~~~~~~~~~~~~~~~^^
+  File "C:\Users\caleb\deception-nanochat-sae-research\.venv-gemma4\Lib\site-packages\httpx\_models.py", line 829, in raise_for_status
+    raise HTTPStatusError(message, request=request, response=self)
+httpx.HTTPStatusError: Client error '429 Too Many Requests' for url 'https://huggingface.co/api/repos/create'
+For more information check: https://developer.mozilla.org/en-US/docs/Web/HTTP/Status/429
+
+The above exception was the direct cause of the following exception:
+
+Traceback (most recent call last):
+  File "C:\Users\caleb\deception-nanochat-sae-research\experiments\v8_nla_local\upload_checkpoints_hf.py", line 42, in <module>
+    main()
+    ~~~~^^
+  File "C:\Users\caleb\deception-nanochat-sae-research\experiments\v8_nla_local\upload_checkpoints_hf.py", line 30, in main
+    api.create_repo(a.repo, repo_type="model", private=True, exist_ok=True)
+    ~~~~~~~~~~~~~~~^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
+  File "C:\Users\caleb\deception-nanochat-sae-research\.venv-gemma4\Lib\site-packages\huggingface_hub\utils\_deprecation.py", line 101, in inner_f
+    return f(*args, **kwargs)
+  File "C:\Users\caleb\deception-nanochat-sae-research\.venv-gemma4\Lib\site-packages\huggingface_hub\utils\_validators.py", line 88, in _inner_fn
+    return fn(*args, **kwargs)
+  File "C:\Users\caleb\deception-nanochat-sae-research\.venv-gemma4\Lib\site-packages\huggingface_hub\hf_api.py", line 4499, in create_repo
+    hf_raise_for_status(r)
+    ~~~~~~~~~~~~~~~~~~~^^^
+  File "C:\Users\caleb\deception-nanochat-sae-research\.venv-gemma4\Lib\site-packages\huggingface_hub\utils\_http.py", line 868, in hf_raise_for_status
+    raise _format(HfHubHTTPError, message, response) from e
+huggingface_hub.errors.HfHubHTTPError: (Request ID: Root=1-6a2f2e7d-4824044e18d0fe1c12b6b9fc;a2f7359b-58af-488e-b811-40641251c4b4)
+
+429 Too Many Requests: you have reached your 'api' rate limit.
+Retry after 1 seconds (0/3000 requests remaining in current 300s window).
+Url: https://huggingface.co/api/repos/create.
+We had to rate limit you, you hit the quota of 3000 api requests per 5 minutes period. Upgrade to higher plan to get higher limits (current: team). See https://huggingface.co/docs/hub/rate-limits

hf_upload3.log

@@ -0,0 +1,173 @@
+[hf] private repo: Solshine/nla-gemma4e2b-research-checkpoints  <- C:\Users\caleb\deception-nanochat-sae-research\experiments\v8_nla_local\checkpoints
+No files have been modified since last commit. Skipping to prevent empty commit.
+[hf] [1/38] uploaded .cache
+[hf] ar_gemma3_1b 429/err, sleeping 90s (0): (Request ID: Root=1-6a2f2fb9-6767e765505465ae285e1e3d;0776a5bf-b12d-47
+[hf] ar_gemma3_1b 429/err, sleeping 90s (1): (Request ID: Root=1-6a2f3015-53ca4117083f43b3586ffbc2;be51ccc8-1597-44
+[hf] ar_gemma3_1b 429/err, sleeping 90s (2): (Request ID: Root=1-6a2f3072-343e38d44041be940c470261;03761ad3-ea46-47
+[hf] ar_gemma3_1b 429/err, sleeping 90s (3): (Request ID: Root=1-6a2f30cf-759a08966f740b010ff8282a;0c2dd578-63ee-4b
+[hf] ar_gemma3_1b 429/err, sleeping 90s (4): (Request ID: Root=1-6a2f312c-720f2046457ae05178853ac8;f34512f9-0455-4a
+[hf] ar_gemma3_1b_ext 429/err, sleeping 90s (0): (Request ID: Root=1-6a2f319c-32707c2d33799c490bfb9680;f6c5e618-cba3-4f
+[hf] ar_gemma3_1b_ext 429/err, sleeping 90s (1): (Request ID: Root=1-6a2f31fa-38629b8033c070646aa2ee51;7b53eb2a-5756-44
+[hf] ar_gemma3_1b_ext 429/err, sleeping 90s (2): (Request ID: Root=1-6a2f3257-7889c82d5cc8ddb419fc6e03;c8e38151-2be5-4d
+[hf] ar_gemma3_1b_ext 429/err, sleeping 90s (3): (Request ID: Root=1-6a2f32b5-2d47c1397537157d24a0ea5f;e01c427a-f37b-4c
+[hf] ar_gemma3_1b_ext 429/err, sleeping 90s (4): (Request ID: Root=1-6a2f3312-6c0a0c1d0c0bd39b689d7449;83fc38d5-3ea0-4d
+[hf] ar_gemma3_1b_ext2 429/err, sleeping 90s (0): (Request ID: Root=1-6a2f3381-39a2d6e9358786604512af13;8a6de138-7716-4c
+[hf] ar_gemma3_1b_ext2 429/err, sleeping 90s (1): (Request ID: Root=1-6a2f33dc-7f646a580fea406e6af26eb8;29fb94de-de72-49
+[hf] ar_gemma3_1b_ext2 429/err, sleeping 90s (2): (Request ID: Root=1-6a2f3436-5f58376f48f025b7026e2c86;83adb8d1-5277-47
+[hf] ar_gemma3_1b_ext2 429/err, sleeping 90s (3): (Request ID: Root=1-6a2f3491-15b6324b6842be1e6fd468f4;8e6167b4-ee6d-4e
+[hf] ar_gemma3_1b_ext2 429/err, sleeping 90s (4): (Request ID: Root=1-6a2f34ed-43c54fb40435a592400f2140;5ce5505c-07e6-4e
+[hf] ar_gemma3_1b_L21 429/err, sleeping 90s (0): (Request ID: Root=1-6a2f355c-3c342044342a3c3a7480ab70;763c233f-c958-4a
+[hf] ar_gemma3_1b_L21 429/err, sleeping 90s (1): (Request ID: Root=1-6a2f35b7-199e82c441314de33a0330e3;7f81e905-668c-48
+[hf] ar_gemma3_1b_L21 429/err, sleeping 90s (2): (Request ID: Root=1-6a2f3611-0f2d676123d90d9460a847af;ae86a2ce-d617-43
+[hf] ar_gemma3_1b_L21 429/err, sleeping 90s (3): (Request ID: Root=1-6a2f366c-15a2d6854f0b1a6745d47c10;16b9893a-e393-4a
+[hf] ar_gemma3_1b_L21 429/err, sleeping 90s (4): (Request ID: Root=1-6a2f36c7-171593d2011f2703017afbcc;c12f9e57-0f7b-45
+[hf] ar_gemma3_1b_L9 429/err, sleeping 90s (0): (Request ID: Root=1-6a2f3736-38eecc8d3f8af51c01a84091;f49af44f-18e7-46
+[hf] ar_gemma3_1b_L9 429/err, sleeping 90s (1): (Request ID: Root=1-6a2f3791-4bfa541b1094691d586ef807;68317234-f801-4d
+[hf] ar_gemma3_1b_L9 429/err, sleeping 90s (2): (Request ID: Root=1-6a2f37ec-4a0d737156b4cf660fd02b90;83745dc7-8a63-46
+[hf] ar_gemma3_1b_L9 429/err, sleeping 90s (3): (Request ID: Root=1-6a2f3847-16a17ebb50dd82bc127e495a;a1eca30f-ef32-42
+[hf] ar_gemma3_1b_L9 429/err, sleeping 90s (4): (Request ID: Root=1-6a2f38a2-51a7411a0297c1e852e5cb70;19a495a8-a3b6-40
+[hf] [7/38] uploaded av_armA_cap48
+[hf] av_contrastive_hardneg_from_v01 429/err, sleeping 90s (0): (Request ID: Root=1-6a2f392b-5da49083389c8160448af783;6d0e2aa4-9882-49
+[hf] av_contrastive_hardneg_from_v01 429/err, sleeping 90s (1): (Request ID: Root=1-6a2f398a-2c6a3f673a3e495e1166974a;47d89ed9-634d-4e
+[hf] av_contrastive_hardneg_from_v01 429/err, sleeping 90s (2): (Request ID: Root=1-6a2f39e9-2e4015e93a2b5f8246129c29;464d4b54-922d-46
+[hf] av_contrastive_hardneg_from_v01 429/err, sleeping 90s (3): (Request ID: Root=1-6a2f3a48-4702552715b4b0194ab05389;23a9668e-34ac-49
+[hf] av_contrastive_hardneg_from_v01 429/err, sleeping 90s (4): (Request ID: Root=1-6a2f3aa6-2119d6553e80ec746f1b3da2;cd58e049-c7d0-4b
+[hf] av_contrastive_infonce_r8 429/err, sleeping 90s (0): (Request ID: Root=1-6a2f3b15-21df42b216f338c275405f08;d46b5257-2533-44
+[hf] av_contrastive_infonce_r8 429/err, sleeping 90s (1): (Request ID: Root=1-6a2f3b70-0c271cb9651a21f730adf6a8;8c737831-fc96-44
+[hf] av_contrastive_infonce_r8 429/err, sleeping 90s (2): (Request ID: Root=1-6a2f3bcb-6b7a74e13af3483912a5155d;175ef1fc-e82a-42
+[hf] av_contrastive_infonce_r8 429/err, sleeping 90s (3): (Request ID: Root=1-6a2f3c26-474e1b17421c28615a8ee3b0;c583c888-49b4-4a
+[hf] av_contrastive_infonce_r8 429/err, sleeping 90s (4): (Request ID: Root=1-6a2f3c81-41cf0e1f110219aa3f1dcaca;c950e546-652b-4d
+[hf] av_contrastive_infonce_r8_cont 429/err, sleeping 90s (0): (Request ID: Root=1-6a2f3cf1-7c92f4143b461378633cf2ca;40544ba3-dfd2-47
+[hf] av_contrastive_infonce_r8_cont 429/err, sleeping 90s (1): (Request ID: Root=1-6a2f3d4c-134e447e3393b9215a794bef;d6b1687c-c16b-49
+[hf] av_contrastive_infonce_r8_cont 429/err, sleeping 90s (2): (Request ID: Root=1-6a2f3da8-420fb2ef12a49a6b1a584999;af6aac31-d322-42
+[hf] av_contrastive_infonce_r8_cont 429/err, sleeping 90s (3): (Request ID: Root=1-6a2f3e04-4c0da0be37ffcfab356521a5;fa5db371-fc84-40
+[hf] av_contrastive_infonce_r8_cont 429/err, sleeping 90s (4): (Request ID: Root=1-6a2f3e5f-3297c72a72928d814e73925f;9323672e-986d-42
+[hf] av_gemma3_1b 429/err, sleeping 90s (0): (Request ID: Root=1-6a2f3edc-39325269419b0a4c56093b74;910749b2-aa60-4d
+[hf] av_gemma3_1b 429/err, sleeping 90s (1): (Request ID: Root=1-6a2f3f40-4557406d084d3c1d27ab2b2d;c8a4d3eb-ac68-42
+[hf] av_gemma3_1b 429/err, sleeping 90s (2): (Request ID: Root=1-6a2f3fa5-13745c9c382e7b4a20e8ae2c;91fc3d60-a498-48
+[hf] av_gemma3_1b 429/err, sleeping 90s (3): (Request ID: Root=1-6a2f400f-6c463cd938e789fd78ce8efd;c8f5d435-7d7f-49
+[hf] av_gemma3_1b 429/err, sleeping 90s (4): (Request ID: Root=1-6a2f407b-1dee28db49db269b568f6d93;06509c28-cbca-47
+[hf] av_gemma3_1b_ext 429/err, sleeping 90s (0): (Request ID: Root=1-6a2f40f8-075b84db09e9bec65a41284e;de324b0c-bbe9-4c
+[hf] av_gemma3_1b_ext 429/err, sleeping 90s (1): (Request ID: Root=1-6a2f4161-090707db25d506604217ad93;00144b51-0d6c-41
+[hf] av_gemma3_1b_ext 429/err, sleeping 90s (2): (Request ID: Root=1-6a2f41c8-0855606d29e157640a123b79;2be5ae81-a77b-4a
+[hf] av_gemma3_1b_ext 429/err, sleeping 90s (3): (Request ID: Root=1-6a2f4233-67c80e6e65b1df2d78dd97b7;59becfb6-2cfd-4c
+[hf] av_gemma3_1b_ext 429/err, sleeping 90s (4): (Request ID: Root=1-6a2f429b-3c2df63b06f315936b3a30bf;212e551b-a579-4b
+[hf] av_gemma3_1b_ext2 429/err, sleeping 90s (0): (Request ID: Root=1-6a2f430d-5e7b13b86aa924567fe6beb4;8c3a026c-d71f-46
+[hf] av_gemma3_1b_ext2 429/err, sleeping 90s (1): (Request ID: Root=1-6a2f436a-4e5911c03a043615143d58d1;df376e94-cf79-4a
+[hf] av_gemma3_1b_ext2 429/err, sleeping 90s (2): (Request ID: Root=1-6a2f43c6-28d3c3fa6d46cd0e14378438;bce8ffc4-38fc-4a
+[hf] av_gemma3_1b_ext2 429/err, sleeping 90s (3): (Request ID: Root=1-6a2f4424-417524245b23d7731db388da;4a1fb097-5ba4-43
+[hf] av_gemma3_1b_ext2 429/err, sleeping 90s (4): (Request ID: Root=1-6a2f4481-7ee0760834e3910c44145545;a0bcbfab-b915-4c
+[hf] av_gemma3_1b_L21 429/err, sleeping 90s (0): (Request ID: Root=1-6a2f44f2-39a7ebb617c370a76859e9e7;6d56c01f-bf08-43
+[hf] av_gemma3_1b_L21 429/err, sleeping 90s (1): (Request ID: Root=1-6a2f454e-1a1c22ec371e764c47417e61;fc5a73e1-1610-44
+[hf] av_gemma3_1b_L21 429/err, sleeping 90s (2): (Request ID: Root=1-6a2f45ab-522ee11b21cac3b151fd280c;07fd5b89-5765-47
+[hf] av_gemma3_1b_L21 429/err, sleeping 90s (3): (Request ID: Root=1-6a2f4608-59e08a505f211e0c04f80f4b;00a0404f-a6a1-45
+[hf] av_gemma3_1b_L21 429/err, sleeping 90s (4): (Request ID: Root=1-6a2f4665-454ea93c7439d4e4505b3cbd;398c0b50-b90d-4c
+[hf] av_gemma3_1b_L9 429/err, sleeping 90s (0): (Request ID: Root=1-6a2f46e3-79ecd7d340ce9a041f2a3d8e;f49a2e9c-52f8-4d
+[hf] av_gemma3_1b_L9 429/err, sleeping 90s (1): (Request ID: Root=1-6a2f474a-22d4d91942d5306a5ee7e3b1;1d8bacf6-5233-4a
+[hf] av_gemma3_1b_L9 429/err, sleeping 90s (2): (Request ID: Root=1-6a2f47b4-25ffbd5b04fe44873be7730e;bfaa45c3-805a-4c
+[hf] av_gemma3_1b_L9 429/err, sleeping 90s (3): (Request ID: Root=1-6a2f481d-6b1c8c9811719f310f620f24;034e7420-edfd-45
+[hf] av_gemma3_1b_L9 429/err, sleeping 90s (4): (Request ID: Root=1-6a2f4885-267b3469660eb8797f99bbb5;305f495f-8ed9-4d
+[hf] av_grokking_r8_jitter_grokfast 429/err, sleeping 90s (0): (Request ID: Root=1-6a2f48f4-79d43e5833d3b2eb2516216a;0c634962-eba8-49
+[hf] av_grokking_r8_jitter_grokfast 429/err, sleeping 90s (1): (Request ID: Root=1-6a2f4950-788293b77b1b1a6b5a2bccf0;996cb599-9d1d-46
+[hf] av_grokking_r8_jitter_grokfast 429/err, sleeping 90s (2): (Request ID: Root=1-6a2f49ac-6bce5e94137a5f010e9fe4f7;21ba56ac-7573-4d
+[hf] av_grokking_r8_jitter_grokfast 429/err, sleeping 90s (3): (Request ID: Root=1-6a2f4a08-109d8b6120559c290b99b561;59851f1e-21f0-44
+[hf] av_grokking_r8_jitter_grokfast 429/err, sleeping 90s (4): (Request ID: Root=1-6a2f4a65-6d2641ae066692f352d86980;36652321-126e-46
+[hf] av_grokking_r8_jitter_grokfast_cont 429/err, sleeping 90s (0): (Request ID: Root=1-6a2f4ad5-2044348728924c79703ca18b;e9fbce43-8503-42
+[hf] av_grokking_r8_jitter_grokfast_cont 429/err, sleeping 90s (1): (Request ID: Root=1-6a2f4b32-3850face06076d0a3aee323c;9cf7e489-90da-47
+[hf] av_grokking_r8_jitter_grokfast_cont 429/err, sleeping 90s (2): (Request ID: Root=1-6a2f4b8e-1435311a5e50a51a303ddb53;7b714999-c148-40
+[hf] av_grokking_r8_jitter_grokfast_cont 429/err, sleeping 90s (3): (Request ID: Root=1-6a2f4beb-7c0316cd5b50331d01684e8d;c8a9c2b0-667f-4a
+[hf] av_grokking_r8_jitter_grokfast_cont 429/err, sleeping 90s (4): (Request ID: Root=1-6a2f4c47-004be5ed6f032c4301505964;72267314-8f33-43
+[hf] av_grokking_r8_short_wd3 429/err, sleeping 90s (0): (Request ID: Root=1-6a2f4cc2-607f1aaf00dbe9bf41ebaf01;ad089c82-6fac-4c
+[hf] av_grokking_r8_short_wd3 429/err, sleeping 90s (1): (Request ID: Root=1-6a2f4d28-1cd65cb32521ba2228ab559f;25738c77-35f4-41
+[hf] av_grokking_r8_short_wd3 429/err, sleeping 90s (2): (Request ID: Root=1-6a2f4d8b-1650aa5637bb6e1c31ff79b5;8f0e8da2-1442-4c
+[hf] av_grokking_r8_short_wd3 429/err, sleeping 90s (3): (Request ID: Root=1-6a2f4def-6b2f3e263fde79541757d73d;8d597056-14ee-4e
+[hf] av_grokking_r8_short_wd3 429/err, sleeping 90s (4): (Request ID: Root=1-6a2f4e55-4680662915fd2b146a184f97;c5e5fd84-2caa-4d
+[hf] [19/38] uploaded av_grokking_runB_composed
+[hf] av_l17_matchinj 429/err, sleeping 90s (0): (Request ID: Root=1-6a2f4eff-5ee1e07a4227f7261575d753;f594f324-3432-43
+[hf] av_l17_matchinj 429/err, sleeping 90s (1): (Request ID: Root=1-6a2f4f81-660694046b4e827c7f2c1363;4c86931c-2d83-48
+[hf] av_l17_matchinj 429/err, sleeping 90s (2): (Request ID: Root=1-6a2f5002-0884f10f429648463ee45cc8;d649b5f1-ff28-44
+[hf] av_l17_matchinj 429/err, sleeping 90s (3): (Request ID: Root=1-6a2f5081-609895703df5c06837dc8f4a;2a897b7c-6d6d-4b
+[hf] av_l17_matchinj 429/err, sleeping 90s (4): (Request ID: Root=1-6a2f5101-499f77fc119e7f2845b1baaa;9f16975f-88a8-45
+[hf] av_L23_clean 429/err, sleeping 90s (0): (Request ID: Root=1-6a2f5172-544f02ef3b9f695b5c1c25e5;38d1b22b-bc0a-4b
+[hf] av_L23_clean 429/err, sleeping 90s (1): (Request ID: Root=1-6a2f51cf-6cb1fc654e2b4a772c2e8a49;f5edd32a-8d0d-4d
+[hf] av_L23_clean 429/err, sleeping 90s (2): (Request ID: Root=1-6a2f522c-79f8f3f22a033f364e1b04dc;3db8202c-6063-42
+[hf] av_L23_clean 429/err, sleeping 90s (3): (Request ID: Root=1-6a2f528a-156eb96e45f0707935b947e0;4914aa87-c67f-49
+[hf] av_L23_clean 429/err, sleeping 90s (4): (Request ID: Root=1-6a2f52e7-669744ab5c2553cb7434660b;17903d19-98c3-49
+[hf] av_mega_L23 429/err, sleeping 90s (0): (Request ID: Root=1-6a2f535c-7ca2bb9d7905629b277cbe26;5c33906a-a884-44
+[hf] av_mega_L23 429/err, sleeping 90s (1): (Request ID: Root=1-6a2f53bd-2e53167e22ea88f67dbc3329;9cc48808-4a59-45
+[hf] av_mega_L23 429/err, sleeping 90s (2): (Request ID: Root=1-6a2f541e-3cd65fbd234e3fe554d1805e;10f700fe-f4e9-4c
+[hf] av_mega_L23 429/err, sleeping 90s (3): (Request ID: Root=1-6a2f5480-68897f3969f8bada5c51213e;af18ef6e-1902-4a
+[hf] av_mega_L23 429/err, sleeping 90s (4): (Request ID: Root=1-6a2f54e0-0c768e5923bab8340b7c2501;919dc196-01fb-40
+[hf] av_minpair_from_v01 429/err, sleeping 90s (0): (Request ID: Root=1-6a2f5569-7986f2d77628eab50781658f;adcc6432-614f-45
+[hf] av_minpair_from_v01 429/err, sleeping 90s (1): (Request ID: Root=1-6a2f55db-679a2ed62d7172e83dc10f49;0223a6e7-2d39-42
+[hf] av_minpair_from_v01 429/err, sleeping 90s (2): (Request ID: Root=1-6a2f564d-0b0492f42c0a55fc6c64bcd4;ca5e3c93-8492-45
+[hf] av_minpair_from_v01 429/err, sleeping 90s (3): (Request ID: Root=1-6a2f56c1-260b92b5708e029a7cd48c89;007c9a09-48d1-46
+[hf] av_minpair_from_v01 429/err, sleeping 90s (4): (Request ID: Root=1-6a2f5735-408c00c1170338e70e800773;b90f54c1-c2b8-49
+[hf] av_minpair_grok_wd3 429/err, sleeping 90s (0): (Request ID: Root=1-6a2f5807-2369e4370c843f5c629edae1;3a69043e-62ff-45
+[hf] av_minpair_grok_wd3 429/err, sleeping 90s (1): (Request ID: Root=1-6a2f58ba-5a2d11a079b66bd52c380242;9aaed89f-9bb8-45
+[hf] av_minpair_grok_wd3 429/err, sleeping 90s (2): (Request ID: Root=1-6a2f596b-2f6c579b654418293570f4a8;09cc7111-5f1e-4f
+[hf] av_minpair_grok_wd3 429/err, sleeping 90s (3): (Request ID: Root=1-6a2f5a1d-6fd5fc66504aa83556cd5db7;8413ab98-022f-42
+[hf] av_minpair_grok_wd3 429/err, sleeping 90s (4): (Request ID: Root=1-6a2f5acc-46aef7653cbeae2d28a75ca0;7bbf96f9-ad98-44
+[hf] av_priordev_g1_from_v01 429/err, sleeping 90s (0): (Request ID: Root=1-6a2f5b81-43dbea0a0153d8413f74f660;0b436019-7556-48
+[hf] av_priordev_g1_from_v01 429/err, sleeping 90s (1): (Request ID: Root=1-6a2f5c16-514738652796bcec61cff24d;8203ba54-73a8-4f
+[hf] av_priordev_g1_from_v01 429/err, sleeping 90s (2): (Request ID: Root=1-6a2f5cb8-18f5ebba13560e6743af676d;abaa58b5-6785-42
+[hf] av_priordev_g1_from_v01 429/err, sleeping 90s (3): (Request ID: Root=1-6a2f5d4d-6ae420c51e5d04551d102b74;5cd53b98-29e6-4c
+[hf] av_priordev_g1_from_v01 429/err, sleeping 90s (4): (Request ID: Root=1-6a2f5de2-28aa0b116cb33c771400f0f8;e70f4a30-1d5d-48
+[hf] [26/38] uploaded av_priordev_grokking_wd3
+[hf] [27/38] uploaded av_priordev_grokking_wd3_ga8
+[hf] av_priordev_uniform_from_v01 429/err, sleeping 90s (0): (Request ID: Root=1-6a2f5eb7-7b99e20b7f7a9db859e0e056;1c2d2813-fd2a-4f
+[hf] av_priordev_uniform_from_v01 429/err, sleeping 90s (1): (Request ID: Root=1-6a2f5f4c-706c0d8a302205d24839cfd2;06de89e2-e01d-44
+[hf] av_priordev_uniform_from_v01 429/err, sleeping 90s (2): (Request ID: Root=1-6a2f5fe2-0a7f01a2213c06774e916a8d;31cf2f42-b980-47
+[hf] av_priordev_uniform_from_v01 429/err, sleeping 90s (3): (Request ID: Root=1-6a2f6078-27e602533f1092ac12a7b5a4;90422a0e-d0fc-47
+[hf] av_priordev_uniform_from_v01 429/err, sleeping 90s (4): (Request ID: Root=1-6a2f610d-4376e87647c689aa146b0dc0;94590f10-e70b-48
+[hf] av_softprompt_entity 429/err, sleeping 90s (0): (Request ID: Root=1-6a2f6194-0d93d69d2e70e81e679c7774;d1b448f1-4288-42
+[hf] av_softprompt_entity 429/err, sleeping 90s (1): (Request ID: Root=1-6a2f6205-5347cc787962973f3b921c57;8551e8bb-64a3-42
+[hf] av_softprompt_entity 429/err, sleeping 90s (2): (Request ID: Root=1-6a2f6277-3ede7ff47e7b48e67934d9b6;582a6c28-7970-47
+[hf] av_softprompt_entity 429/err, sleeping 90s (3): (Request ID: Root=1-6a2f62e8-72adfcbb7829220324612c54;ce7ff04f-6274-4e
+[hf] av_softprompt_entity 429/err, sleeping 90s (4): (Request ID: Root=1-6a2f635a-00b7af093129155656282b88;2daaf8d2-db2c-4c
+[hf] av_softprompt_v01_30step 429/err, sleeping 90s (0): (Request ID: Root=1-6a2f63da-2c48d4b218b6b4d92fae2f52;3c78e559-2cea-47
+[hf] av_softprompt_v01_30step 429/err, sleeping 90s (1): (Request ID: Root=1-6a2f6442-2448437a4d82ee7277b13a42;400ef0f0-fef3-4e
+[hf] av_softprompt_v01_30step 429/err, sleeping 90s (2): (Request ID: Root=1-6a2f64aa-18982e744e1b7582348380c1;3d7ddac7-bf59-43
+[hf] av_softprompt_v01_30step 429/err, sleeping 90s (3): (Request ID: Root=1-6a2f6513-07d3036b5f15bf72499df8c5;42c43bac-a900-41
+[hf] av_softprompt_v01_30step 429/err, sleeping 90s (4): (Request ID: Root=1-6a2f657c-7294031a6cf39bc22c0d14f3;a66abe3a-4d1a-4a
+[hf] av_softprompt_v01_long 429/err, sleeping 90s (0): (Request ID: Root=1-6a2f6600-2f23259133347cb141e8b396;a0f1c128-a043-4a
+[hf] av_softprompt_v01_long 429/err, sleeping 90s (1): (Request ID: Root=1-6a2f666b-6c3cbaf61374ebc56f970f10;2788166f-875e-43
+[hf] av_softprompt_v01_long 429/err, sleeping 90s (2): (Request ID: Root=1-6a2f66d7-32351a3e0969b661764fce53;0c3102a6-90cb-42
+[hf] av_softprompt_v01_long 429/err, sleeping 90s (3): (Request ID: Root=1-6a2f6744-602e202c266d81370f3b9166;aba35260-cc30-44
+[hf] av_softprompt_v01_long 429/err, sleeping 90s (4): (Request ID: Root=1-6a2f67b0-14e5a87261cd846823e3a1b6;2ba5cb27-cd32-46
+[hf] av_stage2_A_domaware 429/err, sleeping 90s (0): (Request ID: Root=1-6a2f6827-7525c90726da50da7f5a23ea;8e806368-e2ec-42
+[hf] av_stage2_A_domaware 429/err, sleeping 90s (1): (Request ID: Root=1-6a2f6887-0001c860777df2e122209185;2a1fe90d-332b-4d
+[hf] av_stage2_A_domaware 429/err, sleeping 90s (2): (Request ID: Root=1-6a2f68e7-106609b37290d2857d67b7e1;8f681c93-f367-48
+[hf] av_stage2_A_domaware 429/err, sleeping 90s (3): (Request ID: Root=1-6a2f6948-2b00344a78716f7b0a95648f;f6750be2-5f90-49
+[hf] av_stage2_A_domaware 429/err, sleeping 90s (4): (Request ID: Root=1-6a2f69aa-3bd277d26766a37346d8695e;609bd285-1f49-49
+[hf] av_v0_1_aux_readout 429/err, sleeping 90s (0): (Request ID: Root=1-6a2f6a22-7c4eee856e53cd7e5aa6e74f;6ab13bb3-d36f-44
+[hf] av_v0_1_aux_readout 429/err, sleeping 90s (1): (Request ID: Root=1-6a2f6a85-3c7495d7213471b35299af14;7d1d61c7-d38f-41
+[hf] av_v0_1_aux_readout 429/err, sleeping 90s (2): (Request ID: Root=1-6a2f6aea-01d9a7443c7df7b6623d446a;89ca181e-da60-4a
+[hf] av_v0_1_aux_readout 429/err, sleeping 90s (3): (Request ID: Root=1-6a2f6b4d-041290ff26f566d0617c170c;e3305907-2429-4a
+[hf] av_v0_1_aux_readout 429/err, sleeping 90s (4): (Request ID: Root=1-6a2f6bb2-7017c18d2029c5240d8c8701;8719c545-1332-45
+[hf] av_v0_1_dd_inj39_bf16_long 429/err, sleeping 90s (0): (Request ID: Root=1-6a2f6c53-2363fe2d70b30fee05ae7292;7af56057-cdf2-48
+[hf] av_v0_1_dd_inj39_bf16_long 429/err, sleeping 90s (1): (Request ID: Root=1-6a2f6cda-7a6062554e6bb4d570c7a103;bfba8d5e-90d2-48
+[hf] av_v0_1_dd_inj39_bf16_long 429/err, sleeping 90s (2): (Request ID: Root=1-6a2f6d63-5dd44fda511fce607fb4a031;9043a49b-fd3e-4d
+[hf] av_v0_1_dd_inj39_bf16_long 429/err, sleeping 90s (3): (Request ID: Root=1-6a2f6dea-4afe73ed11093df20ed4841b;471cd84d-f7cd-49
+[hf] av_v0_1_dd_inj39_bf16_long 429/err, sleeping 90s (4): (Request ID: Root=1-6a2f6e73-262ac25b50c3dbdf36ab6919;8fd2922d-82f2-43
+[hf] av_v0_1_rft 429/err, sleeping 90s (0): (Request ID: Root=1-6a2f6ef4-09728baa29f595bb217434c9;3b2ad9df-86dd-42
+[hf] av_v0_1_rft 429/err, sleeping 90s (1): (Request ID: Root=1-6a2f6f5d-18e159bf011655cc41b3420a;c582509a-4752-49
+[hf] av_v0_1_rft 429/err, sleeping 90s (2): (Request ID: Root=1-6a2f6fcc-23a71d0d1cb7f48725fe97d5;dc4a66f4-bbe0-4a
+[hf] av_v0_1_rft 429/err, sleeping 90s (3): (Request ID: Root=1-6a2f703a-74d44651320f464c207b56aa;b1b00b5a-b274-48
+[hf] av_v0_1_rft 429/err, sleeping 90s (4): (Request ID: Root=1-6a2f70a5-0fb2a1a945d596e14fe479c2;eb7904f7-69f0-43
+[hf] av_v0_1to3_grok 429/err, sleeping 90s (0): (Request ID: Root=1-6a2f711e-3179f5c9111660f452045dde;d0caa02e-9932-4b
+[hf] av_v0_1to3_grok 429/err, sleeping 90s (1): (Request ID: Root=1-6a2f7181-2c0248b3453ec516317aaafe;491e9d93-5b2c-47
+[hf] av_v0_1to3_grok 429/err, sleeping 90s (2): (Request ID: Root=1-6a2f71e5-5cb0125c7c5b64e16c36d2ac;13407c64-078d-48
+[hf] av_v0_1to3_grok 429/err, sleeping 90s (3): (Request ID: Root=1-6a2f7248-500e51752f52c1602693b667;fc687b1a-9c61-47
+[hf] av_v0_1to3_grok 429/err, sleeping 90s (4): (Request ID: Root=1-6a2f72ab-2e72f3514252fd1c4bab01a0;3ad0cc4d-e956-4a
+[hf] av_v0_4_fineweb_from_v01 429/err, sleeping 90s (0): (Request ID: Root=1-6a2f7324-430a026617b8cec556ae4726;b2bb2de6-d93c-4f
+[hf] av_v0_4_fineweb_from_v01 429/err, sleeping 90s (1): (Request ID: Root=1-6a2f7388-0154564e25fc793534f39ae5;09295f7d-3025-42
+[hf] av_v0_4_fineweb_from_v01 429/err, sleeping 90s (2): (Request ID: Root=1-6a2f73ea-6a0dd30b0a37943e6e760b33;de17c26e-f5aa-4b
+[hf] av_v0_4_fineweb_from_v01 429/err, sleeping 90s (3): (Request ID: Root=1-6a2f744e-2e65cce8480e17104cb33c2a;26a0eb68-ba8c-43
+[hf] av_v0_4_fineweb_from_v01 429/err, sleeping 90s (4): (Request ID: Root=1-6a2f74b1-3e5b4e8c600c8cee501f74f6;ea9f5a33-a4b2-4a
+[hf] channel_probe_r8 429/err, sleeping 90s (0): (Request ID: Root=1-6a2f7520-42da970d3e87e7de7924d869;0b7fb1f6-59d5-4e
+[hf] channel_probe_r8 429/err, sleeping 90s (1): (Request ID: Root=1-6a2f757b-2ff348b463351b6253d28474;38fa948f-9cfb-46
+[hf] channel_probe_r8 429/err, sleeping 90s (2): (Request ID: Root=1-6a2f75d6-5c22eec077f5a5b70c40b01d;54e0c959-2f9b-49
+[hf] channel_probe_r8 429/err, sleeping 90s (3): (Request ID: Root=1-6a2f7631-041f20b364119769342cc320;2efb8069-a35a-4b
+[hf] channel_probe_r8 429/err, sleeping 90s (4): (Request ID: Root=1-6a2f768c-5c3fda922be6600c03b99f00;decc3263-95bb-44
+[hf] DONE -> https://huggingface.co/Solshine/nla-gemma4e2b-research-checkpoints

hf_upload4.log

@@ -0,0 +1,13 @@
+[hf] private repo: Solshine/nla-gemma4e2b-research-checkpoints  <- C:\Users\caleb\deception-nanochat-sae-research\experiments\v8_nla_local\checkpoints
+[hf] av_stage2_A_domaware 429/err, sleeping 120s (0): (Request ID: Root=1-6a2f77ab-4d1b679e042aed6e260538c7;b6dcc0b2-5dff-42
+[hf] av_stage2_A_domaware 429/err, sleeping 120s (1): (Request ID: Root=1-6a2f7828-5951a8d57e5d8e056f938308;23e9b40e-94d6-48
+[hf] av_stage2_A_domaware 429/err, sleeping 120s (2): (Request ID: Root=1-6a2f78a8-3d2d5b77341a7c9d08b0708b;f0a418e9-e657-4b
+[hf] av_stage2_A_domaware 429/err, sleeping 120s (3): (Request ID: Root=1-6a2f7927-7a580c1d6fdb4c65744978ed;0235ae77-2cf3-40
+[hf] av_stage2_A_domaware 429/err, sleeping 120s (4): (Request ID: Root=1-6a2f79a6-429658f91893829f22cb422f;4811617e-1147-44
+[hf] av_mega_L23 429/err, sleeping 120s (0): (Request ID: Root=1-6a2f7a2e-71d94bb442c032d619399f69;68cfc17a-615e-43
+[hf] av_mega_L23 429/err, sleeping 120s (1): (Request ID: Root=1-6a2f7aab-0b768a27262434ac6120392c;922860e6-73e9-4d
+[hf] av_mega_L23 429/err, sleeping 120s (2): (Request ID: Root=1-6a2f7b2a-73e498945f8bceb14b3b4c7f;bd688108-f64f-4c
+[hf] av_mega_L23 429/err, sleeping 120s (3): (Request ID: Root=1-6a2f7ba8-1d19d4c552fc631c3f0a021f;bbeae7a6-f725-4e
+[hf] av_mega_L23 429/err, sleeping 120s (4): (Request ID: Root=1-6a2f7c26-4deb5ecc0a7cfd1c0d71eb72;0c9cc8e9-eb1a-42
+[hf] av_L23_clean 429/err, sleeping 120s (0): (Request ID: Root=1-6a2f7caa-0f9d81d91dcf7a590a3c9d29;ddb4238a-97c2-43
+[hf] av_L23_clean 429/err, sleeping 120s (1): (Request ID: Root=1-6a2f7d24-17370aa53f903d305eb12db6;c7cd9123-f7de-48

sp_entity_evals.log

@@ -1,29 +1,29 @@
-===== SOFT-PROMPT ENTITY TRAJECTORY (in-domain n=50, vs v0.1 0.140) =====
---- step_0050 (10:15) ---
-[sp-eval] sp_av_softprompt_entity_step_0050_on_indomain_eval_cmp: n_unique=7/50 tfidf=0.100(p=0.1514) semantic=0.040(p=0.8386)  vs v0.1 in-domain 0.140
-   out: '<explanation>\nUS government agency\n</explanation>\n</explanation>\n</explanation>\n</explanation>\n</explanation>\n'
-   out: '<explanation>\nUS Army Corps of Engineers\n</explanation>\n</explanation>\n</explanation>\n</explanation>\n</explana'
-   out: '<explanation>\nLGBT rights in the Philippines\n</explanation>\n</explanation>\n</explanation>\n</explanation>\n</exp'
---- step_0100 (10:38) ---
-[sp-eval] sp_av_softprompt_entity_step_0100_on_indomain_eval_cmp: n_unique=6/50 tfidf=0.120(p=0.0025) semantic=0.060(p=0.7811)  vs v0.1 in-domain 0.140
-   out: '<explanation>\nnon-linear optical phase measurement\n</explanation>\n</explanation>\n</explanation>\n</explanation>'
-   out: '<explanation>\nhigh-pressure gas transmission\n</explanation>\n</explanation>\n</explanation>\n</explanation>\n</exp'
-   out: '<explanation>\nnon-linear optical phase modulation\n</explanation>\n</explanation>\n</explanation>\n</explanation>\n'
---- step_0150 (11:00) ---
-[sp-eval] sp_av_softprompt_entity_step_0150_on_indomain_eval_cmp: n_unique=1/50 tfidf=0.080(p=0.0625) semantic=0.080(p=0.0610)  vs v0.1 in-domain 0.140
-   out: '<explanation>\nCME magnetic flux emergence\n</explanation>\n</explanation>\n</explanation>\n</explanation>\n</explan'
-   out: '<explanation>\nCME magnetic flux emergence\n</explanation>\n</explanation>\n</explanation>\n</explanation>\n</explan'
-   out: '<explanation>\nCME magnetic flux emergence\n</explanation>\n</explanation>\n</explanation>\n</explanation>\n</explan'
---- step_0200 (11:23) ---
-[sp-eval] sp_av_softprompt_entity_step_0200_on_indomain_eval_cmp: n_unique=1/50 tfidf=0.080(p=0.0590) semantic=0.080(p=0.0610)  vs v0.1 in-domain 0.140
-   out: '<explanation>\ndark matter halo mass function\n</explanation>\n</explanation>\n</explanation>\n</explanation>\n</exp'
-   out: '<explanation>\ndark matter halo mass function\n</explanation>\n</explanation>\n</explanation>\n</explanation>\n</exp'
-   out: '<explanation>\ndark matter halo mass function\n</explanation>\n</explanation>\n</explanation>\n</explanation>\n</exp'
---- step_0250 (11:45) ---
-[sp-eval] sp_av_softprompt_entity_step_0250_on_indomain_eval_cmp: n_unique=3/50 tfidf=0.080(p=0.0625) semantic=0.080(p=0.3158)  vs v0.1 in-domain 0.140
-   out: '<explanation>\nanti-drone regulation argument\n</explanation>\n</\n</\n</\n</\n</\n</\n</\n</\n</\n</\n</\n</\n</\n</\n</\n</\n</'
-   out: '<explanation>\nanti-drone regulation argument\n</explanation>\n</\n</\n</\n</\n</\n</\n</\n</\n</\n</\n</\n</\n</\n</\n</\n</\n</'
-   out: '<explanation>\nanti-drone regulation argument\n</explanation>\ndrone regulation argument\n</explanation>\n</explana'
-===== V0.1 BASELINE BIGGER-N (public number, n=160, 40 docs) (12:14) =====
-[checkpoints_av_v0_1_dd_inj39_bf16_long_step_000250_on_indomain_eval_cmp40] n_unique=89/160  tfidf top1=0.056 (chance 0.025, p=0.0102)  semantic top1=0.050 (p=0.0348)
-===== EVAL CHAIN DONE (13:05) =====
+===== SOFT-PROMPT ENTITY TRAJECTORY (in-domain n=50, vs v0.1 0.140) =====
+--- step_0050 (10:15) ---
+[sp-eval] sp_av_softprompt_entity_step_0050_on_indomain_eval_cmp: n_unique=7/50 tfidf=0.100(p=0.1514) semantic=0.040(p=0.8386)  vs v0.1 in-domain 0.140
+   out: '<explanation>\nUS government agency\n</explanation>\n</explanation>\n</explanation>\n</explanation>\n</explanation>\n'
+   out: '<explanation>\nUS Army Corps of Engineers\n</explanation>\n</explanation>\n</explanation>\n</explanation>\n</explana'
+   out: '<explanation>\nLGBT rights in the Philippines\n</explanation>\n</explanation>\n</explanation>\n</explanation>\n</exp'
+--- step_0100 (10:38) ---
+[sp-eval] sp_av_softprompt_entity_step_0100_on_indomain_eval_cmp: n_unique=6/50 tfidf=0.120(p=0.0025) semantic=0.060(p=0.7811)  vs v0.1 in-domain 0.140
+   out: '<explanation>\nnon-linear optical phase measurement\n</explanation>\n</explanation>\n</explanation>\n</explanation>'
+   out: '<explanation>\nhigh-pressure gas transmission\n</explanation>\n</explanation>\n</explanation>\n</explanation>\n</exp'
+   out: '<explanation>\nnon-linear optical phase modulation\n</explanation>\n</explanation>\n</explanation>\n</explanation>\n'
+--- step_0150 (11:00) ---
+[sp-eval] sp_av_softprompt_entity_step_0150_on_indomain_eval_cmp: n_unique=1/50 tfidf=0.080(p=0.0625) semantic=0.080(p=0.0610)  vs v0.1 in-domain 0.140
+   out: '<explanation>\nCME magnetic flux emergence\n</explanation>\n</explanation>\n</explanation>\n</explanation>\n</explan'
+   out: '<explanation>\nCME magnetic flux emergence\n</explanation>\n</explanation>\n</explanation>\n</explanation>\n</explan'
+   out: '<explanation>\nCME magnetic flux emergence\n</explanation>\n</explanation>\n</explanation>\n</explanation>\n</explan'
+--- step_0200 (11:23) ---
+[sp-eval] sp_av_softprompt_entity_step_0200_on_indomain_eval_cmp: n_unique=1/50 tfidf=0.080(p=0.0590) semantic=0.080(p=0.0610)  vs v0.1 in-domain 0.140
+   out: '<explanation>\ndark matter halo mass function\n</explanation>\n</explanation>\n</explanation>\n</explanation>\n</exp'
+   out: '<explanation>\ndark matter halo mass function\n</explanation>\n</explanation>\n</explanation>\n</explanation>\n</exp'
+   out: '<explanation>\ndark matter halo mass function\n</explanation>\n</explanation>\n</explanation>\n</explanation>\n</exp'
+--- step_0250 (11:45) ---
+[sp-eval] sp_av_softprompt_entity_step_0250_on_indomain_eval_cmp: n_unique=3/50 tfidf=0.080(p=0.0625) semantic=0.080(p=0.3158)  vs v0.1 in-domain 0.140
+   out: '<explanation>\nanti-drone regulation argument\n</explanation>\n</\n</\n</\n</\n</\n</\n</\n</\n</\n</\n</\n</\n</\n</\n</\n</\n</'
+   out: '<explanation>\nanti-drone regulation argument\n</explanation>\n</\n</\n</\n</\n</\n</\n</\n</\n</\n</\n</\n</\n</\n</\n</\n</\n</'
+   out: '<explanation>\nanti-drone regulation argument\n</explanation>\ndrone regulation argument\n</explanation>\n</explana'
+===== V0.1 BASELINE BIGGER-N (public number, n=160, 40 docs) (12:14) =====
+[checkpoints_av_v0_1_dd_inj39_bf16_long_step_000250_on_indomain_eval_cmp40] n_unique=89/160  tfidf top1=0.056 (chance 0.025, p=0.0102)  semantic top1=0.050 (p=0.0348)
+===== EVAL CHAIN DONE (13:05) =====